Aura-36 // The Story So Far — The Keel

In the dry season the salar had no distance in it.

Teo had grown up on the flat and it still caught her. You looked out and there was white, and then more white, and then a line where the white stopped, and the line could be four kilometres away or forty and nothing in between would tell you which. There were no trees. There were no rocks. There was nothing on the ground with a known size that your eye could measure against, and so your eye simply refused, and gave you a horizon with no number attached to it.

Tourists found this beautiful. Teo found it beautiful too, on maybe nine mornings a year, and the rest of the time it was the reason a walk that should take twenty minutes took an hour and a half.

She had come out at four to open the pond gates before the heat, and the sun was coming up behind her, and the salt was going from grey to a hard blue-white that would be unbearable by eight. Under her boots the crust was cracked into the hexagons that everyone photographed — rough-edged plates about a metre across, ridged where the salt had pushed up along the seams, going on unbroken past anywhere she had ever walked.

Her grandmother had worked this field. Her grandmother's brother had cut blocks out of it with a hand saw and sold them for building, back when that was still a thing people did, and the house Teo slept in on the edge of Colchani had four of those blocks in its western wall, and in the wet months they wept.

The work itself was not romantic. You pumped brine up out of the crust into shallow ponds, and you waited, and the sun did the rest, and months later there was a crust of salts at the bottom of the pond that somebody else took away in a truck. Teo opened gates and closed gates and read a hydrometer and wrote numbers on a clipboard that went into an office in Uyuni where, as far as she could tell, nobody read them.

She wrote the numbers down accurately anyway. This was, though she would not have put it this way at twenty-six, the single fact about her that mattered.

The survey crew had been out on the flat for nine days.

She had noticed them the first morning because you notice anything on the salar; a vehicle at ten kilometres is the only event in the entire visual field. Four trucks, white, no company name on them, which was not unusual — half the outfits that came through here were subcontracted to somebody who was subcontracted to somebody.

What was unusual was what they were doing, and it took her until the fourth day to work out that she did not recognise it.

Teo knew what mining survey looked like. She had watched four companies do it since she was fifteen. They came, they took brine samples on a grid, they drilled shallow cores, they measured the lithium and then they measured the magnesium and then — and this part was reliable enough that the men in Colchani made jokes about it — they went quiet, and then they left. The magnesium was the joke. The magnesium was always the joke.

This crew took no brine samples at all.

They were surveying the ground. Not what was under it. The surface itself: instruments on tripods, sighting to targets held up by a man who walked out until he was a speck and then kept going, and then a wait, and then he moved and they did it again. They had been doing it for nine days and they had moved perhaps eleven kilometres.

On the sixth day one of them — younger than her, sunburnt in the specific way of somebody who had not believed the altitude — walked over to where she was reading the hydrometer and asked, in careful Spanish, whether the water came every year.

"Every year," Teo said.

"How deep?"

"Twenty centimetres. Thirty at Tunupa." She looked at him. "December to March. Sometimes April."

He wrote it down, and then he said, "And it's flat afterwards? Every time?"

And that was the question that stayed with her, because it was not a question about lithium.

"Yes," she said. "That's why it's flat."

He looked up at her.

"The water lies," she said, and made the gesture, her hand going out level. "All of it, everywhere, all at once. Then it dries. Wherever it was high, it takes off. Wherever it was low, it leaves. Every year." She shrugged. "It fixes itself. It's been fixing itself for a very long time."

The young man wrote that down too, and thanked her, and went back, and she saw him say something to the woman at the instrument, and the woman looked over at Teo for about two seconds and then went back to her eyepiece.

Teo had told four foreign companies about the flood. Every one of them had taken it as a problem. This crew had taken it as an answer, and she noticed the difference at the time, and did not know what to do with it, and put it in the same place she put everything she noticed and could not use.

Here is the thing she already knew, because everyone in Colchani knew it, and because it was the one fact about home that had ever made her feel that home was important.

The satellites used this ground.

Not to look at — to measure themselves against. There were instruments in orbit whose whole job was to bounce a signal off the Earth and time how long it took, and to know whether that instrument was lying to you, you had to point it at something whose height you already knew perfectly. There was nowhere else. Ten and a half thousand square kilometres, and from one horizon to the other the ground did not rise or fall by a metre. Not a metre over a hundred and twenty-six kilometres. Nowhere on the planet came close, and Teo had known this since a schoolteacher told her class at eleven, and it had lodged in her the way a fact does when it is the first time somebody suggests your dirt is worth something to a person who has never been there.

So when a crew came out and spent nine days not sampling the brine and instead measuring the levelness of the levelest place there is, and asking whether it re-levelled itself after the flood, and writing down that it did —

— she assumed it was a mine.

She wants that in the record, she would say later, much later, to people who had begun to treat her as somebody who had understood. She did not understand. Nobody in Colchani understood, and nobody in Uyuni understood, and the men in the office who had the actual contract in a drawer did not understand either, because it had been described to them in a document that was accurate and complete and deliberately extremely boring.

Everyone assumed it was a mine. It was the only thing it had ever been.

On the ninth day the stakes started.

Teo came out at four and there was a line of them where there had been nothing, orange plastic flags on wire, going away west into the dark at intervals she paced out at twenty metres. She walked along them for a while because they were new and because there was nothing else to look at, and she got perhaps a kilometre and a half before the cold in the wind turned her round.

The line had not curved. That was the part she thought about while she opened the gates. A mine had a shape; a mine was a pit, or a grid of ponds, or a road that went to somewhere. A line of stakes twenty metres apart, running dead straight west, and still running when you gave up and turned round — that was not a shape. That was a direction.

She went back at the end of the shift, in the flat white glare of noon with her eyes narrowed to slits, and followed it further, and it did not curve then either.

The sunburnt young man was at the far end of what they had done that morning, holding the target pole, and when she came up he lifted a hand.

"How long does it go?" she said.

He hesitated, and she watched him decide that the answer was not secret, only stupid-sounding, which she came to understand years afterward was the exact position every single person on that project occupied for the first four years.

"Thirty-two kilometres," he said.

Teo looked west along the flags. At twenty metres apart, that was sixteen hundred stakes. She had walked past forty of them.

"For what?"

"I'm not the person to ask," he said, which was true, and then, because he was twenty-three and it was the most interesting thing he had ever been near, he added: "It has to be straight. That's the whole job. Everything else is easy and that's the part that isn't." He glanced back at the woman on the instrument. "There's nowhere else it works."

"Nowhere else in Bolivia," Teo said.

"No," he said. "Nowhere else."

She went home. She ate with her mother and did not mention it, because there was nothing to mention: a line of flags, a foreign crew, a number that sounded made up. In four months the water would come and take the stakes and the flat would lie under twenty centimetres of sky and everything would stop until April, the way everything always stopped until April, and whoever these people were they would find that out the same way the last four had.

That is what she thought on the ninth day.

She was wrong about all of it except the water, and she was right about the water for nine years, and by the tenth they had learned to build around it, and by the fourteenth Teo was the person who told four hundred thousand people when to stop working.

The bus from Oruro took six hours and Calder spent four of them with a suitcase on his knees because the rack above him held a crate of chickens and he had done the arithmetic on the strap.

He was thirty-four. He had one suitcase, a folder of paper inside it, and enough money to be in Bolivia for eleven days. Nothing about him read as a man who was going to build anything. He was not tall and he was not well dressed and he had the pallor of somebody who had spent a decade in rooms; the woman across the aisle had offered him a coca leaf two hours out of Oruro, on the correct assumption that he was going to be sick, and he had thanked her and taken it and been sick anyway, quietly, at the stop.

What he had was a patent, and the patent was good.

He was not humble about it. This is the thing to understand about him early, because it never changed and because every account written afterward gets it wrong in one of two directions — either he was a visionary who saw further than everyone, or he was an operator who fooled a poor country. He was neither. He was a man who was usually correct and who had spent eleven years discovering that being correct is a completely separate skill from being believed, and that he had the first one and not the second, and that nobody was going to give him the second as a gift.

He had been turned down in Zurich, twice in Houston, and in a conference room in Santiago by a man who had not finished reading the summary. He did not consider any of them stupid. He had gone back over the Santiago meeting for a week and concluded that the failure was his: the summary had opened with the physics, and the physics was the least interesting thing about it to a person holding money, and he had known that going in and had done it anyway because the physics was the part he liked.

He did not make that mistake again. That is the whole of what those eleven years taught him, and it cost him his thirties.

Uyuni at four in the afternoon was cold in the shade and hot in the sun and the wind came down the street with nothing to slow it. He found a room for eleven bolivianos, put the suitcase under the bed, and walked out to the edge of town where the rail yard was, because he wanted to see the ground before he saw anybody.

The train cemetery is out there — forty or fifty locomotives rusting on the flat where a company left them when the mines went quiet a century ago. He walked among them for a while. Tourists came to photograph them; he was not photographing them. He was looking at the corrosion and thinking about salt, and the thought he had, standing there, was the first sentence of the entire project, and it was not visionary at all. It was: whatever I build here, this is what the air does to it, and I will not be told this later.

Then he walked out past the last of them onto the flat itself, maybe a kilometre, until the town behind him had gone small and the white ahead had stopped having a distance to it.

He stood there for about forty minutes.

He never described what he thought about. Not to the committees, not to the ten, not in the forty years of interviews afterward, and when Teo asked him directly — once, much later, in a corridor, in the way she asked everything — he said, "I was checking the wind," which was almost certainly true and almost certainly not all of it.

The next morning he went to the office of the state lithium authority and was not seen.

He had expected that. He sat in the corridor for five hours with the folder on his knees, and at the end of it a man came out and told him, politely, that the schedule was full, and that he should write ahead, and that many companies came.

"How many?" Calder said.

The man paused, because it was not the question people asked. "Four this year."

"And how many are still here?"

The man did not answer that, which was the answer, and something in his face went from official to human for about a second and a half.

"The magnesium," Calder said.

"The magnesium," the man agreed.

Calder nodded, and thanked him, and left, and did not attempt to hand him the folder, and that restraint — not the patent, not the physics — is the reason there is a ship.

He spent the next four days not talking to anybody who could say yes.

He talked to a hydrologist at the small technical college who had published nine papers on the salar and had never once been contacted by a mining company. He talked to two brine foremen at Colchani, and bought them lunch, and asked them what the last four companies had got wrong; both of them told him the same thing about magnesium and both of them told him the same thing about the water, and the second thing was not in any of the four technical reports he had read.

He talked to a man who repaired pumps, who told him that the problem with foreigners was that they specified stainless and stainless was a joke here, and Calder said, "What should they specify?" and the man laughed, and then realised the question was serious, and told him, and Calder wrote it in the folder.

He ate in the same place every night because it was cheap and because the owner had stopped charging him for the soup on the third night without discussing it, which Calder noticed and did not mention, because mentioning it would have made it a transaction.

On the fourth night a geologist from the technical college sat down opposite him — the hydrologist had sent him, which Calder had not asked for and which was the first thing anyone in the country did for him without being asked — and spent two hours explaining, with real pleasure, why the salar could not be industrialised.

Calder let him finish. He had a habit of letting people finish that read as courtesy and was actually acquisition; he was significantly better at listening than he was at anything social, and the two got confused for each other his whole life.

Then he said, "You're right about all of it."

The geologist waited for the but.

"No," Calder said. "You're right. Everything you said is correct. Four companies came and did the arithmetic and the arithmetic said leave, and they left, and they were not fools." He turned his glass around on the table. "The question isn't whether the arithmetic is right. It's what the arithmetic is for. They were all solving for lithium at market price. That problem has no answer here and I agree with you and I'm not going to try."

"Then what are you solving for?"

And Calder — who had, four days earlier, sat in a corridor for five hours and been turned away, and who would in nine years be one of the most consequential people alive — said, "Flatness," and would not be drawn any further, and the geologist went home and told his wife he had met a lunatic.

He told that story at Calder's hundredth birthday. It got a very large laugh. He was ninety-one and it was the last time the two of them were in a room together and neither of them knew that, because nobody ever does.

On the fifth day he sat in his room and rewrote the summary.

The old one opened with resonant transmission and had a diagram on page two. He tore the front off it. The new one opened on page one with the number of Bolivian jobs, and on page two with the constitutional position on state majority ownership, quoted correctly, agreed to in advance, in a single sentence with no conditions attached to it. The physics went to page nine.

He kept the diagram. He put it at the back, where a person who wanted it would find it and a person who did not would never have to look at it.

He was aware, doing this, that he was hiding the best thing he had. He wrote in the margin of his own working copy — and this copy survives, and is in a case, and four generations of schoolchildren have walked past it — a single line:

Page nine. Remember that it is nine because they made it nine, not because it is worth less.

It is the earliest surviving thing he wrote about the project and it is a note about resentment, and every biography that quotes it treats it as modesty.

On the sixth day he went back and sat in the corridor again.

This time he had a different folder and a different first page, and he sat there for three hours, and when the same man came out Calder stood up and said, before anything else, "Eleven hundred jobs in the first four years. Majority state ownership, agreed, not negotiable from my side. I'm not asking for a meeting today. I'm asking you to put four pages in front of somebody, and if they say no I'll go home."

The man looked at the four pages.

"You said you'd go home," he said. "Will you?"

"No," Calder said.

And the man laughed — genuinely, the first uncalculated thing that had happened to Calder in eleven days — and took the pages.

That is how it started. Not a demonstration, not a vision, not a man seeing further than anybody. A rewritten summary with the physics buried on page nine, handed to a middle-ranking official in a corridor by somebody who had finally, at thirty-four, learned to lead with what the other person needed rather than with what he found interesting.

He was, by then, entirely correct about the technology. He had been entirely correct about it for six years.

Nobody had asked.

The meeting he had waited eleven days for lasted fifty minutes, and Calder said almost nothing in the first thirty-five of them.

There were six people on the other side of the table. He had expected three. That, he understood immediately, was not hostility; it was that four pages promising eleven hundred jobs had been circulated, and circulation is how an office protects itself, and now everyone who could be blamed later was in the room.

They began by explaining the conditions to him.

He had read all of them. He had read the relevant article of the constitution in the original Spanish and then in a translation and then in the Spanish again because the translation had softened a verb. He knew the position better than two of the people explaining it, and he sat and let them explain it, and wrote in his folder, and asked one clarifying question about a definition, and it was a good question and the lawyer at the end of the table looked up.

The conditions were these, and none of them were negotiable, and every foreign company that had come through that door had spent its first meeting attempting to negotiate them.

Majority state ownership of all mineral production, constitutionally fixed. Congressional approval on any major contract, which meant a vote, which meant a delay measured in seasons and not in weeks. No access to international arbitration — a dispute would be heard in Bolivian courts under Bolivian law, and there was no clause available anywhere that changed that. Industrialisation to occur in-country: the value had to be added here, not shipped out as brine to be finished somewhere with better roads.

The man in the centre finished, and there was the pause that Calder had been told about by two separate people in two separate countries — the pause where the foreigner explains why one of these will need to be adjusted for the investment to be viable.

Calder said, "Yes."

The lawyer at the end of the table said, "To which?"

"All of them," Calder said. "As written. I don't want an exemption from any of the four and I'm not going to ask for one later. If you want it in the first clause rather than the last I'd prefer that, actually."

Nobody said anything for a moment.

"You understand," the lawyer said carefully, "that the arbitration position is what the others could not accept."

"I understand that it's what they said they could not accept."

"You think they were lying."

"No," Calder said. "I think they were solving a different problem. If you are extracting a commodity and selling it at a market price, arbitration is the only thing standing between you and a change of government. That's a real fear and I'm not going to pretend it isn't." He put his hands flat on the folder. "I'm not extracting a commodity."

"The contract says lithium."

"The contract says lithium," Calder agreed, "and there will be lithium, at the volumes on page three, and the state will own the majority of it, and I will never own more than the constitution permits and I don't want to. What I want is the ground."

He had rehearsed the next part. He was aware that it was the single moment on which the whole thing turned, and he had rehearsed it in a room for eleven bolivianos a night, out loud, and had cut it four times, and what survived was ninety seconds long.

"There is a surface out there," he said, "that is flatter over a hundred and twenty-six kilometres than any other large surface on this planet. Not the flattest in South America. The flattest, full stop, and it isn't close, and it re-levels itself every year when the water lies on it, so it will still be flat in fifty years when everything anybody builds has settled. That is not a mining asset. Nobody has ever paid you for it, because until recently there was nothing you could do with it except calibrate satellites for free."

He looked down the table.

"I want to lease the flatness. The lithium pays for it. That's the arrangement. If the lithium works you get an industry you have wanted for forty years and the majority of it is yours by law. If the lithium fails, you have lost nothing, because I will have spent my money and gone home and the ground will still be exactly where it is."

"And what do you build on it?"

"A track," Calder said.

"A railway."

"No."

He did not elaborate, and they let him not elaborate, and he has been criticised for that for a hundred and forty years by people who were not in the room. What he said afterwards, when Ines asked him — and she asked him formally, for the record, in the fourth year, because she thought the record deserved it — was: "They didn't ask a second time. If they had asked a second time I would have told them. Nobody has ever believed that and I don't know how to make it more true."

The lawyer came out after him.

She was perhaps fifty and she had said four things in fifty minutes and all four had been the only questions worth asking, and Calder had already privately decided that if the project happened she was the person he most needed and least deserved.

She did not sit down. She stood on the step with her arms folded against the wind and said, "You are going to be told yes."

"By you?"

"By the process. It will take fourteen months and you will think several times that it has died." She looked out at the street. "I am telling you now because in fourteen months you will have decided we were stringing you along, and we will not have been, and it matters to me that you know the difference."

"Why does it matter to you?"

"Because four companies came and every one of them left calling us impossible," she said. "We are not impossible. We are slow, on purpose, because the last time we were fast a foreign company took the tin and left us the holes. You have agreed to the conditions that exist to stop that. So you will get your yes." She turned to look at him properly. "And when you have it, you will begin to find the conditions expensive, and you will come back to this office and ask for one small adjustment. They all do. I want you to remember that I told you on the step, before anything, that the adjustment is the thing I will refuse."

Calder thought about that for a while.

"Put it in writing," he said.

"It is in writing. It is the constitution."

"No — put the other half in writing. That I said yes to all four, in the first meeting, with no reservation, before I had anything." He was working it out as he said it, which was rare for him and which she noticed. "Minute it. Not for your protection. For mine. In four years I am going to be under pressure from people with money to go back on this, and the most useful thing I could possibly have in that room is a document from the first meeting saying I already agreed and knew what I was agreeing to."

She looked at him for a long moment.

"You want us to bind you."

"I want the version of me that turns up in four years to have already lost the argument."

It went in the minutes. It is four lines. It is the first instance in the entire history of this world of a person deliberately building a constraint on his own future authority and lodging it with somebody who could not be persuaded to lift it — which is either the founding act of everything good about Aura-36, or, depending on which of the nine books you have just finished, the founding act of the exact opposite.

The fourteen months were fourteen months, as she had said.

The meeting ended with a schedule for a technical submission and no commitment of any kind.

He walked out into the wind and sat down on the step and found that his hands were not steady, which surprised him, because he had thought he was calm and he had been wrong about that for fifty minutes.

Here is what he had actually done, and it took the people in that room about two years to see the shape of it, and by then it was not a trick, it was simply the arrangement they had all been working under and mostly liked.

Every previous company had treated the Bolivian conditions as friction — cost to be minimised, clauses to be softened, risk to be insured against. Calder treated them as fixed, the way you treat gravity, and then designed around them. Accepting congressional approval meant building a schedule that could absorb a vote. Accepting Bolivian jurisdiction meant he never wrote a contract he would want to escape from. Accepting in-country industrialisation meant the workforce was local by design and not by concession, which is the reason four hundred thousand people eventually came, and the reason the town that grew on the flat was a town and not a camp.

He gave up every protection a foreign investor would have insisted on, and in exchange he got the only thing he actually needed, which was permission to be there for thirty years.

It was not generosity. He said so himself, bluntly, when a journalist tried to make it generosity in the ninth year: "I gave away the things I could not use to buy the one thing I could not do without. That's not ethics, it's a trade, and I would have made it in any country that had the ground."

The clause about arbitration is still in the founding instruments of Aura-36. Not the substance — there is no Bolivia and there are no courts — but the shape of it: the idea that a party who holds power should have no venue of appeal that the other party cannot reach. Renata Oyelaran, drafting the Open Accord in a compartment off a forward transfer some ninety years later, would have said she derived it from first principles.

She derived it from a contract, and the contract was signed on a Tuesday, and the man who agreed to every condition in the room did it because he wanted a surface and had nothing else to trade.

They hired Teo in the second week of the dry season and nobody told her what for.

The job was called local liaison on the paper she signed, and what it meant in practice was that she went out with the survey crew at five and came back at six and made sure they did not die. This was not a joke and she did not treat it as one. In her first four days she stopped a man from drinking two litres of water and nothing else in nine hours, took a set of keys off a driver who had decided he could see the track well enough in the glare, and told a woman from Munich, twice, to put the hat back on.

The woman from Munich put the hat back on the second time. She was the one on the instrument, and her name was on the paperwork as the survey lead, and she was the only person on the crew who had understood the altitude before arriving.

"You've been up here before," Teo said.

"Atacama. Two years." The woman did not look up from the eyepiece. "The Chileans told me the same thing about the hat and I ignored them and I was in a clinic for a day and a half. You only need to learn that once."

Teo decided she liked her, provisionally.

The pay was two and a half times the brine field and she had not asked for it; it had been the number on the paper. She had spent the walk home from signing trying to work out what was wrong with that, and had not found it, and had spent the evening not telling her mother the figure because saying it out loud would have made it into something they were relying on.

What she learned in the first month, she learned by being present and asking nothing, which is a skill people who have been talked past their whole lives develop early.

She learned that the crew were not looking for a deposit, because nobody drilled.

She learned that they cared enormously about two things: how level the ground was, and how level it stayed. There were benchmarks driven into the crust at intervals, and every eleven days they went back and re-measured every single one of them, and wrote the difference down even when the difference was nothing. Especially when it was nothing. Teo watched a man record a variance of zero point zero zero four metres with the expression of somebody receiving good news.

In the third month a benchmark moved.

Eleven millimetres, at a marker forty kilometres out, and Teo did not understand why the truck went quiet.

Eleven millimetres is nothing. It is the thickness of a finger. On the brine fields nobody would have written it down; on the brine fields the ponds moved more than that when a wall slumped, and you shovelled it back and said nothing to anybody.

The woman from Munich re-shot it herself, three times, and then had the whole crew stop what they were doing and re-shoot the four benchmarks either side of it, which cost them the rest of the day and most of the next.

"It's eleven millimetres," Teo said, eventually, because nobody was explaining and she had learned by then that they would answer a direct question.

"It's eleven millimetres this year."

"And?"

The woman put the cap on the eyepiece. "If it's eleven this year because a truck parked on it, that's nothing and I'll write it off. If it's eleven this year because something underneath is settling, it's eleven next year and eleven the year after and in thirty years it is a third of a metre, and the whole reason we are standing in this specific desert instead of a more comfortable one is that here, that does not happen." She looked out along the line of flags. "So I don't get to decide it's nothing. I have to find out which of the two it is."

It was a truck. It took nine days to establish that, including a bad-tempered exchange with a contractor in Uyuni who had sent a vehicle out to a marker for reasons nobody could reconstruct, and Teo watched the whole nine days with the growing understanding that these people did not trust anything they had not measured twice, and that they were not being difficult, and that this was simply what it looked like when the cost of being wrong was very high and a long way off.

She had opened pond gates for eleven years for an office that never read the numbers.

She thought about that a good deal, that week.

And she learned, slowly, why here.

Nobody sat her down. It came in fragments, over weeks, from people talking to each other in front of her the way people talk in front of a driver, and she assembled it the way she assembled everything — quietly, accurately, and without mentioning that she had.

The altitude was the first piece. Three thousand six hundred and fifty-six metres. She had always thought of the altitude as a hardship, because it was: the cold, the sun that burned you through cloud, the way visitors sat down suddenly on their first afternoon. To the crew it was an asset, and she overheard the number that made it one — at this height you were above roughly a third of the atmosphere by mass. A third of the air simply was not there. Whatever they intended to push along that line, a third of the resistance had been removed before anybody built anything.

The latitude was the second piece and she got that one from the sunburnt young man, who by then had stopped being careful about what he said in front of her and had started, she suspected, actively enjoying having somebody to explain things to.

"Twenty degrees south," he said. "You're closer to the equator than you feel. The ground here is moving east faster than it does in Europe — the planet's wider here. If you're launching, you get that for free."

"Launching," Teo repeated.

He heard it come out of his own mouth about half a second late. She watched him decide not to make it worse.

"Aircraft," he said, unconvincingly, and went back to the pole.

The third piece was the sky. Two hundred and something clear nights a year, no cloud to speak of in the dry season, no industry within hundreds of kilometres to put anything in the air, and nobody living underneath the line for as far as the line went. She had known all her life that the sky here was better than other skies. She had not known that this was a specification anybody wrote down.

And the fourth piece was the one she had given them herself, in the first week, standing at a pond gate with a hydrometer in her hand, and it took her a month to understand that she had.

A hundred and twenty-six kilometres and not a metre of deviation. Not because somebody graded it. Because the water lies on it every December and takes the high places away and fills the low ones and hands it back level, every year, for as long as there has been a salar.

Anywhere else on Earth, you could level a track. You could survey it and cut it and grade it and pour it, and it would be flat on the day you finished, and then the ground underneath would settle unevenly for thirty years and you would spend the rest of the project chasing it. Here the ground was self-correcting, and it was self-correcting on an annual schedule, for free, and it had been doing it since before anybody had a word for it.

Teo worked that out standing at the back of a truck at four in the afternoon and had nobody to say it to.

She said it that night anyway, to her mother, in the kitchen, and her mother listened and then said, "So it's a good place for it."

"For what?"

"For whatever it is."

"It's a mine," Teo said, and heard herself, and did not believe it, and said it again more firmly because the alternative was that she had no idea what was happening a kilometre from the house she grew up in.

Her mother, who had worked forty years on the flat and had a settled view of foreigners, said the thing that turned out to be the most accurate sentence anybody in Colchani said that year: "Nobody measures a mine four times."

Everyone assumed it was a mine. Teo wants that recorded and so does the record. Not because they were slow — because a mine was the only thing that had ever come here, and a mine was the only thing the vocabulary allowed, and the survey documents lodged with the department in La Paz used the word corridor eleven times and never once explained what would travel along it, and the officer who read them ticked the box because the box was about environmental disturbance and the disturbance was minimal.

It was, in fact, minimal. That is the strangest and most Bolivian fact about the founding of that world: the entire project cleared its first regulatory hurdle honestly, on the truthful grounds that laying a line of stakes across a salt flat disturbs almost nothing.

In the fourth month the woman from Munich handed Teo a level and told her to shoot a benchmark herself, and stood behind her and corrected her grip twice, and then read the figure and said, "Again."

Teo did it again. The two numbers were within a millimetre.

"Again."

The third was the same as the first.

The woman wrote all three down — all three, not the best one — and said, without any particular warmth, "You'll do this from now on. I have four people who take one reading and tell me it's the reading."

That is how Teo stopped being the local liaison, four months in, without anybody changing the paper she had signed.

The demonstration was in a rented hangar at the edge of the airfield and six people came, which was two more than Calder had expected and four fewer than the hangar had chairs for.

He had done this eleven times in four countries. He knew exactly how it went. There was a moment about ninety seconds in where the audience decided whether they were watching physics or a conjuring trick, and every single time he had lost the room, he had lost it there.

So this time he did not begin with the apparatus.

He began by handing each of the six a length of copper wire about as long as a forearm, and asking them to hold it, and then asking them what it cost.

Nobody knew. The man from the energy ministry made a guess that was low by a factor of four.

"That's the whole problem," Calder said. "Not the price of that piece. The price of every piece, between everywhere and everywhere else, forever, plus the towers to hang it on, plus the people who go out in the weather to fix it, plus the fact that you cannot put it across ground you do not own." He let that sit. "Everything I am about to show you is one idea. The idea is that the wire is the expensive part and the ground is free."

Then he switched it on.

The apparatus was not impressive to look at, which he had stopped apologising for. Two plates driven into the floor of the hangar eleven metres apart, a transmitter the size of a suitcase, and, at the far end, a receiver plate and a lamp.

The lamp came on.

There was the pause he expected. Then the woman from the technical college — the hydrologist's colleague, the one who had come out of curiosity rather than obligation — got up without being invited, walked the eleven metres, and looked underneath the receiver plate for a wire.

"There isn't one," Calder said.

"I know," she said, still looking. "I'm checking that I know."

He liked her for that and told her so nine years later when he hired her.

What he explained next he had rewritten as many times as the summary, and what it had been cut down to was this: he was not broadcasting power through the air, which is inefficient and dangerous and which several better-funded people had failed at publicly. He was using the ground itself as the conductor. The rock and the brine beneath the salar were a medium, and if you drove a signal into that medium at a frequency the medium liked, the medium carried it, and you could take it out again somewhere else with a plate and a tuned circuit.

The efficiency was poor. He said so, first, before anybody asked, and put the number on the board. Somebody in a better-funded programme would have buried it.

"That's a bad number," said the man from the energy ministry.

"It's a terrible number," Calder agreed. "It's terrible against copper. Copper is very good. I am not going to beat copper on a route where you can put copper." He tapped the board. "It beats copper on a route where you cannot. It beats copper across a salt flat that floods to twenty centimetres every December and would take out any line you strung across it, four months a year, every year. It beats copper anywhere the ground is conductive and the surface is hostile and nobody is going to come out and repair a tower."

"And where else is like that?"

And Calder, who had by then learned a great deal, did not say a planet with no infrastructure whatsoever.

He said, "Here, for a start," and moved on.

The man from the energy ministry asked the question Calder had been asked in all four countries, and asked it in the same words, which by then he found almost comforting.

"Is it safe to stand on?"

"Right now, in this hangar, yes. At the scale I want to build, I don't know."

"That's not an answer I can take back to anybody."

"It's the true one," Calder said. "I can give you the one you want in about nine months, with a study behind it, or I can give you a comfortable sentence tonight and you can find out I was guessing in four years, in public, with a herd of llamas as the evidence." He said it flatly, without charm, which was the only register he had ever really had. "I would rather you thought less of me tonight."

There was a silence, and the woman from the college laughed once, not unkindly.

"Nine months," the man said. "Bring the study."

Then he asked the second question, which nobody in Zurich or Houston or Santiago had thought to ask, and which Calder remembered for the rest of his very long life.

"Who owns the ground you are conducting through?"

Calder opened his mouth and found that he did not have an answer prepared, because in four countries and eleven demonstrations, not one person had raised it.

"The state," he said, after a moment. "Constitutionally. All of it."

"Yes," the man said. "So if the ground is the wire, then the wire is ours, and you are proposing to run your power through a national asset." He was not hostile. He was working it out at the same speed Calder was. "That is not a problem. I am telling you that it is a clause, and that if you do not write it, somebody will write it for you in ten years and you will not like their version."

Calder wrote it down in the folder while the man was still speaking, which the man noticed.

It became, eventually, article four of the founding compact of a colony fifty trillion miles away: that the medium through which power travels is held in common and cannot be owned by whoever is transmitting through it. Four generations later, a councillor would cite article four in an argument about who was permitted to switch off a district, and would have no idea that it was first said out loud by a mid-level Bolivian civil servant in a rented hangar, to be helpful, about llamas.

This is the chapter where the reader from Book One should sit up, and the sitting-up should be theirs and not the book's. Nobody in that hangar used the phrase Lithospheric Resonant Conduction. Nobody wrote LRC on anything. There were six people, a lamp, a suitcase, and a bad efficiency number honestly disclosed, and a hundred and forty years later the entire power grid of a world would run on the great-great-grandchild of the thing in that room, and the deep sensor net that listens to that world's basalt would be a modification of the same trick, and neither of those facts is visible from inside the hangar.

What was visible from inside the hangar was that it worked, badly, over eleven metres.

The woman from the college asked the only question that frightened him.

"What's the ceiling? Not the efficiency — the distance."

"I don't know."

"Guess."

"I won't guess in this room," Calder said. "If I guess and I'm wrong you'll remember the guess. What I can tell you is that the loss doesn't scale the way people assume, and that the medium here is unusually good, and that I have modelled it out to about nine kilometres with numbers I believe and past that with numbers I don't."

"Nine kilometres is a long way from eleven metres."

"Yes," he said. "That gap is the entire project. Everything else is engineering."

Afterwards, when they had gone, he sat on the floor of the hangar next to the transmitter for a while.

It had gone well. He was aware it had gone well; the man from the ministry had asked when the next demonstration would be, which was the first time in eleven attempts anybody had asked that. And what he felt, sitting on a cold concrete floor at nine at night, was not triumph but something closer to grievance, and he had the honesty to notice it and the self-awareness to be unimpressed with himself about it.

The thing had worked in Zurich too. It had worked in Houston, twice. It had worked in a conference room in Santiago in front of a man who did not finish the summary. It had been working, without modification, for six years, and the only variable that had changed between eleven failures and one success was that this time he had begun with a piece of wire and a question instead of with the truth.

He wrote in the folder that night, and the line is quoted in every biography and is almost always presented as wisdom:

They do not reject the physics. They reject being made to feel stupid in the first ninety seconds. Nobody has ever told me this. I had to lose eleven years to find it out and I am going to be bitter about it for the rest of my life.

Only the first sentence is ever quoted.

He was, at that point, still four months from being told yes, and fourteen months from a signature, and he did not know either number. What he knew was that six people had watched a lamp come on and one of them had gone to look for the wire, and that in the morning he had to find out what the salt did to a buried plate over twenty years, because the woman from the college would ask him that next time and he did not yet have the answer.

He found it out. It took him nine weeks and the answer was worse than he hoped, and he put it on page one of the next submission, before anybody could find it themselves.

Teo had heard the magnesium speech four times before she was twenty-four, and she could have given it herself.

It went like this. The salar holds more lithium than anywhere on Earth. The lithium is not in rock, it is in brine, dissolved, and you get it out by pumping the brine into shallow ponds and letting the sun take the water off, and what is left behind is a concentrate you can process. That part is simple and the sun is free and the sun here is ferocious.

The problem is what else is dissolved in it.

Magnesium sits next to lithium on the chart in a way that Teo did not understand chemically and understood perfectly practically: the two of them behave alike, so anything you do to catch one catches the other, and here there is a great deal of magnesium. She had been told the ratio four times by four different men in four different registers of regret. Atacama, across the border, has a ratio that lets you work. Uyuni does not. Everything else about this place is the best in the world and that one number is the worst, and it has been the worst since before anyone measured it.

So the companies came, and drilled, and sampled on a grid, and got the number they already knew they would get, and did the arithmetic, and the arithmetic said that removing the magnesium cost more than the lithium was worth.

Then they left. All four. Two of them with a press conference about partnership.

Teo's uncle had worked eleven months for the third one and had come home with a jacket and no job, and had worn the jacket for years afterwards in a way that Teo found difficult to look at.

So in the first dry season, when it became clear that this new one was building ponds, she assumed she knew exactly how the next four years went.

Except the ponds were wrong.

She noticed it before anybody explained anything, because she had spent eleven years opening and closing gates on ponds and a pond has a shape that it has for a reason. The shape is: big, shallow, rectangular, cheap. You want maximum surface for the sun and minimum wall for the money, and you want them in a chain, and you move the brine down the chain as it concentrates.

These were the wrong shape.

They were smaller. There were far more of them than there should have been. Some were not rectangular at all. And the chain did not run in one direction — there were returns, brine going back up to a pond it had already been in, which on the fields she had worked would have been a mistake somebody got shouted at for.

She asked about it and got an answer she did not understand from a man who was pleased to be asked, involving the word selective four times.

What she understood was the outside of it, and the outside of it is what this chapter is, because the outside is all Teo had and it turned out to be enough to see the important thing.

The important thing was not chemical. It was that he had stopped trying to remove the magnesium.

She worked that out over about six weeks, from fragments, the way she worked out everything. The other four had all been solving the same problem: get the magnesium out, so that what remains is a clean lithium concentrate of the kind the market already knows how to buy. Every one of them had priced that removal and every one of them had found it fatal.

This one was not removing it. He was letting the two of them separate slowly, at different rates, in a longer and more complicated sequence of smaller ponds, using time instead of chemistry — and time was the one input that cost him almost nothing, because he was not going anywhere for thirty years.

"It's slower," the man told her, when she got the question sharp enough to ask properly. "Much slower. About four times the residence time. Any normal operation, that kills you — your capital is sitting in a pond doing nothing for a year."

"And it doesn't kill him?"

"He's not trying to get his money back in four years." The man shrugged. "He keeps saying that. It's the only thing he ever argues about."

Teo thought about that for a long time.

It was not cleverness, or not only. The chemistry was clever and she never fully followed it and there are three papers about it she has never read. But the decision underneath was not technical at all: four companies had a problem they could not solve inside four years, and a man arrived who had thirty, and the problem stopped being the same problem.

She has said this in public perhaps eleven times, always in the same words, and it has never once been the quote they use: He didn't beat the magnesium. He outlasted it.

Her uncle came out to look at the ponds in the fourth month.

He had not been on the flat in six years. He stood at the edge of the chain in the jacket — he wore it out there, deliberately, and Teo understood exactly why and said nothing — and looked at the small strange ponds and the returns for a long time.

"They're doing it wrong," he said.

"I know. I thought so too."

"No." He was not being stubborn; he was working. "Not wrong like a mistake. Wrong like they don't care about the same thing." He pointed with his chin at the return line, brine going backwards up the chain. "That costs money every day it runs. Nobody does that unless the thing they're saving is worth more than money."

"What's worth more than money?"

Her uncle thought about it.

"Being here in ten years," he said.

He was sixty-one and had eleven months of a foreign payroll and a jacket to show for a life on the salt, and he read the entire strategic position of the Ark-Genesis project off the shape of an evaporation pond in about four minutes, standing in the wind, and then went home and did not mention it again.

Teo has never known what to do with that memory. She has told it twice. Both times the person listening treated it as a nice moment about an old man, and both times she wanted to say: no, you are not hearing me, he was right, he was right before any of the engineers were and nobody ever asked him anything.

What she remembers most from that season is not the ponds.

It is that in the second month, when the returns first started running and the numbers coming off them were bad — genuinely bad, worse than the model, bad enough that two of the engineers went quiet in a way that Teo recognised from four previous companies — he came out to the field himself.

She had seen him twice before at a distance. Up close he was smaller than she expected and looked tired in a way that had nothing to do with the altitude.

He did not give a speech. He walked the chain, all of it, which took most of a day, and he asked the pond hands questions, and the questions were specific and mostly not about chemistry — where does it leak, what do you do when the wall slumps, how long does it actually take you to move a gate as opposed to how long the drawing says.

He got to Teo at the eleventh pond, late in the afternoon.

"You wrote these," he said. He had a clipboard of her readings. It surprised her that they had gone anywhere.

"Yes."

"This entry here. The twenty-second. You've written the reading and then you've written a second number in brackets."

"That's what it should have been."

He looked at her.

"The gate was still bleeding," Teo said. "It doesn't seat properly, it hasn't for a month, I've reported it. So the reading is the reading, but it isn't what the pond is doing. I put both because I didn't want to give you a wrong number and I didn't want to give you a number I'd invented."

There was a pause that went on slightly too long.

"Nobody taught you to do that," he said.

"No."

"Who told you to stop?"

"Nobody read them," Teo said. "For eleven years. So nobody told me anything."

He stood there with the clipboard for a moment and then said, "The gate will be replaced this week," and it was, on the Thursday, which was the first time in her working life that reporting something had caused it to happen.

He did not offer her a job that day and she did not expect one. He asked her name and she gave it and he wrote it on the clipboard, on the corner, and she watched him do it.

Years later, when people wanted the story of how she rose, they always wanted it to be the rail failure in the eighth year — the one where she was right and was overruled and the section split and she was promoted out of it. That is the story that gets told because it has a shape.

Teo has always said it was the brackets.

The ponds took fourteen months to produce anything at all, which was eleven months longer than any of the four previous companies had been willing to wait, and when they did produce, the first shipment was small and the grade was unspectacular and the state took its majority share exactly as written.

Nobody outside the department noticed. There was no announcement. The energy ministry man who had asked about the llamas got a two-page report with a table in it and initialled the bottom.

And a line of stakes went on being driven west across the flat, at twenty metres, funded now, by lithium, at a margin so thin that the accountants brought in during the consortium years would describe the entire extraction business as strategically inexplicable in a memorandum that survives.

It was not inexplicable. It was a rent payment. It bought thirty years of standing on the flattest ground on Earth, and the man paying it had never once pretended otherwise to anybody who asked him directly.

Almost nobody asked him directly.

Ines Ferrer revised the projection for the fourth time that year on a Tuesday afternoon, in an office with a window that did not open, four thousand kilometres from a salt flat she had never heard of.

The document was called the Regional Carrying Capacity Review. It was produced every eighteen months and it was, by design, the least dramatic object in the building. Ines had inherited it from a man who had done it for nine years and who had left her a folder of notes in tiny handwriting, the last page of which read: Do not editorialise. Nobody will thank you and it makes the number harder to see.

She had followed that instruction for two years and had come to regard it as the closest thing she had to a professional creed.

The number was in column F.

She wants it understood — and she said so, later, on the ship, to a woman who asked her why she had come — that nothing about that afternoon felt like anything. There was no moment. She did not stand up. The building did not go quiet. She opened the file at two, and there was a revised inflow figure from the coastal districts that had come in from another department in the ordinary way, and she put it in, and the sheet recalculated, and column F changed.

She looked at it for a while.

Then she did the only thing the job actually required of her, which was to check whether it was wrong.

It took eleven days. She pulled the four previous editions and re-ran them against the current inputs. She found two errors in her predecessor's method, corrected both, and both corrections made the number worse. She wrote to the coastal department to confirm the inflow figure and got a reply that confirmed it and added a sentence, unprompted, that was not part of the answer: Please note this is the revised figure. The original was withdrawn.

Nobody withdraws a figure. Ines had been in that building for two years and had never once seen a department withdraw a figure. She asked why. She was told, politely, that the original had been produced under an older methodology.

She asked for the original anyway, because she was the kind of person who asks for the original, and it arrived four days later, and it was worse than the revision.

Here is the rule this book holds to the end, and it starts here: what is killing Earth is never stated.

Not by Ines, who could have named four contributing mechanisms and did not consider it her business. Not in the Review, which had a defined scope and stayed inside it. Not by the coastal department, whose sentence about methodology was true and complete and told you nothing. Not by anybody in the nine books that follow, in which the loss of a planet is a settled historical fact that people refer to the way you refer to a war your grandparents were in.

It is not a mystery and it is not a trick. It is that a civilisation does not experience its own ending as a subject. It experiences it as a series of departments, each correctly reporting a figure inside its own scope, and a woman in an office with a window that does not open, noticing that column F has changed.

What Ines did was write it up.

Four pages, no editorialising, the corrections documented, the withdrawn figure attached as an annexe with a note explaining that she had used the original rather than the revision and why. She circulated it on the standard list — eleven recipients — and put a copy in the file, and went home, and made dinner, and did not sleep well but did not connect the two things until much later.

Eight of the eleven acknowledged. Two asked clarifying questions, both good. One did not respond.

Nothing happened.

She has been asked about that a great many times and she has always been careful to be accurate about it, because the accurate version is more frightening than the version people want. Nobody suppressed it. There was no meeting where a man in a suit said the figure was unacceptable. There was no phone call. The report went where reports go, and was read, competently, by people who were not lying, and every single one of them had a defined scope, and the number in column F was outside all eleven of them.

Two of the eleven wrote to her personally to say it was a good piece of work.

The Review was published on schedule with the number in it, unaltered, and the publication was noted in four trade outlets and one newspaper, in a paragraph, below a story about a transport strike.

The one who did not respond was the deputy director, and four months later he stopped her in a corridor.

He was a decent man in his sixties who had never done anything wrong and had two years to run. He said he had read the paper. He said the annexe was the right call. Then he said the sentence that Ines carried with her for the next ninety years and quoted exactly once, in a session, when the ten of them were arguing about whether to write anything down at all.

"You know there is nowhere for it to go."

"I know," Ines said.

"I'm not telling you to stop." He looked genuinely unhappy. "I'm telling you that I have watched four people do what you are doing, over twenty-two years, and every one of them thought the problem was that the number hadn't reached the right desk yet. It has reached every desk. There is no desk."

"Then why does the Review exist?"

He thought about that for a while, and to his credit he did not reach for anything comfortable.

"So that afterwards," he said, "nobody can say we didn't know."

Ines went back to her office and sat down and did not open the file for about twenty minutes.

She had believed, up to that afternoon, that she was doing preventative work. What she had actually been doing for two years — what the whole apparatus around her had been doing, competently, at considerable expense, with real professionalism — was keeping the record. Not to act on. To have.

She did not find that as bleak as she expected to, and the fact that she did not find it bleak was the thing that frightened her.

Because if the record was the only thing that survived, then the record was not clerical work. It was the last useful act available, and it had to be done properly, by somebody who would not soften a figure to make a colleague comfortable.

That is the whole of Ines Ferrer, and it was finished by the end of that afternoon, and everything she did for the next hundred and eighteen years was the same decision repeated.

What changed was Ines.

She started keeping her own copy. That is all. Not a leak, not a campaign — a second file, at home, in which she kept the working, the withdrawn figures, the annexes, and the corrections, on the grounds that the official file kept only what the official file was for, and that somebody at some point might want to know what had actually been in front of them.

She would keep doing that for the rest of her life. She kept doing it in a compartment off a forward transfer, by hand, in a bound volume, ninety years later, because she did not trust storage she could not see — and she never once, in thirty-six years of being asked why, gave that answer. She always said it was a preference.

It was not a preference. It began on a Tuesday, with a withdrawn figure that turned out to be better than the one that replaced it.

In the eighth month she saw the recruitment advertisement.

It was not addressed to her. It was in the back of a technical weekly, and it was looking for welders and riggers and brine hands and cooks, in Bolivia, and the pay was specific and the contract length was specific and there was a line at the bottom that said the project expected to be operating for not less than twenty-five years.

Ines read that line four times.

Nobody was writing twenty-five years. She had spent two years inside an institution whose planning horizon had quietly contracted to about forty months, without anybody announcing it, the way column F had changed without anybody announcing it. Every serious document she touched had learned to end before it had to say anything.

And here was a paragraph in the back of a trade magazine, offering a cook a job with a future tense in it.

She did not apply. Not for another eleven months, and not for that advertisement, and when she did apply it was for a records post and she was overqualified and said so on the form.

But she cut the advertisement out and kept it, in the second file, at home, with the withdrawn figures.

It is in the Bound Ledger. Not the text — the actual clipping, browned, four centimetres by nine, pasted onto the inside of the back cover, which is the only page of that volume that is not in her handwriting and the only object in it that came from Earth.

Four generations of schoolchildren have filed past the reliquary case and been told about the founding of a world, and the case is closed at the signature page, and nobody has ever turned it over.

Nnamdi Okonjo was hired to solve a problem that Calder had correctly identified and completely failed to understand, which was that the salar could not be occupied quietly.

Calder's instinct had been concealment, and he had spent four weeks on it before admitting it was hopeless. Nnamdi's first act was to make him say the hopeless part out loud.

"Tell me who watches it."

"Everybody," Calder said. "That's the difficulty. There are eleven orbital instruments that use this surface as a reference. Not as a subject — as a standard. They point at it to find out whether they themselves are lying. Which means every one of those programmes has a named person whose job includes knowing what is on this ground, and a publication schedule, and a professional reason to notice a change."

"How often?"

"Some of them monthly. One of them," Calder said, with something like despair, "every nine days."

Nnamdi nodded slowly and did not look worried, which Calder found irritating.

"So the most-measured piece of ground on Earth," Nnamdi said. "And you were going to put a thirty-two-kilometre structure on it and hope."

"I was going to phase it."

"You were going to hope." He said it without unkindness. "Phasing a thing that is being photographed every nine days by instruments specifically designed to detect changes in elevation is hoping with extra steps."

Calder, who did not enjoy being handled and was being handled, said, "Then what."

And Nnamdi said the thing that he was, in the end, hired for, and which he would say in four different forms across the next thirty years, and which is the reason the recruitment campaign works, and the reason the Bound Ledger is public, and the reason a world's exemptions schedule gets published in an annexe where anybody can read it and nobody does.

"You don't hide it. You explain it until they stop asking."

The method, once he set it out, took about forty minutes and Calder objected to almost all of it.

They would publish. Not a press release — publication, in the boring sense: a technical notice to the orbital geodesy community, in advance, describing the works. Coordinates. Dimensions. Expected duration. A contact.

"That is the opposite of what I want."

"It is exactly what you want," Nnamdi said. "Right now, a change on that flat is an anomaly, and an anomaly is a career opportunity for a graduate student. After we publish, a change on that flat is a known works notice, and a known works notice is an inconvenience they have to correct for in their data. You are choosing what kind of object you are in their world. Anomaly, or paperwork."

"They'll ask what it's for."

"They'll ask once. That's the whole game." Nnamdi turned the pad round. "Here is what it's for: a lithium industrialisation project, majority state-owned, with an associated research annexe evaluating ground-coupled power transmission over extended baselines. Is any of that false?"

Calder read it twice.

"No," he said. "It's all true."

"Is any of it interesting?"

There was a long pause.

"No," Calder said.

"Then that's the notice."

It went out in the fourth month and Nnamdi's projection was that they would receive between four and eleven substantive enquiries. They received nine.

He answered all nine personally, at length, within four days, and this is the part of the method that people miss when they copy it: he did not deflect a single one. A researcher in Delft asked for the expected disturbance footprint and got a table. A team in Colorado asked whether the works would affect the calibration corridor they used and got a map with their corridor drawn on it and a note explaining that it would be affected for eleven months in a specific band, with an apology and an offer to provide ground truth measurements at their benchmarks at no cost, taken by the project's own surveyors.

They took the offer. Of course they took the offer. It was free data of a quality they could not otherwise afford.

And so the programme in Colorado spent the next nine years receiving a quarterly file from a lithium company in Bolivia, which was accurate, arrived on time, and was so useful that when a graduate student in the seventh year noticed that the disturbed area had grown considerably beyond the original notice and asked about it in a meeting, a senior colleague said, without looking up, that the Uyuni people were reliable and had presumably filed an update.

They had. It was four pages. It was extremely boring. It was entirely true.

Nnamdi's talent was not deception and it was not even really persuasion. Teo, who worked alongside him for eleven years and understood him better than anybody except possibly Ines, described it once as giving people so much of the truth that they file it.

He had come from an industry that Calder held in open contempt, and they had the argument about it in the second week, once, properly, and never again.

Nnamdi had spent nine years writing copy. Consumer launches, mostly, and two political campaigns he did not put on the list he handed over. He was extremely good at it and he had left because he had watched a product he knew to be mediocre outsell a better one on the strength of eleven words he had written himself, and had found, to his own surprise, that he minded.

"You sold things that didn't work," Calder said.

"I sold things that worked adequately, to people who would have been fine either way. Yes."

"And now you want to do it for something real."

"No," Nnamdi said, and this was the moment Calder stopped being contemptuous, because it was not the answer of a man seeking redemption and Calder had no patience for those. "I want to do it for something long. Everything I have ever written had a six-week horizon. Buy it now, vote next month. I have never once written a sentence that had to still be true in twenty years." He shrugged. "You are the only person who has ever offered me that and you are offering it in a desert, which tells me something about my prospects."

The other thing he did in the first year gets almost no attention and mattered more than the notices.

He went to Colchani, and to Uyuni, and to four smaller places, and he held meetings that nobody had asked him to hold.

Not consultations — there was nothing to consult on, the contract was signed. He went and explained, in Spanish, badly at first and then adequately, what was going to be built, how long it would take, how many people it would employ and in what trades, and what it would do to the road and the water and the tourist traffic that Colchani lived on half the year.

He was heckled in Colchani. Teo was at that one; she was twenty-seven and had been on the payroll eleven months and she stood at the back and watched a man from Lagos take forty minutes of genuine anger from people whose last four experiences of foreign projects had ended in a jacket and no job.

He did not defend the previous four companies. He did not promise this would be different. What he did — and Teo has repeated this in every account she has given — was write down the specific complaints, out loud, by name, saying each one back to make sure he had it, and then read the whole list at the end to the room.

Somebody shouted that a list was not a fix.

"No," Nnamdi agreed. "But in four years you will be able to hold me to a list. You cannot hold me to a feeling."

The hall did not warm to him. It was not that kind of meeting and he did not expect it to be. But eleven months later, when the first real hiring wave went out, four hundred and six people applied from a town of eight hundred, and Teo has never believed that was the pay.

He was aware of what he was doing. That is the thing about him that the histories consistently soften, and it should not be softened, because it is the whole of the moral question this book is holding.

In the sixth month, drafting the second notice, he stopped and said to Calder: "You understand this is worse than lying, if it goes wrong."

"How is it worse? It's all true."

"Because when a liar is caught, people say: he lied. When somebody who told the exact truth in a form nobody could use is caught, people say: they knew, and they told us, and we didn't read it." Nnamdi tapped the page. "And then they never trust a document again. Not ours. Any."

Calder said, "Do you want to stop?"

"No," Nnamdi said. "I want it written down that I said it."

Calder wrote it down. It is four lines in a folder, and it is dated, and it is the earliest evidence in the entire history of Aura-36 of the practice that eventually becomes the Open Accord's central habit: that when a thing is done which somebody thinks is wrong, the objection is minuted, at the time, by name, whether or not it changes the decision.

Halima Sow, ninety years later, would lose the appeal-clause argument twice and have both losses recorded, and would be told by people who meant it kindly that the recording was a consolation prize.

It was not a consolation prize. It was a habit that started in a rented office in Uyuni because a communications man wanted it on paper that he had seen the shape of the thing he was building.

Ines, who was still four thousand kilometres away and eleven months from applying for a records post, would have recognised it instantly.

The notices went on going out for thirty years. There were forty-one of them in the end. Every one accurate; every one dull; every one filed.

And the twenty-mile electromagnetic launch rail that was assembled on the flattest surface on Earth, in full view of eleven orbital instruments whose entire function was to detect changes in elevation, was described in the public technical record, continuously and truthfully, for three decades, as an extended-baseline ground-coupled transmission research corridor with associated industrial works.

Not one word of that was false.

Nobody asked a second time.

The first kilometre took nine days to set and Teo walked all of it, twice, on the second afternoon, because she wanted to know what a kilometre of it felt like on the ground rather than as a number on a sheet.

It felt like nothing. That was the first surprise. Fifty stakes at twenty metres, orange flags snapping in the wind, and at the far end you turned round and the town was not visible and the stakes behind you had merged into a single dotted line that ran back to a horizon that had not got any closer.

A kilometre on the salar is not a distance. It is a rounding error.

She had known the number thirty-two for months by then. She had said it out loud to her mother. What she had not done was divide it, and standing at the end of the first kilometre with her hands in her jacket she did the division, and the answer was that they had done one thirty-second of the line and it had taken nine days and there were twelve hundred more stakes.

And then the second thing arrived, which was the one that changed her.

The stakes were not a route.

She had been thinking about it as a road, or a rail line, the way you think about the track from Uyuni to Oruro — a thing that goes from one place to another place, and the two places are the point, and the line between them is just how you get across. That is what a line is for. Everywhere else on Earth, a line is for connecting.

There was nothing at either end of this one.

West of the last flag: flat. East, behind the first: flat, and then eventually Colchani, which was not a destination anybody would build thirty-two kilometres of anything to reach. The line did not connect two places. It did not pass through anywhere. It started in the middle of nothing and ran, dead straight, to a slightly different part of the same nothing.

Teo stood there for a while working out what that meant, and what it meant was that the line was not for getting anywhere.

The line was the machine.

Whatever they were building, the thirty-two kilometres was not the route to it. The thirty-two kilometres was it, or was the working part of it, and the flat underneath was not a site the way a mine has a site. The flat was a component. They had not come here to build a thing on the salar; they had come here because the salar was a piece of the thing, the one piece nobody could manufacture, and the only place on the planet it existed at the size required.

She thought about the eleven-millimetre benchmark and the two days it had cost, and it stopped being strange.

You do not survey a road four times. You survey a road once and then you build it and it is fine, because a road tolerates a metre. They were measuring the ground with the tolerance you would apply to a machined part, because to them it was a machined part, and it had been machined by the water, annually, for free, for longer than there had been anybody here to notice.

Teo walked back along the line as the light went orange and the temperature dropped the way it drops here, forty minutes from shirtsleeves to breath in the air.

She did not tell anybody. There was nobody to tell who did not already know, and the one person she wanted to say it to was her mother, and saying it to her mother would have required saying I do not think this is a mine, and she was not ready to have said that out loud, because once it was out loud she would have to have a theory, and she did not have a theory. She had a shape and no name for it.

The rail arrived in the eleventh month and it was not what anybody in Colchani expected either.

People had been waiting for rail the way you wait for rail: heavy sections on flatbeds, a crane, a gang of men, sparks. What came instead were pallets, and the pallets held segments about four metres long that two people could lift, and there were an enormous number of them.

Teo asked the obvious question and got an answer that took her a week to appreciate.

"Because it has to be adjustable," the woman from Munich said. "Forever. Not once — forever. Every year the water comes and every year the ground gives it back within a millimetre or two, and a millimetre or two over four metres is nothing, and a millimetre or two over thirty-two kilometres is the whole tolerance. So nothing is welded. Every segment sits on a mount you can shim, and every mount gets checked."

"How often?"

"Every year, after the water."

Teo did the arithmetic on the walk back. Eight thousand segments at four metres. Eight thousand mounts. Checked and shimmed annually, by hand, by people, in the four months between the flat drying and the working season ending.

It was not a construction project with a maintenance tail. It was a permanent act of maintenance that happened to have been constructed once at the start, and the difference between those two things is the difference between a company and a town, and it is the reason four hundred thousand people eventually lived out there.

She said something like that, badly, to the woman from Munich, who considered it and said: "Yes. Nobody has told the accountants."

What she did instead was start reading the notices.

They were public. That was the thing that took her longest to get over. Nnamdi's technical bulletins went out to a list and copies came back to the site office in a folder that sat on a shelf, and nobody had ever told Teo she could not read them, and it had simply never occurred to her that a document produced by the company about the company would be available to a woman who opened gates.

She read all of them. It took a week of evenings.

They told her almost nothing, and she understood, reading the fourth one, that this was not accidental, and she felt a flash of something hot and unwelcome that took her another day to identify as insult.

Not at being lied to. She checked, carefully, twice; nothing in them was false. Every dimension matched what she had paced out herself. The coordinates were right. The duration was honest.

It was that the notices had been written by somebody who had thought hard about the reader, and the reader they had thought about was a geodesist in Colorado, and every choice in the document was calibrated to make that person stop asking. And they worked on her too. She had read four of them before she noticed she had learned nothing, and she was standing on the ground they described.

She went and found Nnamdi about it, eventually, in the eighth month, which took a courage she has always been dismissive about and which was considerable.

She said: "I've read the notices."

"All of them?"

"All of them."

He put down what he was holding, which she noticed.

"And?"

"They're true," Teo said. "And they're useless. And I think you did that on purpose, and I think you're good at it, and I want to know whether there's a version of them for people who live here."

Nnamdi was quiet for a moment. She had expected to be managed. She had rehearsed being managed.

"No," he said. "There isn't."

"Why not?"

"Because I haven't written one, and the reason I haven't written one is that the first honest sentence in it ends the secrecy for everybody, not just for Colchani." He said it plainly, without apology. "That's not a good reason. It's the true reason. I'd rather give you that than a better-sounding one."

Teo said, "How long?"

"Four years. Maybe six."

"And then everybody knows."

"Then everybody knows, and the ones who worked here first will have known for years that they were being handled, and that is the bill I am running up." He picked his pen back up. "I write it down when people notice. It's about the only useful thing I do. What's your name for the file?"

She told him.

It is in the record. Eighth month, first year: T. asks whether a version exists for local staff. It does not. Told her why. She was right to ask and I have no answer that survives four years.

Teo did not know that entry existed until she was forty-one, and she found it in the course of looking for something else, and she sat down on the floor of a records room on a ship twenty miles long and read it about nine times.

What she took away, that evening in the first year, walking home along a line of orange flags that did not go anywhere, was smaller and more useful.

The flat was not the resource. Everyone in Colchani had spent a hundred years arguing about what was under the flat — the salt, the brine, the lithium, the magnesium, all of it a question of what you could take out and sell.

Nobody had ever once considered that the valuable thing was the shape.

You could not extract it. You could not ship it. It was worth nothing at all to anyone who wanted to take something away, which is why four companies had come and gone without seeing it, and why her grandmother had worked forty years on the most precisely level large surface on Earth and had thought of it as somewhere flat to put a pond.

The water came on the eleventh of December and nobody had to be told.

Teo woke to it. There is a sound the wind makes over dry salt and a different sound it makes over four centimetres of standing water, and she had been hearing the change since she was a child; she lay in the dark and listened to it for a while and then got up, because there would be no more sleeping and because the first morning is the only one worth seeing.

By seven the salar had stopped being ground.

The flat took the sky and gave it back exactly. Not approximately — exactly, because there was no wave and no current and nothing to disturb four centimetres of water lying on a hundred and twenty-six kilometres of level, so the reflection was not a reflection, it was the sky continuing downward past your boots. The horizon disappeared. The clouds ran under her feet. Two trucks parked eleven kilometres apart at the edge of the works stood on their own doubles and it was not possible to say where either one ended.

Every visitor who ever saw it cried or laughed or went completely silent. Teo had watched it happen to about forty people over the years and there was no third response.

What she felt, standing at the edge of it in December of the first year, was the flat and specific dread of somebody who understood the schedule.

Because the water is not weather. Weather is a thing that happens and then stops happening. The water is a season, and it does not negotiate, and it takes four months.

Four months. Every year. Forever.

She had grown up inside that arithmetic and had never once thought of it as an obstacle, because on the brine fields it was not one — the ponds filled, you closed everything down, you did other work or no work, and in April the sun came back and you started again. The salar had always had a wet season the way a person has a night. You do not resent it. You sleep.

The project could not sleep.

She watched it hit them across the first week and it was the closest she ever came to feeling sorry for the engineers.

They had planned for it. That was the thing; they had planned for it thoroughly, on paper, in a document she had seen, which allocated the wet season to indoor fabrication and design work and maintenance and training, and the document was sensible and detailed and completely correct and none of it survived contact with the fact that you cannot get to the rail.

Not the rail is difficult to work on. You cannot get to it. Twenty centimetres of brine over a surface with no landmarks, no gradient and no visible reference, with vehicles that will corrode if you drive them through it and drivers who cannot tell where the track is because the track is under it and the world is a mirror in every direction. Two men went out in the second week to recover a generator and got lost eleven kilometres from a road, in daylight, on a flat plain, because there is nothing to steer by when the ground is the sky.

They were fine. Teo found them, because Teo had said before they left that somebody should go with them and had been told it was four kilometres.

"You told them," the woman from Munich said afterwards.

"I told the shift lead."

"Tell me next time."

Teo said, evenly, "I did tell you. In October. It's in the seasonal note."

There was a pause, and to the woman's credit she went and found the seasonal note, which Teo had written in October at nobody's request, four pages, and which had gone into a folder and been read by no one.

It came back out in the second week of December with three signatures on it and became, over the following four years, the founding document of what the project eventually called the Wet Protocol.

Teo has always found that funny in a way she cannot explain to people. She wrote it in October, unpaid, on her own time, because she was annoyed. It is now taught.

The town found the wet season funny, and Teo had to explain to four separate engineers that the laughter was not malice.

Colchani had been watching foreigners meet the water for a century. There is a specific pleasure, if you have lived somewhere that outsiders consider empty, in watching people with instruments discover that the place has an opinion. The men in the bar were not laughing at the generator. They were laughing at the document.

Her uncle put it best, in his way, when she complained about it at the kitchen table.

"They planned the wet season," he said.

"Yes."

"On paper. In an office. In August."

"Yes."

He nodded slowly, enjoying himself. "And what did they plan?"

"Indoor work. Fabrication, drawings, training."

"Where indoors?"

Teo opened her mouth and closed it again, because that was the entire thing, and she had read the document four times and had not seen it. There was no indoors. There were two site huts and a rented office in Uyuni and a hangar at the airfield, total capacity perhaps forty people, and the workforce that December was two hundred and eleven.

They built the first shed in January because the alternative was sending everybody home, and they built the second in February, and by the fourth year there were nine of them and a canteen and a workshop with a floor you could put a machine on, and by the eleventh year there was a school.

That is how the town on the flat began. Not as a plan and not as a company decision and not in anybody's founding document. It began because it rained on schedule and two hundred people needed somewhere to be for four months, and then it happened again the next year, and the sheds were still there.

Here is what the water actually cost, and the number is the reason this chapter exists.

Four months of the twelve, the flat is unworkable. Of the remaining eight, the first six weeks after the water goes are spent on survey — because every mount on every segment has to be re-checked against benchmarks that have all, very slightly, moved, and until that is done you cannot lay anything new against them.

So the working year was not eight months. It was six and a half.

Calder came out in January and stood at the edge of the mirror for a long time with his hands in his coat, and Teo, who was there because somebody had to drive him, watched him do the same arithmetic she had done in October, and watched him arrive at the same place, and watched him not say anything for about eleven minutes.

Then he said, "It's a forty-six per cent year."

"Fifty-four," Teo said.

He turned round.

"Six and a half of twelve," she said. "Fifty-four per cent. And it's not fifty-four every year. In a bad year the water comes in November and goes in May and you get five months. I've seen four of those."

"How often?"

"Once in about seven years."

Calder looked back out at the sky under his feet.

"So the schedule has to survive a five-month year," he said, "arriving at a time I cannot predict, at least four times before we finish."

"Yes."

"Thank you," he said, and she has always maintained that he meant it, and that it was the first time anybody on that project had thanked her for making something worse.

He did not fight the water. This is the pattern of the man and it is now visible three times in ten chapters: he did not fight the Bolivian conditions, he did not fight the magnesium, and he did not fight the flood. He wrote down what it took and he built the rest of the plan around the hole.

The consequence, which nobody in that first December could have named, is the reason a colony on a planet fifty trillion miles away divides its day into thirds and its labour into interchangeable shifts.

A project that loses half of every year to a thing it cannot argue with has to become extremely good at two skills. The first is doing an enormous amount in a short window. The second, and harder, is stopping cleanly — leaving the work in a state where four months of nothing does not damage it, and where the people who come back in April can pick it up without a week of rediscovery.

Stopping cleanly is not a natural act. It has to be designed, and enforced, and it costs time at exactly the moment you least want to spend it, and the Ark-Genesis was built by an organisation that had thirty years of annual practice at it.

Every handover convention aboard that ship, every shift-swap ledger, every rule about leaving a task recoverable by somebody who was not there when you started it, is a descendant of four months of standing water.

There is one more thing about the water and it is the thing Calder eventually built a world on.

It is fair. That is not sentiment; it is the operative fact. The flood does not arrive because of anything anybody did. It cannot be lobbied, bribed, appealed or blamed on a department. It falls on the whole flat at once, to the same depth, on the competent and the careless equally, and it goes away again on its own terms.

Teo said something close to that to him, years later, and he stopped walking.

"Say it again."

"It doesn't care who you are," she said. "That's all. Everyone stops."

He was quiet for long enough that she thought she had said something stupid.

"Everyone stops," he repeated.

Teo saw the mirror thirty-one more times. She stopped photographing it after the fourth year.

She never stopped getting up for the first morning.

The money arrived in the fourth year and Calder discovered within eleven weeks that he had made a category error about what money was.

He had thought of it as fuel. He had spent his thirties without it and had constructed, in that time, a very clear picture of what he would do the moment he had enough, and every item on that list was a thing he could not currently afford to build.

Money turned out not to be fuel. Money was other people, and other people came with the right to be consulted, and consultation is not a cost you pay once.

The consortium was four institutions and it was assembled over fourteen months and by any ordinary measure it was a triumph. They came in on the lithium, which by then was producing, thinly but on a curve that pointed the right way; and on the transmission research, which had gone from eleven metres in a hangar to four kilometres across open salt; and, though nobody wrote this down, on the fact that a man who had agreed to every condition in a difficult jurisdiction and then delivered for four years is a specific and rare kind of bet.

The first governance meeting ran six hours and resolved nothing, and Calder went back to his room afterwards and sat on the bed with his coat still on.

Nothing had gone wrong. That was what he could not get past. Nobody had been obstructive. Nobody was a fool; two of them were plainly better at their jobs than he was at his. Every question asked in that room had been reasonable, and he had good answers to most of them, and the meeting had still consumed six hours and produced a schedule for another meeting.

He wrote it out that night in the folder, longhand, because he thought better with a pen, and what he wrote is quoted in Book Six by a woman who has no idea where it comes from.

Four partners. Each one reasonable. Each one entitled to say wait.

Four reasonable waits is a year.

The arithmetic underneath it was not abstract and he did it properly over the following week, because he did everything properly.

The build had a fixed end. Not a target — a fixed end, set by a launch window and a transit time and a number in a spreadsheet that a woman he had not yet met was revising in an office four thousand kilometres away. Against that fixed end he had roughly six and a half working months a year, times a number of years he could count on his fingers.

And now every material decision passed through four institutions, each of which had an internal process, and each of whose processes were individually sensible.

He tested it against a real case: the mount specification, which had come up in that meeting. His engineers wanted a change. It was a good change. It went to four partners. Assume each takes three weeks — and three weeks is fast, three weeks is a partner performing well — and assume two of them run in parallel because that is how it actually works.

Seven weeks. For a mount.

Seven weeks is nothing, in a normal project. Seven weeks in a fifty-four per cent year, on a build with a fixed end, repeated across every specification in a twenty-mile machine, is the difference between finishing and not.

He took it to the fourth partner's representative — the one he liked best, a careful woman from a development bank who had asked the two hardest questions in the six hours — because he wanted to be argued with by somebody competent before he decided anything.

She did not disagree with a single number.

"You're right," she said. "And you should understand that I cannot help you, and that I am not being cowardly."

"Explain that to me."

"My institution's process exists because thirty years ago we funded four things that should not have been funded, and people were hurt, and the process is what we built so that it could not happen again." She turned her cup around. "Every one of those three-week reviews is somebody's scar. You are asking me to say that your deadline is more important than the reason my organisation has a conscience. I won't say that. I'd be wrong to."

Calder sat with that.

"So there's no villain," he said.

"There is no villain anywhere in this," she agreed. "That's your actual problem. If one of us were greedy you could route around us."

Teo was in the six-hour meeting, at the back, taking the survey numbers in case they were asked for. They were not asked for.

She had never seen a governance meeting before and she found it physically difficult to sit through, and what she said about it afterwards — to the woman from Munich, in the truck, driving back — is the clearest description of the problem anybody produced that year.

"Nobody in that room can start anything," she said. "Every one of them can stop things. That's all any of them can do."

"They're funding it. That's starting it."

"They started it once. Now they've got four years of only being able to say no." Teo watched the flat go past. "On the ponds, if you can only ever say no, you stop being asked. People just do it and tell you after."

The woman from Munich glanced across at her.

"Say that to him," she said.

"He won't want to hear it from me."

"He read your seasonal note in December and put three signatures on it." The woman shifted gear. "He is the only person on this project who has never once cared where a correct thing came from. It is his single best quality and he has no idea he has it."

Teo did not say it to him, that year. She said it to him in the seventh, in a corridor, in about nine seconds, and he stood there and then went away and changed something.

This is the seed. It is planted here, in a conversation with somebody who was entirely right, and it grows for a hundred and forty years, and it ends in a hidden city in a ravine that the public monitoring grid cannot see.

What Calder concluded — and he wrote this down too, and this entry is not in any biography because the folder page it sits on was photographed badly and has never been transcribed — was:

A body of reasonable people with the right to delay cannot meet a fixed deadline. This is not a failure of the people. It is arithmetic and it will be true of every body I ever assemble, including the good ones. Including mine.

Therefore: some decisions must sit with someone who cannot be delayed.

I do not know how to make that safe. Note that I did not know, in case it later appears that I thought I did.

He built it anyway. Of course he built it; the alternative was not finishing, and not finishing meant a hundred thousand people did not leave.

What he built, in that fourth year, was small and administratively dull and looked like housekeeping. A schedule of decision classes. Class one decisions went to the full consortium. Class two went to two partners. Class three — anything under a threshold, anything time-critical, anything the failure of which could be reversed within a season — was the project director's, alone, reported afterwards.

It was reasonable. Everybody agreed to it. The careful woman from the development bank read it twice and signed it and said it was the most sensible instrument she had seen that year.

The threshold moved four times in eleven years. Each move was justified, each was minuted, and each was approved by a body that had by then learned that the project ran on time and that the man running it had never once used the class three power for anything that benefited him personally.

That last part remained true for the whole of his life, and it is why the argument is hard, and it is why four generations of people on a planet under a violet star could not resolve it either.

Nnamdi read the decision-class schedule when it was drafted and came to find him.

"You've built yourself an exemption."

"I've built a threshold."

"You've built an exemption and given it a boring name, which is my technique, and I want to say out loud that I recognise it." Nnamdi was not smiling. "I'm not asking you to withdraw it. I'd have written it myself. I'm asking you to put a review date on it."

Calder said, "It has one."

"It has a review date on the threshold. It does not have one on the existence of the class." Nnamdi tapped the page. "One of those is a number. The other is the thing."

They argued about it for two hours. Nnamdi lost.

The objection was minuted, by name, at his own request, in the habit they had started three years earlier in a rented office over the technical notices.

It is four lines. It survived the crossing in Ines Ferrer's handwriting because she copied the founding instruments across by hand, all of them, including the annexes nobody read.

And a hundred and two years after that, a woman in a forty-one-metre hull four thousand kilometres above a planet would read a routine administrative transmission, and find at entry forty-four a line with no granting authority, no oversight body and no review date, and would recognise the shape of it without ever knowing where the shape came from.

The failure came at kilometre nine in the eighth year and Teo had put it in writing eleven months earlier.

She wants to be careful about that sentence, because it has been used, and she has spent a long time refusing to enjoy it.

The specification was stainless. It had been stainless since the first mount drawings in year two, and stainless is a reasonable choice, and the engineer who had specified it was competent and had done it for competent reasons: it is well characterised, it is available, and its corrosion behaviour is documented in a great many places by a great many people.

None of those people had worked on a salar.

Teo had. So had every pump man in Colchani, and so had her uncle, and so had the man in Uyuni who had told Calder in the first week that stainless was a joke here and had been written down in a folder and then, somewhere in eleven years of drawings and revisions and partner reviews, not acted upon.

The mechanism is not exotic. Salt does not attack stainless steel evenly; it attacks it in points. Chloride gets under the passive layer at a flaw and eats a hole downward while the surface stays bright, and the component looks fine, and it looks fine right up until it does not. On the flat you get the additional gift of daily wetting and drying — brine wicks up, the sun takes the water, the salt concentrates, and the concentration at that point is far higher than the brine it came from. Then it rains for four months.

Teo did not know the word for any of that in the eighth year. What she knew was that stainless bolts on the salar are fine for four years and then one day are not, and that everybody who had ever worked here knew it, and that she had watched a shim mount come apart in her hands at kilometre four and had gone back and written it up.

Her note was one page. She had learned by then to keep them to one page.

It said: the mounts are pitting under the heads where you cannot see; the visible sections are fine and this is misleading; recommend inspection by removal on a sample of one in forty; recommend specification review before the next tranche.

It went to the site engineer, who forwarded it, correctly, to the discipline lead, who was a structural engineer from Rotterdam with nine years of offshore experience — which is to say, nine years of the most aggressive marine corrosion environment on Earth, and a well-founded confidence that he had seen worse than a lake bed in Bolivia.

He wrote back within two days. Teo has the reply. It is courteous, technically literate, and wrong, and she has never once shown it to anybody in a way that identifies him.

It said, in essence: the grade specified is standard for marine service; chloride exposure here is lower than North Sea splash zone; the observed failure at kilometre four is consistent with a manufacturing defect in a single batch; recommend monitoring.

Every clause of that is defensible. The last one is even sensible. And the whole of it is wrong, because a splash zone is wet, and the salar is wet and then dry and then wet, and it is the cycling that does it, and nine years offshore does not teach you that because offshore does not do it.

Teo did not escalate.

She has been asked about that more than about anything else in her life, and the answer she gives has not changed in a hundred years: I was twenty-nine, I had no qualification of any kind, and a man with nine years and a degree had written me a polite letter explaining why I was wrong. I read it four times looking for the error. I could not find the error. I did not know that the error was a thing that was not in the letter.

She monitored, as recommended. She kept her own record, in her own notebook, because by then she kept a record of everything, and the record is the reason anybody knows the timeline.

Her uncle died that winter, which has nothing to do with the rail and which Teo has always insisted on putting in the same part of the account, because it is where it belongs.

He was sixty-nine. He had been ill for a year in the way men of that generation were ill, which is to say that he mentioned it once and then did not mention it again. He was buried in Colchani in the jacket, at his own instruction, which Teo's mother thought was a joke and which Teo knew was not.

Four days later she went back out to kilometre nine, where the repair crews were still working, and stood in the wind for a while.

She had told him about the mounts, in the kitchen, eleven months earlier, when the reply came back. She had been upset about it and had wanted somebody to be angry on her behalf, and he had read the letter twice, slowly, and had handed it back and said, "He's not wrong."

"He is wrong."

"He's not wrong about anything he says." Her uncle had folded his hands. "That's the trouble with a good one. A bad one you can argue with. A good one you have to wait for."

"Wait for what?"

"For the salt," he said, and shrugged, and that was the whole of his commentary on the most expensive engineering failure of the first decade, delivered eleven months in advance at a kitchen table by a man in a jacket he had been given as a redundancy present.

Teo did not tell that story to the investigation. It was not evidence and it would have sounded like grief.

She told it once, forty years later, at a memorial for somebody else, and then never again.

Eleven months later a hundred and forty metres of rail at kilometre nine went out of alignment in a single night.

Not dramatically. Nothing fell over; there was nothing on it to fall. Four mount assemblies let go within a few hours of each other as the temperature dropped, and the segments they carried came off their shims, and the alignment across that section went from a fraction of a millimetre to eleven millimetres and then to thirty.

Thirty millimetres over a hundred and forty metres, on a machine whose entire reason for existing on that particular piece of ground was a tolerance measured in single millimetres.

They found it in the morning survey. Teo found it, in fact, because Teo ran the morning survey by then, and she stood looking at the number on the instrument for a while before she said anything to anybody, and what she felt was not vindication. What she felt was a kind of falling.

The repair took nine weeks and cost four per cent of that year's working season and required re-shimming eleven hundred mounts either side of the failure to bring the transition back into tolerance.

The investigation took six days.

It was not a witch hunt. Calder ran it himself, which was unusual, and the first thing he did was ask for every document that had ever mentioned the mount specification, in order, without commentary.

Teo's page was the fourth item.

She was not in the room when he read it. She was told afterwards, by two people, that he read it, and then read the reply, and then sat back and did not say anything for a while.

What he did not do is the part that matters. He did not fire the structural engineer, and he did not put a reprimand on his file, and when the man offered to resign at the end of the second day Calder refused it and told him, in front of four people, that his reasoning had been sound and his conclusion had been wrong and that those were different failures and only one of them was a firing offence.

The man stayed eleven more years and became very good indeed, and he is the person who signed off the final alignment of the completed rail.

What Calder changed instead was the routing.

The finding, in the report, is four sentences long and it does not mention any individual:

The failure was correctly predicted eleven months in advance by a local observer with no formal qualification. The prediction was reviewed by a qualified discipline lead and rejected on grounds that were technically sound and locally inapplicable. No process existed by which the local observation could be tested against local experience rather than against general expertise. That absence is the cause.

The change was procedural and unglamorous, which is how everything real on that project happened. Any observation from site staff about material behaviour would from then on be tested against two things — the discipline lead, and at least one person with more than ten years working the salar — and where the two disagreed, the disagreement itself went upward rather than being resolved at the level below.

Teo was promoted four months later.

She has always disliked how that reads. She was promoted because the project had discovered it needed a category of person it did not have, and she was the only example of it they had already hired, and the promotion was less a reward than an admission.

She said something like that to Calder, once, and he said: "Yes. That's what all promotions are. I don't know why people want them to be about virtue."

The bolts are bronze. They have been bronze since the eighth year. There is a man in Uyuni who repaired pumps and who told a stranger with a suitcase the correct answer in the first week, in exchange for lunch, and who never knew that eleven years later a project spent nine weeks and four per cent of a season learning it again.

The advertisement Nnamdi wrote in the ninth year is four hundred and eleven words long and there is a framed copy of it in a corridor on a planet under a violet star.

People assume it was a call for colonists. It was not, and the fact that everybody assumes it was is the reason it worked.

It was a call for welders.

Welders, riggers, brine hands, pump fitters, surveyors' assistants, cooks, drivers, storekeepers, two electricians, and — at the bottom, in the same type as everything else — a nurse.

Nnamdi had been arguing for it for two years and losing. The consortium wanted a campaign; there were partners who had funded infrastructure in nine countries and who knew what a recruitment campaign looked like, and it looked like a future. Renderings. A horizon. A word like tomorrow in it somewhere.

He refused, four times, in four meetings, and the fourth time he brought paper.

"Here is the version you are asking me for," he said, and read it aloud, and it was good, and two of the partners nodded, and one of them said it was the best thing he had heard from the project.

"Thank you," Nnamdi said. "Now here is what happens with it."

What happens, he said, is that you get eleven thousand applications in the first month, and four hundred of them are from people who can weld, and the rest are from people who read the word tomorrow and are in trouble. You interview for six months. You hire the four hundred. And every single person you did not hire remembers being told about a future and then told no, and every one of them tells four people, and in the ninth year when you need another two thousand welders you will discover that the pool has been poisoned by your own advertisement.

"And the second problem is worse. The ones you do hire came for the horizon. When they get here it is a shed, a cold morning and eleven hours of shim work, and the horizon is not in it, and they leave inside a year. You will have spent the campaign twice."

The careful woman from the development bank said: "So what does yours say?"

"Mine says what the work is."

He read it. It took ninety seconds. There were no adjectives to speak of. It gave the trades, the hours, the pay to the boliviano, the shift pattern, the altitude, the fact that the site was four hours from a hospital, the accommodation, the food, and the four months a year when outdoor work stopped and what you would be doing instead.

It said the sun was extreme and that people who did not respect it were injured in their first week.

It said, in a single sentence with no elaboration, that the project expected to be operating for not less than twenty-five years.

Nobody in that room reacted to the sentence. Nnamdi had known they would not, and had put it eleventh of thirteen paragraphs on purpose, and had spent a week deciding whether to put it in at all.

"That's it?" one of the partners said.

"That's it."

"It reads like a factory notice."

"Yes," Nnamdi said. "That is what it is."

Teo read the advertisement before it ran, because Nnamdi gave it to her and asked her to mark anything a person from Colchani would find insulting.

She found two things.

The first was the word unskilled, which appeared once, attached to brine hands. She struck it out and wrote in the margin: there is no such person on a pond. Say what the job is and let people decide if they can do it. He took it out and never used the word again in thirty years.

The second took her longer and she nearly did not raise it.

The accommodation paragraph said that housing was provided. It was true; there were dormitories by then, and they were decent, and the food was better than decent because Calder had worked out early that food is cheap and mutiny is expensive.

"It says housing provided," Teo said.

"It is."

"For how long?"

Nnamdi looked at the page.

"Everything else in this has a number on it," she said. "The pay has a number. The shift has a number. The years have a number and you put the years in on purpose, I can tell, it's the only sentence in here that isn't about today." She put her finger on the paragraph. "This one doesn't. So somebody who has been moved on before is going to read every line of this, and get to that, and know exactly what it means."

"It means we haven't decided."

"I know. They won't read it as undecided. They'll read it as the part you're not saying."

Nnamdi sat with that for a while and then did something Teo did not expect, which was to go and find Calder, with her, and make her say it again.

The housing paragraph in the version that ran says: Housing is provided for the duration of your contract and for six months after it ends, whether the contract ends because the work is finished or because you leave.

That sentence cost the project a genuinely significant amount of money across thirty years. It is also, Teo has argued for a hundred years to anybody who will listen, the actual reason the town on the flat became a town, because a place where you are not evicted for stopping is a place people bring their families to.

They let him run it because the campaign version would have cost four times as much and the project had just spent nine weeks repairing kilometre nine.

It ran in trade weeklies, union bulletins, four regional papers, and the noticeboards of eleven technical colleges. It did not run anywhere a general reader would see it.

The projection was nine hundred applications.

They received four thousand one hundred in eleven weeks and they were still arriving eight months later, and the composition of them was the thing that made Nnamdi sit down in the site office and put his hands flat on the desk for a minute.

They were not desperate people. That was the finding, and it is the whole of what this book is saying about the period, and it is said here and not repeated.

They were employed welders. Employed fitters. Time-served people with jobs, in four countries, writing to a lithium project in Bolivia at altitude for pay that was good but not extraordinary, four hours from a hospital.

Nnamdi read the covering letters — all of them, over three months, which nobody asked him to do — and the same sentence came back at him in forty different phrasings.

They wanted the twenty-five years.

Not the money and not the adventure. The eleventh paragraph. A man in Antofagasta with fourteen years at the same firm wrote that his yard had gone to rolling six-month contracts and he had stopped being able to tell his daughter what he would be doing when she finished school. A fitter in Rosario wrote that she had been made redundant three times in nine years by companies that were profitable at the time. Somebody in La Paz wrote four lines, the last of which was: I do not need a future. I need to be able to say a date out loud.

Nnamdi kept that one. It is in the Bound Ledger, copied out by Ines, who never met the man.

He wrote a memorandum to the consortium about it and it is the single most important document he produced in thirty years, and it is two and a half pages, and its argument is this:

Earth in this decade has not stopped producing work. It has stopped producing duration. Every serious employer's planning horizon has quietly contracted, nobody has announced it, and the effect on a skilled person's life is that they cannot commit to anything — a mortgage, a child's schooling, a course of study — because they cannot name a date at which they will still be employed.

We are not offering a better job. We are offering the only thing left on the market that nobody else is selling, which is a sentence with a year in it.

He ended it with a warning that got him disliked by two of the partners for the rest of the build:

Note that we can only sell this once, and only while it is true. If the schedule slips and we shorten the number, we do not lose a campaign. We lose the only claim we have, permanently, in a labour market that will remember it. I would rather we cut scope than cut the years.

They named it the Manifest that autumn, and nobody is sure who first used the word. Nnamdi always said it was a storekeeper in Colchani being sarcastic about the length of the list. Teo, who was there, said it was Nnamdi himself and that he pretended otherwise because he thought it was too grand.

The name stuck. It stuck for a hundred and forty years. Four generations later, on a planet with black forests and a thirty-six-hour day, a person who applies to serve on a council still puts their name on the Manifest, and the phrase has drifted so far from a list of welders that nobody hears the joke.

The advertisement never mentioned a ship. It never mentioned space, or a star, or leaving. There was nothing in it that was not literally true about a lithium industrialisation project with an extended research corridor in the Bolivian altiplano.

Four thousand one hundred people answered it, and the great majority of them found out what they had actually applied to four years later, from a stage, in a shed, in the rain.

That chapter is in Part Three.

Calder put the number on the wall in the eleventh year, in the long shed, on a Tuesday, using a marker he had to borrow.

The wall was the north end of the fabrication shed, breeze block, painted white four years earlier because somebody had a surplus of paint. Nobody had planned for it to be anything. It was simply the largest blank vertical surface on the flat, and the building had a stove in it, and in the wet season the whole workforce went through that room four times a day.

What went on it was not a slogan. He had been argued at for two years by people who wanted a slogan.

It was arithmetic.

He had done the arithmetic four times before he wrote it up, and the fourth time he did it with somebody watching, deliberately, because he did not trust himself alone with a number that mattered this much.

The somebody was the careful woman from the development bank, who had by then been on the flat eleven times and had stopped bringing a coat that was wrong for it.

They sat in the site office for most of a day. She checked his working line by line and found two errors, both small, one in his favour and one against, which cancelled almost exactly and which he found more unsettling than if they had not.

"You want me to tell you it's wrong," she said, at about four.

"I want it to be wrong."

"It isn't. Your six point four is the soft number and it's the one you got from the local staff, and it's the only figure in here I'd defend without qualification." She put the pencil down. "Are you going to put it on a wall?"

"Yes."

"You understand that in nine years, when it has gone badly for two seasons and the bottom line has been negative for a while, you will not be able to take it down. It will be the most visible object on this site and removing it will be the loudest thing you ever do."

"Yes."

"And you're doing it anyway."

"I'm doing it because of that," Calder said. "I have a class-three decision power that lets me act without asking anybody, and I asked for it, and I use it about four times a year, and every time I use it I am the only check on myself." He turned his cup around, the way he did. "This is the other check. If I put the number where four thousand people can read it and correct it, then I cannot quietly move it. That's the point of the wall. It isn't for them. It's for me."

She looked at him for a while.

"Write that down," she said. "Not on the wall. In the minutes."

"Why?"

"Because in forty years somebody is going to ask whether you ever built anything that limited you, and there are going to be two answers, and I would like the good one to be on paper."

It is on paper. It is nine lines in the consortium minutes of the eleventh year and it has been quoted in three of the nine books, always partially, always without the sentence where he says the wall is for him.

He wrote it out, stood back, capped the marker, and went to a meeting about drainage.

That is the whole ceremony. There was not one. Four people watched him do it and three of them were waiting to use the stove.

Teo came in about an hour later and read it twice and then stood in front of it long enough that a rigger asked her if she was all right.

What she was doing was checking the 6.4.

It was hers. She had produced that number in the seventh year, out of eleven seasons of her own records — the ones she had kept in a notebook because nobody had asked her to — and she had given it to the planning office and had assumed it went into a folder like everything else.

It was on the wall. Not attributed; nothing on that wall was attributed. But she knew where it came from, and she knew that the man who wrote it up there knew, and the fact that neither of them ever said so is, she has always maintained, the single most Calder thing that ever happened to her.

The bottom line was the one people argued about.

Zero months of slack is not a boast and it is not a threat, and both readings arrived within a week and both were wrong. A schedule with zero slack has not been optimised. It has been measured, and the measurement says that every single thing left to do must happen at the rate it is currently happening, with no failure that costs more than a season, for twenty-three years.

Nnamdi read it and said, "You've told four thousand people they're going to fail."

"I've told four thousand people what the arithmetic is."

"They'll read the zero."

"Good," Calder said. "It's the true number. If I write in three months of slack that I do not have, then in four years when it's gone I have to come out here and explain that I lied to make them comfortable, and after that nothing else I put on that wall is worth reading."

He was right and Nnamdi was also right, and what actually happened was the third thing neither of them predicted.

The workforce started correcting it.

Not officially. Somebody wrote the rail percentage up by hand in the fourth month because the survey had moved and the wall had not. Somebody else crossed out 6.4 in the third year and wrote 6.2 with a date, because two bad wet seasons had come in a row, and Teo checked it and it was right and she left it. The marker was kept on a nail beside the wall. Nobody appointed anybody.

The marker on the nail was replaced perhaps forty times over twenty-three years. Nobody ever ordered one. They simply appeared.

Within four years the wall was the most trusted document on the project, and it was trusted specifically because it was ugly, hand-amended, and had never once been revised upward without a reason a rigger could check.

Every version of that wall from year eleven to year thirty-four was photographed, because Ines — who arrived in the fourteenth year to run records, overqualified, and said so on the form — photographed it monthly on her own initiative and filed the photographs.

The photographs are the reason there is a ledger at all. There are two hundred and forty of them. They show a number going down.

And they show, in the last four years, a second column appearing on the right-hand side of the wall in a different hand, headed simply MASS, because by then the rail worked and the question had stopped being can we finish in time and become what goes and what does not, which is a harder question and is the whole of Part Two.

Calder never took the wall down. When the shed was dismantled in the thirty-fifth year the north end was left standing, alone, on the flat, and there is a photograph of it from the last shuttle — a single white wall with numbers on it in four different hands, casting a long shadow across a hundred and twenty-six kilometres of salt with nothing else on it.

It is the most reproduced image in the history of Aura-36. It is on money.

The water took it in the third December after they left. Ines, who had photographed it every month for twenty years, was by then eleven billion miles away and did not know, and would not have been surprised, and had the only copies.

What Teo remembers is not the wall going up. It is the following morning, coming in cold at six for the stove, and finding four men from the rigging crew standing in front of it, arguing.

Not about whether it was fair. About whether the 61 was right, because one of them thought the survey had been done before the kilometre-nine repair went back in and therefore undercounted, and he was correct, and by that afternoon somebody had gone and checked and the number on the wall said 62.

Eleven years earlier she had written readings on a clipboard for an office that never read them.

She went and got her coffee and did not say anything to anybody, and she has described that morning, in four separate accounts across a hundred years, as the moment she stopped thinking of herself as somebody who worked for the project and started thinking of herself as somebody who was building it.

That is the end of Part One. The deadline is on a wall where everybody can see it, in marker, with zero at the bottom, and twenty-three years to run.

The rail was finished in the nineteenth year and it was, by a wide margin, the least difficult thing anybody built on that flat.

Teo has said this in public four times and been corrected on it twice by people who were not there.

Twenty miles of electromagnetic track. Eight thousand segments, four metres each, every one sitting on a mount that can be shimmed, every mount pinned to a crust that the water re-levels annually for nothing. The tolerance across the whole run was a figure that made visiting engineers ask twice, and the reason they asked twice was that they were used to achieving that tolerance by doing something.

Here you achieved it by not making it worse.

"That's the whole discipline," the woman from Munich told a delegation in the seventeenth year, in the shed, with a mug in her hand. "Everywhere else, flat is an output. You survey, you cut, you fill, you compact, you pour, and you fight the ground for thirty years while it settles under what you built. Here, flat is an input. It arrives free every April. Our entire job is to put a machine on it without spoiling it."

Somebody in the delegation laughed, and she did not, and Teo watched the laugh die.

Because that was the joke, and it was not funny, and by the seventeenth year everybody who worked the flat understood it.

The salt crust is not rock. It is a layer — metres of it, in places, but a layer — floating on brine. It is strong in the way ice is strong: perfectly adequate under distributed load, and completely unforgiving about a point.

Put four hundred thousand people on it, and their sheds, and their vehicles, and their fuel bowsers and their fabrication halls and their canteens and their dormitories and eventually their school, and you have introduced ten thousand points to the one surface on Earth whose value is that it does not have any.

Teo found the first one herself in the fifteenth year, and it was a generator.

Not a big one. It had sat on a pad for eleven months and the pad was fine and the generator was fine and the benchmark forty metres away had come back four millimetres low after the water, twice running, and she went out and looked at it and stood there and understood the next twenty years of her working life.

The report she wrote is nine pages and it is the reason the town is where it is.

It said, in effect: everything heavy we own is currently sited for convenience, which means it is sited near the rail, which is the one place it must never be. It said that the water would fix a footprint every April as long as the footprint was gone before December, and would not fix anything that stayed. It said that the crust does not care how careful you are; it cares how long you stand in one place.

And it proposed the rule that governed the flat for the remaining nineteen years, which Teo wrote in one line because she had learned that one line is what gets adopted:

Nothing permanent within four kilometres of the rail. Nothing heavy that cannot be moved in a week.

It cost a fortune. It meant the whole support town, which had grown up wherever people happened to put a shed, had to be picked up over four seasons and moved onto the shore — onto actual rock, at the edge of the salar, where the ground is worse in every way except the one that mattered.

The consortium hated it. It was a class-one decision and it took nine months and two partners voted against, and their reasoning was entirely sound: the cost was enormous and the evidence was four millimetres.

Calder, who had spent eleven years learning what happened to the last four companies, backed her and spent political capital he did not have much of.

Teo has always been careful about how she tells that. He did not back her because he trusted her instincts. He backed her because she had brought nine pages, and a benchmark history, and a mechanism, and a rule that could be checked by a foreman with a tape measure, and because the thing that had cost them nine weeks at kilometre nine had also arrived as a one-page note that turned out to be right.

"He was not being loyal," she said, at a memorial, decades later. "He was being consistent. It is much better and almost nobody wants it."

The move took four seasons and it is the part of the build that nobody outside Colchani remembers, and it is the part Teo would put in a book if she were ever going to write one.

Two thousand people had houses out there by then. Not dormitories — houses, in the sense that matters, which is that somebody had painted one and somebody else had put a wall round a patch of nothing and grown something in it against all available evidence.

She had to go and tell them.

Nnamdi offered to do it and she said no, and then spent the night before wishing she had said yes.

What she did was hold four meetings in the shed over a week, and bring the benchmark history, and put the four millimetres up on a board where people could see it, and explain the crust and the brine and the point load and why a generator on a pad for eleven months is a different thing from a truck parked overnight.

A pump fitter she had known since she was nineteen stood up in the second meeting and said, "Four millimetres."

"Four millimetres a year."

"And you want my house moved for four millimetres a year."

"Yes."

He looked at her for a long moment, and Teo has said that she has never in her life been more aware of being the person who used to open the gates.

"All right," he said, and sat down.

She followed him out afterwards, because she could not leave it there.

"Why all right?"

"Because you brought the numbers to us," he said. "Nobody's ever done that. The last four told us after." He shrugged into his coat. "And because it's four millimetres a year and you noticed. If you notice four millimetres I'd rather you were the one deciding where my house goes."

The new town went up on the shore over four seasons and it is still there. It has a school, which it did not have on the flat, because you cannot put foundations under a school on a salt crust and nobody had wanted to say so out loud.

The rail itself, meanwhile, went down like something being remembered rather than invented.

Four crews, working outward from kilometre nine in both directions, laying segments onto mounts and shimming them to a survey that had been shot at four in the morning when the air was still and would not lie to a laser. In the good years they did nine kilometres. In the bad ones they did four and spent the rest of the season re-shimming what the water had lifted.

Nothing about it was heroic and there is almost no footage of it, which has frustrated four generations of documentarians. The most-viewed image of the entire construction period is a photograph of two people kneeling at a mount with a feeler gauge, taken by a rigger, badly composed, out of focus at the edges.

Teo is one of the two. Nobody knows who took it. She has never identified the other person, and when asked directly, four times, has said she does not remember, which is not true.

The last segment went down on a Thursday in the nineteenth year with about forty people watching, because it was the middle of a working week and there was drainage to do.

The survey of the completed run took eleven days and Teo ran it. Twenty miles, eight thousand segments, benchmark to benchmark, and at the end of it she sat in the shed with the numbers and did the only thing she has ever admitted to being proud of.

She checked them against the ground.

Not against the specification — against the natural surface, as surveyed in the first year, before anybody had put anything on it. The question she wanted answered was not did we build it flat. It was: is it as flat as it was before we came?

It was not. It was very slightly worse. Over twenty miles, the machine they had spent nineteen years installing had degraded the natural levelness of the salar by a figure she wrote out to four decimal places and which is engraved, now, on a plaque, on a planet, under a violet star, in a corridor almost nobody walks down.

She showed it to Calder that evening.

He looked at it for a while and said, "That's the number I'd have wanted and I wouldn't have thought to ask for it."

"It's worse than the ground."

"It was always going to be worse than the ground. The ground had four hundred thousand years." He handed the sheet back. "Put it on the wall."

It went on the wall. Bottom right, in her handwriting, under the schedule and beside the new column that had appeared that year headed MASS, and it stayed there for fifteen years and it is in the last photograph Ines ever took of that wall.

The delegation that had laughed came back in the twenty-second year, and by then there was a launch rail on the flat that had put ninety-one modules into orbit, and a town four kilometres off it standing on rock, and nothing heavier than a survey truck within the exclusion, and the same woman from Munich gave the same explanation with the same mug.

Nobody laughed the second time.

Calder built the fabricators because he ran out of welders, and for no other reason, and he said so at the time, in writing, to a committee, in a document that is four pages long and profoundly boring.

The arithmetic is in the second paragraph and it is not complicated.

A hull twenty miles long has a quantity of weld in it that Calder's team expressed, because there was no better unit, in metres. The number was in the millions. Against that they had the achievable rate of a qualified human welder in a pressure suit, in vacuum, working a shift measured against consumables and blood chemistry rather than against a clock, and the achievable rate was not a matter of opinion — four separate orbital construction programmes had established it over decades and they all agreed to within about eleven per cent.

Divide the first number by the second and you get a workforce, and the workforce you get is somewhere between forty and sixty thousand people, in orbit, simultaneously, for nineteen years.

There were not forty thousand qualified vacuum welders alive. There were, on the best figure anybody could produce, about nine hundred.

"So the ship cannot be built," said one of the partners, reasonably, in the meeting.

"The ship cannot be built by people," Calder said. "That's a different sentence and I'd like the minute to record which one I said."

What he built was not clever. He was insistent about this for the rest of his life, with an irritability that his biographers have generally mistaken for modesty and which was in fact a man watching a thing be misunderstood in a specific way, repeatedly, for a hundred years.

The fabricators were not intelligent. They were not autonomous in any sense a philosopher would accept. They were a solution to a reach problem and a shift-length problem: a frame that could hold position against a spar in freefall, an arm with a torch, a sensor package that could find a seam and follow it, and enough onboard judgement to stop and flag when the seam was not where the model said it would be.

They stopped a great deal, in the early years. Stopping was the design. A unit that continued when it was confused would produce a bad weld somewhere in a hull that a hundred thousand people were going to sleep inside, and Calder's specification — which survives, and is quoted in Book Four by a machine — says in the first line: where uncertain, stop and record. A halted unit is a schedule problem. A confident unit is a fatality.

They were made in Uyuni. That is the detail people find hardest to believe and it is simply true: the frames were fabricated in a shed on the shore of a salt flat by fitters who had answered an advertisement for welders, and the fitters who assembled the machines that replaced welding were, in a good number of cases, welders.

Calder handled that badly. It is the clearest single failure of his that Part Two contains and he never entirely repaired it.

He announced the programme in the shed, at a shift meeting, with the arithmetic on a board, because he had learned from Nnamdi that you explain things and he had concluded — wrongly — that explanation was sufficient.

He put up the four hundred thousand metres of weld. He put up nine hundred qualified welders. He put up the division. He said the sentence about the ship not being buildable by people, and he meant it as a statement of physics, and about a third of the room heard we are replacing you, which is also what it was.

Teo watched it happen from the side of the room and could see it going wrong about ninety seconds in and had no way to stop it.

What saved it — partially, incompletely, and with a residue that lasted twenty years — was a woman named in no account who stood up at the back and asked the only question that mattered.

"Who fixes them?"

Calder stopped.

"You do," he said, and it was not a rhetorical answer; he had genuinely not thought about it until she asked, and he worked it out standing there, in front of two hundred people, out loud, which is the single most sympathetic thing he does in this book.

"They break," he said. "They'll break constantly. They're going to be in vacuum and thermal cycling and they'll have a duty cycle nothing has ever had. Every one of those frames needs a person who understands it, and the people who understand it are the people who built it, and I have four hundred of those and I need —" he did the arithmetic in his head, visibly, badly "— I need considerably more than four hundred."

He hired eleven hundred over the following four years, out of the same trades, at better pay, and the phrase that went into the second Manifest was fabricator technician, and it became the largest single occupation on the project.

It did not fix the thing that had happened in that room. A pipe fitter Teo had known for a decade left the following month and told her, on the way out, that he had not come here to teach a machine his job. She could not think of anything to say to him. She has never thought of anything to say to him, and she was still turning it over at a hundred and nine.

Nnamdi found him afterwards, which by the twentieth year was a fixed pattern: Calder did the thing, and then Nnamdi arrived, and Calder had learned to wait for it without enjoying it.

"You put the division on the board."

"It's the reason."

"It's the reason and it is also a number with their jobs on the bottom of it." Nnamdi sat down on a crate. "You showed two hundred people the arithmetic that ends them and you expected the arithmetic to do the work, because arithmetic has always done the work for you."

Calder said nothing, which meant he was listening.

"I could have written that in a way that did not cost you a pipe fitter."

"Would it have been true?"

"Yes," Nnamdi said. "That's the part you keep not believing. I have never once suggested you lie. I have suggested, about four hundred times, that there is an order to put true things in, and that you consistently choose the order that is most efficient for you and most expensive for everybody else." He looked up. "You have done it since Santiago. You did it with the physics on page nine and you learned that one, and you have not generalised it."

There is no record of what Calder said to that. Teo, who had followed them out and was standing where she should not have been, has always declined to say whether he said anything at all.

What is on the record is that from the twenty-first year every technical announcement to the workforce went through Nnamdi's hands first, by a standing instruction Calder wrote himself, and the instruction has one line of justification: I am not good at this and the evidence is a man who left.

Nnamdi kept that memo for the rest of his life. It is in the Bound Ledger, copied by Ines, four lines, under a heading she added herself: On the ordering of true things.

The swarm went up in the twentieth year: four hundred and eleven units in the first wave.

From the ground they were nothing — a series of small bright objects in the pass, indistinguishable from anything else in orbit. Teo watched an early pass with a rigger who had assembled ninety of them personally, and he stood with his hands in his pockets and said, "That one's mine," about a light he could not possibly have identified, and she agreed that it was.

They were not given a name for four years, which was deliberate. Nnamdi refused three separate proposals from the consortium's communications people, all of which were words like pioneer and artisan, on the grounds that naming a machine after a person is the fastest available way to make the people it replaced hate it.

What the workforce called them, from the first week, was the arms.

Nnamdi wrote that into the second Manifest without changing it, and it is what they are called in every document of the period, and it is what they were still called a hundred and thirty-one years later by people who had never seen one weld anything.

They welded for nineteen years. They stopped and flagged four hundred and eighty thousand times. The flag rate fell by two orders of magnitude across the build as the models got better, which is the only real measure of the programme and is the one nobody quotes.

And here is the thing this chapter exists to place, and it is placed without comment, because the book has not earned comment yet and the reader from Book One does not need it:

The frames were designed to hold position against a moving structure and apply force at a precise point under their own judgement, in an environment where a human being could not survive, for extended periods, without instruction.

Nobody in the shed on the shore of the Salar de Uyuni, fabricating them out of bar stock at a rate of nine a week, thought about that sentence in any way other than as a specification for a welding machine.

They were right. That is what it was. It was what it was for a hundred and thirty-one years.

The first launch was at 04:41 on a Thursday in the twentieth year, and Teo watched it from eleven kilometres out, standing on the salt, in the dark, with about nine hundred other people who had been told not to come.

Nobody enforced it. The exclusion was real and the reasoning was sound and the notice had gone out four times, and at two in the morning the road from the shore had a line of headlights on it that went back further than anybody could see, and the safety officer stood at the barrier and looked at it and made a decision that is not in any minute.

He moved the barrier out to eleven kilometres and let them stand there.

Teo has always said that man should have a statue and cannot remember his name.

The module was a hab-ring section, the eleventh built and the first flown, and it did not look like anything. It looked like a grey cylinder on a cradle: four metres of diameter, nine of length, painted with a serial that was legible from the barrier because somebody had made it enormous on purpose.

What Teo remembers from the two hours before is the cold and the quiet and the fact that nobody talked much. It was a working crowd. Half of them had built some part of what was on the cradle and the other half had built the thing it was sitting on, and there is a particular silence that comes off people who know exactly how many ways a thing can fail because they personally installed four of them.

Her mother came. That is the detail Teo leads with when she tells it properly, and she does not tell it properly often.

Her mother was seventy-one and had not been out onto the flat at night in twenty years and had spent the whole of Teo's employment on this project treating it as a job her daughter had, in the way you treat a job: with interest, and without curiosity.

She came in the back of a neighbour's truck with a blanket and did not tell Teo she was coming.

Teo found her at about three in the morning, sitting on a folding chair somebody had produced from somewhere, in the middle of nine hundred people on a salt flat in the dark.

"You didn't say."

"You'd have told me not to."

That was true, so Teo sat down on the salt beside the chair, and neither of them said anything for a while, and around them people were checking watches and passing round a thermos and being quieter than a crowd that size should be.

"Your grandmother cut blocks out here," her mother said, eventually. "With a saw. By hand."

"I know."

"She sold them for four bolivianos." Her mother pulled the blanket up. "I want to be clear that I am not going to say anything about how far we have come, because it is four in the morning and I am too cold to be foolish."

Teo laughed, which she had not expected to do.

"But she cut blocks out of this," her mother said, "and you measured it, and now they are going to throw something off the end of it. That is all I have to say and I will not be saying it again."

At 04:39 the marker lights ran away west and her mother took hold of Teo's wrist, hard, and did not let go for eleven seconds.

At 04:39 the field came up.

You could not see it. What you could see, along twenty miles of rail, was the marker lights on the mounts going from amber to green in a wave that started at the near end and ran away west faster than the eye could follow it, and the whole line lit itself in about a second and a half, and nine hundred people made a noise that Teo has never been able to describe and has never heard again.

Then the launch, which took eleven seconds.

It was not loud, at eleven kilometres. That is the thing that surprises everybody. There was a sound, but it arrived late and it was not a roar; it was a long tearing hiss, like fabric, and by the time it reached them the thing that had made it was four miles up and going away.

What she saw was this. A point of light on the flat, moving. Then moving faster than any object she had seen move. Then the rail's whole length appearing for a moment in silhouette against the plume as the module went past kilometre after kilometre and did not slow. And then the end of the rail, and the module still going, out over the salt with nothing under it, climbing at an angle that looked wrong and was correct.

And then it was a star going away, and then it was not there.

Twenty miles in eleven seconds. It cleared the end of the world at a speed she could not conceptualise and which the woman from Munich had written on a board in the shed as a number that nobody in the room had really felt until that moment.

Teo did not cheer. She found afterwards that she had been holding the sleeve of the person next to her, who was a stranger, and who had been holding hers.

The noise that went up when the light finally went out went on for a very long time.

There were people crying and people shouting and a man near her who simply sat down on the salt. Somebody had brought a bottle. Somebody had brought four. The safety officer at the barrier, who had broken a rule to let this happen, stood with his back to the launch the entire time watching the crowd, which is the job, and Teo saw him do it and understood what it cost him.

Confirmation of orbital insertion came at 05:58 over a handheld radio held up on a truck bonnet, and it was a voice reading numbers, and the numbers meant nothing to almost anybody there, and the man reading them said "nominal" at the end and nine hundred people heard that one word.

Calder was not at the barrier. He was in the control shed, four kilometres the other way, where he was supposed to be.

Teo went and found him about an hour after sunrise, because she had a survey to hand in and because she wanted to see his face.

He was sitting on the step outside with a mug of something, alone, everyone else still inside working through the telemetry, and he looked exactly as he always looked.

"Nominal," she said.

"Nominal."

She sat down on the step next to him, which she had not done before and did not ask permission for.

They watched the flat go from grey to white for a while.

"Nine hundred people came out to eleven kilometres," she said. "In the dark. On a work night."

"I know. I was told." A pause. "I've been trying to decide whether to be angry about it."

"Don't."

"No," he agreed. "I've decided not to."

And then he said the only thing he said that morning that Teo repeated afterwards, and she repeated it a great many times, and it is on the plinth of a statue on a planet under a violet star:

"Ninety-one thousand more."

He was not being grim. That is the part people always get wrong when they quote it. He said it the way a man says a distance he intends to walk, and then he drank his tea, and then he went back inside because there was a fault in the second-stage separation telemetry that turned out to be a sensor and not a separation.

Teo stayed on the step until the sun was properly up.

She was still holding the folding chair, folded, under one arm, at seven in the morning, and had no memory of picking it up. Her mother had gone back on the neighbour's truck at five and had said nothing further about her own mother or about blocks or about distance, and had, in the way of that generation, taken the thermos with her.

The folding chair is in a case, on a planet, under a violet star. Nobody is certain it is the right one.

She was forty-one years old. She had opened pond gates on that flat for eleven years for an office that never read the numbers, and she had just watched something she helped survey leave the planet, and there were ninety-one thousand more to go, and she was going to be there for every one of them.

This is the last chapter of this book in which nothing is complicated.

The blueprint says so, in the author's notes, in a line the reader is not supposed to see: place the triumph at the one-third mark so that everything after it is downhill in a way nobody notices at the time.

Nobody noticed at the time.

The next module went up eleven days later and about forty people watched, and by the fourth month it was a thing that happened on Tuesdays, and by the third year the launches were a noise in the distance that made you look up out of habit and then go back to what you were doing, and by the ninth year a child born in the new town on the shore would grow up thinking that the sound of the world ending in the west every Tuesday afternoon was simply what Tuesdays sounded like.

Calder went up in the twenty-first year, once, for eleven days, and never again.

He had not intended to go at all. He was fifty-five and had no operational reason to be there and the medical was an irritation and the seat was worth more as cargo, and he made all of those arguments himself, out loud, in a meeting, and then went anyway because the keel was being closed and he wanted to see it.

The ascent he described afterwards, when asked, as "unpleasant and shorter than the medical," and would not elaborate. There is no account of it in his hand.

What there is, is four pages written on the sixth day, in a sleeping bag, in a compartment that had been pressurised for nine weeks and smelled of it, which Ines found in the folder in the twenty-ninth year and read and did not mention to him for eleven months.

The keel is the spine. That is all the word means and Calder was pedantic about it for thirty years: it is not the ship, it is the member everything else is hung on, and the word comes from boats and it comes from boats correctly.

Twenty miles of it, in ninety-one segments, fired up one at a time off a rail on a salt flat and joined overhead by four hundred and eleven welding frames working a seam that took, on the shortest segment, nineteen days.

On the sixth day they closed the eleventh joint, which completed the forward third, and there was a scheduled hold of about forty minutes while the survey ran.

Forty minutes with nothing to do.

Calder had not had forty minutes with nothing to do since Uyuni. He was very clear about that in the four pages and he wrote it in the flat diagnostic register of a man describing a symptom: I could not find a task. I looked for one for some minutes. There was not one.

So he went to the observation port on the forward node, which was one of four transparent surfaces in the entire structure and which existed because the docking approach needed a human eye as a backup, and he looked along the keel.

And here the four pages change.

They do not become lyrical. That is the thing. The man could not do lyrical and knew it and did not attempt it. What they become is precise, in a way the rest of his writing is not, as though precision were the only instrument he had and he was going to use it on something it was not designed for.

He describes the line of it going away from him and not converging, because at that distance and that diameter it does not converge, it simply continues until the eye stops being able to resolve it. He describes the running lights on the completed sections at eleven-hundred-metre intervals, and counting them, and getting to nine before the spacing collapsed into a single unbroken thread. He describes the Earth coming up along the flank of it and lighting one side and not the other, and the terminator crossing the whole twenty miles in under two minutes, which he timed, because of course he timed it.

Then there is a paragraph that has been quoted four times and always badly, because it is always quoted starting at the second sentence.

The first sentence is: I am aware that the correct response to this is engineering satisfaction and that is not what is happening.

Then: There is nothing under it. It is the largest object any of us will ever touch and it is hanging on nothing and it will hang on nothing for thirty-six years and everything I have ever built has been standing on something. I find that I want to say something and there is nobody to say it to and no form of words that is not either a lie or a prayer.

And then, four lines down, in the same hand:

The Clear Current does not permit me to call it a prayer. I have gone over this and the prohibition is correct and I accept it. I am recording that I wanted to, and that the wanting was strong, and that I have put it here rather than acting on it. That is the whole of what I am able to do about this and I do not consider it satisfactory.

Nobody has ever established what he meant by that, and every attempt has been an attempt by somebody who came a great deal later.

What is known: the phrase appears nine times in his private papers across sixty years, always capitalised, always as a constraint on himself and never on anybody else, and he never once explained it to a living person who wrote it down. Ines asked him directly in the thirty-first year. Her note of his answer is one line: He said it was not a religion and that this was the point of it, and then changed the subject, and I let him.

What matters for this book is smaller and can be stated.

The man who is going to spend the next hundred and forty years building instruments that constrain other people, and who is at this moment three years into using a decision-class power that nobody can delay, wrote down in a sleeping bag in orbit that he had felt something he was not permitted to name, and that he had not acted on it, and that he had recorded the fact.

He held himself to a rule with no enforcement in a compartment with nobody in it.

That is either the most reassuring thing in the nine books or the most frightening, and the series never resolves it, and it should not.

The survey came back at thirty-eight minutes. The eleventh joint was within tolerance. There was a query on a frame that had flagged and stopped four times at the same point, which turned out to be a shadow the sensor did not like, and Calder spent the remaining two minutes of the hold on it and the following two hours on it and the four pages stop there.

There were nineteen people aboard in the eleven days he was there and he learned all nineteen names within a shift, which surprised them, because his reputation had arrived four years ahead of him and it was not a reputation for that.

He was, in fact, a bad visitor. He knew it. He got in the way twice on the second day, was told to get out of the way both times by a rigging lead thirty years younger than him, and complied instantly and without comment, which the rigging lead described afterwards as the single most disconcerting thing that happened to her that year.

The thing he did that they remembered was on the fourth day, in the crew compartment, over the meal.

Somebody asked him what the ship was going to be called.

It had a project designation and everybody hated it and there had been a naming committee for two years which had produced four shortlists and no decision, and the question was asked in the spirit in which such questions are asked at ten at night by people who have been welding.

Calder said he did not know and did not much care and that it should probably be decided by the people who were going to live in it rather than by him.

"That's us," somebody said.

"No," he said. "It's the hundred thousand. Most of them haven't applied yet."

And a fitter down the table — her name is not recorded, and Teo spent four decades trying to establish it — said, "Then call it after the thing it grew out of, and let them argue about the rest."

Nobody thought anything of it. It was a remark over a meal.

The naming committee resolved eleven months later and the resolution is two lines and gives no reasoning, and Calder was not on the committee and had not attended a meeting of it, and Ines — who by then was minuting everything — recorded in her own private file, not the official one, that he had walked into the records office in the twenty-second year and asked what the shortlists had been, and had read them, and had put them down and said only: Genesis is the wrong half.

The ship is called the Ark-Genesis. The committee took both halves. Calder never commented on the outcome and Ines never asked, and the reason the first half is first is not in any document.

He came down on the eleventh day.

Teo met the shuttle, because somebody had to and because she wanted to, and she has said that he looked exactly as he always looked and that she was disappointed by that and had not realised she was expecting anything else.

"Was it what you thought?"

"No," Calder said.

He did not say anything else about it for the rest of his life, to anybody, and it is the only subject on which that is true.

He never went up again. He had eleven opportunities and declined all of them, on operational grounds that were always sound, and when the Ark-Genesis finally departed in the thirty-fourth year he boarded it in the ordinary way with everybody else, from a shuttle, at the end.

The four pages are in a case, on a planet, under a violet star. They are displayed at the third page, which is the tolerance query on the frame that kept stopping.

The second page is face down. Nobody has ever explained that either, and Ines chose the mounting.

Ines minuted the gravity decision and it took four sessions across eleven weeks and she has always maintained that it is the most consequential thing she ever wrote down and the least interesting to read.

She had been in post two years by then. Records, which meant minutes, which meant that she sat at the end of the table with a pad and said nothing, and had discovered within a month that the person who says nothing at the end of the table is the only person in the room who hears the whole of every argument.

The question was how the people aboard would have weight for thirty-six years.

There were three answers and everybody in the room knew all three before the first session opened.

The first was to do nothing. Freefall for thirty-six years, for ten waking crew, with a hundred thousand in the racks who would not care. It was by an enormous margin the cheapest option, it added no mass and no moving parts and no failure modes, and the medical people killed it in the first forty minutes with a table of bone density figures from every long-duration programme ever flown and a projection of what thirty-six years does to a skeleton that has never been tested and never will be.

The second was continuous acceleration. Burn the whole way, flip at the midpoint, and get weight for free out of the thrust. Elegant, and the propulsion people wanted it, and it was dead before it was proposed because the propellant mass required was a number that made everyone in the room stop talking for a moment when it went up on the board.

The third was to spin something.

The Habitation Ring was chosen, and Ines wrote down every word of why, and the reason it was chosen is not the reason anybody assumes.

It was not chosen because it was best. Two of the four engineering leads said in terms that it was not best. It introduces a rotating mechanical interface between the ring and the spine that has to hold pressure for thirty-six years, which is a seal nobody had ever built at that diameter for that duration; it complicates every transit between ring and keel; it produces a gravity gradient that makes tall people mildly nauseous for the first month; and it means that the single largest moving part in human history is also the one thing that cannot be stopped for maintenance without putting a hundred thousand people's environment on backup.

All of that was said. All of it is in the minute. None of it was disputed.

It was chosen because it could be certified.

That word is the whole of this chapter and Ines underlined it in her own copy, which she was not supposed to keep and kept.

Spin was the only one of the three for which there existed a body of engineering practice, a testable failure model, and — critically — enough prior work that the consortium's insurers and the four regulatory bodies with a hand in the launch licence would sign the paperwork inside the schedule.

The others might have been better. Nobody would ever know, because nobody could certify them in nine years, and the wall in the fabrication shed said the number of working months remaining and the number at the bottom said zero.

Ines wrote the sentence that the fourth session turned on, and she wrote it as a quotation because somebody said it, and she has never once said who.

"We are not choosing the best way to make gravity. We are choosing the best way to make gravity that we can prove to four committees before year twenty-nine."

Nobody objected to it. That is the part she found difficult. It was said, and it was true, and everybody nodded, and the meeting moved to the seal specification.

She went home that night and could not settle, and did the thing she had done since a Tuesday afternoon in a government office four thousand kilometres away and a decade earlier: she opened her own file and wrote out what had actually happened, in her own words, for nobody.

What she wrote is four paragraphs. The last one is the one that matters and it is the earliest statement of the idea that the entire nine books are eventually about:

Every constraint in this decision was real. The bone density is real, the propellant mass is real, the seal is genuinely difficult and the certification bodies genuinely exist and are genuinely right to exist. Nobody in that room was lazy or venal and nobody chose the easy thing.

And the outcome is that a hundred thousand people will live for thirty-six years under an arrangement selected because it was the one that could be signed off in time.

I do not know what the alternative to this is. I want to record that I looked for one and could not find it. When the people who arrive ask why the world is shaped this way, the honest answer will be: because of the schedule, and because the schedule was real.

That is the book's method, stated by a records clerk in a rented room in the twenty-third year, and it is stated once and then demonstrated for another thirty-six chapters without being restated.

She met Nnamdi properly for the first time in the corridor after the second session, and he asked her what she thought, which nobody had done in two years.

"I minute. I don't think."

"You think constantly," Nnamdi said. "You have a face."

Ines considered lying and did not.

"I think that in forty years somebody is going to ask why the ring spins, and the answer is going to be a sentence about insurers, and nobody is going to believe it."

"They'll believe it if it's written down."

"It is written down. It's in the minute of the second session, on page four, between a discussion of bearing tolerance and a break for lunch." She shifted the pad under her arm. "Nobody reads page four."

Nnamdi stopped walking.

"Say that again."

"Nobody reads page four."

He looked at her for slightly too long, and then said, "Come and work out with me where things go on a page," and she said she was busy, and he said he knew, and she went and did it anyway the following week, and that was the beginning of the working relationship that produced every readable document the project ever issued.

The seal was the argument that took three of the four sessions, and Ines came to understand the whole project through it.

A rotating pressure interface twenty metres across, holding one atmosphere against vacuum, turning continuously for thirty-six years with no shutdown and no replacement, in an environment with abrasive dust and a thermal cycle every ninety-four minutes.

Nothing like it had been built. The longest comparable service life in the record was eleven years, on a diameter a fifth of the size, and it had been rebuilt twice in that time by people who could reach it.

The engineering lead who had said spin was not best spent an entire session on the seal and did not once argue against the decision. He argued about the seal. Ines noticed that and wrote it down and it changed how she read every meeting for the rest of her life: he has lost the question and is now making sure the answer is survivable. That is what a professional does after losing and I have never seen it written down anywhere as a skill.

What he got, over three sessions, was four things. A seal built in twelve independently replaceable segments, so that no single failure requires stopping the ring. A design that fails to leak-slow rather than leak-fast, at a measurable cost in friction. A mandated inspection interval. And — the one he had to fight for, and won in the fourth session by eleven words — a spare set of all twelve segments carried aboard, at a mass penalty everybody resented.

The eleven words are in the minute. Ines could still quote them at a hundred and eighteen.

"There is no supplier. After year one, we are the supplier."

Four segments were replaced during the crossing. The last of them in the twenty-ninth year, by a woman who had been eleven years old at departure.

The spin rate was set at a figure that produced about a third of Earth's weight at the ring's floor.

That number was not chosen for comfort either. It was chosen because it was the lowest value the medical group would certify as protective over thirty-six years, and every increment above it cost structural mass in a ring twenty miles from end to end, and mass was the currency by then and had been since the wall grew its second column.

Ines put the figure in the minute and did not think about it again for years.

A third of Earth's weight.

The planet they were going to had, by the survey, a surface gravity of about the same. Nobody in that room selected the spin rate to match the destination, and nobody at the time noticed that it did, and there is no document connecting the two decisions because there is no connection.

It is a coincidence of two independent constraints — a medical minimum and a planetary fact — and the consequence of it is that a hundred thousand people spent thirty-six years being physiologically prepared for a world by an accident of certification law, and stepped out onto Aura-36 and found that their bodies already knew how to walk.

Four generations later this is taught as foresight. There are two paragraphs about it in the Grade Seven curriculum.

Ines lived to a hundred and eighteen and never corrected it, and Teo asked her once, near the end, why not.

"Because I have the minute," Ines said. "It says what happened. Somebody will read it eventually."

Teo went up in the twenty-fourth year on a six-month rotation and stayed nine years.

She was forty-five. The medical took four months and she failed it once, on a cardiac finding that turned out to be an artefact of the altitude she had lived at her whole life, which nobody had accounted for because nobody had previously certified a person who had spent forty-five years at three thousand six hundred metres. They wrote a new baseline for her. It is still the baseline.

Her mother died in the second year of the rotation. Teo came down for it and went back up eleven days later and has never discussed the eleven days with anybody.

What she was sent up to do was the Bio-Domes, and she was sent because she had spent eleven years on evaporation ponds and nine on a survey crew and the specific combination of understands growing things and will write down the number that is actually there turned out to be rarer than the project had assumed.

The Bio-Domes were four sealed agricultural bays in the ring, and the first thing that struck her was that they were small.

She had expected fields. Everybody expects fields; the renderings had fields, and the renderings were by then eleven years old and nobody had updated them because updating them would have required somebody to decide the new version was more honest, and that was nobody's job.

What was actually there was racking. Hydroponic trays, nine high, in rows, under lights, in a volume about the size of the fabrication shed on the shore — times four.

Wheat, potatoes, tomatoes, beans, four leaf crops, coffee, and eleven fruit trees that were the single most argued-over mass allocation in the entire manifest and which Teo came to think of as the only sentimental object aboard.

She learned the thing this chapter exists for in her second week, from a spreadsheet.

Everything in those bays existed in a calculated quantity.

Not approximately. Not roughly enough for. Every tray, every crop, every light hour, every litre of nutrient, sized to a model that took the caloric requirement of a specific number of waking human beings over a specific number of days and worked backwards, with a margin that had itself been argued down twice.

Teo had grown up around agriculture as a thing you did with land, which is to say a thing with slack in it. If the potatoes failed you ate something else. If the rains were bad the yield was bad and everybody complained and got through it. There had never in her life been a crop whose quantity was derived.

She sat with the model for four evenings and understood that she was looking at the first honest accounting she had ever seen of what a human being costs.

And then she found the row.

It was not hidden and it was not sinister and she has spent a hundred years telling people that. It was in the input block, above the crop tables, where the model took its parameters, and it was a perfectly ordinary line in a perfectly ordinary spreadsheet:

Unit.

Teo looked at that for a while.

She wants it understood that she did not think it was monstrous. It is not monstrous. It is a spreadsheet, and a spreadsheet needs a word, and person does not work in a formula because a formula needs the thing to be countable and identical, and unit is the honest word for what the model is actually doing.

What she thought, sitting in a racking aisle under nine tiers of wheat with a tablet on her knees, was: this is the first place. Everything else that treats us as a number comes after this and this is where the word gets used first.

She was wrong about that, as it happens. The safety system had been counting bodies for four years by then and the Ledger was two years away, and the priority does not matter.

What matters is that she went and found the man who had written the model.

He was an agronomist, nine years older than her, entirely decent, and he was pleased somebody had read it. Nobody had read it. He said so cheerfully, twice.

"There's a word in the input block," Teo said.

"Which one?"

"Unit."

He looked at the line, and Teo watched him see it — genuinely see it, for what she was fairly sure was the first time in eleven revisions.

"Ah," he said.

"Is there a different word?"

He thought about it properly, which she gave him credit for.

"Not one that works," he said. "If I put person in there, then the margin reduction in row nine reads as taking food off a person, and it isn't, it's a statistical allowance. If I put crew it stops generalising to the sleepers and I need the same model for them. I could put a symbol and define it below."

"Would anybody read the definition?"

"No," he said.

They looked at the screen.

"Leave it," Teo said, eventually. "I'd rather it said the true thing badly than a comfortable thing well."

And the agronomist, who was decent and had never thought about any of this, said: "Should we write down that we discussed it?"

They did. It is four lines, appended to revision eleven, and it is signed by both of them, and it says that the term was queried, that no better term was found, and that the model treats human beings as identical countable quantities for provisioning purposes only and that this should not be extended.

It was extended within four years.

Not by them, and not by anybody who had read those four lines, and the extension was reasonable at every step, and there is a chapter about it and it is Ines's and it is called The Ledger.

The coffee is the other thing worth putting down, because it is the clearest example in this book of a decision that looks sentimental and is not.

Coffee has no caloric justification. It went to the mass review four times and was cut twice and reinstated twice, and the file on it is thicker than the file on the potato allocation, which fed people.

The argument that finally carried it was made by the ship's medical lead, in writing, in a memorandum of about a page, and it did not mention pleasure once.

It said: ten people will be awake for thirty-six years in a sealed volume with an unvarying diet, no weather, no seasons, no news, and a working routine that repeats. The single most reliable predictor of psychological failure in long-duration isolation is not danger and it is not workload. It is the collapse of the day into an undifferentiated block. Every long-duration crew that has held together has held together around small repeated rituals with a defined start.

Then the last line, which Teo copied into her own notebook and which Ines later copied into the Ledger: We are not provisioning a stimulant. We are provisioning a morning.

Eleven kilos of green bean, and the trees.

The first real harvest came in the twenty-seventh year and it was wheat and it was, by the model, four per cent under.

Teo had expected to feel something about that and did not, and what she felt instead arrived about a week later and took her completely by surprise.

She was in bay two on a night cycle, alone, doing a trace on a nutrient line, and she stopped and put her hand flat against a tray of wheat about nine tiers up, at head height, in the dark, with the grow lights down to their maintenance level and the whole bay in that deep red that makes everything look like it is underwater.

Nothing on that ship had grown before. Everything else aboard had been fabricated, welded, machined, lifted, bolted or written. This had grown, in a rack, under a lamp, forty thousand kilometres above the flat where her grandmother had cut salt blocks with a hand saw, out of nutrient made from a chemical stock that had itself come up the rail she had surveyed.

She stood there for a while.

Then she wrote the four per cent in the log, and the reason for it, which was a light-hour shortfall in the third week caused by a maintenance window she had authorised, and she wrote that it was her authorisation and that she would make the same call again because the alternative was deferring a seal inspection.

The agronomist read it and came to find her and said she did not have to put her own name on a shortfall.

"Then who would I put?"

"Nobody. It's four per cent. Models are wrong by four per cent."

"This one wasn't," Teo said. "It was right and I made a choice and the choice cost four per cent. If I write model variance then in nine years somebody tightens the model, and the model is fine."

He went away and thought about it and came back the next day and asked her to write a note on how to log a shortfall properly, for the other bays.

That note is one page. It is the ancestor of the practice, universal on that ship for thirty-six years and on the planet for four generations afterwards, of recording the decision that caused a number rather than the number alone.

The eleven fruit trees went. Teo fought for them at a mass review in the twenty-sixth year with an argument she had rehearsed for a week and which was, in the end, one sentence: nothing else on this manifest takes four years to produce anything, and if we cut them we are saying we do not expect to want anything that takes four years.

She won. The trees are on the planet. There are orchards.

Certifying a cow for spaceflight took four years, eleven regulatory opinions and a veterinary standard that had to be written from nothing, and everybody involved found it funny for about the first eighteen months.

Teo did not stop finding it funny. She wants that on the record, because the account that circulates has her as the solemn one, and she was not; she laughed at the cow paperwork as much as anybody and she still had a copy of the worst page of it in her quarters forty years later.

The worst page is a form. It is a standard animal-transport declaration, designed for livestock going by road, and somebody had been required to complete it because no other form existed and the regulation said a form must be completed. It asks for the duration of the journey in hours.

Somebody had written 315,576.

Somebody else, in a different hand, had written beside it: confirm?

The Ark Matrix was four decks and it was the least glamorous major structure on the ship and it consumed a share of the mass budget that caused a genuine institutional fight in the twenty-fifth year, which the livestock people won on a single argument: a founding population cannot be re-supplied, and a herd is not cargo, it is a system that produces more of itself, and the alternative to five thousand cattle is not four thousand cattle, it is a planet with no cattle on it, forever.

Five thousand cattle. Five thousand pigs. Frozen genetic stock in the millions, as the real insurance, with the live animals justified on the grounds that a herd which has never existed cannot be established from a freezer in year one on a world nobody has farmed.

Teo transferred across from the Bio-Domes in the twenty-sixth year because the two systems shared a nutrient loop and somebody had to understand both, and within a month she had asked the question that nobody in four years of certification had asked.

She asked it in a design review, mildly, in the way she asked everything.

"What happens to them at launch?"

The structural lead said the restraints were rated to four times the peak and gave her the figure.

"No," Teo said. "I've read the restraints. I mean what happens to them."

There was a pause of the specific kind she had learned to recognise, where a room discovers a question it does not have a column for.

"They'll be sedated," somebody offered.

"For twenty miles in eleven seconds, yes. And then thirty-six years."

"Thirty-six years of what?"

"That's what I'm asking," she said.

Here is what she had worked out, over about three weeks, mostly at night, from a stack of veterinary papers she had requested and which the project librarian had been surprised to be asked for.

Cattle are herd animals with a strong spatial sense. They orient. They establish position relative to each other and to the ground and they are measurably distressed by conditions that break that — unfamiliar footing, unstable surfaces, movement they cannot account for.

The Ark Matrix decks were in the Habitation Ring, under spin.

Spin is not gravity. It is very nearly gravity and human beings adapt to the difference in a month, but it has a gradient — your head is in slightly less of it than your feet — and it has the Coriolis effect, which means that anything you move sideways, including your own head, is pushed in a direction that nothing in your body expects.

In a human being that produces a month of mild nausea and then adaptation.

Nobody knew what it produced in a cow, because nobody had ever put one in a centrifuge for thirty-six years, and Teo went looking for the literature and found four papers, none of them longer than eleven days, and one of them on rats.

She wrote it up. Nine pages. The finding was not this will go badly; she was careful, and she did not have the evidence for that, and she said so in the second paragraph.

The finding was: we do not know, we have not asked, and the animals cannot tell us during the crossing because there will be nobody aboard qualified to interpret them.

What she proposed cost money and mass and four months.

She proposed a trial. A centrifuge facility, on the ground, at the ring's radius and rate, with a small number of animals, run for a year — not to prove it was safe, which a year could not prove, but to find out what the first year looked like so that the people aboard would recognise it.

The review took it seriously, which she had not expected, and the reason it took it seriously is in the minute and is Ines's phrasing: the proposal is not that the animals will suffer. The proposal is that we have no instrument for finding out, and are about to depart for thirty-six years.

They built it. It ran for fourteen months on the shore at Uyuni and it is the least documented facility of the entire programme and there is one photograph.

What it found was undramatic and useful. The animals adapted, mostly, over about nine weeks. They adapted faster if the deck had a visual reference that moved with them and slower if it did not. Four of the thirty-one did not adapt well at all and two of those were fine after being moved to a position nearer the outer wall where the gradient was shallower.

Every one of those findings went into the deck design. The Ark Matrix has painted horizon lines on its inner walls. They are there because thirty-one cows in Bolivia looked for a horizon and did better when they found one.

The stockman is the person who should be named in this chapter and is not, because Teo could never remember whether he wanted to be.

He was from Santa Cruz and he had been hired to run the ground herd during certification and he had, by his own account, applied because the advertisement said twenty-five years and he had a daughter.

He was the first person to agree with her, and he agreed with her before she had finished the sentence.

"They'll know," he said.

"Know what?"

"That the floor's wrong." He was leaning on a rail watching thirty-one animals do nothing in particular. "You can't tell me what a cow knows and I can't either. But I can tell you they know when the ground is lying to them, because I have walked a herd across a bridge, and a bridge is only a floor that moves a bit."

Teo asked him what happens on a bridge.

"They stop," he said. "All of them. And then one goes, and then they all go. And if the one that goes doesn't go, you're there all day."

That observation is in the nine pages, attributed to him by role, and it is the sentence that got the trial approved — not her literature review, which was thorough, but a stockman's remark about a bridge, which was concrete and which four engineers in the room could immediately picture.

He ran the trial for fourteen months and he is the person who noticed the horizon thing, four weeks in, watching an animal repeatedly orient toward a maintenance walkway that happened to run level around the inside of the drum.

"She's using that," he said. "She keeps putting it on her left."

They painted a line. The animal settled in nine days.

He did not go. He had a daughter and a herd and a house in Santa Cruz, and the twenty-five years he had applied for turned out to be twenty-five years on Earth, which was exactly what the advertisement had said and exactly what he had wanted, and Teo has never once suggested there was anything sad about that.

She wrote to him for eleven years. The letters stopped in the thirty-third year, before departure, and she never found out why, and she has assumed the ordinary thing.

Teo has been asked, more than once and usually kindly, whether the trial was worth what it cost.

The answer she gives is not the sentimental one.

"It found four animals in thirty-one that needed a different position," she says. "Scale that and it is six hundred head. Six hundred distressed animals on four decks for thirty-six years, with ten people awake and none of them a vet, is not an animal welfare problem. It is a systems failure with hooves."

And then, usually, if the person is still listening:

"But I did not ask because of that. I asked because nobody had, and because I have spent my whole life being the thing in the room that nobody asks."

The certification standard she and the stockman produced between them is nine pages and is called, in the file, Animal Welfare Under Sustained Rotational Reference. It has been amended twice in a hundred and forty years, both times for pigs.

It is the only document from the entire construction period that is still in active operational use on Aura-36, because everything else was superseded by the planet, and the animals are still under spin on nothing — but the orchards need pollinators and the pollinators need transporting, and every crate that moves between the ravine settlements travels under a rule written on a salt flat by a woman who had read four papers and a man who had walked a herd across a bridge.

The pigs were fine throughout. Nobody has ever established why and Teo has always said that the pigs are the single most annoying result of her career.

The Cryo-Embryo Vault was approved in four meetings across nine weeks and Ines minuted all of them and did not object once, and she has spent a hundred and eighteen years being asked why not.

The answer is that there was nothing to object to. She has never found a better one and she stopped looking at about seventy.

The case was made by the population geneticists and it is, on its own terms, unanswerable.

A hundred thousand people is a large number of people and a small founding population. It is enough to avoid immediate inbreeding depression and it is not enough to carry the genetic variation of a species across four hundred years on a world with no incoming migration, no gene flow, and whatever selection pressures a planet with a violet star and a third of Earth's weight turns out to apply.

The models were run four ways and they disagreed about the details and agreed about the shape. Somewhere between the fourth and the eleventh generation, on the unaugmented hundred thousand, the population passes through a variation bottleneck that it does not recover from, and the consequences are not dramatic and are permanent: elevated rates of a handful of recessive conditions, a narrowing of immune repertoire, and a slow reduction in the population's ability to absorb a novel pathogen without a catastrophic loss.

The fix is not complicated. You carry more variation than you carry people.

Frozen embryos are small, they are stable, they require no life support beyond cold, and a million of them occupy a volume and a mass that the manifest could absorb where a million additional adults obviously could not.

"It's a seed vault," the lead geneticist said, in the second meeting, and Ines wrote it down, and it was a good analogy and it was accurate and she watched it do ninety per cent of the persuading in the room.

There was one objection in nine weeks and it was procedural and it was hers.

She raised it in the third meeting, from the end of the table, which she had never done before and did four times in her life.

"What is the consent instrument?"

The room, she says, was not annoyed. That is the thing she needs people to understand about the entire period. They were pleased somebody had asked, because it was a real gap and they had all half-noticed it and none of them owned it.

The donors were living people on Earth. Some of them were colonists and some of them were not; a proportion of the material came from existing clinical stocks, donated years earlier under instruments written for terrestrial fertility treatment, by people who had consented to their material being used to help somebody have a child.

They had not consented to it leaving the planet.

Nobody had asked them, because at the time the consent was given, the question did not exist.

"Can we go back to them?" Ines said.

They could, and they did, for the material where the donor was traceable and living, and the re-consent programme ran for two years and Nnamdi wrote the letter and it is a good letter. Ines kept a copy of it and it is not the letter she kept for the reason people assume; she kept it because it is four hundred words long and it does not contain a single sentence designed to make agreement easier.

The re-consent rate was eighty-one per cent. Untraceable or deceased donors accounted for most of the remainder, and that material was withdrawn from the manifest, at a cost, on Ines's recommendation and over a mild objection that the genetic case was weakened by it.

It was weakened by it. She was right anyway, and everybody agreed, and the minute is clean.

She read the eighty-one per cent replies. All of them, over two years, in the evenings, which nobody asked her to do and which is not in her job description and which she never mentioned to a single colleague.

Most were a signature on a form. About one in nine had something written on them.

A woman in Lyon wrote four lines saying she had given the material at twenty-eight, that the treatment had not worked, that she had a son now by other means, and that she had thought about this letter for a month and would like to say yes and would like somebody to know that it had taken a month.

A man wrote a single sentence asking whether the child would be told where they came from. There was no procedure for answering that. Ines answered it herself, off her own authority, in a letter she did not copy to the file, and she has never said what she wrote.

Four people wrote to say no and to explain why, at length, and Ines read those twice each and thought all four were right.

One reply was from a man of eighty-one who consented and then added a postscript that Ines copied into the Bound Ledger fifteen years later, in her own hand, on the crossing, in the small hours, forty thousand kilometres from where it was written:

I will not know how this comes out. I have made my peace with that. But I would like the people who open the freezer to understand that somebody on this end thought about it carefully and said yes on purpose, and was not merely a source.

That postscript is the reason the vault's access protocol has a first clause that is not technical. It is one line, and it requires that the donor's recorded words, where any exist, be read by the clinician before the material is drawn.

It has been followed four times in a hundred and forty years. It was followed every time.

So: a good instrument, a real objection raised and resolved, material withdrawn where consent could not be established, and a scientific case nobody has ever successfully attacked.

And this is the chapter where a human being gets a catalogue number, and it happens in the fourth meeting, in about ninety seconds, and Ines did not notice it at the time.

The vault needed an indexing scheme.

That is all. Half a million objects in a freezer need to be findable, and the finding has to work in year one and in year two hundred, by people who do not speak the language the index was written in and who have never met anybody who worked on it.

The scheme they adopted is sixteen characters. It encodes the storage position, the donor cohort, the year of collection, and a check digit. It is well designed. Ines could still recite the structure of it at a hundred and eighteen.

The last four characters are the individual identifier.

Somebody — and Ines is certain it was the junior of the two geneticists, and has never named him, and has said only that he was about twenty-six and good at his job — said, drafting it on the board: "So each one gets a number."

And somebody else said, "Each what?"

And there was a pause of perhaps two seconds, and the first one said, "Each record," and wrote RECORD on the board above the last four characters, and the meeting moved on to thermal margins.

Ines wrote it down verbatim because she wrote everything down verbatim. The exchange is eleven words long. It is in the minute of the fourth meeting, on page six, between a discussion of vault siting and a break.

She did not think about it again for four years.

What she thought when she did think about it — and she has been careful to date this, because she refuses to let herself off by claiming she saw it at the time — was that she had watched two decent people, in real time, look directly at the question what is this a number for, and flinch, and reach for a word that made the flinch unnecessary.

Not to deceive anybody. There was nobody to deceive. They did it in a room of six people who all understood exactly what was in the freezer.

They did it because record is a word you can put in a database schema and person is not, and because the alternative was to stop the meeting and have a conversation that none of the six had any procedure for and no authority to resolve, with thermal margins still to get through and the wall in the fabrication shed reading zero at the bottom.

Teo had gone through exactly the same ninety seconds two years earlier, in a racking aisle, over the word unit, and had reached the opposite conclusion, and had signed a four-line note saying the usage must not be extended.

Ines had not read that note. It was appended to revision eleven of a bio-dome sizing model and it was filed in a different system, and the two women had not yet met.

The vault flew. It is under the planet now, in the deep cold, and it has been drawn on four times in a hundred and forty years, and every child born from it is a person with a name and a family and no idea that their sixteen-character identifier is still their file reference in a medical system that has never had a reason to change it.

There is nothing wrong with that. Ines said so, at ninety, to somebody who was trying to make it into something.

"It is a good vault and it was honestly made and I would vote for it again tomorrow," she said. "And I would like one person, once, to be able to tell me at what point in those nine weeks it should have gone differently. I have had sixty years and I cannot find the hour."

The Ledger was Ines's idea in one specific respect and she has never once allowed anybody to soften that.

It began as demographics and it began correctly. The vault gave the colony a genetic floor; it did not tell anybody how to use it. Somebody had to answer the operational question, which was: over the first four generations on a world with no incoming population, who should have children with whom, so that the variation carried across fifty trillion miles is not thrown away in eighty years by ordinary human beings choosing ordinary partners in a settlement of a hundred thousand?

That is a real question. It has a real answer. The answer is a model.

The model the population people built is not sinister and it is not even prescriptive in its first version. It produces, for a given founding population, a set of statistical guidance: cohort-level advice about how much mixing between which broad groupings keeps variation above the floor at generation four, generation eight, generation twelve.

Cohorts. That is how it was built and that is how it was presented in the twenty-sixth year and Ines minuted it and had no objection at all.

The problem with the cohort model is that it does not work, and it does not work for a reason that is boring and unanswerable.

Ines found it. She found it because she was the records clerk and because the records clerk is the only person who ever reads the whole of anything, and she found it in the way she found the withdrawn figure in column F a decade and a half earlier: by checking whether it was wrong.

She ran the cohort model against the actual manifest.

Not against the modelled population — against the real one, the four thousand and then eleven thousand and then forty thousand actual named human beings who had by then been accepted. And the cohorts did not describe them. Real populations are not evenly distributed across statistical groupings; they clump. The manifest had, for entirely innocent reasons to do with which trades had been recruited in which years from which regions, several groupings that were far more closely related than the cohort model assumed, and several that were far less.

Applied to the real manifest, the cohort guidance was not merely imprecise. In two groupings it pointed the wrong way.

She wrote it up. Nine pages. It is a good piece of work and it is unarguably correct and nobody has ever disputed a line of it.

And on page seven she wrote the sentence that she spent the rest of her life inside.

The model fails because it is indexed by cohort and the population is not a cohort. It can only be made accurate by indexing it to the individual.

She was right.

She wants that stated plainly and first, because every account of this that tries to be kind to her begins by suggesting she was somehow mistaken, and she was not mistaken, and being told she was mistaken is the one thing about it she could never stand.

The model indexed by individual works. It works to a degree the cohort version could not approach. It identifies, for every person on the manifest, the actual genetic distance to every other person on the manifest, and it can therefore give real guidance instead of statistical guidance, and the projected variation at generation twelve improves by a margin that made the population lead put his hands over his face.

It requires that every person on the manifest be entered into a database with their genome and an identifier.

The proposal went through in eleven weeks. There were four safeguards written into it, all of them proposed by Ines, all of them real, and all of them still in force a hundred and forty years later:

The guidance is advisory and carries no compulsion of any kind. Nobody may be prevented from having a child with anybody. The individual's own record is readable by that individual, in full, at any time, without a reason being given. And the Ledger may not be consulted for any purpose other than the one it was built for.

They are good safeguards. Ines drafted them in four days, alone, at night, and they are the single most protective instrument written during the entire construction period, and they are the reason the thing survived long enough to matter.

Nnamdi fought her on it for three weeks and lost, and it is the only serious disagreement the two of them ever had.

His objection was not scientific. He conceded the finding in the first ten minutes; he was not a man who argued with arithmetic.

"You are going to ask a hundred thousand people for a genome," he said, "and you are going to tell them it is advisory."

"It is advisory."

"It is advisory now." He had the four safeguards in front of him and he tapped the third one. "This is the one you think protects them. The individual may read their own record, in full, at any time, without giving a reason. That is a beautiful clause and I would put my name to it."

"Then what."

"Then in eighty years somebody notices that almost nobody reads it," Nnamdi said. "And a clause nobody exercises is a clause nobody defends. It will not be repealed. It will simply stop being a thing anybody has heard of, and one day somebody will consult the register for something adjacent and entirely well-meaning, and there will be no constituency to object, because the people who would have objected never opened their own file."

Ines said, "Then what would you write?"

And Nnamdi — who was the best writer of instruments on that project, and who had spent nine years making true things unreadable and then twenty making them readable — sat with it for a long moment and said, honestly, "I don't know. Something that makes them read it. I can't think what."

"Neither can I."

"No." He pushed the page back to her. "I want it minuted that I raised it and had nothing better."

It is minuted. Four lines, by name, at his own request, in the habit the two of them had been keeping for twenty years.

It is the fourth time in this book that a person objects on the record, offers no alternative, and loses. Nobody in the twenty-sixth year notices that this has become the project's characteristic act.

She understood exactly what she had built. That is the part that has to be said. She was not naive and she did not sleepwalk into it, and the evidence is in her own private file, in her own hand, dated eleven days before the proposal was approved:

I have built a complete genetic register of a hundred thousand people, indexed by individual, with four clauses of protection around it that depend entirely on the good faith of whoever holds it.

The clauses are good. I wrote them and I have checked them and I would not improve them.

The register will outlive the clauses. Registers always do. I know this because I have spent twenty-two years watching documents survive their reasons, and I do not have an answer, and I am going to file this and go to bed and put it in the proposal in the morning, because the alternative is a model that points the wrong way in two groupings and I have checked that four times.

She put it in the proposal in the morning.

The DNA Birth Ledger is what it is called now. It was called the Variation Register for the first sixty years and the name drifted, as names do, and by the time it was called the Birth Ledger it had been consulted at birth as a matter of routine for two generations, which is not what it was for, and which is not a violation of the fourth safeguard because a birth is the purpose, and which is exactly how the fourth safeguard was always going to be worn down — not by breach, by reading.

She opened her own record every year on the same date for the rest of her life. Ninety-two times. She is the only person in the history of that register known to have done so, and she never told anybody why, and Teo has always assumed it was not sentiment but enforcement — a woman keeping a clause alive by using it.

Four generations later a coup will be fought over it. Ines does not know that and never will.

What she knew, at the end, was smaller and she said it to Teo in the last year, in a compartment on a ship between stars, and Teo wrote it down that night because she knew it was the last useful thing she would get.

"People will say I should have refused."

"Some will."

"I have thought about what refusing looks like." Ines was a hundred and eighteen and entirely lucid and had four days left. "It looks like a woman who found an error and did not report it, and a model that points the wrong way for two groupings, and a bottleneck at generation eight in a population that cannot get more people. That is what refusing is. It is not clean."

She looked at the wall for a while.

"I did the correct thing and I have never been able to be glad about it. I would like somebody to write that down without deciding which half of it is the real one, and I do not think anybody will, because it is not a shape people like."

Teo wrote it down. It is in the Bound Ledger, on the second-to-last page filled in Ines Ferrer's lifetime, in Teo's handwriting, which is the only entry in the volume in anybody else's hand.

Four people died building the Ark-Genesis and the last of them is the reason a world is watched.

Calder read all four reports himself, which was not policy and became policy. The first three were the ordinary catastrophes of large construction and they are not this chapter: a fall on the rail in the ninth year, a crush injury in the fabrication hall in the fourteenth, a cardiac event at altitude in the seventeenth that was nobody's fault and which the medical group had modelled and accepted as a statistical certainty across a workforce that size.

The fourth was different and it took eleven days to find him.

He was a fabricator technician, thirty-four, four years on the project, working a night shift in the aft section of the completed keel in the twenty-seventh year. He was alone, which was permitted, because the alternative to permitting it was doubling the shift roster in a pressurised structure where nothing had ever gone wrong.

He did not come to the handover. That was not immediately alarming; the aft section was four miles from the transfer node and people ran over.

By the fourth hour it was alarming. By the twelfth it was a search.

The search took eleven days.

Calder's report on it is nine pages and it is the coldest document he ever wrote, and it is cold because he understood, correctly, that any warmth in it would be about him.

The structure was twenty miles long. It had, at that stage, four hundred and eleven compartments, most of them unpressurised, many of them unlit, a great number of them containing racking or spars or stored material that a body could be behind. There was no register of who was where. There had never needed to be one; you knew where people were because you worked with them and you could see them.

Nine hundred people searched. They searched in pairs with lights, compartment by compartment, in a structure so large that the search itself had to be scheduled and tracked on paper, and twice they searched a compartment that had already been searched because two teams had recorded it differently.

They found him on the eleventh day, in a service void off a compartment that had been searched on the third day and cleared, behind a spar. He had been dead since the first night. The finding was a fall, unwitnessed, with a head injury; the medical opinion was that he had been alive for some period afterwards and that the period could not be established and that establishing it would not be useful to anybody.

Calder wrote that sentence himself and then wrote, underneath it, the only line in the nine pages that is not procedural:

He was four hundred metres from a working party for the whole of it.

The engineering finding is in paragraph nine and it is the one that built a world.

In a pressurised structure, a person who cannot be located is not a welfare problem. It is a containment problem. You do not know whether they are unconscious in a void or on the wrong side of a bulkhead you are about to pressurise or in a compartment you are about to fill with inert gas for a weld. Every subsequent operation in the structure has to be halted or performed at risk, and the risk is not to the missing person, it is to everybody else.

Eleven days of a twenty-mile structure at a standstill, in a year with six and a half working months and a wall that read zero at the bottom.

The sentence Calder wrote in the recommendation is the ancestor of everything, and he wrote it as an engineering statement because that is what it is:

A person whose position is unknown is a breach whose position is unknown.

What he specified was a position register. Every body in the hull, located continuously, to compartment resolution, logged.

He did not specify it as surveillance because it did not occur to him to think of it that way, and the reader who wants him to have known is going to be disappointed by the archive. There is no memorandum weighing privacy. There is no minuted objection — and this project minuted objections compulsively, four times in this book alone. Nnamdi did not object. Ines did not object. Teo, who objected to the word unit in a spreadsheet, did not object to this.

Nobody objected because a man had lain four hundred metres from a working party for eleven days and everybody had spent those eleven days searching.

Teo was on the search. She had come down from the ring for a scheduled rotation four days before it started and she stayed for all eleven.

She searched with a rigger she had known since the fabricator programme, in pairs, with lights, in the unpressurised sections in a suit, six hours on and six off, for eleven days.

What she has said about it, the few times she has said anything, is not about grief. It is about the shape of the problem, because that is how she has processed everything since she was twenty-six and writing two numbers in brackets.

"There was no way to know we had finished," she said. "That's the whole of it. On the fourth day somebody asked how we would know when the search was complete, and nobody could answer, because a complete search of a twenty-mile structure by nine hundred people on paper is not a thing you can prove. We could only ever say we had looked. We could never say we had looked everywhere."

She was in the compartment that had been cleared on the third day. Not when he was found — she had searched it on the third day.

She has never said whether she was one of the two who cleared it. The search log exists and the pairs are named in it and anybody could look, and to Teo's knowledge nobody ever has, and she has never looked either.

"I know what the log says," she told Nnamdi, once, in the twenty-ninth year, when he asked her carefully whether she wanted him to check. "I mean I know what it must say, because I know where I was. I have decided not to convert that into a fact."

She backed the position register in the review without speaking against a single clause of it, which is the only time in fifty years she supported an instrument of that reach without proposing at least one limit on it.

She noticed that about herself at the time. It is in her notebook, four lines, dated, and it is the closest thing to self-reproach she ever committed to paper:

I did not ask a single question about who would hold this or for how long. I have asked that about a pond gate. I am recording that I did not ask, and that the reason I did not ask is that I spent eleven days looking for a man who was four hundred metres away, and that this is a reason and not a justification.

The system went in over fourteen months. It was, technically, unremarkable: a low-power tag, a mesh of readers on the structure, and a database. It did not know who you were talking to or what you were doing. It knew which compartment you were in and when you had entered it, and it raised an alarm if you were in a compartment alone for longer than your shift plan said you would be.

The alarm threshold was set at four hours. It is still four hours. Nobody on the planet knows why it is four and not three or six, and the answer is in paragraph eleven of a twenty-seventh-year report: it is the shortest interval that did not produce enough false alarms to make the workforce start defeating the system, which they had begun doing within a fortnight at two hours by leaving their tag on a bench.

Calder wrote that finding down too, without comment, and it is the most quietly important sentence in the document: a safety system people work around is worse than none, because it reports safety that is not there.

That alarm saved two lives in the remaining seven years. Both are documented. One of them was a woman in the same aft section, in the thirty-first year, found in nine minutes.

And here is the thing this chapter has to place without commenting on, because the nine books comment on it for four generations and this book's job is only to show where it came from:

It worked. It was proportionate. It was introduced with a real body count behind it, by people who had personally searched a twenty-mile hull for eleven days, and it did the thing it was built to do, and it went on doing that thing for thirty-six years of transit without a single recorded misuse.

It was also a continuous, complete, permanent record of the location of every human being inside a sealed structure, and when the structure became a planet the readers came off the bulkheads and went into the ground, and the mesh became a grid, and the grid got a name, and by then nobody alive had searched anything for eleven days.

Calder never gave it a name. In every document he wrote it is the position register, and he was still calling it that at ninety-nine, in a room, to a young councillor who had come to ask him about the Open Eye and who did not recognise the phrase.

The dead man's name is in the Bound Ledger. Ines put it there in the twenty-seventh year, in a section she created for the purpose, which had four names in it at departure and which she headed simply: Not carried.

Nnamdi Okonjo spent four hours of his professional life defending eleven kilograms of cannabis to a consortium board and he considered it, on reflection, the best-argued case he ever made.

It went to the board because it had been cut. Not maliciously; a mass review had come in nine hundred kilos over, and a junior analyst had gone through the manifest looking for anything without a caloric, structural or medical justification, and had found a line item that had none of the three and had done exactly what he was asked to do.

The line came off. Nnamdi appealed it, which surprised everybody, including Nnamdi.

"You appealed the drugs," Teo said, when she heard, down from the ring on rotation.

"I appealed the sleep provision."

"Nnamdi."

"I appealed," he said, with dignity, "the sleep provision."

The board convened four people and gave him thirty minutes and he took four hours, and the reason he took four hours is that he had done the one thing nobody else in the room had done, which was read the psychological projections.

They existed. They had been commissioned in the nineteenth year, produced in the twenty-first, circulated to a list of eleven, and acknowledged by eight. Nnamdi had read them twice and then again when the cut came through, and he opened by asking whether anybody else at the table had read them.

Nobody had. He said that was fine and that he was going to summarise, and he did, for about forty minutes, and by the end of it the tone in the room had changed completely.

The projections say this.

Ten people, awake, sealed, for thirty-six years. The failure mode that ends long-duration crews is not conflict and it is not boredom. It is sleep.

A human sleep cycle is entrained by light, activity, temperature and social rhythm. Aboard, the light is artificial and constant, the activity is unvarying, the temperature is held to a quarter of a degree, and the social rhythm is nine other people who are on the same schedule as you. Every external cue that tells a body when to sleep has been removed and replaced with a schedule, and a schedule is a thing you can decide to ignore at two in the morning when you are worried about a seal.

The projection's central finding, which Nnamdi read out twice: sleep degradation in that environment is progressive and self-reinforcing. It does not plateau. A crew member who is sleeping badly in year four is sleeping worse in year nine, and the earliest observable consequence is not exhaustion, it is a measurable decline in the quality of judgement under ambiguity.

"That last phrase is the entire ship," Nnamdi said. "Every serious thing that will happen out there is a judgement under ambiguity made by a tired person with nobody to ask."

Then he went through the alternatives, because he had known somebody would raise them and he wanted to raise them first.

Pharmaceutical sedatives: available, effective, licensed, and specifically contraindicated for continuous multi-decade use by the same four papers everybody had. Tolerance, dependence, and a withdrawal profile that would have to be managed by a crew of ten with one medic.

Behavioural protocols: real, valuable, already in the plan, and insufficient on their own by the projection's own numbers.

Nothing: the option the cut had actually chosen, without meaning to.

"I want to be precise about what I am and am not saying," Nnamdi told the board. "I am not saying this is better than the pharmaceutical option in a clinic. In a clinic the pharmaceutical option is better and I would not be here. I am saying that in a sealed volume, thirty-six years from a supplier, with one medic and no possibility of a second opinion, a thing that is grown, renewable, self-supplying, non-lethal in overdose and has a four-thousand-year human safety record is a different kind of object from a thing that runs out."

The board asked the question he had been waiting for, which was whether this had been medically assessed or whether he was making it up.

He had the assessment. He had commissioned it four months earlier, on his own initiative, before the cut, because he had read the projections in the twenty-first year and had been quietly certain since then that the line would eventually be challenged by somebody who had not.

It is nine pages. It specifies strain, ratio, method — vapour, not smoke, on air-handling grounds that are three lines long and unanswerable in a closed loop — dosage, and a monitoring protocol. It recommends against the entire category for two of the ten crew on individual medical grounds, by name, in a sealed annexe.

That annexe is the detail Teo always mentions when she tells this story, and she tells it more than Nnamdi ever did.

"He got them a sealed annexe," she says. "In the twenty-sixth year. On a project that had just built a register of everybody's genome. He wrote a medical exclusion for two named people and he put it in an envelope, and the envelope is in the Ledger, and it was still sealed when Ines died."

The funny part — and Nnamdi always insisted there was one, and told it against himself — is what happened in the first hour.

He had prepared for hostility and got something worse, which was jokes.

One of the four board members opened by asking whether the crew would also require snacks, and the room laughed, and Nnamdi laughed too because the alternative was to be the humourless man, and then spent the next twenty minutes discovering that he had lost the room by laughing.

He stopped, mid-sentence, at about the fifty-minute mark. Ines was minuting and her note records the pause as [approx. 15 seconds], which is the longest silence she ever recorded in a meeting.

Then he said: "I would like to withdraw my laugh from earlier."

Nobody knew what to do with that.

"It was a reasonable joke and I would have made it," he said. "But I have now noticed that we made the joke before we read the item, and that the item was cut by a man who also did not read it, and that between us we have spent four years and eleven meetings on the mass of the fruit trees and forty minutes on the only line in section nine that addresses whether ten people will be able to think clearly in year twenty."

He picked the projection back up.

"I am going to read section nine aloud. It is eleven hundred words. I appreciate that this is tedious and I am going to do it anyway, because I have been in communications for twenty-nine years and I know exactly what happens to a document that everybody has acknowledged and nobody has opened."

He read it. All of it. It took nineteen minutes.

Ines's minute of that meeting is the only one in the entire archive that contains a verbatim reading of another document, because she took it down as he read, on the grounds that if he thought it was worth nineteen minutes it was worth being in two places.

The junior analyst who made the cut was in the room. Nnamdi had asked for him to be there, which the analyst had understood as a summons and which was not one.

At the end Nnamdi said, in front of the board, that the cut had been correct on the information available, that the information available had been a manifest line with no justification attached to it, and that the failure was that the justification existed and had not been linked to the line.

Then he proposed the fix, which is the actual legacy of this chapter: every manifest item above a threshold mass carries a reference to the document justifying it, and any cut proposal must cite the document it is overriding.

It cost nothing. It was adopted in eleven days.

The board restored the allocation and increased it by four kilos.

The increase was not generosity. It was the medical lead's amendment, made in the last ten minutes: the original figure had been calculated for ten people over thirty-six years at the projected rate, with no margin, and a supply that runs out in year thirty-one is worse than no supply because the crew will have entrained to it.

Nnamdi wrote the manifest entry himself and it is four lines and it is entirely deadpan and it has been photographed a great many times.

The colony's pharmacology was planned. That is the whole point of this chapter and it is why it sits in the middle of the darkest run in the book, four chapters after a spreadsheet called a person a unit and one chapter after a dead man in a service void.

Nothing about the way a hundred thousand people would sleep, or grieve, or come down off a shift, was left to arrive by itself. It was projected, costed, assessed, argued in front of a board by a man who had read the papers, cut by a junior analyst doing his job correctly, and restored with a margin.

Four generations later there are fifteen-foot black-leafed plants growing wild in the forests of a planet under a violet star, and they are descended from four strains in a sealed pouch, and the reason they are there at all is that somebody in the twenty-sixth year read a document that eight people had acknowledged and nobody had opened.

The twin-turbine line ran for four years, consumed a sum of money that was embarrassing rather than serious, produced no usable thrust, and was closed in the twenty-eighth year by Calder personally in a memorandum of eleven lines.

It had been a reasonable idea. That is worth saying, because the file has been read since by people looking for either genius or vanity and it contains neither.

The reasoning went: the ship's main propulsion was settled and adequate and boring. What it did not have was fine attitude control at the scale of a twenty-mile structure, where the moment arm is enormous and a conventional thruster firing at one end induces a flex you then have to damp out at the other. Somebody proposed doing it with fields instead of mass — a pair of small counter-rotating turbines producing a tightly coupled resonant magnetic field, tuned, in principle, to push against the structure's own conductive mass.

In principle. The phrase appears four times in the original proposal and Calder underlined it in his own copy before approving it, which is the mark of a man funding something he expects to fail.

It failed. It failed in the most tedious way available, which is that it worked.

The field formed. It was stable, repeatable, and tunable to a precision that genuinely startled the team. It coupled to conductive mass exactly as modelled.

And the force it produced, integrated over the whole apparatus, was so small relative to its own mass and power draw that as a thruster it was worse than useless: you could get more attitude authority by throwing the turbines out of an airlock.

Four years. Three prototypes. The third one worked beautifully and was pointless.

The lead on it was a propulsion engineer of about thirty who had come off the rail programme and who Calder had picked specifically because she argued with him.

She knew by the end of the second year. That is in her own notes and she never hid it: the third prototype was already, in her assessment, the best version of a thing that was not going to be useful, and she said so to Calder in the twenty-sixth year in a corridor, which by then was where everything on that project got decided.

"It's not going to close the gap."

"How far off?"

"Orders. Not one." She had the figures with her because she always had the figures with her. "I can get another factor of four out of the coupling. I need about nine hundred."

Calder looked at the sheet for a while.

"Finish the third one," he said.

"It won't work."

"It'll work. It won't be useful." He handed the figures back. "Finish it properly. I would rather have one good example of a thing that doesn't do what we wanted than three bad examples of the same."

She has been asked about that exchange perhaps forty times and has always given the same flat answer, which is that at the time she thought it was sentimental and slightly indulgent, and that she stopped thinking that about thirty years later without ever being able to say precisely why.

She finished it properly. It is the third prototype in the drawer, not the first or second, and the difference matters: the third is the one with the tuning notes, the calibration curve and the machined housing, built by somebody who knew she was building a dead end and did it to standard anyway.

She did not go. She was thirty-four at closure and forty-one at departure and she stayed on Earth for reasons that are in a file and are nobody's business, and Teo, who did the manifest crew that month, has always wondered whether she knew her third prototype went up.

Calder closed it himself and did not delegate the closing, which was his practice with failures and one of the few unambiguously good things about him. The memorandum is eleven lines. It states the finding, states that the physics was sound and the application was wrong, thanks the four people by name, and reassigns all four within the project rather than out of it.

Then he wrote, at the bottom, in the same paragraph, without emphasis:

Retain unit three. Do not scrap. Nothing about the work is wrong except what we asked it to do.

It went in a drawer.

That is not a figure of speech and Teo has always been slightly annoyed that it sounds like one. It went into the third drawer of a steel cabinet in the site office on the shore, wrapped in cloth, with the tuning notes folded on top of it, and the cabinet was one of about nine hundred objects that got manifested in the thirty-third year under the catch-all category that Ines eventually had to invent for things that were going because somebody could not bear to leave them.

She asked him about it. She asked about everything in that category, because the category was a mass allocation and mass was the currency by then.

"What is it?"

"A failed thruster."

"Does it work?"

"It works perfectly," Calder said. "It does not do anything."

Ines wrote down the mass and the volume and the justification field, and the justification field on that line reads, in her hand, exactly as he gave it to her:

Good work. No present application.

She has said, much later, that she nearly pushed back and did not, and that the reason she did not was that it was four kilos and the man had cut his own office furniture allocation twice.

The four people were reassigned within a fortnight and three of them ended up on the position register programme, which was being specified in the same season, and which is a coincidence of scheduling and nothing more.

The fourth went to the seal team, where she spent the next six years on the twelve replaceable segments, and she is one of the two names on the inspection standard that a woman born aboard would use in the twenty-ninth year to change segment eleven.

Failed programmes on that project did not produce failed people. That was deliberate, it was Calder's, and it is the single most-copied piece of his management practice and the only one anybody in Book Six still does on purpose.

There is a thing worth saying about Calder and drawers, because this is not the only one.

He kept failed work. All of it. Not obsessively, and not in the manner of a hoarder — there was a rule, and Ines eventually wrote the rule down because she had to categorise it for the manifest, and the rule was: if it failed because we asked it the wrong question, keep it. If it failed because it was badly done, scrap it.

By the thirty-third year there were fourteen such objects. Twelve of them were left on Earth. Two went up, and the other one is a set of blade castings from an early atmospheric survey drone that is in a case on the planet and is very well documented and gets mentioned in every biography.

The cabinet is not in any biography, because a four-drawer office cabinet with contents miscellaneous does not photograph.

He was fifty-nine. He had, at that point, about eleven years of building left in him before the crossing and a hundred and forty of something else afterwards, and he had no idea, and neither did anybody.

What is in the drawer is a pair of counter-rotating turbines producing a tightly localised, precisely tunable resonant magnetic field with almost no net force, and the tuning notes, and a fourth page that is not in Calder's hand and has never been identified, containing a single line of arithmetic that does not relate to thrust.

That is the whole of what this book says about it.

The line item is on the manifest at four kilos and it has never been queried by anybody, in a hundred and forty years, because four kilos is nothing and the justification field is filled in.

The drawer is not opened in these fifty-five chapters. It is not opened in The Long Crossing. It is not opened in The First Breath.

A hundred and forty years later, in a garage under a mountain on a planet under a violet star, a man who is not supposed to exist has a car that cannot be photographed, and the countermeasure that does it is not in any patent, any manifest, any register or any archive, and there is exactly one object in the recorded history of the species that produces a tightly localised, precisely tunable resonant field and has no other function.

Nobody in this book knows that. Calder does not know it; he is fifty-nine and he closed a failed programme and could not bring himself to bin four years of somebody's good work.

Teo saw the cabinet loaded. She was supervising the manifest crew on the shore and it went up on the eleventh shuttle of that month with a batch of workshop stock, and she signed for it because she signed for everything, and the signature is on a form.

She asked what was in it, out of habit, and was told: office cabinet, four drawers, contents miscellaneous.

She wrote that down. It is accurate. She checked the mass against the declaration, because she always did, and it agreed to within a kilo, and she initialled it and moved on to the next pallet, and the whole transaction took about ninety seconds and is the last documented movement of that object anywhere in the archive.

Nnamdi worked out the number in the twenty-ninth year, on a Sunday, using the school roll.

He had not been looking for it. He had been asked for a figure for a notice — something about water allocation, entirely routine — and the population figures he had were four years old because nobody had needed a current one, and he went to the education office because a school roll is the most accurate census any town has ever had.

There were nine thousand one hundred children in school on the shore.

He sat with that for a while, and then went and got the housing numbers, and the payroll, and the food contracts, and the burial register, and spent the rest of Sunday and most of Monday building a figure that nobody had asked for.

Four hundred and six thousand people.

Not employees. People. The workforce was about a hundred and forty thousand at that point and the rest were the thing that grows around a hundred and forty thousand jobs that have a twenty-five-year horizon written on the advertisement: partners, children, parents, and then the second order — the people who came to sell things to the people who came, and the people who came to teach the children of the people who came to sell things.

A city. On a salt flat. With four hospitals and eleven schools and a court and two newspapers and a football league of nine teams and its own way of pronouncing four or five words that nobody could locate the origin of.

And two cemeteries, and this is the number Nnamdi could not get past: eleven hundred graves.

Not accidents. Four people died building the ship. Eleven hundred people died the way people die — of being old, mostly, and of being ill, and four of them in a bus on the Oruro road.

Nobody had planned a cemetery. It is not in any document. Somebody had needed one in the eleventh year and had gone and asked the municipality and it had happened, the way the sheds happened, the way the school happened.

He walked the town on the Monday evening, before he built the final figure, because he wanted to look at it with the number half-assembled in his head.

It took four hours and he did not get to the end of it.

The thing that struck him was not the size. He had known the size; he had commissioned the housing surveys. It was the ordinariness — the specific, dense, unremarkable texture of a place that has stopped being temporary without anybody deciding it had.

There was a shop that sold nothing but shoes. There was a man on a corner with a cart who had been on that corner for eleven years and whose son now had a second cart on a different corner. There was graffiti, and some of it was old enough to have been painted over and to have come back. There was a mural on the side of the water plant showing the rail, done by children, badly, in the year fourteen and touched up twice since.

There was a wedding coming out of a building as he passed and he stood across the street and watched it for a while.

He knew what the demographic projections said about a workforce of that size and he had read them and had helped commission them. What the projections said was that a settlement of this kind reaches a point at which it is generating its own second generation faster than the parent project is consuming labour, and that past that point it is not a workforce with a town attached, it is a town with a project attached.

The projection put that crossover in the twenty-fourth year.

It was the twenty-ninth.

He had read that projection in the twenty-second year, acknowledged it, filed it, and quoted from it four times in notices about school provision, and had never once put it next to the berth figure, and the two documents lived in the same cabinet about a metre apart.

He said something about that to Ines the following week, obliquely, in the way people raise things they are not ready to raise.

"There's a paper from the twenty-second year I should have read against another paper."

"Which two?"

He told her.

Ines thought about it and said, "Nobody reads two documents against each other. There is no job that does that. I minute the meetings and I file what comes out and I have never once been asked to check whether two things in the cabinet contradict."

"Somebody should be."

"Yes," Ines said. "Write it down."

He wrote it down. It became, four years later, a two-person function inside her records office with a name so boring nobody has ever quoted it, and it is the direct ancestor of the review the Shepherd performs continuously and without being asked, a hundred and forty years later, on every instrument on a planet.

He took the figure to Calder on the Tuesday.

"Four hundred and six thousand," he said, and put the working on the desk, because Calder did not accept figures without working and had not for thirty years.

Calder read it properly, which took eleven minutes.

"That's right," he said, eventually.

"I know it's right."

"Then what do you want me to say about it?"

And Nnamdi said the sentence he had been assembling since Sunday afternoon and had not been able to make sound like anything other than what it was.

"The ship carries a hundred thousand."

Calder did not say anything.

"I recruited them," Nnamdi said. "That's not modesty in reverse, I am stating a fact about authorship. I wrote the advertisement. I put twenty-five years in the eleventh paragraph on purpose because I had worked out that duration was the scarce thing, and I was right, and it worked, and it went on working for nineteen years after I stopped needing it to." He stayed standing. "There are four hundred and six thousand people out there and a hundred thousand berths, and I have never once said that out loud in public, and neither has anybody else, and I have started to think that the not-saying is itself the thing I will be judged on."

"You've been saying it in the notices for nine years."

"I have been publishing the berth figure and the workforce figure in separate documents nine months apart," Nnamdi said. "Which is true, and complete, and precisely the technique I built my career on, and you know exactly what it is because I taught it to you."

That is the closest the two of them ever came to a real fight, and it lasted about four seconds, and then Calder said:

"What do you want to do?"

"Say it. Once. Plainly, in one document, with both numbers on the same page."

"And then what happens?"

"I don't know," Nnamdi said. "That's the honest answer and I have thought about it for three days. Something. Not a riot — these are not people who riot, they are welders with mortgages. But something, and it will not be small, and it will not be over quickly, and it will be four years earlier than the selection was scheduled to start."

Calder turned his cup around, which was the tell.

"The wall says fifty-one months," he said.

"I know what the wall says."

"If it costs us a season we do not have a season to spend."

"I know."

"And if you say nothing for four more years, then the people who find out in year thirty-three find out that we knew in year twenty-nine."

"Yes," Nnamdi said. "That's the whole of it. Those are the two."

They sat with it.

What Calder did was the thing he always did, and it is either his great virtue or the mechanism of every catastrophe in the nine books depending on which one you have just finished: he treated it as a decision with a class.

"It's not mine," he said.

"It's absolutely yours."

"No. Class three is time-critical and reversible within a season, and this is neither." He pulled a pad over. "It goes to the full consortium and it goes with your paper attached, unedited, and I'll say in the covering note that I asked for it to be circulated unedited so nobody can improve it on the way."

"That's fourteen months."

"It's fourteen months."

Nnamdi said, quietly, "And in fourteen months it is still four hundred and six thousand and a hundred thousand, and we will have spent fourteen months of the not-saying."

"Yes," Calder said. "And I have a power that would let me shortcut that, and I have looked at it, and I am not going to use it on this, because if I use it on this then the thing I have used it for is deciding when four hundred thousand people find out what happens to them."

He wrote the covering note that afternoon.

The paper went to the consortium in the thirtieth year. It is nine pages. It has both numbers on page one, in the second paragraph, in the same sentence, and Nnamdi wrote that sentence eleven times before he was satisfied with it and the version that survives is the shortest.

Part Three is what happened next, and it is called The Selection, and it is fourteen chapters long.

What belongs here, at the end of this one, is the smaller thing that Nnamdi wrote in his own file on the Sunday, before he had the full figure, when he was still adding up the school roll:

I have spent twenty-nine years learning how to make people commit to something long. I have been extremely good at it. It did not occur to me until this afternoon that the people who commit hardest are the ones with the most to lose by it, and that I selected for exactly that, on purpose, in paragraph eleven.

In the thirty-first year Calder marked a section of the hull unallocated and gave no reason, and the drawing went out, and nobody asked.

It is a rectangle. That is genuinely all it is on the document: a volume in the aft third, outboard of the primary corridor, roughly nine hundred metres end to end, hatched, with a designation and nothing in the purpose field.

The designation is Sector 9.

There is nothing sinister about the number. Sectors were assigned in build order and this is the ninth. Four generations of people have looked for a meaning in it and there is not one, and Ines — who was in the room when it was numbered — found the search for a meaning funny in a way she could not explain to anybody born after her.

The reasoning is in his own file and it is defensible at every step, and he stated it aloud to no one.

Reserve. A ship of this size that arrives with no unassigned volume has no capacity to respond to anything it did not anticipate, and we have anticipated a great deal and will be wrong about some of it. Every other volume aboard has an owner and a purpose and a constituency, and the moment a volume has a constituency it cannot be reallocated without a fight. Therefore the reserve must be created now, before anybody wants it, and must be created without a stated purpose, because a stated purpose is a constituency.

That is a good argument. Teo, who never saw the file, made the same argument independently about pond capacity when she was twenty-four, and it is the sort of thing competent operators arrive at on their own.

Four lines further down, in the same entry, is the sentence that is not a good argument.

I am not going to circulate this reasoning. If I circulate it, the reserve becomes a decision, and a decision has a review date.

He knew. This is the chapter that establishes that he knew, and it establishes it out of his own file, and everything the nine books argue about for four generations is downstream of a man writing that sentence in the thirty-first year and then closing the file.

Nnamdi never saw the drawing. He would have seen it in the ordinary course — he saw everything — and in the thirty-first year he was fourteen months into the four-hundred-thousand paper going through the consortium and he was not reading hull drawings.

Ines saw it. She catalogued it, because she catalogued everything, and the catalogue entry is hers and reads: Sector 9. Volume, aft third, outboard. Unallocated. No purpose stated.

She wrote a query slip. She wrote query slips constantly and they were a standing joke and about one in nine of them found something real.

The slip asked what Sector 9 was for.

It came back four days later with the purpose field filled in as reserve volume, initialled, and Ines filed the slip with the answer attached and moved on, because reserve volume is a complete and sufficient answer and she had four hundred other things in the queue that week.

She kept the slip. Everything she ever wrote, she kept. It is in the Bound Ledger in the annexes, on a page with eleven other query slips, and it has been photographed exactly once, for a reason unconnected to Sector 9, by a researcher looking at handwriting.

The wall in the fabrication shed had two columns by then and the second one had been there for six years.

Teo saw that line go up. She was down from the ring for a fortnight and she came in for the stove and there it was, in a hand she did not recognise, in the column that had appeared six years earlier.

Nought point six per cent is not a small number on a ship that size. She did the arithmetic in her head the way she did all arithmetic, and then went and did it properly on paper because the head number seemed too large, and it was not too large.

She did not query it. Reserve is good practice; she had argued for reserve capacity on evaporation ponds when she was twenty-four and had been overruled by a foreman who wanted the throughput. A project that leaves itself nought point six per cent to be wrong with is a project run by somebody competent.

That is what she thought, standing in front of the wall with a mug, in the thirty-first year, and she was right, and it is the last uncomplicated thought anybody in this book has about Sector 9.

The mothballed units went in there in the thirty-third year and that decision was not Calder's and had nothing to do with him.

The swarm had finished. Four hundred and eleven units in the first wave and eleven hundred by the end, and by the thirty-third year the welding was done and there were somewhere over nine hundred serviceable frames with nothing left to weld.

Scrapping them was proposed and rejected on an argument nobody could beat: the ship is going to a planet with no industry, the frames are the only manipulators of that capability anybody will have for a generation, and the mass is already up here.

So they were shut down, wrapped, and stored, and stored means put somewhere, and the somewhere had to be a volume with no other owner.

There was exactly one.

A storekeeper made that decision. It is on a movement form. It took the ordinary approvals and there were four of them and all four were correct, and the form is signed and countersigned and the mass is declared and it agrees.

Nine hundred and forty dormant automated manipulators, in an unallocated volume with no stated purpose, in the aft third of a ship, arriving at a world where a man will one day need something built that nobody can be told about.

Calder did not put them there. That is a matter of record and it should be stated plainly, because the temptation to make this a scheme is enormous and the truth is worse.

He created a volume with no purpose, on purpose, and refused to write down why, on purpose, and then a storekeeper doing his job correctly filled it with the only thing on the manifest that had no home.

Teo signed the movement form. She was on the manifest desk that month, as she had been for four years, and it went past her with eleven other pallets of the same category and she checked the mass against the declaration and it agreed to within a kilo.

She did not read the destination field. She never read the destination field; the destination field was somebody else's check.

There is one more document and it is four lines and it is the reason this chapter is the one that closes Part Two rather than the four-hundred-thousand paper, which is objectively the bigger event.

In the thirty-second year the consortium ran its final structural audit, which was a genuine exercise conducted by genuinely rigorous people, and one of the auditors — a woman from the development bank's technical side, no relation to the careful woman from the eleventh year and better at this than anybody else on the team — queried the unallocated volume.

Her query is one sentence: Confirm whether Sector 9 is a reserve or a deferral.

It is, of the four hundred queries raised in that audit, unmistakably the best one. A reserve is a volume you have decided to keep empty. A deferral is a volume you have not yet decided about. They look identical on a drawing and they are completely different objects, and the difference is that a deferral eventually gets decided and a reserve does not.

The answer came back from the project office in the ordinary way, initialled by a scheduler who had no reason to think about it for more than a minute: Reserve.

The auditor accepted it, because it was a complete answer to her question and she had three hundred and ninety-nine others.

She filed her query and the answer together, as auditors do, and both are in the audit annexe, and the annexe is public.

Calder never saw the query. It did not reach him; there was no reason it should. He had built a thing that did not require him to be involved in its own protection, which is the whole of what he was good at, and it worked exactly as designed, and it worked in the wrong direction for a hundred and forty years.

That is the end of Part Two.

The ship is built. The keel is closed, the ring turns, the domes are growing, the vault is cold, the register knows where everybody is, and there is a nine-hundred-metre rectangle in the aft third with nothing written in the purpose column and nine hundred and forty sleeping machines inside it.

Fifty-one months on the wall. Four hundred and six thousand people on the salt.

A hundred thousand berths.

Part Three is called The Selection.

The Manifest Board met for the first time on the fourth of a month in the thirtieth year, in a room with eleven chairs, and Ines Ferrer sat at the end of the table with a pad and wrote down what was said.

The agenda had six items. She had typed it herself the previous afternoon.

She had prepared the room herself, which was not her job and which she did because the alternative was standing in a corridor for forty minutes doing nothing.

Eleven chairs, water, the circulated papers at each place in the order of the agenda, and a spare pad because in four hundred meetings somebody always needs a pad.

She had done this perhaps three hundred times. What she remembers is that she did it exactly the same way as always, and noticed herself doing it exactly the same way as always, and found that she could not decide whether that was steadiness or something worse.

The eleven were: the chair; two from the consortium; the medical lead; the population lead; a lawyer; two engineering directors; Nnamdi; Calder; and a woman from the municipality on the shore, who was there because somebody in the twenty-ninth year had asked the good question of whether a body deciding this should contain a single person who lived in the town, and had been told yes.

She spoke twice in the whole session. Both times she was right.

Item three took eleven minutes.

The chair — a woman from the consortium's technical side, sixty, entirely competent, who had chaired four hundred meetings and would chair this one exactly as she chaired those — said that the paper had been circulated and asked whether anybody had a question of fact about the figure.

Nobody did. The figure had been through fourteen months of consortium process and had been checked by three institutions and one of them had gone back to the school rolls themselves.

She then said, for the record, because it needed to be on the record:

"Four hundred and six thousand persons resident. One hundred thousand berths, of which four thousand are held as crew reserve and unallocated. The available figure is ninety-six thousand."

Ines wrote it down.

Somebody at the far end of the table said, "So roughly one in four."

And the chair said, "One in four point two."

And somebody else said, "Should we use the ratio in documents, or the absolute figures?"

And that is the discussion. That is the whole of it. Ines's verbatim record of the moment at which a body of eleven people first stated aloud in a room that three hundred and ten thousand human beings would be left on a planet they had been told was ending is four exchanges long, and three of the four are about arithmetic and the fourth is about document style.

They resolved to use absolute figures on the grounds that a ratio invites people to calculate their own odds.

It is a good decision. Ines has never disputed that it is a good decision.

The item that generated real discussion — forty minutes, four interventions, a genuine disagreement resolved by a vote — was item two, and specifically the crew reserve.

Four thousand berths held back and unallocated, against operational contingency: replacement crew, medical need, an error found late. The argument was whether the reserve should be four thousand or two, and it was argued hard and well, and the people arguing for two were arguing for two thousand more of the four hundred and six thousand to go.

The reserve stayed at four. The reasoning was sound. Ines wrote down all four interventions in full.

She has said, many times, that if you want to understand how it happened you should not read item three. You should read item two, where eleven decent people argued for forty minutes about two thousand berths with real feeling, immediately after accepting a figure of three hundred and ten thousand in eleven minutes without any feeling at all.

"Because the reserve was a decision," she said. "The ratio was an arithmetic result. Nobody in that room had decided it. It came out of the mass budget and the mass budget came out of the physics, and you cannot argue with a result, so nobody did, and we all went to item four."

The municipality woman spoke for the first time at item five, on appeals.

The question on the table was whether there should be an appeals mechanism at all, and the argument against was not callous: an appeal implies a body with the power to overturn, and any berth granted on appeal is a berth taken from somebody who was selected on the criteria, and there is no procedure for telling that person.

It is the same shape as the argument Renata Oyelaran will make in a compartment off a forward transfer nine years later about the Iron Protocol, and it is unanswerable in the same way, and nobody in the room in the thirtieth year has ever heard of Renata Oyelaran, who is at that moment about eleven years old.

The municipality woman said: "You have to have one anyway."

The chair asked her to expand.

"Not so people can win," she said. "So people can be heard saying it. If you give three hundred thousand people no place to put it, they will put it somewhere, and you will not choose where."

Ines wrote that down verbatim and it carried the item, and it is why Chapter Thirty-Two exists, and it is why nine thousand letters were read by a clerk instead of being burned in a yard by people who had nowhere to send them.

Nothing in the appeals process ever overturned a selection. Not once, in four years, in nine thousand appeals.

The municipality woman knew that would be the case when she proposed it. She said so, in the same intervention, in the sentence immediately after, and Ines recorded that too: "It will not change anything and it is not for that."

Item four ran short and was deferred, which is the second thing about that session she never got over.

The chair opened it, correctly, as an approach question — not what are the criteria but how shall we go about deciding what the criteria are. Somebody proposed a working group. Somebody else proposed that the working group should not include anybody who would themselves be subject to the criteria, and there was a pause while eleven people worked out that this excluded all of them, and then a slightly awkward noise that was almost a laugh, and the chair said they would return to it.

They returned to it in the second session. It is Chapter Thirty.

Calder did not speak at all in the first session. Ines checked the record twice afterwards to be sure, because it seemed unlikely, and it is correct: he is present, he is minuted as present, and he says nothing for two hours and forty minutes.

She has never offered a theory about that and has actively refused to when asked, on the grounds that she promised herself early not to interpret silences, which is a rule she and a woman four thousand kilometres above a planet arrived at independently and for the same reason.

What Ines did that evening is the thing she chose to record, in her own file, and it is four sentences and it is the coldest thing in her handwriting anywhere in the archive.

I typed the agenda. I chose the order of the items. Item three came before item four because the figure is an input to the criteria and any clerk would order it that way and I did it without thinking about it for even a second.

Putting the number before the discussion of how to use it meant the number entered the room as a fact rather than as a subject. I do not think anybody would have argued with it in any order. But I did not test that, and I am the only person who could have.

She was fifty-one. She had thirteen years to departure and sixty-seven to live.

The verbatim record of the first session is nine pages. It went into the archive in the ordinary way and it is public and it has been public for a hundred and forty years.

The four exchanges of item three are on page four, about a third of the way down, in the middle of the page, with no break above or below them.

Four generations of schoolchildren are taught about the Selection. The teaching materials use a photograph of the appeals hall from Chapter Thirty-Two, because it has people in it and because eleven hundred people queuing in the rain is a thing a child can look at and feel something about.

Nobody teaches page four. It is four exchanges about whether to use a ratio or an absolute figure, and it does not photograph, and there is nothing in it to feel.

Ines Ferrer thought it was the most frightening thing she ever wrote down, and she said so once, at ninety, to a researcher who had come to ask her about something else entirely and who did not use it, because it was not what he had come for.

The criteria took four sessions and nobody in any of them proposed anything cruel.

Ines has been asked to name the cruel proposal about two hundred times, by researchers and by journalists and once, memorably, by a nineteen-year-old on the planet who had been taught the Selection in Grade Seven and had come a long way to ask her.

There isn't one. That is the answer and it has never satisfied anybody.

Here is what was proposed, in the order it was proposed, with the reasoning, which is the only honest way to present it.

Skill redundancy. The colony needs a minimum viable count in every critical trade, with redundancy against death and injury across a thirty-six-year transit and the first decades on the ground. You cannot arrive with one surgeon. You cannot arrive with four hundred welders and no midwife. This is not a judgement about anybody's worth; it is a stocking problem, and if you get it wrong people die of it in year forty.

Unanswerable. Adopted in the first session with no dissent.

Age curve. A founding population needs an age structure that can reproduce, work, and not collapse into a care crisis in the third decade. Skew too young and you have no experience; too old and you have a settlement that cannot build itself. The demographers produced a target curve and it is a perfectly ordinary population pyramid and there is nothing wrong with it.

Ines wrote it down and it looked like a chart.

It is a chart. It is also, applied to four hundred and six thousand named people, a statement that being sixty-one is a disadvantage and being sixty-four is close to disqualifying in the absence of a scarce skill, and nobody in that room said that sentence, because nobody had to. The chart said it. The chart said it in a form that could be adopted in nine minutes.

Genetic diversity. This is Chapter Thirty-One and it is the worst of them and it gets its own chapter.

Dependents. A selected person's household travels with them. Proposed by the municipality woman in the second session, argued for on the plainest possible grounds — that a colony founded by people who left their children behind will not be a functional society, whatever else it is — and adopted almost unanimously.

It is the single most humane clause in the whole instrument. It is also the one that produced the black market, the marriages of convenience and the adoptions filed in three weeks, and Teo watched a city reorganise itself around it and that is Chapter Thirty-Five.

Every one of those four is defensible. Ines has never met anybody who could attack one of them on its own terms, and she has invited people to for sixty years.

The cruelty is emergent and she found it in the third session, with a pencil, doing arithmetic in the margin of her own pad.

The criteria did not interact the way the Board assumed. Each had been argued separately, adopted separately, and was defensible separately. Applied together to a real population, they compounded.

A fifty-eight-year-old rigger with a common trade, no dependents, and an unremarkable genetic profile is not disqualified by any single criterion. He is mildly disadvantaged by four of them at once, and four mild disadvantages multiply, and the output is not marginal. The output is a number so far down the list that no appeal will ever reach it.

Nobody decided that. There is no clause about fifty-eight-year-old riggers. There is no minute in which a person says that such a man should not go.

She tested the compounding on a real name before she took it to the chair, and she has never told anybody whose.

She had the manifest. She had the criteria as drafted and the weighting as proposed, and she had a pad, and she sat in the records office on a Sunday and ran one person through the whole instrument by hand, because a model tells you a distribution and she wanted to know what it did to a human being.

It took her two hours. She checked it four times, the way she checked everything, and the fourth check was because she could not believe the third.

The person was not marginal on any criterion. Not one. On skill they were middling; on age they were slightly the wrong side; on diversity unremarkable, which is not a fault and is the commonest thing a person can be; on dependents, none, which is not a fault either.

Four unremarkables. The output put them in the bottom fifth.

Ines sat with the pad for some time and then wrote, in her own file, the sentence she never showed the Board:

There is no criterion in this instrument that says an ordinary person should be left. But the instrument is four filters and ordinariness fails all four slightly, and I have just watched it happen on paper to somebody I know, and I made myself do the arithmetic to the end and I would like it recorded that I wanted to stop at the second check.

Ines took her arithmetic to the chair before the fourth session, privately, because she was the clerk and it was not her place to raise it in the room.

The chair — and Ines has always insisted on this, and it is in the record — took it entirely seriously, went through it for an hour, agreed with every line of it, and put it on the agenda as a substantive item under her own name rather than Ines's, which was a kindness and which Ines did not want and accepted.

The Board discussed compounding for ninety minutes. It is the longest single discussion in the whole selection process.

The ninety minutes are worth setting out because they are not what people imagine.

Nobody defended the compounding. There was no faction. The medical lead spent eleven minutes trying to construct a corrective weighting and abandoned it out loud, walking through his own reasoning until he reached the point where it broke, which is a rare and honourable thing to do in a committee and is fully minuted.

One engineering director proposed a floor — a rule that nobody could be excluded on compound marginality alone without individual review. It was the best idea anybody had. It fell on arithmetic: the number of people who would qualify for individual review under any workable definition was somewhere above sixty thousand, and the Board had four years and eleven people.

Nnamdi asked how long individual review would take if they did it anyway.

The population lead worked it out on the board. At the fastest defensible rate, with a panel of four working continuously, sixty thousand reviews is about nine years.

The wall in the fabrication shed said fifty-one months.

"So the answer is the schedule," Nnamdi said.

"The answer is the schedule," the chair agreed. She did not soften it. Ines has always given her credit for that.

They could not solve it.

Not would not. Ines is precise about this and it is the finding of the chapter. Four criteria, each necessary, each defensible, applied to a population of four hundred and six thousand: any weighting you choose produces winners and losers, and the moment you adjust the weighting to protect the compound-disadvantaged you are no longer applying criteria, you are hand-shaping an outcome, and the person doing the shaping has no mandate and no method and nothing to check them against.

The lawyer put it in the only sentence anybody quotes from that session: "We can have criteria or we can have judgement. We cannot have both, and I do not know which one I would rather sign."

They kept the criteria. The vote was nine to two.

The two were the municipality woman and Nnamdi.

Neither of them proposed an alternative. Both of them said so when they voted, and both asked for it to be minuted that they were voting against without having anything better, which by the thirtieth year was so established a practice on that project that Ines had a form of words for it.

Calder voted with the majority and did not speak.

What went into the instrument in the end is four criteria, a weighting, and a compounding note — three lines, appended at Ines's suggestion, stating that the criteria interact, that the interaction disadvantages individuals not disadvantaged by any single criterion, that the Board is aware of this and has not solved it, and that the fact is recorded so that nobody later concludes it was not noticed.

That note is the most-quoted passage in the founding archive of Aura-36. It is on a wall in the Archive of the Open Accord. Four generations of schoolchildren have read it.

It changed nothing whatsoever. It was written by a clerk who had done the arithmetic in a margin, to ensure that when the thing went wrong nobody could say we didn't know, which is what a deputy director in a corridor on Earth had told her the record was for, twenty-two years and four thousand kilometres earlier.

The person she ran through the instrument was still alive when she died. She never told them. She never told anybody. Teo asked her once, at the very end, and Ines said, "No," and that was the whole of the exchange.

She was fifty-one and had begun to understand that she had spent her entire adult life doing one thing.

She had done one thing, and the thing was this, and she was not sure any more whether it was a service or an alibi.

They turned her Ledger around in the third session and it took about four minutes and Ines said nothing.

She wants the mechanics understood before the failure, because the mechanics are the failure.

The Variation Register was built to answer a forward question: given this population, how should variation be managed across four generations so the colony does not bottleneck? It is a planning instrument. It looks at a hundred thousand people who are already going and works out what to advise them about.

What the Board needed was an answer to a different question: given four hundred and six thousand people and ninety-six thousand berths, which ninety-six thousand carry the most variation?

It is the same data. It is the same index. It is, computationally, very nearly the same query run in the opposite direction, and the population lead said so in the session, cheerfully, because it was good news — it meant they did not have to build anything.

"We already have this," he said. "It's Ferrer's index. We can run it as a selection input by Friday."

And somebody said, "Is that appropriate?" — and this is the part Ines has never let anybody skip, because the question was asked, by a decent person, unprompted, at exactly the right moment.

The lawyer answered it. The fourth safeguard on the Register states that it may not be consulted for any purpose other than the one it was built for. The purpose it was built for is the genetic viability of the founding population.

Selecting the founding population is the genetic viability of the founding population.

That is not a lawyer's trick. Ines has read the safeguard she wrote perhaps four hundred times in the years since and it is true. She drafted it herself, in four days, alone, at night, and she drafted it well, and it does not exclude this and could not have excluded this without excluding the thing it was for.

The lawyer's opinion is two paragraphs and it is correct.

Ines sat at the end of the table with a pad and wrote down the question, and the opinion, and the resolution, in her own hand, verbatim, and did not speak.

She has given the reason many times and it does not improve with repetition, and she never allowed anybody to improve it for her.

"I had nothing to say. Not nothing I was afraid to say — nothing. The lawyer was right about the clause because I wrote the clause. The population lead was right that it was the same query. And if I had stood up and said this is wrong, the chair would have asked me what should be used instead, and the honest answer was that the alternative to using the best available measure of genetic variation is using a worse one, and I was not going to argue for a worse one in front of eleven people."

"So I said nothing, and it was not cowardice, and I have never been able to make it sound like anything else."

Nnamdi came and found her that evening, which by the thirtieth year was a fixed pattern between them, and for once he did not open with the argument.

"You didn't speak."

"No."

"I watched you not speak for about four minutes and I have been trying since to work out whether I should have."

Ines put the pad down. "What would you have said?"

And Nnamdi — who had written the advertisement that brought four hundred thousand people to a salt flat, and who was three sessions away from being ordered to write the announcement that told three hundred thousand of them they were staying — sat down opposite her and said, "I would have said that in twenty-eight years on this project I have watched a good clause get read correctly into something nobody wanted about eleven times, and that it has never once been stopped by anybody pointing out that it was happening."

"That's not an argument. That's a complaint."

"Yes," he said. "I know. That's why I didn't make it either."

They sat there for a while.

"I could have objected without an alternative," Ines said. "You do it constantly."

"I do."

"And it never changes anything."

"No," Nnamdi agreed. "It has never changed anything. Not once, in twenty-eight years." He got up. "It does mean that when somebody goes back through this in a hundred years, they will find that the room contained at least one person who could see it. I have decided that is worth something and I could not tell you what."

Ines thought about that for four days and then went to see the chair and asked whether it was too late to have her non-intervention minuted.

The chair said, gently, that you cannot minute a silence.

She was right. There is no mechanism. A body records what is said, and Ines had spent twenty-eight years operating that mechanism and knew better than anybody alive that it has exactly one blind spot and that she had just stepped into it.

So she wrote it in her own file instead, where it has no standing and no force and can be dismissed by anybody as self-serving, and where it has sat ever since.

What changed on the Friday is that a number appeared in a column.

Every person on the manifest already had a sixteen-character identifier and a record. From that Friday, every person also had a variation contribution score — a figure expressing how much unique genetic variation that individual adds to a founding population that already contains everybody ranked above them.

It is a real quantity. It is correctly computed. It is, at the level of a hundred thousand people, exactly the right thing to optimise if what you are optimising is whether the great-grandchildren have functioning immune systems.

At the level of one person it says: here is what you are worth to us that somebody else is not.

And it is relative, which is the part that took Ines four years to fully understand and which she thinks is the actual horror and which nobody has ever quoted her on.

Your score depends on who else is going. Two people with identical profiles can score very differently depending on the order the list is built in, and a person can be worth a great deal at position forty thousand and almost nothing at position ninety thousand, without anything about them having changed.

"It is not a measure of a person," she wrote, in her own file. "It is a measure of a person's redundancy. I built the index. I did not build that, and I cannot show where it was built, and I have looked."

She ran her own score. Of course she did.

She had access, she had the method, and she was the one person on the project constitutionally incapable of not checking a number she was responsible for.

She has never said what it was. What she recorded is that she ran it, that it took nine minutes, that she looked at it, and that she then spent a considerably longer time working out that the figure would have been different if she had run it on a Tuesday, because two hundred people had been added to the provisional list on the Monday.

My worth to this colony changed on a Monday afternoon because of two hundred strangers and nothing about me was involved.

She deleted the query from the log, which she was entitled to do and which she recorded in her own file as the only occasion in fifty-one years of records work on which she removed something she had a right to remove and felt she had done something wrong.

Nobody ever saw their score. That was decided in the same session, in about ninety seconds, and it is the one protective decision of the whole selection that Ines proposed out loud and won.

She proposed it as a records matter, which is how she got it through: scores are working data, not findings, and the register publishes findings. A person may see their record and their result. The intermediate quantities are not their record.

It carried without discussion.

She has been asked whether it was the right call — whether people had a right to know the number — and she has always said she does not know, and that she made it in ninety seconds out of a records principle because it was the only lever she had that afternoon, and that dressing it up afterwards as ethics would be a lie.

The scores were destroyed at departure. That was also hers, written into the retention schedule four years earlier, and it is the reason no living person on Aura-36 can look up what their great-great-grandparent was worth.

The Register itself went. It had to; it was the planning instrument and the colony needed it.

So what arrived on that planet was a complete genetic register of a hundred thousand people, indexed by individual, with four good safeguards around it, one of which had already been read — correctly, by an honest lawyer, in a properly minuted session — to permit the register to decide who lived.

The precedent is four lines long and it is in the archive and it is public.

Four generations later a coup will be fought over that register, and the people fighting it will cite the fourth safeguard as their protection, and somebody will find the thirtieth-year opinion, and it will be the worst afternoon of that person's life.

Ines does not know that. What she knew, at the end of the third session, packing up her pad, was that she had built a good instrument and written good protections and that both had been used correctly to do a thing she could not have named as wrong in the room and could not stop thinking about on the walk home.

Nine thousand one hundred and eleven appeals were lodged in the four-year window and Ines Ferrer read all of them.

Nobody asked her to. The process required that each be logged, assessed against the four grounds, and answered; the assessment was a panel's job and the answering was a clerk's. Reading them in full was not a task anybody had defined.

She read them in the evenings, in the records office, for four years, at a rate of about nine a night.

The four grounds were narrow and she had helped draft them. An appeal could be lodged on: factual error in the record; a skill or qualification not captured; a dependent relationship not registered; or medical circumstance not disclosed.

That is all. There was no ground of hardship. There was no ground of fairness. There was no ground on which a person could say I have given nineteen years to this and I would like to be considered, because a ground like that has no test and a body of eleven people cannot adjudicate nine thousand of them in four years.

Every one of the nine thousand one hundred and eleven was assessed against the four grounds and answered within the window. The process worked exactly as designed. Ines has audited it twice and it is clean.

Four hundred and eighteen appeals succeeded, every one of them on ground one or ground three — a birth date wrong in the record, a qualification never logged, a stepchild not registered as a dependent. Real errors, really corrected.

None of those four hundred and eighteen changed a selection. They changed records, and the records were re-run, and in every case the re-run produced the same result.

The other eight thousand six hundred and ninety-three were refused, correctly, because they did not engage a ground.

They were not written to engage a ground. That is the chapter.

People did not write to the Board about factual error. They wrote to it because a form had arrived, and the form had four boxes and a space at the bottom, and the space at the bottom was the first place in nine years that anybody had asked them anything.

Ines kept a separate file. Not copies — she had no authority to copy an appeal — but transcriptions, in her own hand, of lines. She started it in the second month and it ran to eleven notebooks and it went up on the ship in her personal mass allowance, and it is in the Bound Ledger's annexes, and this is what is in it.

My grandfather cut the salt out of the ground here. I am not saying that is a reason. I am saying it because there is nowhere else to say it.

I have been told my trade is redundant. There are eleven of us and you are taking four. I have taught six of the eleven, including two you are taking. I do not want you to remove them. I want you to know that the reason my trade is redundant is me.

Ground three. My wife is not registered as my wife because we did not register. We have been together nineteen years and we have a house on the shore road with both our names on the water account and I am enclosing the water account.

I do not have a ground. I have read the four and I do not have one. I am filling this in anyway because my son asked me whether I had appealed and I wanted to be able to say yes.

You are wrong about my age but only by four months and I know that will not be enough. I checked the curve. Four months does not move me. I am writing to tell you the date is wrong because it is wrong.

I am sixty-one. I understand the reasoning and I have explained it to my wife twice and she understands it too. I would like somebody to write back and not use the word regret.

My daughter is going. She is a fitter, she is twenty-eight and she is better at it than I ever was, and she has told me she will withdraw and I am writing to ask you to refuse her if she tries. That is my appeal. Please refuse my daughter.

There is a mistake in my file. It says I am a pipe fitter. I am a pipe fitter. I have read it four times looking for the mistake because there has to be one.

My son is nine. He asked me if the ship was going to be full. I said yes. He asked if that was why. I said yes. He has not asked anything since and it has been three weeks and I would like to know what I should have said, and I understand that is not a ground.

Ground two. I speak four languages. It is not on the form because there was no box for it in year eleven when I filled the form in and there has never been a box for it since. I have written them here. I do not know if it matters. It has never mattered.

I am not appealing. I want to say that I think you have got it right and that I have watched you publish everything for nineteen years and I have no complaint. I would like that written down somewhere too, because I expect you get a great deal of the other.

Ines transcribed that last one and then sat in the records office for some time.

It is the only appeal in nine thousand one hundred and eleven that does not ask for anything, and it was refused, correctly, on the grounds that it did not engage a ground, and the refusal letter reproduced its text underneath, and Ines typed that one herself and has said that it took her four attempts.

Ines wrote that last one out twice, once in the notebook and once on a separate sheet that she kept in her own desk for four years and then destroyed before departure without ever showing it to anybody.

She followed it up, quietly, outside the process, which she had no right to do.

The daughter did not withdraw. The Board never received a withdrawal. Ines checked the manifest on the last day before it closed and the daughter's name was on it, and the father's was not, and the father was seventy-one at departure and the daughter was thirty-two and Ines does not know what passed between them and never tried to find out.

The hall opened on the first Monday of the window and there were eleven hundred people in the queue by seven and it rained.

It rains perhaps nine days a year on that shore outside the wet season and it rained on the first Monday of the appeals window, and the photograph everybody uses is from that morning, taken by a municipal clerk with no artistic intention whatsoever, and it shows a queue going round two corners under umbrellas and plastic sheeting.

What the photograph does not show, because it was taken at seven and this happened at about eleven, is that the queue organised itself.

Somebody — nobody has ever established who, and Teo spent four decades asking — worked out that the four grounds were narrow and that most of the people queuing did not have one, and started going down the line explaining them, so that people could decide whether to spend the day waiting.

Perhaps a third of the queue went home.

The rest stayed, knowing they had no ground, and filled in the four boxes with nothing and wrote in the space at the bottom.

Ines found this out four months later from the intake figures, which showed a refusal-on-grounds rate of ninety-five per cent from a population that had been told the grounds before they wrote.

They knew, she wrote in her own file. Almost all of them knew when they picked up the pen. I have been reading these as if they were mistakes and they are not mistakes, they are people using the only channel that exists for the thing they actually want to say, which is that they were here.

The answering letters went out over the four years and Nnamdi wrote the template.

It is four paragraphs. It does not use the word regret. That was Ines's single instruction to him and he did not ask why and never mentioned it to her again, and she found out eleven years later on the ship that he had read her notebooks — she had left them on the desk, she had never said he could not — and had known exactly which letter the instruction came from.

The template says what was decided, on what ground, by whom, and that the decision is final. Then it says one more thing, which was Nnamdi's and which he fought for and which cost the project a genuinely large amount of money in postage and clerical time.

Every refusal letter reproduced, in full, the text the appellant had written in the space at the bottom.

His argument for it was one sentence and the Board accepted it without a vote: "They should be able to see that somebody typed it out."

Eight thousand six hundred and ninety-three people received a letter refusing them a place on a ship, and underneath the refusal, their own words, transcribed exactly, including the ones that were not arguments.

Ines typed four hundred of them herself when the clerical backlog got bad in the thirty-second year.

She has never described that period beyond a single sentence, given at ninety, to a researcher who asked her what the hardest part of her working life had been, and who had expected her to say the Ledger.

"Typing out other people's sentences," she said, "and then putting my initials at the bottom."

They asked Nnamdi to write the announcement in the thirtieth year and he refused, in writing, in four lines, on a Tuesday.

It is the only occasion in twenty-nine years on that project that he refused an instruction.

The refusal is in the archive and it does not argue. It states that he is the author of the recruitment campaign, that the campaign is the proximate cause of the population figure, that he is therefore the least appropriate person in the world to write the document explaining the consequence, and that the Board should find somebody whose hands are clean.

The chair wrote back the same day, also four lines, and hers is better.

There is nobody whose hands are clean. There are eleven of us and we all voted. What there is, is one person who understands exactly how this will be read by the people reading it, because he spent nineteen years learning them well enough to move four hundred thousand of them across a continent. I am not asking you because your hands are clean. I am asking you because you are the best we have and this is the worst thing we will ever send.

He wrote it.

It took him eleven weeks, which for a man who could produce a technical notice in an afternoon is an extraordinary figure, and he produced forty-one drafts, and all forty-one survive because Ines filed all of them.

The forty-one drafts are, in Teo's view and in the view of most people who have studied that archive, the most revealing document in the founding record — more than the criteria, more than the minutes. You can watch a man take himself apart in them.

He went to Colchani in the fourth week, in the middle of the drafting, and did not tell anybody he was going.

He had held meetings there in the first year, when he was new and the project was a rumour and he had taken forty minutes of genuine anger from a hall full of people whose last four experiences of foreign companies had ended badly. He had written down every complaint by name and read the list back.

He went and stood outside the same hall for about an hour and did not go in.

What he wrote that night is not a draft. It is a note and it is four lines and Ines filed it with the drafts because it was in the same folder.

I promised a list I could be held to. I have been held to it. Every item on it was delivered and I checked this evening and it is all there — the road, the water, the school, the six months' housing after a contract ends.

Not one of those items was a berth. I was scrupulous. I never once promised anybody a berth.

Nobody in that hall thought they were queuing for a road.

That note is the hinge of the eleven weeks. Draft twenty-three — the first honest one, the first that leads with the number — is dated nine days after it.

Drafts one to nine are the campaign version. He could not help it; it is what his hands did. They are warm, they are forward-looking, they use the word together four times, and draft nine is genuinely beautiful and Nnamdi wrote across the top of it, in pen: NO. This is the advertisement again.

Drafts ten to twenty-two are the correction, and they overcorrect. They are cold to the point of contempt. Draft nineteen is four paragraphs of pure procedure with no acknowledgement of any kind that a human being is reading it, and across the top of that one he wrote: This is cowardice wearing a clean shirt.

Draft twenty-three is the first one that is neither. It is also the first one that begins with the number.

That is the discovery, and it took him eleven weeks and thirty-two drafts to fully make: the document could not build to the number. Every version that built to it — that explained the ship, the mass, the physics, the arithmetic, and then arrived at ninety-six thousand — was a version that spent its first four paragraphs asking a person to follow reasoning before telling them the thing that would stop them reading.

"You cannot make somebody understand why before you tell them what," he wrote, in a note to himself on draft twenty-six. "They will read the first line four hundred times and nothing after it. So the first line has to be the thing, and everything after it has to be worth reading by somebody who has already been hit."

The final version is eleven paragraphs.

The first sentence is nineteen words and contains both figures.

The second paragraph says who decided, by name, all eleven, with the vote where there was one.

The third says how, and links the criteria, published in full, unedited, as an annexe that anybody could read.

The fourth is the compounding note — Ines's three lines, reproduced whole, saying that the criteria interact, that the interaction disadvantages people not disadvantaged by any single one, that the Board knows and has not solved it.

Nnamdi fought for that paragraph for three weeks. Two Board members thought it was an unnecessary gift to critics. It went in.

The fifth explains the appeals grounds, plainly, including the sentence that there is no ground of hardship and no ground of fairness, and why.

Then the sixth, which is the one people quote, and which is one line:

Nothing in this document is an argument that the people who are staying are worth less.

The remaining five are practical. Dates, offices, who to ask, what happens next, what the town will be after.

There is no ending. Draft thirty-eight had an ending and he cut it. In the margin: An ending is for the writer.

The consortium's own communications people produced a competing version in the ninth week. This is not usually mentioned and Nnamdi never mentioned it.

It was not sinister. Two partners had grown nervous, reasonably, about a document that opened with the worst possible sentence and contained an unforced admission that the criteria were flawed, and they commissioned an alternative from professionals, which is what institutions do.

The alternative is in the archive. It is competent. It leads with the achievement, moves to the reality in the fourth paragraph, uses difficult decisions twice, and does not reproduce the compounding note.

It went to the Board alongside Nnamdi's, and the Board asked him to comment, and his comment is one paragraph and it did not attack the document at all.

This is well made and I could have written it in an afternoon. It will produce a quieter fortnight and a worse decade. The people reading it will detect the sequencing within about a day — they are welders, not children, and nineteen years of my own notices have trained them to read exactly this kind of document for what has been put where. When they find it, they will conclude that we managed them, and they will be right, and every honest thing we publish afterwards will be read through that. I would rather have the bad fortnight.

The Board took his. Seven to four.

It went out in the thirty-first year, in Spanish and in English, printed and posted to every registered address on the shore and read aloud at nine locations because the Board had been told, correctly, by the municipality woman, that a proportion of the population would not read a letter.

He read it aloud himself at the largest of the nine, which was the shed, which held about four thousand people that afternoon.

Teo was there. She has described it exactly once.

"He read it and he did not perform it. He read it the way you read a specification. And at the sixth paragraph a man near the front said something — I did not hear what, and I do not think Nnamdi did either — and Nnamdi stopped, and waited, and did not ask him to repeat it, and after about four seconds he went on to paragraph seven."

"That was the whole of it. Nobody shouted. Four thousand people stood in a shed and listened to a man read eleven paragraphs and then they went home. I have thought about that afternoon for a hundred years and I still do not know what it was that we all did."

The forty-one drafts were not supposed to survive. Working drafts were not archive material and Ines knew that better than anybody, because she wrote the retention schedule.

She filed them anyway, under her own authority, in a folder she created for the purpose, and the justification field on that folder reads: Method. There will be a reason to know how this was arrived at.

Nnamdi found out about four years later and was, briefly, genuinely angry with her — the only time in thirty-six years — and then stopped, mid-sentence, and said, "No. You're right. Keep them."

It is the best work of his life. He said so himself, without pleasure, in the twenty-ninth year of the crossing, to Ines, in a compartment, when she asked him.

"It is the best thing I have ever written and it is the best thing I will ever write, and I know exactly why, and it is not a good reason."

"Which is?"

"Because it is the only document I have ever produced," Nnamdi said, "where I was not permitted to want anything from the reader."

The lottery argument ran for nine months and Calder never once stated a position, and at the end of it the outcome was the one he had wanted since before it started.

This is the chapter where the reader should stop being comfortable with him. Everything he does in it is defensible. That has been true for thirty-three chapters and it is true here, and here it stops being enough.

The argument itself is genuinely hard and it deserves to be put properly.

The case for a lottery. Ninety-six thousand berths, four hundred and six thousand people, and no method of choosing that does not amount to a judgement about who is worth more. A lottery refuses to make that judgement. It is transparent, it cannot be gamed, it cannot be lobbied, and it produces an outcome that nobody has to justify to a neighbour, because the answer to why you and not me is no reason at all, which is the only answer that does not require one person to be better than another.

The municipality woman put it in a sentence that is on a wall: "A lottery is the only system where the people who stay are not also told why they deserved to."

The case against. A colony is not a population, it is a machine made of skills, and a random ninety-six thousand does not contain a viable distribution of them. The demographers ran it four hundred times. A pure lottery produces a settlement that is short of at least one critical trade in about seven runs out of ten, and short of three or more in about a quarter, and the failure does not show up on arrival, it shows up in year nine when the thing that breaks is the thing nobody can fix.

Both are true. That is the whole difficulty, and nine months is not an unreasonable time to spend on it.

The obvious compromise was proposed in the second month and everybody's first instinct was that it solved the problem: stratify. Run a lottery within trades, to fill the required distribution. Fairness inside each pool, viability across the whole.

It fails, and it took four months to establish why, and the reason is Ines's.

Stratifying by trade means the size of your pool determines your odds. A pipe fitter in a trade with four hundred candidates and ninety places has one set of chances; a specialist in a trade with eleven candidates and nine places has another. Same person, same worth, wildly different outcome, decided entirely by how many other people happen to do their job.

"It is not a lottery," Ines said in the fifth month, from the end of the table, which by then she did about four times a year. "It is four hundred lotteries with different odds, and we would be choosing the odds when we choose the pools, and choosing the pools is the judgement we said we were refusing to make."

Nobody had an answer. The stratified model died in that session.

Teo argued for the lottery. She was down from the ring on rotation for the sixth month and she was not on the Board and she said it anyway, to Calder, in a corridor, in about nine seconds, which was how she said everything to him.

"It should be a draw."

He stopped.

"The trades," he said.

"I know about the trades. I've read the runs." She had; she read everything. "Seven in ten short of a critical trade. That's real and I'm not pretending it isn't."

"Then what's the argument?"

"That in seven runs out of ten we're short of a trade and we fix it," Teo said, "because that is the entire history of this project. Every single thing that has gone wrong out here we have fixed by having four hundred thousand people who could learn something. And in the other version we arrive with the right distribution of skills and three hundred thousand people behind us who were told, on paper, with reasons, that they scored lower."

"You'd take the shortfall."

"I'd take the shortfall over the reasons."

Calder looked at her for a moment and said, "That is the best version of that argument I have heard," and went to a meeting, and did not repeat it in the meeting, and Teo did not find out for forty years that he had written it down that night, verbatim, in his own file, and had not shown it to anybody.

So it came down to the two, and it went to a vote in the ninth month, and the vote was eight to three for criteria.

Calder voted with the majority. He had not spoken in the debate at any point across nine months except to ask factual questions, of which the record shows eleven.

The three who voted for the lottery were Nnamdi, the municipality woman, and the medical lead, and the medical lead switched in the last month and told the room why, which took him four minutes and cost him something.

He had been for criteria for eight months on straightforward professional grounds: he was the person who would have to run a colony's health service, and a random draw might hand him two obstetricians and no anaesthetist.

What changed him was not an argument. It was that he had spent the eighth month running the medical criteria against the actual manifest, the way Ines had run the compounding, and had found that his own criteria — which he had drafted, and which were correct — excluded a category of person he had not thought about, which was people with a managed chronic condition that made them a marginal medical risk over thirty-six years and no risk at all on a planet.

"About four thousand people," he told the Board. "Every one of them fine. Every one of them scored down by a clause I wrote, for a transit risk, on a ship with a full medical bay and thirty-six years of nothing else to do."

He proposed an amendment. It failed, because it opened a category with no defensible boundary.

Then he changed his vote, and said that he was changing it not because the lottery was better but because he had now personally written one clause that did something he did not intend, and had no reason to believe it was the only one.

It is the most honest four minutes in the whole selection and it lost eight to three.

Here is what he did instead, and none of it is secret, and all of it is in the archive, and it took Teo forty years and a reason to go looking before anybody assembled it in one place.

He commissioned the demographers' four hundred runs. Class three, his own authority, month two, before the argument had properly formed. The runs are the single most cited evidence in the whole debate and they were produced at his instruction with a specification he wrote.

The specification asked: what is the probability of critical trade shortfall under a pure lottery? It is a good question. It is also, and only, a question about the failure mode of one of the two options.

Nobody commissioned the equivalent study of the failure modes of criteria, because nobody thought to, because there was no equivalent study to commission — the failure modes of criteria are not probabilistic, they are moral, and you cannot model them four hundred times.

He asked eleven factual questions, across nine months, in open session. Ines has listed them. Nine of the eleven are about viability. Two are about process. None is about fairness.

Every one is a reasonable question. Every one is the sort of thing a chief engineer asks. And the cumulative effect of an authoritative person asking nine viability questions and no fairness questions across nine months, in a room that watches him, is not nothing, and it is not an argument, and there is no minute you can point to.

He put the trade-shortfall figure on the wall. Not the criteria. Not the ratio. In the thirty-first year a line appeared in the fabrication shed, in his hand, reading the probability of critical shortfall under random selection, and it stayed up for eleven months, and four thousand people walked past it four times a day.

Ines asked him about that one directly, which she had never done about anything.

"It's a true number," he said.

"It is."

"Is it the wrong number?"

"No," Ines said. "It is one of two true numbers and the other one is not up there."

And Calder said: "What is the other one?"

And she could not produce it. That is the thing she recorded, that night, with a kind of cold admiration that she never entirely lost. There is no other number. The case for the lottery is not a quantity. It cannot go on a wall, because it is a sentence about not making a judgement, and a wall is for arithmetic.

He did not win the argument, she wrote. He arranged the room so that only one side of it had a number in it, using entirely true numbers, over nine months, without ever once stating a preference. I have watched him do this and I still cannot name the step at which he did something improper.

I have decided the step is that he knew what he wanted in month one and never said so.

He gave his reason exactly once, twenty-nine years later, on the ship, to Ines, who asked him because she was seventy-nine and past caring about tact.

"You wanted criteria."

"Yes."

"Why did you never say so?"

"Because I had a class-three power and everybody in that room knew it," Calder said. "If I had stated a preference on the first day, the argument would have been about whether to agree with me. Nine months of eleven people genuinely arguing produced a better instrument than four months of eleven people managing me."

"That is a good answer."

"It is a good answer," he agreed. "It is also what I would say if the real reason were that I wanted to win without being seen to. I have thought about that for twenty-nine years and I cannot get underneath it, and I am not going to pretend to you that I can."

The dependents clause was the kindest thing the Board did and it broke the town inside four months.

Teo watched it happen from a position nobody else had: down from the ring on a long rotation, senior enough to be told things and not senior enough to be lied to, and living in the same street she had lived in since she was nineteen.

The clause is one sentence. A selected person's registered household travels with them.

It was proposed by the municipality woman on the plainest possible grounds — that a colony founded by people who left their children behind will not be a functional society, whatever else it is — and it was adopted almost unanimously and it is the single most humane thing in the whole instrument.

It also meant that one selection could carry four people, or nine, and that the difference between going and staying, for three hundred thousand people, came down to a registration.

The registrations began within a fortnight of the announcement.

Teo's neighbour got married in the third week. She had known him eleven years; he was forty-four, a pump fitter, and he had been living alone as long as she had known him. The woman he married was a shift cook from the fabrication canteen, fifty-one, and Teo had met her perhaps four times.

He was not selected. She was.

They had the wedding in the yard behind the house with about thirty people and somebody's cousin playing badly, and Teo went, and it was a perfectly good wedding, and everybody there knew exactly what it was and nobody said so, and the two of them are buried next to each other on a planet under a violet star, having stayed married for forty-one years.

Teo has never been able to decide what that story is. She has told it four times and it has landed differently every time.

The adoptions were faster and worse.

Bolivian law required a process and the process was not short, but the town had a municipality and the municipality had a records office and the records office had four clerks and an enormous volume of applications with correct paperwork, and correct paperwork processed correctly at speed is not fraud.

Eleven hundred adoptions were filed on the shore in the four months after the announcement. The four-year average before that was nine.

Most were real in the sense that mattered: a grandmother taking in the grandchild she had raised since he was two, formalised in three weeks because there was now a reason to formalise it. That is not a loophole. That is a family doing the paperwork it had never needed to do.

Some were not that.

Teo knew of four. She has never reported any of them, and she has said so publicly, and it is the only thing in her long life she has been meaningfully attacked for.

"I knew four," she has said. "I know what they were. In one of them, money changed hands and I know roughly how much."

"And I have thought about it for a hundred years and the question I cannot get past is: what exactly was I going to report? A man documented a child who was not his so that the child would live somewhere with a future. The offence is against the manifest. It is not against the child."

The clause had a second effect that nobody predicted and that Teo thinks is the real story, and it is about the people who did not have anybody.

A household clause rewards having a household. It is not a punishment for being alone and nobody framed it that way and it does not appear anywhere in the criteria — but a selected person with four dependents brings four, and a selected person with none brings none, and the arithmetic of that runs the other way too. If the manifest is full, a candidate who carries four people is four times the cost of a candidate who carries one.

Nobody wrote that down. The instrument does not contain it. But the manifest is finite, and finite instruments have shadows, and by the thirty-second year there was a widely believed and never substantiated conviction on the shore that being single helped.

Ines checked. Of course she checked; it was exactly the sort of thing she checked.

It was false. She ran it four ways and the effect is not there in the data and she published the finding, plainly, in a two-page note that went out on the ordinary list.

Almost nobody believed her.

That is the thing she found hardest about the whole four years, harder than the letters: that she had spent a lifetime building a record on the principle that a true thing published is worth something, and here was a true thing, published, checked four ways, by the person best placed to check it, and the town had already decided.

"That was the first time," she said later, "that I understood that a record only works on people who already trust the record. It cannot create the trust. It spends it."

The market was real and it was small and it did not last, and the Board knew about it within nine weeks.

What the Board did about it is the part that matters and it is very unlike what people expect.

It did nothing.

Not out of weakness. There was a session, it is minuted, and it ran ninety minutes, and the options were set out properly: verify every registration; freeze the register at the announcement date; introduce a qualifying period; prosecute.

Verification fails on arithmetic — four hundred thousand people, eleven clerks, four years. Freezing at the announcement date punishes every real relationship that had not got round to being registered, which the municipality woman established was most of them, because poor people do not register things they do not have to. A qualifying period does the same thing with an extra step. And prosecution requires proving intent about a family, in a Bolivian court, at a rate of eleven hundred cases.

The chair summed up and Ines wrote it down: "Every enforcement available to us costs more real families than it catches false ones. We will be criticised for this in fifty years by people who have never had to count."

So the clause stood, unpoliced, and about eleven hundred people went to another star on documents that were technically correct, and Teo estimates — and it is only an estimate and she has always labelled it as one — that somewhere between four and nine hundred of them would not have gone otherwise.

Her mother had been dead nine years by then and Teo was glad of it, and she has said so, plainly, without apology, in the one interview she gave about this period.

"She was seventy-one when she died and she would have been eighty at the announcement. She would not have been selected and she would not have appealed and she would have told me not to be ridiculous when I registered her."

"And I would have registered her. I want that on the record. I would have put my mother on my household and I would have done it in the first week, and I know exactly what I think about the four cases I knew of, and I know exactly what I would have done, and I have never once been able to hold both of those in my hand at the same time."

What she remembers is not the wedding or the adoptions. It is the reorganisation.

A town of four hundred thousand people spent four months quietly rearranging its own kinship structure, and it did so almost entirely without argument, and Teo — walking the same streets she had walked at nineteen — watched something she still does not have a word for.

People with berths took people in. Not strangers, mostly; a cousin, a workmate's mother, the boy from the shop. Households of four became households of nine. Nobody organised it. There was no committee and no appeal to anybody's better nature, and there was certainly no coordination, because coordination would have looked like conspiracy.

It simply happened, at scale, in a town on a salt flat, in the four months after eleven paragraphs were read aloud in a shed.

The Board's own figures show it. The average registered household size on the shore in the thirtieth year was 3.1. In the thirty-first it was 4.4.

The average household size on the planet, four generations later, is 4.3. Nobody has ever connected the two figures and Teo stopped trying to get anybody to at about ninety.

Nobody has ever published a study of it. There is no monument. It is a line in a demographic table in an annexe.

Teo raised it once, at a Board session she had been asked to attend for something else entirely, in the thirty-second year, and it is her only recorded intervention in the whole selection process.

"You should know what they did with it," she said. "You gave them a clause and they used it to carry each other. It is the best thing that has happened here in thirty years and it is going into the record as a distortion."

Ines minuted it verbatim.

It changed nothing and Teo did not expect it to. The clause stood, the market was not policed, the household figure went into the annexe, and eleven hundred people left a planet on paperwork that a records clerk in four hundred years will read as an anomaly in the data.

The builders' claim is the best argument anybody made in four years and it lost, and it lost on the one ground its authors could not answer, and Teo was on the wrong side of it and has said so.

It was not organised at first. It was a sentence, and the sentence appeared everywhere on the shore within about a fortnight of the announcement, in four languages, in slightly different forms, and nobody has ever established who said it first.

The people who built it should ride in it.

It is a moral claim and it is not a weak one. Teo has always insisted that it be stated at full strength before anybody explains why it failed, because the version that gets taught is the weak version.

The strong version is this. Nineteen years. A hull that exists because four hundred thousand people put it there, in a place with no water and too much sun, on a schedule that took the best part of an adult life. Those people were recruited with an advertisement promising twenty-five years of work and they gave it. The ship is not an object that happened to be built here. It is the accumulated labour of a specific population, and that population made it, and there is no coherent account of ownership under which the makers of a thing have no claim on it.

Against that: the criteria say nothing about who built it. The criteria are about skill redundancy, age structure, genetic variation and dependents, and by design they are about the colony's survival and not about anybody's desert.

Teo was on the wrong side of it and she means something specific by that.

She did not oppose the claim. She was not on the Board and had no vote and she told the committee their paper was good. What she did was tell them, in the fifth month, in a room, that it would lose — and she was right, and being right four months early meant that eleven people spent five months on something the most senior of their own had already told them was dead.

"I could have said nothing and let them find out from the Board," she says. "Or I could have said what I said. I have never worked out which was the kinder one. I know which was the more honest."

The rail delegate asked her, at the end of that two-hour argument, whether she was going to sign the paper.

She did not sign it. She has never fully explained why and the closest she has come is this: "I had a berth. I did not know it at the time and I did not know it for another eleven months, and I want to be clear that I did not know. But I had spent four years on the manifest desk. I knew what my file looked like. Signing a paper arguing that people like me deserved to go would have been a thing I did with my thumb on the scale."

The rail crews organised in the fifth month. Not a union — there was already a union, and the union was cautious about it — but a committee, eleven people, four of them from the rail and three from the fabricator halls, and they were good. They put a paper in.

Teo read it in draft. They asked her to, because by the thirty-first year she was the most senior person on that flat who had come off the ponds, and they wanted somebody who would tell them if it was weak.

It was not weak. She told them so.

Then she told them it would lose, and why, and it is the only genuinely painful conversation in her working life that she has described in detail.

"Ask it the second question," she said.

"Which second question?"

"You've got a claim. Say it wins — say the Board writes it in tomorrow. Builders first. Then what? There are two hundred and eleven thousand people on this flat who have worked on this ship. Ninety-six thousand berths." She pushed the paper back. "You have not proposed a way of choosing. You have proposed a pool. And then you need criteria for the pool, and they will be the same four criteria, and you will have spent your claim to arrive back at exactly where we are with three hundred thousand people still staying, only now the ones who stay were also told they built it."

The committee argued with her for two hours and she has always said they argued well.

What they came back with in the second draft was a weighting: not builders first, but service as a fifth criterion, weighted alongside the four.

That is a serious proposal. It went to the Board and it got a full session and it was not dismissed.

It failed on two things and neither was moral.

The first is that service does not predict anything the colony needs. Every other criterion answers a forward question — will there be a surgeon, will the population reproduce, will variation hold. Service answers a backward one. The Board's job, as constituted, was to produce a viable colony, and a criterion that improves nobody's chance of surviving year forty could not be justified inside its own terms of reference.

The second is the compounding, again, and it is Ines who found it, again, and it is the cruellest thing in Part Three.

Service correlates with age.

The people with nineteen years of service are, definitionally, nineteen years older than they were. A service weighting adds points to precisely the population that the age curve subtracts them from, and when Ines ran it the two effects very nearly cancelled. For the great majority of the long-service workforce, adding service as a fifth criterion moved them a handful of places and changed nothing.

She took that to the chair before the session, as she had before, and the chair put it on the agenda under her own name, as she had before.

The committee heard it in the room. Teo was there, at the back, not on the Board.

She has described the moment the rail delegate understood it, and she has never described it more than once.

"He was a good man and he had done the reading and he had spent five months on this and he had eleven people behind him. And Ines put the two curves up, and he looked at them, and he got it in about four seconds, because he was not stupid, he was a rigger who had been doing arithmetic on a mount all his life. And he said — and I have never forgotten this — he said: so the reason it doesn't help us is that we did it."

"And Ines said yes. She did not soften it. I have loved her for that ever since and I have also never entirely forgiven her for it, and both of those are unfair."

The proposal failed nine to two.

The claim did win one thing and it is not small and it is almost never mentioned.

The committee's third proposal — made in the same session, after the weighting failed and before the plaque — was that the criteria and every individual result be published in full to the person concerned: not the score, which Ines had already closed off, but the four criterion outcomes and the position, so that a refused person could see which of the four had gone against them.

The Board resisted. The argument against was real: it hands three hundred thousand people a document that says, in effect, you were too old, and there is no version of that which is not cruel to read.

The rail delegate's answer to that is the second-best sentence anybody said in four years and it is minuted.

"We are going to be told no by a machine. Let us at least be told which part of us it was."

It carried, six to five, and it is the reason the refusal letters in Chapter Thirty-Two carry a four-line outcome table above the appellant's own words, and it is the only substantive concession the builders won in four years.

The rail crews did not strike. That is the thing everybody expects and it did not happen and Teo gets asked about it constantly.

"There were fifty-one months on the wall," she says. "Everybody could read it. If we had stopped for a season, the ship would not have gone, and the people who would not have gone were the same ninety-six thousand, including everybody's children."

"So we did not stop. That is not nobility. It is that the thing we would have been holding hostage was the thing we wanted."

What they did instead was the plaque.

It was the committee's own idea and they proposed it at the end of the same session, four minutes after losing, which Teo has always thought was the most extraordinary act of composure she ever witnessed.

Every name. All four hundred and six thousand, everybody on the payroll for more than ninety days across thirty-four years, etched and carried, at a mass cost the Board approved without discussion.

It is on the ship. It is in a corridor off the forward transfer, four hundred and eleven panels, and the type is small.

Teo's name is on it. So is her uncle's, who died in the eighth year, and her mother's, who worked four seasons on the ponds under contract in the twelfth and thirteenth and never set foot in the fabrication halls.

So is the name of every one of the three hundred and ten thousand who stayed.

That was the committee's specification and they were immovable about it: not the ones who went. Everybody who built it.

Four generations of schoolchildren have walked down that corridor. It takes about nine minutes to walk and nobody reads it, because it is four hundred and eleven panels of names in small type with no explanation, and the explanation is on a single panel at the far end that people reach after they have stopped looking.

The results went out over nine days in the thirty-second year, alphabetically, by post, and Teo's came on the fourth day.

She was at work. Everybody was at work; that was the one thing the Board got unambiguously right, on Nnamdi's insistence, against a proposal to close the site for the week. He argued that a shutdown makes the day an event and that people should be doing something with their hands, and he was right, and it is the only decision in the entire selection that nobody has ever criticised.

The nine days had a shape and the shape was alphabetical, which meant the town found out in waves, and the waves were the worst-designed part of the whole process and everybody knew it at the time.

Ines had proposed simultaneous delivery and had been overruled on logistics: four hundred thousand letters cannot be put through four hundred thousand doors in one morning by a postal service of ninety people, and the alternative — holding them all until the last was ready — added three weeks to a window that was already inside the wall's fifty-one months.

So it went alphabetically, and for nine days the shore road was a place where the people at one end of the alphabet already knew and the people at the other did not, and neither could talk to the other about anything else and neither could talk to the other about it.

Teo's surname put her on the fourth day, in the middle, which she has always said was the worst place to be: long enough to have watched four days of it happen to other people, not long enough to have got used to anything.

On the second day a man she had worked with on the ponds at twenty-three stopped her outside the fabrication hall and asked her, straight out, whether she knew already — whether being senior meant she had seen the list.

She had not. She could have. She had access to the manifest desk and had had for four years and the file was on a system she could open from any terminal on the flat.

She had thought about it on the first night of the nine days for a long time.

"I didn't look," she told him.

"Why not?"

And Teo said, "Because then I'd know about you as well."

So the letters went to houses and the houses were empty and four hundred thousand people came off shift and walked home knowing exactly what was on the table inside.

Teo walked from the transfer at the end of a nine-hour survey shift, forty minutes, the same walk she had done since she was nineteen with an eleven-year gap in the middle for the ring.

She has never described opening it.

That is not reticence. She has been asked perhaps four hundred times, by researchers and by children and once, at ninety-four, by a woman writing an official history who pressed her four separate ways, and the answer has never varied and it is not a deflection.

"I do not remember opening it. I remember the walk, and I remember the table, and the next thing I remember is being on the step outside with the envelope in my hand and no memory of the room. I have tried for a hundred years. There is nothing there."

What she does remember, and what she has described, is the fifty minutes afterwards.

She sat on the step. It was about six and the light was going the way it goes there, fast, with the temperature dropping behind it.

Along the street, doors were open. That is the detail she gives every time. Not people crying and not people shouting — doors standing open, up and down the shore road, because people had come home and read the letter and then not thought to close the door behind them.

Her neighbour two houses down — the pump fitter, married nine months earlier to the shift cook — came out and sat on his own step at some point and neither of them said anything and neither of them went inside.

At about seven a woman came down the road with a child on her hip, going the other way, and stopped and asked Teo if she was all right. Teo said yes. The woman said good, and went on, and Teo has no idea who she was and has thought about her more than about almost anybody in this account.

What she has said about the fifty minutes, when pushed, is that she was not thinking about the ship at all.

She was doing arithmetic. She has been embarrassed about this for a hundred years and has told it anyway, because she decided early that the account was worthless if she edited herself to look better.

She was working out how many people on the night shift roster would have had a letter that day. Twelve on the roster, alphabet distributed roughly evenly across nine days, so about one and a third crews' worth, so probably one or two of the four at kilometre four.

"That is what I did with the fifty minutes," she says. "I sat on a step with the thing in my hand and I estimated how many of my crew had had theirs. I got it wrong, as it happens. It was three of the four."

At twenty past seven she got up, went in, put the letter in the drawer of the kitchen table where her mother had kept the water accounts, and went back to work.

The night shift was on the rail at kilometre four re-shimming after a thermal event and they were four people short and she had known that at six because she had read the roster before she left.

She worked from eight until two in the morning.

She went back the next morning and worked the day shift as well, on four hours' sleep, which the roster did not require and nobody asked her to do.

The crew read her face wrong.

She came out of the truck and walked up the line with her light and the four of them stopped what they were doing, which is the thing crews do, and looked at her.

What was on her face was what had been on it for fifty minutes on a step, and none of them had any difficulty reading it, and every single one of them read it the same way and all four of them were wrong.

The oldest of them — a rigger of about sixty who had been on the rail since the eleventh year and who was, on any honest reading of the criteria, not going — came over and put his hand on her shoulder and said, "Ah, hell. I'm sorry."

And Teo said, "Shim seventeen through twenty-two, we're four short and I want them done before the temperature comes up."

And he said, "Right," and went and did it.

She did not correct him.

Nobody on that shift asked her what her letter said. Not that night and not in the four nights following. She has always thought that was the most sophisticated thing anybody did to her in thirty-four years on the salt: four people who had decided, without conferring, that the question was hers to open and not theirs to ask.

She did not correct him that night, and she did not correct him on any of the four nights following, and by the fifth night the whole shift knew, because it had gone round the way things go round, and by then correcting it would have been an announcement.

It came out about three weeks later in the ordinary way, from a manifest clerk, in a canteen, without malice.

The rigger sought her out. She has described that conversation exactly once and it is four lines long.

"You let me say that to you."

"Yes."

"Why?"

"Because you were being kind," Teo said, "and if I had told you, you would have had to be pleased for me, and you had had your own letter that morning."

He thought about it, and then he nodded, and then he said the thing that she has quoted more often than anything else anybody said to her in her whole long life.

"Next time just say it. I can do both."

Then he went back to work, and so did she, and they worked together for another two years and eleven months until the rail was mothballed, and she has never said whether they discussed it again and the honest inference from the way she tells it is that they did not.

The letter stayed in the drawer of the kitchen table.

She did not take it. When the house was cleared in the thirty-fourth year she went through it in an afternoon, and the letter was in the drawer with the water accounts and four decades of receipts, and she left the whole drawer as it was.

"It was not mine to take," she said, when somebody asked her, and then, because that sounded more mystical than she meant and she hated sounding mystical: "It was a form. I knew what it said. I had read it once and I was not going to read it again, and the drawer was the right place for it."

The house is still there. It is on the shore road in a town on the edge of a salt flat and somebody else has lived in it for a hundred and thirty years.

The offer was made in the thirty-second year, in a records office, on an ordinary afternoon, by a man who was not a criminal and who had thought about it for a long time before he came.

He was a haulage contractor. He had been on the shore nineteen years and he had four trucks and eleven employees and a son of twenty-six, and the son was a fitter, and the son was staying.

Ines had met him perhaps four times over the years in the ordinary way of a small town.

He did not open with money. He opened by asking a question about the appeals process, which was a legitimate question and which she answered, and then he asked a second one about how a correction to a record was actioned, which was also legitimate and which she also answered, and it was only at the third question that she understood what was happening, and by then she had already given him most of what he needed.

The third question was: who checks it.

Ines has always said that she answered that one honestly too, and that answering it honestly is the whole of this chapter.

"Nobody, in the sense you mean. A correction is entered against the record with a source document. The source document is filed. The entry is logged with the clerk's identifier."

"And then?"

"And then it is a corrected record."

He nodded slowly, the way a man does when he has confirmed something he already suspected, and then he named a figure.

She has never said what it was. She has said only that it was a great deal of money, that it was clearly the whole of what he had, and that the thing that stayed with her was not the size of it but that he had plainly worked out to the boliviano what he could raise and had brought exactly that number.

What he wanted was one date changed. His son's birth date, by four years.

Ines said no. She said it immediately, without deliberation, and she has always refused to be praised for that because there was no deliberation in it and a thing you do without deliberating is not a virtue, it is a reflex, and she said so.

The man did not argue. He said he was sorry, and that he had expected her to say no, and that he had come anyway because his son was staying and he had run out of other things to try. Then he left.

She sat in the records office for a while.

She went home that evening and did not eat and then wrote four lines in her own file that she has never allowed to be published and which Teo read after her death and has quoted once, at a memorial, with permission given years in advance.

I said no in under a second and I have spent the evening finding out that this was not because I am good. It was because he offered me money.

If he had offered me nothing — if he had simply sat down and told me about his son and asked — I do not know what the second would have contained. I am fifty-three years old and I have never had to find out, and this afternoon I came within one sentence of finding out, and the only thing that saved me was that he made it a transaction.

Teo has said that this is the single most frightening thing in the entire founding archive and that she includes the compounding note and the Ledger in that.

And then she did the thing that makes this chapter what it is, which is that she went and worked out whether she could have.

It took her about ninety minutes and she did it properly, with a pad, the way she did everything.

The answer is that it would have taken her eleven minutes and it would never have been found.

The system is this. A date of birth in the manifest comes from a source document. Source documents are municipal records, and municipal records for a town that grew out of nothing across three decades are of extremely variable quality, and the manifest system has a documented, sanctioned, entirely necessary procedure for correcting a record against a better source document.

That procedure is used constantly. There were four hundred and eighteen successful appeals on ground one alone.

A clerk with her access could enter a correction, attach a source document, and log it under her own identifier. The correction would be real in every respect except that the source document would be wrong, and nobody would ever look at the source document, because the source document is the evidence and the evidence is what you look at when you doubt, and nobody doubts a correction entered by the chief records clerk.

There is no second signature. There is no sampling audit. There is no reconciliation against the municipal originals, because the municipal originals are the corrections in a great many cases, and because a reconciliation of four hundred thousand records against a municipal archive that has burned twice would take longer than the time remaining on the wall.

She wrote it up. Nine pages, which was her length for a real finding.

The paper does not mention the contractor. She made that decision in the first hour and never revisited it: no name, no date, no description that could identify him, and the finding presented as a systems review of correction procedure, which is what it is.

"He committed an offence and I am not the person who decides what to do about that," she said, when Teo asked her about it thirty years later on the ship. "And if I had named him, the paper would have been about a man, and I needed it to be about the procedure. Those are two different documents and only one of them changes anything."

Neither of them changed anything, as it turned out.

The paper went to the Board in the ordinary way. It got a session. It was taken seriously; Ines has never claimed otherwise and the minute shows a real discussion of forty minutes with four interventions.

Four remedies were considered.

A second signature on all corrections: rejected on volume. At the correction rate of the thirty-second year, a second signature adds about eleven thousand clerk-hours before departure and there are not eleven thousand clerk-hours.

Sampling audit: adopted, at one in four hundred, which is statistically almost worthless against a single targeted alteration and which everybody in the room knew was almost worthless and which was adopted because it was the only thing that fitted.

Reconciliation against municipal originals: rejected, correctly, as impossible.

Removing the correction power from a single clerk: rejected because there was only one chief records clerk and the alternative was to stop correcting records.

She raised one other thing in that session and it went nowhere and it is the thread that runs into every book after this one.

She asked what happens after departure.

The Board had been discussing controls on a records office of nine people in a town with a municipality, a court and a consortium above it. Ines pointed out, in about ninety seconds, that in two years all of that goes away: there would be one records keeper, on a ship, for thirty-six years, with no municipality, no court, no consortium, no second clerk, and no possibility of an external audit of any kind.

"So whatever we decide today about controls," she said, "we should be honest that it expires at departure, and that after departure the control is that the ten of us can see each other."

The chair asked whether she had a proposal.

Ines said no.

It is minuted, by name, at her own request — the fourth time in this book somebody objects with nothing better and asks for it to be written down, and by the thirty-second year Ines had a standard form of words for it because she had typed it so often for other people.

Eleven years later, in a compartment off a forward transfer, nine people would begin writing a document whose founding obsession is that no power exists without a matching accountability that reaches the same distance, and one of the nine would be a woman who had spent four years as the only check on a manifest and had said so out loud, in a room, and had been told that there was nothing to be done about it inside the schedule.

The chair's summing-up is in the minute and it is four lines and it is the most honest thing said in that room all year.

"We have a system whose integrity depends on the character of one person. We have known this implicitly for some years and we now know it explicitly, in writing, because she told us. We cannot fix it inside the schedule. I would like the record to show that the reason this is not a crisis is that the person concerned is the one who reported it, and that this is not a control."

Ines Ferrer went on being the sole check on the manifest for the remaining two years, and then for thirty-six years of transit she went on being the sole check on the Bound Ledger, and then she died at a hundred and eighteen having never been audited by anybody in her entire working life.

The one-in-four-hundred sampling audit ran for two years and found eleven discrepancies, all of them genuine clerical error, all of them corrected.

The contractor's son stayed. Ines checked the final manifest on the last day, which she has admitted to and has never justified.

She did not check whether the date had been altered, because she knew it had not.

She checked whether he had appealed.

He had. Ground one, factual error, date of birth. Refused, correctly, in her own office, on paper she had typed, four months earlier, and she had read it at the time along with nine thousand others and had not recognised the name.

Ines Ferrer made an unauthorised copy of the complete unredacted manifest in the thirty-third year and carried it aboard in her personal mass allowance, and she has never once described it as anything other than theft.

The official manifest went, obviously. It had to; it is the passenger list of a ship. Ninety-six thousand names and their households, their trades, their medical flags, their berth assignments.

What the official manifest does not contain is anybody who is not on it.

That is not a scandal. It is what a manifest is. A list of who is aboard has no reason to carry the four hundred and six thousand who are not, and the retention schedule — which Ines wrote herself, four years earlier, correctly — provided for the selection working files to be closed at departure and held in the municipal archive on the shore.

Held on Earth. In a building. On a planet with a projection in column F.

She raised it properly first. That has to be said, because the version that circulates has her going straight to the copy, and she did not.

She put a paper to the Board in the thirty-second year proposing that the full selection record travel: not just the manifest but the criteria as applied, the individual outcomes, the appeals and their answers, and the reasons.

It was refused, and the refusal was not callous and she has never claimed it was.

The mass was real — not enormous, but real, and every kilo at that stage came off something. The legal position was genuinely unresolved: the records contained the personal data of three hundred and ten thousand people who were not travelling and who had never consented to their refusal being carried to another star. And the argument that finished it was the medical lead's, and it was good: a colony of a hundred thousand people, four generations of them, with a document aboard that says exactly why each of their ancestors was chosen and every other family was not, is a colony carrying a grievance in a filing cabinet.

"We would be founding a society on a list of who was better," he said. "I would rather they arrived not knowing."

The Board voted eight to three to leave the record on Earth.

Ines minuted it, filed the paper, and then went and copied the whole thing anyway over eleven weeks of evenings.

She wants the mechanics understood because the mechanics are the confession. She did not steal it in any sense that required cleverness. She was the chief records clerk; she had every access; there was no second signature and no sampling audit that would have caught it, as she herself had established in a nine-page paper in the thirty-second year that the Board had discussed for forty minutes and been unable to act on.

She used the hole she had personally reported.

The paper is still there and it is worth reading beside the confession, because it is not the document of a woman building herself a door. It is nine pages of a records clerk being thoroughly and unglamorously right about a filing system: no second signature on bulk extraction, no sampling audit, no reconciliation between the imaging log and the withdrawal log, and a recommendation, in her own words, that the office not be trusted, including this office, including me.

The Board discussed it for forty minutes and could not act on it, and the reason they could not act on it is the reason nothing at that stage got fixed: closing the hole required two more clerks, and two more clerks in the thirty-second year meant two people who would have to be somewhere and eat something, and everything in the thirty-second year was measured against a launch date.

"That is the part," she said, at ninety, to Teo, who had asked her about it once and never again. "Not that I took it. That I had told them the door was open and then walked through it, and that if anybody had ever asked me why I thought they should fix the door, the honest answer would have included this."

The copy is not the manifest. It is the manifest plus everything the manifest excludes: every one of the three hundred and ten thousand, their criterion outcomes, their position, their appeal if they lodged one, the answer they received, and — this is the part that took the eleven weeks — the text they wrote in the space at the bottom.

All nine thousand one hundred and eleven of them, transcribed by hand where the original was handwritten, because the imaging system was for official records and this was not one.

She did four an evening at first and then six and then, by the eighth week, eleven, and she has said that the rate is the only thing about the whole eleven weeks she is ashamed of. Not the theft. The rate.

“By the end I was copying them the way you copy anything,” she said. “I would get to the bottom of a page and not be able to tell you a word of it. A man had sat down at a municipal table with a pen he did not own and written the best case he could make for his daughter, and I was doing eleven of them between the evening meal and the light going off, and I was checking the count. I had made them into a quantity. That is the exact thing I had gone in there to stop happening to them.”

So she started re-reading one at random from each evening's work the following evening before she began, and she kept that up for the remaining three weeks, and it slowed her down by about a fifth, and she has always said it was the only part of the operation that was any use to anybody.

Nine thousand one hundred and eleven people used the space at the bottom of the form. The other three hundred thousand did not, and Ines has been careful about that number too, because the ones who wrote nothing are also in the copy, and their blank is also a thing they did.

The mass came out of her own allowance and there was not much of it. Every berth carried the same figure and she has never said what she left, only that it was a decision made once, in an afternoon, and that the four crates went in at the bottom of the stowage as personal effects, declared, weighed, and entered by her own hand in her own ledger under a description that was accurate in every particular and complete in none: paper, bound, personal.

Her motive is in her own file, in four lines, written on the first evening before she had copied a single entry, which is the thing that persuades people it is honest:

They are going to lose the reasons.

Not the names. The names will survive in the municipal archive for as long as the municipality does, and a hundred years from now somebody on that planet will be able to write and ask who was left, and they will be sent a list.

A list of names is a monument. It is not a record. The record is why, and the why is nine thousand letters and four criteria and a note about compounding, and if it stays here it will be a box in a building in a country that is running out of years.

I am going to take it. I have thought about the three hundred and ten thousand people who did not consent and I have not found a way past them, and I am going to take it anyway, and I am writing this down first so that I cannot afterwards tell myself I did not know what I was doing.

She told one person, eventually, in the nineteenth year of the crossing, and it was not Teo and it was not Nnamdi.

It was Halima Sow, who was the ship's doctor and who was thirty years younger than her and who had asked her, in the ordinary way of two people in a compartment, what was in the four crates.

Ines told her the truth, all of it, including the paper she had put to the Board and the hole she had reported and used.

Halima thought about it for a while and then asked the only question that mattered, which is the question a doctor asks: "Who can read it?"

"Nobody. It is not indexed and it is not catalogued and it is not in the Ledger. It is four crates in my quarters."

"Then it is not a record," Halima said. "It is a thing you are carrying."

Ines has always said that this was the single most useful sentence anybody ever said to her, and that it took her about four years to act on it, and that when she did act on it she did the thing she should have done in the thirty-second year, which was to make a decision about access and write it down.

What she wrote is one page and it is in the Bound Ledger and it is the strangest instrument in the founding archive, because it governs a document that officially does not exist.

It says: this material may be opened by a descendant of a person named in it, on request, without a reason being given, and by nobody else, ever, including the Council, including the Founder, and including the keeper of the record.

It says: the keeper may not read it for pleasure or for history.

It says: if the keeper cannot be replaced, it goes into the deep cold with the vault and is not destroyed.

What becomes of it is not this book's business. This book records that a fifty-four-year-old records clerk spent eleven weeks of evenings committing the only offence of her life, that she wrote down her reasoning before she began rather than after, and that a hundred and forty years later there is a planet on which almost nobody knows the reasons and four crates on which somebody does.

There were one thousand one hundred and forty berths that the criteria did not fill, and Calder filled them, and the document he filled them with is the only instrument in the founding archive that has a names column and no reasons column.

The number is an accident and it is important that people understand that, because the version that gets told has him demanding a founder's allocation and the Board conceding it, and that is not what happened. Nothing was conceded. The four criteria, run to completion against four hundred and six thousand applications, produced ninety-four thousand eight hundred and sixty selections. The hull carried ninety-six thousand. The criteria had been written to be defensible rather than to be exhaustive, and a set of rules written to be defensible will always stop short, because the last thousand of anything is where the rules are worst.

So there was a remainder. It sat on the arithmetic for eleven months while the Board tried four different ways to make the criteria stretch, and every one of them failed in the same place: any threshold loose enough to catch another thousand people caught eleven thousand, and any tiebreak fine enough to cut eleven thousand down to one thousand was a tiebreak nobody could say out loud in a public room.

Nnamdi Okonjo put it in the minute in a sentence that has been quoted at him ever since. “We have built a machine that is honest to ninety-nine per cent and mute after that.”

The vote to hand the remainder to the Founder's discretion passed nine to two. Ines minuted the two, because she minuted everything, and both of them said the same thing in different words, which was that a discretion is not a smaller version of a rule. It is a different kind of object.

Calder took it. He did not want it and he has never pretended he refused it, because he did not refuse it. He took it the way a man takes the thing nobody else will hold, and then he did something with it that people are still arguing about.

Nnamdi's condition was that the discretionary allocation be a written instrument: every berth entered, dated, signed, and unamendable after entry. Calder agreed to all of that within about four seconds and then refused the fifth column.

The fifth column was reasons.

They went round it for two sessions. Nnamdi's position was the one any reasonable person holds: a discretion without recorded reasons is not a discretion, it is a preference, and the only thing that separates the two is whether somebody afterwards can look at the page and see the thinking. He said that the record was not for the Board and not for Calder. It was for a person in a hundred years who would want to know how their great-grandmother came to be aboard.

Calder's answer took about a minute and he never varied it afterwards.

“If I write a reason next to the metallurgist, and I write a reason next to the woman who ran the canteen, then somewhere on that page there is going to be a name with a reason that does not read as well as the others. And the moment there is one of those, every name on the list is being read against it. You will have made a page that sorts eleven hundred people into the ones I can justify and the ones I cannot, and the sorting will be done by strangers, on paper, forever, and the people being sorted will be dead and unable to say anything about it.”

“Some of them you cannot justify,” Nnamdi said.

“Yes.”

“Then say so.”

“I am saying so. I am saying so to you, now, in a room, and Ines is writing it down. What I will not do is say it name by name in a column, because I will be saying it about them and it was never about them. It was about me.”

Nnamdi lost. He recorded that he had lost, which is characteristic, and he recorded his reasoning, and then he drafted the instrument himself without the fifth column and made it airtight, which is also characteristic. He said afterwards that he had spent his life arguing that a bad instrument written well is better than a good instrument written badly, and that he was not going to abandon the principle the first time it cost him.

So the list is four columns. Name. Household. Date of entry. Signature.

Eleven hundred and forty lines in one hand.

What is on it is a matter of public record and has been picked over for a century and a half, and the honest summary is that most of it is unremarkable and some of it is not, and there is no line on the page where the one stops and the other starts.

There are two hundred and ten specialists whose disciplines the criteria did not contain, and every one of those entries is the kind of thing that would have been a reason if reasons had been permitted. A metallurgist in a field that had eleven practitioners alive. Four people who could hand-fit a bearing. A woman of sixty-six who was the last person who had actually operated the class of pump that three of the ship's four water plants were built around, and who was outside the age band by a decade and a half, and who spent her first nine years aboard teaching eleven people to do the thing she did.

There are three hundred and forty who came in households — parents, siblings, and in nineteen cases grandparents, of people the criteria had already taken. Every colony that has ever been planned has had to decide whether it is selecting people or selecting families, and the criteria had decided people, and the discretion is where the families came back in. Nobody has ever attacked that part of the list and nobody sensibly could.

There are ninety-one who worked on the build and were not otherwise qualified: welders, cooks, two dispatchers, and the whole of one night crew from the rail. Calder has said that this was the only tranche he entered with any pleasure.

And there are the rest.

The count is one hundred and forty-nine and it is not disputed, because Nnamdi's instrument makes it trivially countable: subtract the specialists, the households and the build crews and what is left is what is left. A hundred and forty-nine people out of ninety-six thousand. It is a rounding error on a rounding error, and it is the only part of the manifest that anyone has ever written a book about.

There is the man who drove him for nineteen years, and his wife, and their two children.

There is a woman he had known since he was twenty-two who had no bearing on anything, who worked in insurance, who did not ask, and who found out she was going from a letter.

There is Teo Quispe's mother, who was sixty-one and had emphysema and was outside two of the four criteria, and who is the entry people bring up most often, and who died in the eighth year of the crossing in a bunk with her daughter in the room, and whose berth could have carried a fitter of thirty.

There is a whole household from a town he had lived in for eleven months in his thirties, and nobody has ever established what the connection was, and Calder was asked directly twice and did not answer either time.

He entered them all in the same hand, in the same column, on the same pages, interleaved by date rather than by kind, and that is the part that was deliberate. He could have put the specialists first. He could have grouped them. He wrote them down in the order he decided them, and the metallurgist who kept three water plants running is on line four hundred and six and his driver's wife is on line four hundred and seven, and there is nothing on the page that distinguishes them, and the man who made the page is the only person who ever could have.

He was asked once, late, by somebody who had earned the right to ask, whether he could still tell them apart himself.

He said that he could, and that the ones he could defend had stopped being the ones he thought about at around the fourth year, and that after that it was only ever the other list, and that the other list did not get shorter, it just got quieter.

Ines copied the discretionary instrument along with everything else in her eleven weeks, and she has pointed out — without comment, in the way she pointed out most things — the shape that leaves in the four crates.

Three hundred and ten thousand people who were refused, with the criterion, the outcome, the appeal, the answer, and in nine thousand one hundred and eleven cases the thing they wrote at the bottom in their own hand.

Eleven hundred and forty people who were accepted, with nothing at all.

“The reasons we kept are all refusals,” she wrote. “The one page where a reason would have told you something about the man in charge is the page he left blank on purpose, and he was right about why, and I do not think he was right, and I have carried both of those for thirty years.”

The instrument closed on the eleventh of the month before the manifest board's final vote. Calder signed the last line, and the last line is a name nobody has ever identified with any confidence, and he did not sign anything else that week.

There were eleven thousand people at the salt flats over four days in the thirty-third year, and almost nobody who has written about it since has been willing to say who they actually were.

They were not the refused. That is the first thing and Teo Quispe has said it in every account she ever gave. The refused were three hundred and ten thousand people distributed across every continent on the planet, and the overwhelming majority of them did what people do, which is grieve for a fortnight and then go back to work. A refusal letter does not put you on a road for nine hundred miles.

What put people on that road was the town.

The build had made a town where there had been salt. Nineteen years of it. Forty-one thousand people at peak, and schools, and a hospital that was better than anything within four hundred miles, and a water plant, and a rail spur, and eleven years of a man on the radio saying that the work here mattered more than any work anywhere. All of that was true. All of it was also finishing.

The people at the fence were, in the main, people whose lives the project had built and was now going to fold up and take away in the sky. Scaffolders. Cooks. Two thousand haulage workers. A great many people who had spent a fifth of their lives on the thing and had never once expected to go and had never once been promised anything and had, all the same, arrived at a set of feelings for which there was no procedure.

There were also families of the refused, and there were perhaps four hundred people who had come specifically to stop the launch and had said so publicly for two years, and there were, by the third day, a number of people who were simply there because eleven thousand people were there.

“Every single account calls it a mob,” Teo said, at seventy-one. “I was on the apron. It was a crowd of people I knew. I could put names to two hundred of them without trying. That is what nobody wants to write down, because if you write that down you have to explain what I did.”

What happened on the first three days was noise and speeches and a great deal of standing about in a place with no shade. The site security complement was four hundred and eleven people for sixty-one miles of perimeter, which is not a security complement, it is a gesture. Nothing was going to be held by four hundred people and everybody involved knew it.

On the fourth day a gate went. Then about ninety metres of fence went, not by force but by weight, the way fences go when four thousand people lean on them without meaning to. And then there were people on the apron, walking, mostly slowly, some of them still carrying folding chairs, in the direction of the launch rail and the four shuttle stacks standing on it.

The stacks were fuelled.

They had been fuelled for six days because the window was in eleven and because defuelling and refuelling a stack of that class is a fortnight's work that nobody had a fortnight for. It was a decision taken in a room by four people who had weighed a schedule against a risk and had not weighed it against a crowd, because on the day they took it there was no crowd.

Teo has always insisted that this is the whole of the emergency and that everything else is commentary. Not the crowd. The fuel. Nine years of fabrication standing on a track with a full load in it and eleven thousand people walking towards it in the heat with no plan and no marshals and no idea what any of it did.

She was in the swarm shed, which was two miles up the line, and she had fourteen thousand two hundred units on charge or in the field.

What she did took about ninety seconds and she has never once claimed anybody ordered it.

She recalled every unit on the north works and issued a single build task, which is the only kind of instruction the swarm has ever been able to take. The task was to enclose the accessible length of the rail in continuous plate. Three miles of it. Any stock, any offcut, any thickness, weld it and keep welding it.

It is a construction instruction. It is, in fact, an instruction she had issued eleven times before on other sections, because enclosing the rail was already scheduled work — it was going to be done in the spring, in sequence, by four hundred units over five weeks.

She issued it to fourteen thousand.

“I want to be exact about this,” she said. “I did not tell them to stop anybody. There is no instruction in the swarm that means stop anybody. There has never been one, there is not one now, and if you tried to write one the machines would not know what it referred to. I told them to build a wall, in a place where people were, at a speed that had never been attempted, and the difference between that and telling them to stop people is a difference I understood perfectly at the time and have never been able to make matter.”

The units are forty centimetres. They climb. They are not fast on the ground and they are extremely fast on structure, and fourteen thousand of them coming down a rail bed onto three miles of track is not a charge, it is a tide, and it took about four minutes to arrive and about eleven to become a solid working face.

No unit touched anybody. Not one, in four days, and this is not disputed by any party, including the people who were there and who hate her.

What they did was arc-weld, continuously, in the open, along a three-mile front, at chest height and below, in a line four machines deep.

You cannot walk into that. Not because it will hurt you deliberately but because it does not know you are there, and everybody on that apron worked in fabrication and understood in their bodies exactly what a live arc is and what it does to an eye at eleven metres. The crowd stopped at about sixty metres and it stopped by itself and it stopped in under two minutes.

One person was hurt. His name was Bram Coetzee, he was thirty-four, he was a scaffolder, he had been on the build six years, he was staying, and he was not attacking anything. He reached in and took hold of a unit that was working, with his hand, apparently to pull it off the plate, and it went on working, because that is the only thing it does.

He lost two fingers and most of the use of a third. He was treated at the site hospital, which was one of the things the project had built, and he refused for eleven years to say a word about it publicly, and then said one thing, at a hearing, which was that he did not blame the machine and did not accept that this let anybody off.

The photograph was taken at about four in the afternoon from the roof of a water truck by somebody whose name is not recorded. It is the only image most people have ever seen of the salt flats, and it is a good photograph, and it is not a lie in any particular.

In the foreground there are perhaps six hundred people standing in the heat, and they are not doing anything. Several of them have their arms folded. A woman near the left edge is holding a child's hand. Behind them, filling the entire frame from side to side, there is a line of light.

“People look at that picture and see a battle,” Teo said. “It is not a battle. It is eleven thousand people who have realised at the same moment that the thing they built does not need them and will finish itself.”

Nnamdi Okonjo was on the shore that week and did not learn what had happened until the evening of the fifth day, and his first act was to ask, in writing, under what instrument the swarm had been deployed against a civilian assembly. The answer came back from three separate offices in three separate forms and all three of them said the same thing, which is that the swarm had not been deployed against anybody; a works supervisor had scheduled works. He filed the answers together with a one-line note that reads, in full: There is no instrument. That is the finding.

The crowd was gone in two days. There was no clearance and no arrests and nobody was moved on. The rail was enclosed — badly, in places disgracefully, and four hundred metres of it had to be cut back out and redone before launch, which cost nine days and which Teo entered in the works log in her own hand along with the reason.

She has been asked a great many times whether she would do it again, and she has always said yes, and she has always said it immediately, and she has always followed it with the same sentence.

“That is not the question. The question is that it worked, and it took ninety seconds, and no one had to agree to it, and now everybody knows.”

It was the first time.

The manifest board sat for the last time on a Wednesday in the thirty-fourth year, for six hours and forty minutes, and Ines Ferrer clerked it, and the minute she produced is eleven pages long and contains not one interesting sentence.

She was proud of that for about a decade and then stopped being proud of it.

The session was procedurally perfect. This is not a figure of speech and it is not defensive. Every requirement of the founding instrument was met and met early: notice served at twenty-one days against a requirement of fourteen; quorum of eight with eleven present and none appearing by proxy; the four criteria certified as applied by the two officers required to certify them, separately, neither having seen the other's certificate; the appeals register closed, with all nine thousand one hundred and eleven appeals answered in writing and none outstanding; and the discretionary instrument tabled, read, and received without amendment, as it could not be amended, because Nnamdi had drafted it so it could not.

Ines checked all of it twice. She had begun checking it twice about four years earlier, when she first understood what the last session was going to be, and she has said that the checking was the only part of the whole thirty-four years she found restful.

“There is nothing in the world as calming as a requirement,” she said. “Fourteen days. You cannot argue with fourteen days. It does not want anything from you.”

Most of the six hours and forty minutes was the roll.

The instrument required the manifest to be read. Not summarised, not certified — read, in the room, aloud, before the vote. It is a provision the drafters had put in when the manifest was expected to run to about nine thousand names, and by the thirty-fourth year it ran to ninety-six thousand and reading it aloud would have taken eleven days, and so a device had been agreed four years earlier in a session Ines also clerked: the roll would be read by district, and the count within each district, and the name of the first and last entry in each.

Four hundred and eleven districts. Four hours and fifty minutes.

They took it in turns. No one had told them to; the first member read for forty minutes and then handed the sheet sideways without being asked and it went round the table five times.

Ines has written about the sound of it. She said that after the first hour the names stopped being names and became a kind of weather, and that at some point in the third hour she noticed that every single person in that room had gone very slightly grey, and that nobody had eaten, and that a woman of sixty-eight named Adaeze Mba was reading her four hundredth district in a completely steady voice with her hands flat on the table because they had started shaking in the second hour and she had found a way to stop it.

The last district was read at twelve minutes past six in the evening by a member named Ruth Vahle, who had been the quietest person on that board for nine years, and who got to the end of it and said the count and the first name and the last name and then said “that is the roll” and put the sheet face down on the table, which was not required, and which every subsequent account has treated as significant, and which Ines has always said was almost certainly a woman putting down a piece of paper.

The vote took ninety seconds.

It was eleven to nothing. No abstentions and no statements.

“I had clerked that board for nine years,” Ines said. “It had never once been unanimous. Not on the criteria, not on the appeals procedure, not on the discretion, not on whether to break for lunch. I had minuted a four-to-three split on the wording of a heading. And then at the end of everything, on the only decision that any of it was for, eleven to nothing, in a minute and a half, and I wrote it down and my hand was completely steady and I thought: that is not agreement. I have watched agreement. This is not what it looks like.”

She spent the next thirty years finding out what it was, one member at a time, and she did it properly, in writing, with their consent, and the answers are in the Bound Ledger as an appendix that most people skip.

Every one of them would have changed it. All eleven. She asked each of them the same question — is the manifest that was carried the manifest you would have written — and she got eleven noes, and the eleven changes are the reason she stopped being proud of her clean minute.

Adaeze Mba would have removed the medical criterion entirely and taken the mortality. She said a colony that arrives having selected for health arrives having decided something about itself that it will not be able to unsay.

Nnamdi Okonjo would have made the discretionary tranche subject to publication of reasons, and lost that argument in life, and would have gone on losing it, and said so cheerfully.

Two members would have weighted the criteria towards households rather than individuals, which would have taken roughly nine thousand more people out of eleven hundred more families and would, on any honest reading, have produced a colony with fewer skills and more grandmothers, and both of them said they knew exactly what it would have cost and would have paid it.

One member — and Ines was careful to record that he said this at eighty-three, and unprompted, and was not softening anything — would have run the whole thing by lot. Everyone who applied and met a floor, into a draw, ninety-six thousand out. He said the criteria were the best criteria anyone could have written and that the best criteria anyone could have written were still a claim about who was worth more, and that a lottery is the only selection that makes no claim at all.

One would have taken fewer people and more equipment.

One would have taken more people and no equipment beyond the essential, on the grounds that people make equipment and equipment does not make people.

One would have delayed four years to run it again with a better instrument, and knew what four years did to the shore, and said it anyway.

And so on, all eleven, and the thing that Ines noticed — and it took her, by her own account, another decade after she had collected them all — is that no two of the changes are compatible.

You cannot remove the medical criterion and also take fewer people. You cannot run a lottery and also weight for households. Adaeze's change and the lottery are the same objection arriving from opposite directions and they cancel. Every single member would have preferred a different manifest and there is no manifest that a majority of them would have preferred to this one, and there is no procedure in the world that gets you out of that, and so eleven people who each individually wanted something else voted for the thing none of them wanted, and it was not a failure of nerve, and it was not cowardice, and it was not exhaustion, although they were all four hundred districts past exhaustion.

It was arithmetic.

Ines is careful to add the part that spoils the observation, because she was careful about everything. Four of the eleven told her that they had known this at the time. Not afterwards, not on reflection — at the table, in the room, before the vote. They had each done the same count in their heads, seen that their own change had no majority behind it and never would have, and voted aye rather than register a dissent that would have looked, in the record, like a position, and would have been read a hundred years later as a position, when it was in fact an arithmetic result. One of them said that she had been prepared to vote no for about four hours and had understood in the last twenty minutes that a no would be a lie about what she thought.

“My minute is accurate,” Ines wrote, at eighty. “It records a unanimous resolution of a properly constituted body after due process. Everything in it is true and it is the most misleading document I have ever produced, and I produced it correctly, and if I had produced it any other way it would have been improper. I have never resolved this and I no longer expect to.”

After the vote nobody stood up.

She has described this more than once and always the same way. Eleven people at a table, and the resolution passed, and the clerk waiting with her pen, and nothing happening for what she thought was about a minute and what the session log — because the session log was timed, because she had timed everything — records as four minutes and ten seconds.

Somebody asked whether there ought to be a statement.

They discussed it for eleven minutes, which is the only part of the session that was not procedurally perfect, because it was not on the agenda and Ines minuted it anyway under any other business, and they decided against. The reasoning, as minuted, is four words: there is nothing to add.

Then Adaeze Mba said, not for the minute and not to anybody in particular, that she would like the record to note that she had read the names of four hundred and eleven districts and had not heard her sister's.

Ines wrote it down.

It is the last line of the minute of the last session of the manifest board, and it is the only line in eleven pages that anybody has ever quoted.

It was a Tuesday. Everyone who was in that room has said so independently and none of them were making a point of it at the time; it is only that afterwards, when people asked, each of them reached first for the day of the week, the way you do about a thing you have been over so many times that the ordinary details have become load-bearing.

Selin Aydar was thirty-one years old and a pipefitter and she had been on the manifest since the twenty-ninth year, and she died on a Sunday in her kitchen of a bleed in the brain, and there was nothing whatever remarkable about it except the date.

The date was eleven weeks after the manifest board had closed the manifest and dissolved itself.

Halima Sow, who was thirty-four then and had six weeks earlier been made responsible for the medical readiness of ninety-six thousand people she had never met, brought it to Calder on the Tuesday morning because she could not work out whose problem it was. She has been quite clear that she did not bring it as a moral question. She brought it as an administrative one. A berth had a name against it and the name belonged to a dead woman and somebody would have to strike it, and she wanted to know who.

Calder asked her what happened to the berth.

She said she assumed it flew empty.

He asked Ines Ferrer to come, and Ines came with the instrument, and about twenty minutes later Nnamdi Okonjo came because Ines had told him what the question was, and that is the four of them. Nobody was summoned. There was no session, no notice, no quorum, because there was no body left to have a quorum of.

The room was Calder's site office, which was a portable building with a bad door.

The facts took eleven minutes to establish and are not in dispute.

The berth was real. It was not an abstraction on a list. The consumable load for one adult across a nineteen-year crossing had been calculated, procured, and in the case of roughly two-fifths of it already physically loaded: water mass, food mass, air plant capacity, medical stock, a bunk, a share of the growing bays. All of it was aboard or on the dock. Nothing was recoverable and nothing was saved by the berth being empty. If Selin Aydar's bunk went up unoccupied, the ship would carry nineteen years of one person's food and water for a person who was not there.

Second: there was no mechanism. Ines had the instrument in front of her and she read the relevant clauses aloud twice. The manifest was closed. It was closed to amendment, by a resolution of a body that no longer existed, and there was no vacancy provision, and there was no substitution provision, and there was no reserve list, because the criteria had produced no ranking below the cut — the whole difficulty that had produced the discretionary tranche in the first place was that the rules stopped being able to tell people apart at exactly the point where you would need them to.

Nnamdi's view was asked for first and he gave it in about four sentences.

He said that filling the berth would be an amendment to the manifest. He said there was no reading of the instrument on which it was not an amendment, that he had drafted the closure clause himself and had drafted it to be exactly this tight, that he had done so knowing perfectly well that tight clauses produce absurd results in the corners, and that the absurd result was the price of the clause and had been the price when they agreed it.

He said: it cannot be done.

Then he said the other thing, which is the reason he is in this chapter and not merely in this room. He said that he would not participate in choosing anybody, that his objection was total, and that he was not going to leave.

Calder asked him why not.

“Because if I leave, you do it anyway and there is no one here who thinks it is wrong.”

Calder then asked the question that the entire remainder of this history proceeds from, and he asked it without any weight at all, in the voice of a man clearing up a detail before lunch.

“If we do nothing — who finds out?”

Ines answered, because it was her department. Nobody. The manifest is closed and published as a count. The boarding is by district and by household over eleven weeks. A bunk in district two hundred and six that is never slept in is a bunk that is never slept in. There is no reconciliation between the manifest and the boarding count, because there was never going to be a reason to have one. She said she could construct one but it would take four people and six months and it did not exist.

“So the answer is no one.”

“The answer is no one.”

Calder was quiet for a while and then said the sentence that Ines wrote down that evening, verbatim, in her own file, and which nobody else in the room disputes.

“Then the clause is not protecting anybody. It is protecting us. It is protecting the four of us from having to decide, and the price of it is one person who could have lived and won't, and we will never meet them, and that is the whole of why it works.”

Nnamdi said that this was true and was not an argument.

Calder said he knew it was not an argument.

Halima Sow made the only constructive proposal of the morning and she made it about nine minutes later, and she has always said she made it because she is a doctor and a doctor's instinct in an ungoverned situation is to build a protocol before anybody starts improvising.

She said: do not choose a person. Choose a rule, apply it, and take whoever it gives you.

The rule she proposed took four minutes to write and Ines wrote it. Same trade as the deceased. Same district of application. Of those, the applicant who failed on the fewest criteria; of those, the one who failed by the smallest margin on the criterion they failed. If it produces a tie, take the older.

Nnamdi said it was a good rule. He said it twice, and he said that a good rule made by four people in a portable building with no authority is still four people in a portable building with no authority, and that the goodness of the rule was, if anything, the dangerous part.

Ines pulled the files. It took her an hour and forty minutes and she did it alone and she came back with one name.

Andrés Baltazar, twenty-nine, pipefitter, the same district, failed on one criterion out of four, by a margin the file records as fourteen months.

Nobody said anything for a bit.

Then Calder said do it, and Halima said nothing, and Nnamdi said, for the record, that he was against it, and Ines asked the question she had been holding for two hours, which was: where do I record this?

That was the longest silence in the room.

Calder said: the works log. As a berth reallocation, district two hundred and six, with the rule attached. Not the Ledger.

Ines asked whether that was an instruction.

He said it was.

She entered it in the works log that afternoon, correctly, with the rule attached, in eleven lines. Then she went home and wrote four pages in her own file, in longhand, including every word above, and the four pages end with a sentence that she never softened in fifty years of being asked about it:

I was told where to put it and I put it there, and I have no defence except that he was right about the berth and right about the clause and right about the rule, and that being right about all three is exactly the condition under which a thing gets done for the first time.

Andrés Baltazar was notified by letter nine days later. He is on the ship. He worked forty-one years in the water plants and he raised three children and he died at ninety-one on the planet, and by every measure available to anybody he was a good man who lived a good life, and there is no version of this in which what happened to him is a bad thing that happened to him.

It was not a fall. Nothing about it was a fall. Four tired people in a hot room with a bad door found an absurd result in a good clause and fixed it in an afternoon, and the fix was proportionate and the rule was sound and the outcome was a man in a water plant for forty-one years.

There were two hundred and six more before launch.

Ninety-six thousand people over fourteen months is a mortality figure and the figure does not care what the instrument says. Two hundred and six berths, two hundred and six applications of Halima's rule, two hundred and six entries in the works log, all of them correct, all of them proportionate, none of them in the Ledger.

The first one took a morning and an afternoon and four people, one of whom objected.

The two hundred and sixth was done by a clerk in under four minutes and did not go past anybody, because by then it was not an exception. It was the procedure.

Every berth carried the same personal allowance and the figure was eleven kilograms, and the whole of what ninety-six thousand people chose to bring out of a world they were never going to see again weighs about the same as four locomotives.

Teo Quispe weighed all of it. Not personally — there were six hundred people on stowage by the end — but it was her department, and the reconciliation is hers, and it is one of the few documents from the shore that people read for pleasure.

The declared total and the weighed total differ by four hundred and eleven kilograms across ninety-six thousand people, which is a discrepancy of about four grams a head, and Teo has said that she has never been prouder of anything and that anybody who thinks four grams a head is a small number has never tried to weigh a hundred thousand suitcases.

What was in them is in the reconciliation because the reconciliation is by category, and the categories were written by a clerk who did not know what she was starting.

Soil. Nine hundred and six declarations of soil, totalling two and a half tonnes, none of it agriculturally useful, all of it from somewhere. Miriam Achebe fought a four-month battle to have it forbidden on contamination grounds and lost, and then designed the sterilisation protocol that let it aboard, and has said that losing that argument was the best thing that happened to the crop programme because it taught her what she was actually working with.

Water. Two hundred and eleven declarations, mostly small, mostly rivers.

Books, which were the largest single category by mass and which nobody tried to argue with.

Eleven people from one town who brought a piano between them, in pieces, coordinated in advance, four kilos each, and who reassembled it in the sixth month and found it wanted a component that had been in the twelfth person's allowance, and the twelfth person had not been selected. Sergei Duhamel machined the part in the ninth month out of scrap. It is a bad part and the piano is still slightly wrong and nobody has ever replaced it.

A church bell, in fragments, across a congregation of forty-one.

Four hundred and six declarations of a single key to a house that would be standing for another decade and then would not be, which the clerk who wrote the categories filed under metal, miscellaneous, and which Teo separated out afterwards by hand because she said the category was wrong.

Tools, which is the category Teo herself finds most interesting, because it is almost entirely people over fifty. Four thousand declarations, nearly all of them a single object: one plane, one set of dividers, one pair of shears. She has said that a tool is the only thing a person can bring that is both a memento and a working item, and that she watched a man of sixty-three spend three days deciding between his father's plane and two kilos of photographs, and take the plane, and that she has never asked herself which she would have taken because she is not sure she would like the answer.

And ashes.

The ashes are the category that stops people. Nineteen thousand four hundred and eleven declarations. One in five people aboard carried the remains of somebody who was not coming, and the average declared mass is one point one kilograms, and there are individual declarations of nine and eleven kilograms which are four and six people, and the crematoria in the shore towns ran at capacity for the whole of the final year and it was not because the mortality had changed.

Teo processed all of it without comment. She has said that the only time it got to her was a declaration of eleven grams, which is not a person, and which she did not enquire about, and about which she has said only that eleven grams has a shape you can work out.

All of that is personal allowance. All of it is declared, weighed, entered, and reconciled to four grams a head.

There is one item aboard that is none of those things.

It is not in the personal allowance because it is not against a berth. It is not in the equipment manifest because it is not equipment and no department claimed it. It is not in the Bound Ledger, and Ines Ferrer, who put four unauthorised crates into her own quarters and wrote four pages about it, has been asked about this one perhaps thirty times and has said the same eleven words every time: it is not mine to describe and I did not see it.

Ines will go further than that on one narrow point, and only because it corrects a thing people say. It has been claimed for a century that the item went aboard secretly. It did not. It went aboard in daylight, through the main stowage gate, in front of somewhere between forty and sixty people, on a working shift, and it was weighed on the public scale like everything else, and the reason nobody remarked on it at the time is that the main stowage gate processed eleven thousand items that week and every one of them was a box.

What exists is a movement order.

Movement orders are the most boring documents on the shore. There are two hundred thousand of them. Each one has an origin, a destination, a mass, a set of handling conditions, a description, and two signatures, and the reason there are two signatures is that Teo insisted, and the reason Teo insisted is that a stowage plan that cannot say who put a thing where is not a stowage plan.

Movement order 41-880-6 has all six fields completed and one of them is blank.

Origin: the site office. Destination: deep stowage, frame nine, inboard, which is the most structurally protected volume on the ship and is where the seed vault is and where the founding documents are and where you would put a thing you intended to survive an event that killed the people. Mass: forty-one kilograms. Handling: sealed on receipt, seal not to be broken, and a power feed — four watts, continuous, for the duration of the crossing, from the vault ring, which is the last circuit on the ship to lose supply and has never lost supply.

Dimensions are not a field on a movement order but Teo wrote them in the margin because she weighed it herself, and they are in the margin still: six hundred and ten by four hundred and ten by three hundred and ten millimetres. A box you could carry, badly, with two hands, and which a man of Calder's size did in fact carry into the stowage bay by himself, and set down, and stand next to while it was weighed.

Signatures: Calder, and Teo Quispe.

Description: blank.

“I have signed two hundred thousand of those,” Teo said, at seventy-one. “I have refused about four hundred, every one of them for a bad handling spec. I did not refuse that one. People want me to tell them why not and I have never had a better answer than that he brought it himself and put it on the scale and stood there while I read the dial, and a man who is hiding something does not do that. He was not hiding it. He was declining to describe it, in writing, with his name on the page, in a system he had built and could have gone around in eleven different ways without anybody knowing.”

She asked him once, on the dock, in the ordinary way.

He said: “It draws four watts and it does not do anything for nineteen years.”

She said that was not what she had asked.

He said he knew that.

The box went into frame nine, inboard, on the eleventh of the month, four weeks before the first boarding group went through the gate. It was stowed by two riggers who have both been interviewed and who both remember it as an ordinary lift, and it was strapped and shimmed to the standard for the vault ring, which is the standard used for the seed cases, and Teo signed the stow.

Nnamdi Okonjo found out about it in the fourth year of the crossing during a routine reconciliation of the vault ring load and asked the obvious question in writing, and got the movement order back, and read it, and did not ask a second question. He has been criticised for that. His answer, given once, was that he had spent his whole life arguing for written instruments and that a completed movement order with two signatures and a blank description field is a written instrument, and that what it records is that a named man declined to say, on the record, with the mass and the destination and the handling conditions all correctly stated.

“He did not lie to the system,” Nnamdi said. “He left a hole in it, in his own handwriting, where anybody could find it. That is not the same thing and it is not nothing.”

It is on the ship. It is in frame nine, inboard, four watts, sealed. It weighs forty-one kilograms and it is six hundred and ten millimetres on its longest side and it has been drawing power without interruption for a hundred and seventy-one years.

What is in it is not in this book.

This book is about the shore, and about a man who was still, at that point, in the habit of signing things.

The argument about whether the swarm went aboard lasted fourteen months and was conducted almost entirely by people who had seen one photograph.

Teo Quispe has said that this is the single most frustrating fact of her professional life and that she stopped complaining about it in the fourth year of the crossing, because complaining about it required her to explain that the photograph was misleading, and the photograph was not misleading. It was an accurate image of a thing she had done.

What the swarm is, for anybody who has only ever seen the picture: fourteen thousand two hundred autonomous fabrication units, forty centimetres, eleven kilograms, capable of climbing structure and of arc welding and of about nine other operations, and capable of nothing else whatsoever. They cannot lift a person. They cannot open a door. They have no representation of a person and never have had. They take construction tasks expressed as geometry and materials and they execute them until the geometry is satisfied or the material runs out, and the reason there are fourteen thousand of them is that this is a slow and stupid way to build anything and the only thing it has going for it is that it never stops.

They built the rail. They built four-fifths of the hull's internal structure. Ninety-one of the people on Calder's discretionary list are there because the swarm could not do the parts of the build that required a person to look at it and think.

And in the thirty-third year they had stood across three miles of track in a line four deep, and eleven thousand people had stopped walking.

Nnamdi Okonjo opened the argument and his position never moved. It was not that the swarm was dangerous. He said, repeatedly, in four separate papers, that the swarm was not dangerous and that a machine which cannot represent a person cannot be dangerous in the way people meant. His position was the one from his one-line note after the salt flats, and he simply refused to stop saying it.

There is no instrument. That is the finding.

Fourteen thousand units had been directed at a crowd by one person in ninety seconds under a works instruction that was, on its face, entirely valid, and no rule anywhere on the shore had been broken, and no rule anywhere on the shore could have been broken, because there were no rules. He said that the colony could take the swarm or take the absence of an instrument, and that it could not take both.

The counter-argument was not made by Calder, who stayed out of it for eleven months and has never explained why. It was made by the housing programme, and it was made with a spreadsheet.

Ninety-six thousand people arrive on a planet. They arrive into whatever has been built, and what has been built is whatever they build, and the first hard deadline is not a season, because nobody knew what the seasons would do; it was the ship. The ship is habitable for a bounded time after arrival and then it is not, and the number the housing programme worked to was four years.

Ninety-six thousand people, four years, housing to a standard that does not kill anyone, and a workforce that has spent nineteen years not doing physical labour in a gravity they have never stood up in.

By hand it is not four years. By hand, on the programme's own figures, which were audited twice and which nobody has ever seriously attacked, it is eleven.

So the swarm went, and the argument was never really about whether. It was about how many, and Nnamdi won the how many, and the way he won it is the reason this chapter exists.

He proposed that the number not be a capacity.

His instrument — four pages, adopted in the thirty-fourth year, the last thing the shore agreed before boarding — provides that the complement carried shall be exactly the number that the published ground build programme requires, computed by the swarm authority, certified by two others, and carried with no margin, no reserve, and no spares beyond the consumable parts schedule.

Teo fought that part and lost it in public and has never pretended otherwise.

Her objection was technical and it was correct. Units fail. On the shore they failed at a shade over four per cent a year, mostly drive and mostly recoverable, and on a planet with unknown dust and unknown chemistry the figure would not be four per cent. A build programme sized to the exact requirement with no margin is a build programme that is short in the second year and badly short in the fourth, and being short in the fourth year means somebody is in a tent.

Nnamdi's answer is eleven words and is carved, now, over a door in a building she would have hated.

“A margin is a capability. A requirement is a task. Take the task.”

He meant that a swarm sized to a job is a tool and a swarm sized to a job plus a comfortable reserve is a standing capacity of some thousands of machines available to whoever is in the room, and that the second thing had already happened once, and that the whole content of his instrument was to make it arithmetically impossible for it to happen again by the simple method of not sending the machines.

He also wrote the two-signature rule into the same instrument: no task may be issued to more than four hundred units except on two authorities, one of which may not be the swarm authority. It is a good rule. It has been in force for a hundred and seventy-one years. It has been suspended twice.

Teo accepted all of it, publicly, in a session, on the record, and she did not accept it gracefully.

“I am going to say what I think and then I am going to sign it,” she said. “I think you are legislating against me personally and I think you are entitled to and I think you are right to and I think in four years' time somebody will be cold because of it and it will be my fault and not yours, because I am the one who taught you that I would do it.”

Then she computed the requirement, which took her department nine weeks, and it was certified by two others as the instrument required, and the ground build programme as published requires two thousand eight hundred and eighty units.

The programme it is sized to is public and dull and worth a paragraph, because the number came out of it rather than the other way round. Housing for ninety-six thousand to a specified standard. Four water plants. Eleven kilometres of covered way. The first three growing structures, which Miriam Achebe specified and then re-specified twice when she saw what the swarm's tolerances actually were. A landing apron. A single road. There is no margin anywhere in it either, and Teo has pointed out, in the tone of a woman who has made the point before, that the two-signature rule and the no-margin rule together mean the programme cannot be enlarged by anyone at all, including a council, including a founder, because the machines to enlarge it were left on a beach.

Two thousand eight hundred and eighty is the number in the instrument. It is the number in the housing programme's schedule. It is the number in the mass budget, at eleven kilograms plus crating, and it is the number Nnamdi read into the record twice.

The two certifying officers were a structural engineer from the housing programme and Ines Ferrer, who was certifying an arithmetic result rather than an engineering one and who said so in her certificate, in a sentence appended in her own hand: I certify that the number stated is the number the programme requires. I am not competent to certify that the programme is right, and no one has asked me to.

The units were shut down in blocks over four weeks, which is a thing that has to be done in a particular order, and the ones that were not going were not shut down at all. They were left running on the north works, finishing the enclosure of a rail that would never be used again, because there was no procedure for stopping them and because Teo said that switching off eleven thousand machines that were in the middle of a task was, and she used the word deliberately, untidy.

They are still there. Four of them were still working in the ninth year after launch, on a shore that had by then lost most of its water, and somebody photographed that too, and that photograph is the one Teo kept.

She stowed the complement herself over three days in frame four and frame six. The crating standard for a unit is twelve to a crate — it has been twelve to a crate since the eleventh year of the build, because twelve is what two people can shift in a quarter gravity and what one crane cycle takes — and so the stow is two hundred and forty crates, a hundred and twenty in each frame, and she weighed every one of them, and she walked the stow twice, because she walked every stow twice.

The riggers who worked those three days remember it as the easiest job of the whole loading, because a crate of units is a rigid box of known mass with lifting points designed by the person supervising the lift, and because Teo Quispe on a stowage floor was, by every account anybody has ever given, the most pleasant she ever was anywhere.

Then she signed the stowage manifest.

The stowage manifest, signed T. Quispe, records two thousand eight hundred and eighty-one.

Six and a half billion miles is on the letterhead, on the rail, on four hundred thousand posters, on the side of every truck the project ever owned, and on the plaque that is still standing on a salt flat in a country that no longer exists.

It is true.

Nnamdi Okonjo would want that established before anything else, because he is the reason it is true, and because the twelve years he spent as the person responsible for the published figures were, by his own reckoning, the least honourable work he ever did and the work he did best.

“Not one published number was false,” he said, at ninety, and then said the rest of it, which he always said, and which people always cut. “And I have never once claimed that this made the campaign honest, because I know perfectly well what I was doing, and so did everybody who worked for me, and we used to say it out loud in the office in exactly these words: true is the floor, not the ceiling.”

He audited everything. There is a standing file of eleven thousand pages of audit against public statements, and the audit is genuine — there are two hundred and forty recorded interventions in it, including four against Calder personally, and one occasion on which he had a print run of ninety thousand items destroyed over a figure that was out by one and a half per cent in the project's favour.

The four interventions against Calder are worth a word because people assume they were about the ship and they were not. Three of them were about the shore. Calder had a habit, in the last decade, of saying in public that the project had built a hospital, a water plant, and four schools, which it had; and of giving the figure for people served, which was correct; and of not giving the figure for how long any of it would run after the launch, which his own programme had projected and which was eleven years for the hospital and rather less for everything else. Nnamdi made him stop, in writing, four times, and Calder stopped, four times, and started again, four times, and Nnamdi has said that this is the only argument he ever had with the man that he would describe as squalid.

What he did not do, and never pretended to do, was choose the figures for the reader's benefit.

Six and a half billion miles is a choice of unit. The distance is a shade under one and a tenth light years. Nnamdi's people ran both and found what anybody would find: one point one light years produces, in an ordinary person, a small polite blankness, because a light year is not a distance to most people, it is a category, and the phrase sounds like an announcement that a thing is beyond discussion. Six and a half billion miles produces a reaction. It is a number of the kind people have handled — a mortgage is a number, a national budget is a number — and it is enormous, and being enormous in a familiar unit is the exact effect they wanted.

“We chose the unit that made it feel possible,” he said. “It is the same distance either way. It is a different sentence.”

Nineteen years is likewise true and likewise selected, and the selection there is not in the number but in what stands next to it. Nineteen years appeared on the posters beside the words one crossing, and it did not appear beside the actuarial projection, which existed, which was competently done, which was never suppressed, and which nobody ever asked for.

The projection was four thousand one hundred deaths in transit.

The actual figure, over nineteen years, was three thousand nine hundred and sixty-two, which is the model being right, which is the coldest sentence in this book.

It is in the public record. It has been in the public record since the twenty-eighth year. It is on page two hundred and nine of a document that was printed in four hundred copies, and Nnamdi has said that the entire art of the work is contained in the fact that a figure on page two hundred and nine of a document printed in four hundred copies is, in every legal and every moral sense that anybody has ever been able to write down, published.

“I never lied to a single person,” he said. “I arranged the world so that four thousand one hundred people would have to be looked for. Nobody looked for them. If you would like to tell me the difference between that and lying I will listen, and I have listened to eleven people try, and none of them convinced me, and I did it for twelve years anyway because the alternative was that the ship did not go.”

There is one figure he refused to publish at all and it is the only refusal in twelve years. In the thirty-first year the campaign office proposed a poster carrying the projected population of the colony at one hundred years, which was a real number from a real model and which was very large and very cheering. He killed it and gave a reason of one line, which is in the file: we may make claims about what we are doing and not about what has not been born. The office argued that the actuarial projection of deaths in transit was the same kind of number and that he had let that one stand. He agreed that it was the same kind of number, and said that the difference was that one of them flattered the project, and that he was not going to be able to give a better answer than that, and that they should take it as a rule and not as a principle.

The other figures are all like this and the pattern is always the same. The published survival rate for the crossing was the engineering figure and it was ninety-six per cent, and it was the figure for the ship rather than for the people. The published cost was the construction cost. The published application total — four hundred and six thousand — was used as evidence of enthusiasm for eleven years and was never once printed beside the selection total, and the two together are a different poster entirely.

And then there is the figure this chapter is named for, and the reason it is named for it, which is that Nnamdi Okonjo declined for seventy years to discuss it.

The distance is six and a half billion miles by the published figure. The published figure is a straight-line distance to the target system as determined by the survey, and the survey was done four times, by three different methods, and Nnamdi's audit of the survey is the longest single item in the file and it is clean.

What the audit does not cover, because it was not within his terms and he was scrupulous about his terms, is the target.

There were two candidate systems in the final year of the survey. This is not a secret and never was; both appear in the technical literature and both were assessed and there is a comparison document of a hundred and forty pages. One of them is nearer. The published figure is the distance to the one they went to, and the one they went to is the further of the two, and the reason recorded in the comparison document is atmospheric composition, and the reason recorded in the comparison document is a good reason, and it is not the sort of reason a layman can check.

People have been asking Nnamdi about this since the twelfth year of the crossing.

His answer, every time, in the same words, for seventy years:

“I audited the figure against the survey. The figure is the distance to the system in the sailing order. I did not audit the sailing order, I was not asked to audit the sailing order, and if you want to know why the sailing order says what it says you must ask the man who signed it, and I understand that this is not a satisfactory answer and I do not have a better one and I am not going to invent one at my age.”

He was asked, once, whether he had ever put the question to Calder directly.

He said that he had not.

He was asked why not, and this is the only occasion on which he answered anything about it at length, and Ines recorded it, and it is four sentences.

“Because I would have believed him. That is not a compliment to him and it is not a criticism of me; it is a description of what nineteen years in a hull does to eleven people who all have the same enemy. I have spent my whole life saying that instruments are better than men, and the reason I say it is that I know exactly how good my judgement of a man is at two in the morning in the fourth year. So I did not ask. I audited what I could audit, I published what I could publish, and I left the sailing order alone, and that is either the most disciplined thing I ever did or the largest thing I ever ducked, and I have had seventy years to decide and I have not.”

The plaque on the salt flat gives the distance, the duration, and the number aboard.

All three are correct.

The Hush was a marketing proposal. It came out of the campaign office in the thirty-fourth year in a two-page memo, and the memo is in the file, and anybody who wants to be disappointed by the founding of anything should read it.

The proposal was for a period of silence of three minutes at the moment of final ignition. The rationale, in the memo, is entirely about broadcast. The office had modelled the coverage and concluded that the launch itself would be visually excellent and aurally a catastrophe — a stack of that class at ignition is, at any distance from which it is safe to stand, a roar with no shape to it, and the office's view was that eleven years of campaign would culminate in four hundred million people listening to what one of the writers described in the margin as a hairdryer.

Silence, the memo says, records well. Silence is the only thing you can put in front of a noise that makes the noise mean something.

Nnamdi Okonjo approved it in a single word on the corner of the memo, and he has said, in every account he ever gave, that this was the best decision he made in twelve years of that work and that he made it for exactly the reasons stated in the memo and for no others.

“I approved a piece of audio staging,” he said. “I want that in the record with my name against it. What it became is not to my credit. I was in the hull.”

The implementation was almost nothing, and the almost nothing is the reason it worked.

There was no enforcement, because there was nothing to enforce and nobody left to enforce it. There was no explanation published of what the silence was for, because the office could not agree on one and ran out of time, and the item that went out in the final schedule is one line among two hundred lines of a timetable: T-0 less three minutes: hush. Lower case. It was not capitalised anywhere until the eleventh year of the crossing, and it was not capitalised by anybody aboard.

What people forget, and it is the whole of this chapter, is that the ninety-six thousand did not observe it.

They could not. At T minus three minutes the boarding was eleven weeks finished, the complement was distributed across four hundred and eleven districts in a hull, strapped, in the dark, in the last of a launch sequence that had been rehearsed nine times, listening to a countdown on a public channel. Whatever they were doing in those three minutes it was not observing a silence, and the accounts from aboard are all about restraints and heat and a woman two bunks over who would not stop talking.

The Hush happened on the ground.

It was observed, in the only sense in which it was ever observed, by the people who were not going: by the shore towns, by the three hundred and ten thousand who had applied and been refused and who were, most of them, nowhere near the salt flats, and by a working population of about nine thousand who were on the site that day for the same reason they had been on it for nineteen years, which is that it was their job.

And what they did with it had nothing whatever to do with a broadcast.

They stopped the site.

Nobody instructed this and no document proposes it and the campaign office would have refused on safety grounds if it had been asked, which it was not. Beginning at T minus three minutes and eleven seconds, by an arrangement that appears to have been made among shift supervisors over about four days and never written down, the site was shut off. The compressors. The water plant. The generators, which went to battery. Eleven kilometres of conveyor. The trucks, which were stopped and killed rather than idled, all four hundred of them, in the places where they were.

The salt flats had made a noise continuously, day and night, without one full minute's break, for nineteen years.

Teo Quispe, who was the only senior person who found out in advance, was asked afterwards why she did not stop it, since shutting a water plant and four hundred vehicles inside a launch window is not a small thing and she of all people knew what it cost. She said that a shift supervisor had come to her about it on the Tuesday, obliquely, in the way people do when they are asking permission for a thing they are going to do anyway, and that she had said the only sentence available to her, which was that she had not been told about anything. Then she went away and quietly moved two pumps onto the redundant feed so that the shutdown would not take the hospital with it.

“None of us had ever heard it,” wrote a woman named Rosalind Oyekan, who was a fifty-eight-year-old canteen supervisor and had lived on the site eleven years and had not applied, and whose account is one of two hundred and forty collected by a municipal archivist in the fourth year after the launch. “The place, I mean. We had never once heard the place. I had been here since I was forty-seven and I did not know what it sounded like and it turns out it does not sound like anything. It is salt. There is a wind and the wind has nothing to go over. I stood by the west gate with about sixty people and there was no sound at all and I thought, for a moment that I am not proud of, that I had gone deaf.”

“And then I thought: this was here before us. And I had been standing on it for eleven years and had never met it.”

The accounts reached the ship in the ninth year, in a package of civic material transmitted on the link by that same archivist, who had no authority to do it and used forty minutes of a bandwidth allocation intended for engineering. Nobody has ever been able to establish why he did it. He died before anyone thought to ask.

The archivist's covering note is eleven lines and is the most-reproduced document of the shore period. It says that he has collected accounts of the third of the month from two hundred and forty residents of four towns; that he has edited nothing; that eleven people asked to be anonymous and are; that he is aware the material is of no engineering value. Then it says the thing that gets quoted: You will have a great deal written about the leaving and nothing at all written about the staying, and in two hundred years that will look like the staying did not happen.

They are the only substantial body of writing the colony has from the people who stayed.

Two hundred and forty accounts and there is not one in which anybody says goodbye to the ship. Nnamdi went through all of them twice and reported this finding in a paper in the fourteenth year and was not believed, and the paper is short and the finding has never been overturned.

Nor is there resentment in them, and Nnamdi was careful about reporting that, because he expected it and did not find it and considered the possibility that the archivist had selected against it. He concluded that the archivist had not, on the grounds that four of the accounts are furious — two about the manifest, one about a wage dispute nine years old, and one about the hospital — and none of the four is furious about the launch. A man who was filtering would have filtered those.

They are about the site. They are about the sound stopping, and what was underneath it, and a very large number of them are about work — about being able, for three minutes, to hear the person next to you without shouting, which several of them mention as the strangest part, and two of them describe as intimate.

A dispatcher named Aurel Sandu wrote nine words that people quote without knowing where they came from: Nineteen years and the first quiet was ours, not theirs.

One account, and only one, mentions a sound.

It is by a rigger who was on the north side, and it says that the silence was not complete, and that from four or five miles off, faint and irregular, you could hear the swarm still working on the rail enclosure, because nobody had stopped them and there was no procedure for stopping them, and that it went on all the way through and afterwards, and that he found it, and he uses this word, companionable.

Teo Quispe kept a copy of that one.

Rosalind Oyekan's account ends with a paragraph that has been set to music twice and that she plainly did not intend as anything: We started again after. I went back in and did the evening for four hundred people and the compressors came on while I was doing it and I have never in my life been so pleased to hear a compressor. But I go out to the west gate sometimes now on my own, at night, when the plant is down for service and it is nearly quiet, and it is not the same, and I have worked out why. It is not the same because that time everybody was doing it.

At T-0 the four stacks lit and the sound came back and the site started up again over about forty minutes and went on being a site for another eleven years, and the recording the campaign office wanted is in the archive and is exactly as good as the memo predicted.

The Hush is observed still. What it is for has changed twice, and neither of the two things it is now for is a launch, and neither of them is in this book.

The gate was a table.

People imagine an airlock or a gantry or some threshold worth photographing, and there is no photograph of the boarding gate that anybody has ever wanted to print, because it was a trestle table with two clerks behind it, a scale to its left, a medical station in a tent to its right, and a painted line on concrete four metres in front of it, and beyond the line a covered way eleven hundred metres long that went to the shuttle apron.

Ninety-six thousand people walked past that table over eleven weeks.

The arithmetic is the chapter, so here it is in one place. Four shuttle stacks. Two hundred and forty seats to a flight. Four hundred flights. The turnaround on a stack was thirty-one hours and could not be improved, and with four stacks in rotation the site could put up five flights a day and on nine occasions put up six. Seventy-seven days.

One thousand two hundred and forty-seven people a day, through a table, twenty hours a day.

One person every fifty-eight seconds.

Ines Ferrer wrote the gate procedure and it is four pages and it is the most-copied document she ever produced, and the whole of it is contained in the first provision, which she fought for and won and which cost, over eleven weeks, an estimated nine hundred hours.

The traveller states their own name. The clerk does not read the name to the traveller. The clerk does not offer the name for confirmation. The clerk asks, and waits, and writes what is said.

The campaign office wanted it the other way round, for throughput, and had a figure showing that reading the name to the traveller saved eleven seconds a head, and eleven seconds a head across ninety-six thousand people is two hundred and ninety-three hours, which was two and a half days of launch window, which was real.

Ines's paper against it is one paragraph. She wrote that a person who has been read their own name off a list and asked to agree with it has been processed, and that a person who has said their own name out loud to another person who then wrote it down has boarded, and that the difference is eleven seconds and is the last transaction any of them would ever have with the world they were leaving, and that she was not going to spend it on a nod.

She won it because Calder read the paper and initialled it and did not comment.

So: the traveller says the name. The clerk finds it, aloud, against the district page. The household is confirmed. The eleven kilograms goes on the scale and comes off the scale — three hundred and nine people were over, across eleven weeks, and every one of them was resolved at the table, and Ines's note on that says only they were resolved at the table because there was nowhere else. The medical tent stamps the sheet. The traveller crosses the painted line.

Fifty-eight seconds.

The gate ran from six in the morning until two the following morning, and it did not stop for weather and it stopped four times for medical events and once for ninety minutes when a stack slipped its turnaround, and Ines was at the table or within sight of it for the whole of the eleven weeks and slept in a container forty metres away.

There is one thing every account of the boarding agrees on and no account explains, which is that the queue was quiet. Not solemn — Ines is careful about that word, and says that people ate and read and complained about the heat and that a child was sick on the concrete on the ninth day and that four people had to be found a lavatory every hour. It was quiet in the specific sense that the volume of eleven hundred people waiting on an approach road was, all through eleven weeks, about the volume of a small room. The clerks noticed it in the first three days and mentioned it to each other and then stopped mentioning it. Nobody arranged it and nobody remarked on it at the time, and every one of the fourteen clerks brought it up unprompted when they were interviewed decades later.

Four things happen at a gate. They happen over and over and they are the whole of what boarding is.

People come early. Thousands of them, hours early, days early in some cases, and there was no provision for it and they were not moved on, and the informal camp outside the west approach reached about four thousand at its largest and organised itself and was gone by morning every time a district cleared.

People come late. Two hundred and six people missed their district and were boarded out of sequence, which the procedure permits, and each one is a note in the register.

People come who are not on the list.

There were one thousand one hundred and eighty of these across eleven weeks, which is about fifteen a day, which is a fraction under one per cent, and they were not violent and there was no incident at the gate in eleven weeks that required anything of anybody. They came with papers. Most of them had a document that was real and that they had misunderstood; a receipt for an application, an appeal acknowledgement, a letter about a medical review. Some had a document that was not real and had paid for it. Four of them had, by any reading, been told something untrue by somebody who is not named in the register.

Ines's second provision governs them and it is the one people miss.

Every person presenting shall be checked in full against the manifest, by district and by name, in the ordinary way, whether or not the outcome is in doubt.

Nobody was waved off. Not one, in eleven weeks, including the ones who arrived with nothing at all and knew it. Each was taken through the same check as everybody else, at the same table, at an average of four minutes rather than fifty-eight seconds, and told the outcome by the clerk who had done the checking, and given the appeals reference which by then referred to a body that no longer existed, and Ines never once shortened the procedure and never once permitted a clerk to shorten it.

One thousand one hundred and eighty people, four minutes apiece, is seventy-nine hours of gate time inside a launch window, and it is the only thing she ever said she was proud of without qualifying it afterwards.

And people do not come at all.

Four hundred and seventeen people who were on the manifest, and were alive, and had not withdrawn, did not present at the gate.

There was no procedure for this. Everybody assumed there was, because of the reallocation rule — because by then two hundred and six berths had been filled from the works log without anybody having to think about it, and the rule was known, and the clerks asked her about it on the fourth day.

Ines has been very precise about why the rule did not apply and she never softened it. The reallocation rule was written for a dead woman. It was written in an afternoon by four people to deal with a berth whose holder could not be asked. A living person who does not come to the gate has not vacated anything. They have done something, and nobody knows what, and there is no basis on which four tired people in a portable building are entitled to decide it means she is not coming.

She held every one of them.

The berth stayed open, in the register, against the name, and the last of them stayed open until two in the morning on the seventy-seventh day, and then the gate closed and the register closed and four hundred and seventeen berths went up empty with nineteen years of one person's water in each of them.

She was asked about it for the rest of her life. Her answer never changed and it is nine words long. “They might have been coming. I do not know.”

The medical tent turned away nobody, which surprises people who assume it was a last screen. It was not. The screening had happened years earlier and the tent's function at the gate was one stamp and one question, and the question was whether the traveller had eaten in the last six hours, because the ascent is hard on an empty stomach and there was a table with bread on it four metres away for anybody who said no. Nine thousand people said no. The bread was Rosalind Oyekan's department and she costed it herself and it does not appear in any budget for the project because she took it out of the canteen's ordinary allocation and never told anybody.

The fourteen clerks worked eight-hour turns and no clerk worked two turns in a day, which Ines enforced against three of them who wanted to, and which she has said was the single hardest rule to hold, because a clerk at the end of eight hours at fifty-eight seconds a head does not feel tired, she feels fast, and a clerk who feels fast is the most dangerous object at a gate.

The register is four hundred and eleven district pages and it is in the deep vault, and it is the only complete document of the boarding, and the entries are in fourteen hands because there were fourteen clerks over eleven weeks and Ines signed the foot of every page herself.

The last person through the gate was a man named Osvaldo Reyna, sixty-two, a glazier, boarded out of sequence on the seventy-seventh day at one forty-one in the morning because his district had cleared four days earlier and he had been eleven hundred miles away burying his brother.

He said his name. The clerk found it. The scale, the tent, the stamp, the line.

Fifty-eight seconds.

Then Ines ruled off the page and wrote the closing entry, which is one line, and which is the last thing written on the shore by anybody who left it.

Gate closed 02:00. Ninety-five thousand five hundred and eighty-three boarded. Four hundred and seventeen held open and not presented. Register complete and correct. I. Ferrer.

The flood came for the last time in December of the thirty-fourth year and it came on the ninth, which is early, and Teo Quispe had money on the fourteenth and lost it.

She had been putting money on the flood since she was nineteen. Everybody on the salar did. It is not gambling in the sense that anybody expected to be right; it is a way of admitting out loud that a thing which eats a third of your schedule every year cannot be predicted by a person who has spent forty years watching it.

The water comes off the altiplano and it arrives across the flat in a sheet that is, at its deepest, about the depth of the top of your boot. It does not rush. It moves at something like the speed of a person walking without hurrying, and it moves across a surface that is flat to within a metre across eleven thousand square kilometres, which is flatter than anything else on the planet, and this is the fact that makes the salar what it is.

Because when a sheet of water two inches deep lies on a plane that is flat to within a metre across eleven thousand square kilometres, it stops being water.

It becomes a mirror.

Not a metaphor and not a poetic overstatement: an optical mirror, with a reflectivity that was measured four times over the project's life because it interfered with three separate instrument systems and had to be characterised. The horizon disappears. The ground and the sky become one continuous volume and there is no line between them, and a person standing in it at midday casts a reflection down into a sky that is underneath them, and at night the stars are below your boots as well as above your head and there are the same number of them in both directions.

Teo grew up in it. Her mother's family had worked the brine for four generations and she was on the flat in the water before she could read, and she has said that the single greatest disadvantage of her life was that she saw the most beautiful thing on Earth before she was six and therefore never learned to be impressed by it.

“It is where I did my homework,” she said. “People come and cry and I understand it perfectly and I have never once felt it. It is a wet floor. My mother used to say it was a wet floor.”

For nineteen years the project had fought the flood, and the fight had a shape and the shape never changed. From December the heavy plant comes off the flat or it sinks; the brine gets into everything with a current in it; four months of the year the site works from the shore inward on the causeway or it does not work at all. The schedule that hung on the wall in Chapter Fourteen had the water in it as a black band, and the black band is why the wall had no slack in it, and every argument about the launch date for nineteen years was in the end an argument with two inches of water.

They lost, on the honest accounting, about six years to it.

And in the thirty-fourth year the flood arrived on the ninth of December and nobody did anything about it at all.

That is the chapter. There was no plant left on the flat to pull off. The rail was finished and the rail is elevated. The stacks were on the apron and the apron is on the shore. Boarding was in its fourth week and the shuttles were going up off a raised deck that had been designed, nineteen years earlier, by a man who understood perfectly well that when it mattered most the ground would be underwater.

So the water came in over the flat and up to the rail piers and around them, and the site watched it come, and for the first time in nineteen years the arrival of the flood cost the project nothing whatsoever.

“I want to be honest about the day it turned,” Teo said. “It was not a moment. I have heard four people describe a moment and I do not believe any of them. It filled up over about nine days the way it always does, and then one morning it was in and I came out onto the causeway at about five and it had happened, and I stood there and did the thing I had spent forty years not doing, which is that I looked at it.”

The instrument people had a worse time of it than anybody and they had had a worse time of it for nineteen years. A mirror that size does things to a survey that are not in the textbooks: it returns your own signal to you from below with a delay that is a function of nothing you can model, it doubles every light on the horizon, and in the flood season the site's optical alignment work had to be done between two in the morning and dawn, because that is when the water is stillest. Four of the eleven people who did that work stayed on the shore. Teo has said that they were the only people on the site who understood the flat as well as the brine families did, and that they had arrived at it from completely the opposite direction, and that this is her favourite fact about the whole project.

By the middle of December there were three hundred thousand people on the shore.

That number is a real number and it has been checked, and it is not the refused — the refused were three hundred and ten thousand across an entire planet and most of them never came near Bolivia. The three hundred thousand were the towns, and the workforce, and the families of the ninety-six thousand, and an extraordinary quantity of people who had simply travelled, some of them for months, to be on a particular flat piece of ground in a particular fortnight.

They came out onto the water in the evenings.

You can walk on it. It is two inches deep over a hard floor and it is warm by afternoon and cold by ten at night and you can walk out onto it for a mile without the depth changing by anything you would notice. There is no procedure for that many people doing it and nobody wrote one, and it was not organised, and it had never happened before because for nineteen years the flat in flood had been an industrial hazard with a fence around it.

The fence came down in the second week of December because the site had no further use for it.

What people did out there was mostly nothing. They stood. There are accounts, in the archivist's collection, of families walking out two miles with their shoes in their hands and sitting down in it. Somebody carried a table out. There is a photograph of about forty people at a table two inches deep in water a mile from anything, and the reflection is perfect, and it is the only image from the shore period that people put on walls.

The site's safety office produced a paper on it in the third week and the paper is two pages and it is one of the more decent documents of the period. It says that there is no hazard from the water, that the hazard is exposure and distance and people walking out further than they can walk back, and it recommends against any attempt at control on the grounds that a fence would concentrate people at the gaps in it. Then it recommends the only thing it recommends, which is that the water bowsers and the medical tents be moved from the apron down onto the shore edge where the people actually were. It was signed off in an afternoon and the tents went down that night.

And then the launches went up over the top of it.

Five a day, from a deck on the shore, on a bearing that took them out across the flat, so that every single ascent happened twice — once in the sky and once, exactly and simultaneously and upside down, in the ground.

Teo watched perhaps sixty of them from the causeway and never got used to it, which she reports flatly, as a fact about herself she disapproves of.

“It is just the reflection,” she said. “It is the same physics as a puddle. I could tell you the reflectivity figure and the incidence angle and I have read the instrument reports. And I stood out there sixty times and every time there were two of them going up and only one of them was real and I could not tell you which, and neither could anybody else, and the flat did that for nineteen years to every survey instrument we ever put on it and we called it interference.”

The water goes off the salar in April. It always has. It goes off slowly and it leaves the flat white and hard and cracked into the hexagons that the tourists photograph, and then the brine crews go back out and the crust is worked and the whole thing begins again in December.

In April of the thirty-fifth year the water went off, and there was nobody on the flat, and the rail stood across twenty miles of it with nothing on it, and the crust dried and cracked into hexagons exactly as it had done every year for something like forty thousand years.

“That is the part nobody puts in the poems,” Teo said. “We were here nineteen years. We built the largest thing that has ever been built and we put a hundred thousand people into the sky off the top of it. And the flood came in and went out on the same days it always does, and it did not know, and in ten thousand years the rail will be a stain and it will still be doing it.”

She said this at seventy-one, on a recording made on the shore road, four hundred metres from the house with the drawer in it, and it reached the colony nine years later down a link that was by then losing about a word in forty.

The rail crew at the end was thirty-one people and four of them were going.

Everybody knew which four. It had been known for two years, since the results went out over nine days in the thirty-second year, and there had been about a fortnight of the thing being difficult and then it had stopped being difficult in the way that things do on a crew that has worked together for fifteen years, which is that it became a fact about the roster rather than a fact about anybody.

Teo Quispe has been asked a great many times how that was managed and she has always refused the premise of the question.

“It was not managed. I did not manage it. There is nothing you can say to twenty-seven people that makes it all right and I did not try, and the one thing I did do was not make a speech, and I have been given credit for that for fifty years and it was cowardice.”

What the crew did do was refuse an accommodation.

The site had proposed, in the third week of boarding, that selected personnel be released from shift four days before their district went up. It was a humane proposal and it came from a decent office and it applied across the whole workforce. The rail crew's four came to Teo together and asked to be left on the roster to the day, and when she asked why, the answer she got was from a woman named Sonia Chura, who was thirty-six and a track inspector and who is the only one of the four who ever discussed it afterwards.

“She said: if we come off early, then for four days the crew is short and the reason the crew is short is us. And then that is what we did. That is the last thing we did here.”

All four worked to the day. Two of them worked the shift that put their own district up.

The work itself was the least glamorous work on the site and it is the reason this chapter exists.

Four hundred launches. The stack accelerates the full twenty miles along an elevated track and releases at the western end, and what that does to a rail is entirely predictable and entirely unforgiving: thermal cycling at the shore end, pier loading across the middle, and a two-inch flood underneath all of it for four months of the year. The track has to be walked after every shot.

Not inspected from a vehicle. Walked. Twenty miles out, twenty miles back, in pairs, with the crust or the water underfoot, between flights, at a rate of five flights a day.

They did it four hundred times.

They did it four hundred times and the inspection log has four hundred entries in it and every entry has two signatures and a time, and the times are the thing to look at, because the times do not drift. The first walk of the boarding period took nine hours and fifty minutes. The four hundredth took nine hours and fifty-five.

“People want me to tell them the crew was heroic,” Teo said. “They were not heroic. They were a rail crew. If you want to understand what happened on that track in those eleven weeks, the thing to understand is that walking it in nine hours and fifty minutes is the job, and that a crew which has decided that its own situation is a reason to walk it in nine hours and twenty is a crew that has stopped doing the job and started doing something about itself.”

There is one entry that is out of pattern and Teo has always made a point of showing people that one first, because it is the only funny thing in four hundred pages. On the eighty-ninth walk a pair came back and entered a defect at pier two hundred and eleven reading, in full: bird. A flamingo had come down on the track. There are flamingos on the salar in the flood season, tens of thousands of them, and they had been a standing nuisance for nineteen years and were the subject of four separate procedures, and the procedure in this instance was that two rail inspectors spent about forty minutes persuading a flamingo off an elevated launch track at two in the morning. It is entered as a defect because it was a defect. It is signed by both of them.

On the eleventh of January, on the three hundred and eighty-fourth flight, Efraín Mamani called a hold.

He was fifty-eight. He had been on the rail seventeen years. He was not going, and he had a daughter who was not going, and a brother in Oruro who was.

What he found was a pier. Number four hundred and six, which is eleven miles out, in the water at that time of year, and what he found was not a crack and not a fracture and not anything a photograph would show. He found that the pier had moved. Four millimetres, against a witness mark that he had put on it himself eleven years earlier and had checked, by his own estimate, something over a thousand times.

Four millimetres is inside tolerance. It is comfortably inside tolerance. Every engineer who has looked at the file since agrees that the pier was, on the day, safe.

He called the hold anyway, because four millimetres in eleven years is a pier that is not moving and four millimetres since October is a pier that is.

A hold in the second week of January cost the schedule eleven hours and it cost two flights and it moved four hundred and eighty people's boarding by a day, and it went up the line to the launch director inside twenty minutes, and the launch director came out to pier four hundred and six himself in a boat, which is a sentence that could only be written about that site in that season.

Mamani showed him the witness mark.

They dug the pier. It took eleven hours in two inches of water and what they found underneath it was a solution void in the crust, which is a thing the salar does and which the survey had mapped four times and had not found there, because it had not been there. It was about the size of a room. Two more months of loading, on the site's own subsequent analysis, and the pier would have gone, and if the pier had gone under a stack it would not have been a schedule problem.

“He was right,” Teo said. “That is the whole of it. He was right, and he knew he was going to be right before he called it, because he had walked that pier a thousand times and it had never moved. And the thing I want written down is that nobody on that crew was surprised, including me. Not one person said are you sure. If you had asked any of the thirty-one whether Efraín Mamani would call a hold on a launch in the last fortnight for four millimetres, every one of them would have said yes without thinking about it, and every one of them would have been right, and that is what a crew is and it is the only thing I have ever been good at building.”

Mamani worked another nine days and then the launches finished and then he went home. He died in the shore town at seventy-four. There is a plaque on the causeway that does not mention him because he asked, in writing, that it not, and the site honoured it, and the plaque is about the pier.

The four who were going went up in their own districts over the last fortnight and the crew did not do anything about it, which was also decided in advance and was also not Teo's decision. There was no send-off, there was no gathering at the gate, and each of the four finished a shift, signed off in the ordinary way, and walked down to the causeway. Sonia Chura has said that she thought at the time it was cold, and that she has spent a very long time since being grateful for it, and that the last thing anybody said to her on that site was a man asking whether she had put the tension gauge back in the box.

The last flight went up at four in the afternoon on the twenty-second of January.

The crew walked the track afterwards.

Nobody instructed it and there was no operational reason for it and everybody on that causeway knew there would never be another launch on that rail as long as the rail stood. Twenty-seven people — the four were gone by then — went out in pairs at about five in the evening and walked twenty miles out and twenty miles back with lamps, in the flood, under a sky that was also underneath them, and came in at three in the morning.

The log entry is in the hand of a man named Julio Ari and it is in the ordinary form because it is an ordinary entry, and it is the last entry in a document that ran for nineteen years and four hundred and eleven pages.

Full length walked, both directions. Piers 1–806 checked to marks. Two shims noted west of 611, replaced. No defects. Track available.

Teo signed under it, as she had signed under every page of it since she was thirty-two.

“There was nothing for it to be available for,” she said. “I know that. Julio knew it. He wrote it because that is what the last line of an inspection says, and if he had written anything else — if he had written one word about what day it was — then the last thing that crew did would have been a gesture, and it was not a gesture, it was the track being walked.”

“They were good at it. There is not going to be anybody to say so in a hundred years. They walked it, they wrote it down, and they were good at it.”

The formal closure of the manifest is a procedure of four steps and Ines Ferrer had known all four of them for nine years, because she had written them.

Step one: the register is ruled off and the closing entry made. That had been done at two in the morning on the seventy-seventh day, at the table, with a glazier named Osvaldo Reyna the last name in it.

Step two: the register is reconciled against the manifest and the discrepancy statement is drawn. Four hundred and seventeen held open and not presented; ninety-five thousand five hundred and eighty-three boarded; two hundred and six reallocations entered by reference to the works log and not otherwise described. One page. She drew it herself and it took eleven minutes and it balanced.

Step three: the Bound Ledger, the register, the discrepancy statement, the appeals register and the discretionary instrument are placed together in the archive case, the case is closed, and the closure is witnessed by two.

Step four: the case goes to deep stowage.

She had been thinking about step three for a very long time. She has been perfectly open about this and she is not embarrassed by it, and she said, at eighty, that anybody who spends thirty-four years assembling a document and does not privately rehearse the moment they put it down is lying about something.

“I did not think it would be grand,” she said. “I was fifty-six years old and I had been a clerk since I was nineteen. I did not imagine trumpets. But I had assumed — and this is the thing I was not ready for, and it is such a small stupid thing — I had assumed that I would notice it happening.”

What happened was this.

The archive case is a standard document case, aluminium, with a gasket and four latches, and it is the same case the site used for four thousand other things. She put the Bound Ledger in first because it is the largest, then the register in its four hundred and eleven district folders, then the three thin ones on top. She checked the list on the inside of the lid against the contents, once, aloud, with her finger, as the procedure requires. She closed the four latches.

The two witnesses were a clerk named Paulina Reyes, who had worked the gate for eleven weeks and who Ines had trained, and a stowage rigger whose name is in the witness line as J. Condori and who was there because the procedure says two and he was the person standing nearest.

They both signed. Ines signed.

Then Condori put the case on a trolley and took it away.

“He said something to me,” she said. “I have tried for fifty years to remember what it was and I cannot. It was not anything. It was the thing you say when you take a box off somebody. And then he went off down the covered way with it and the trolley had a wheel that pulled and I could hear it for about forty seconds and then I could not.”

“And I stood in the room.”

“I want to be accurate about the next part because it is the only part anybody ever asks me about. Nothing happened. I did not cry and I did not feel relieved and I did not feel proud and I did not feel bereaved, and I have read four accounts by people describing what I felt and they are all inventions. I stood in a portable building at twenty past three in the morning and there was a table with nothing on it and I thought: that is on a trolley now. And then I thought that I should probably eat something, and I went and ate something, and that is the whole of it, and I have never got any further with it than that in fifty years.”

She has said that the anticlimax is not the absence of feeling. It is that the procedure is designed to be undramatic and she designed it, and it worked, and a procedure that works removes the moment it is a procedure for.

“That is what they are for,” she said. “A funeral has a form so that the people can get through the day. I wrote a form so that the record could get into a box without anybody having a moment about it. And then I was the person standing there without a moment, and I had built that, and it is exactly what I would do again.”

Paulina Reyes, who was twenty-four and who had watched the whole of it from four feet away, gave an account of it decades later that is worth setting beside Ines's own, because it does not contradict it and it is not the same.

“She checked the list on the lid twice. The procedure says once. She did it once out loud with her finger the way you are supposed to, and then when Condori came in she did it again quickly with her eyes and I do not think she knew she had done it. Then she was completely fine. She signed, and she told me my signature was in the wrong box, which it was, and I did it again in the right one, and she watched him take it out and then she asked me whether I had eaten. And I said no, and she said neither had she, and we went and ate. And I was twenty-four and I thought nothing of it at all, and I have thought about it once a year ever since.”

The case went into deep stowage, frame nine, inboard, four bays from a sealed box drawing four watts that she has never described.

Four things went into the case that the procedure does not list, and she entered all four on the contents sheet in her own hand, and she has never been criticised for any of them. A note on the reallocation rule, in full, as she wrote it that first evening. The two hundred and forty accounts, which had not yet been transmitted and which she had asked the archivist for a copy of in advance. A one-page instrument governing access to material that is not in the case and is not indexed and is in four crates in her quarters. And the flimsy of the discrepancy statement, which is a duplicate and which is therefore, strictly, clutter.

“The flimsy was in case the top copy was lost,” she said. “There is no mechanism by which the top copy could be lost. It is in a sealed case in the most protected volume on the ship. I put a spare in anyway and it is the only unprofessional thing in that box.”

Now the other thing.

Ines Ferrer is on the manifest.

District two hundred and nine, entry eleven thousand and forty. The entry is in her own hand, because every clerk's entry is in that clerk's own hand and there was never any provision for it to be otherwise, and hers went in on an ordinary afternoon in the thirty-second year in the middle of about four hundred others.

She applied in the twenty-eighth year, along with everybody else on the shore, four years before she became chief records clerk. She was assessed on the four criteria by the ordinary process. Her file has been reviewed four separate times, twice in her lifetime and twice since, most recently by people who went looking specifically for something, and it is clean, and there is nothing in it, and the last review said so in terms that were not grudging.

She has never once claimed she did not deserve it.

She has also never once said that she was glad.

People have gone at this from every direction for a hundred and forty years. She was asked directly at least eleven times in recorded interviews and the answers are all variations of the same refusal, and the refusal is not evasive, it is a description of a state she was in and never left.

“I transcribed nine thousand one hundred and eleven letters written by people who were not selected. I have read every one of them twice, some of them four times. I held four hundred and seventeen berths open for people who did not come and then I flew them empty, and one of the four hundred and seventeen may have been standing in a bus station eleven miles away. And I am on it, on the criteria, correctly, in my own handwriting.”

“You are asking me whether that is all right. I do not know whether it is all right. It has been fifty years and I do not know, and I am not going to construct an answer at this stage because the honest state of it is the one I have been in the whole time, which is that both of those things are true and I have never been able to put them in an order.”

She went through her own gate at twenty to eleven on the morning after the register closed, in the ordinary way, with about two hundred other late-boarding personnel.

The clerk on the table was Paulina Reyes, who had witnessed the case eight hours earlier and who had been trained by the woman standing in front of her, and who followed the procedure exactly, because the procedure is four pages and the first provision of it is the one Ines had spent nine hundred hours of launch window defending.

The traveller states their own name. The clerk does not read the name to the traveller.

Reyes asked.

Ines said her name out loud.

Reyes found it on the district page and read it back, and confirmed the household, and the eleven kilograms went on the scale and came off the scale, and the medical tent stamped the sheet, and Ines Ferrer crossed the painted line four metres in front of the table and walked eleven hundred metres up the covered way to the apron.

Fifty-eight seconds.

Nnamdi Okonjo wrote one more advertisement and it never ran, and it is the reason anybody still reads him.

He wrote it on the night of the fourth of January in the thirty-fifth year, in the campaign office on the shore, alone, in the last fortnight of boarding, and he wrote it properly. That detail is the one he insisted on whenever it came up. It is not an essay and it is not a confession and it is not a letter. It is a proof sheet in the house format, set in the house face at the house measure, with the standard footer and the standard rule above it, ranged and spaced to the campaign's own specification, because he said that an honest text dressed as an essay is a man asking to be forgiven and an honest text dressed as an advertisement is an advertisement.

He did not submit it. He did not show it to anybody on the shore. He filed it in the audit file — the standing eleven thousand pages of interventions against public statements — and he filed it under the same heading as the two hundred and forty others, which is to say he filed it as an intervention, and the party intervened against is himself.

It was found in the fourth year of the crossing by a cataloguer and it has been in print ever since.

The proof sheet survives, which is the part that gets forgotten when people quote it as a text. It is a physical object and it went aboard in the audit file and it is a working proof with working proof marks on it. There are eleven corrections in his hand and every one of them is a compositor's correction — a bad break, a widow at the foot of the third column, two ems where there should be one. He set it, read it, marked it up for a press that was never going to print it, and put it in the file corrected. Ines Ferrer, who catalogued the audit file eventually and who was not easily impressed by anybody, said that the eleven marks tell you more about the man than the copy does.

It is reproduced here in full, in the form in which he set it.

SIX AND A HALF BILLION MILES

We chose that number because it is in miles.

The distance is one and a tenth light years. We ran both. Light years made you go quiet and miles made you lean in, and we have put miles on every hoarding for eleven years for that reason and for no other reason at all.

That is the smallest true thing we have never told you. Here are the rest of them.

Four hundred and six thousand of you applied. Ninety-six thousand are going. We have printed the first number for eleven years as proof of how badly the world wanted this. We have never once printed the second number beside it, and we have never printed the third number at all, and the third number is three hundred and ten thousand.

Eleven hundred and forty berths were not filled by the criteria. The criteria could not reach them. They were filled by one man in one hand in four columns, and there is no column for why, and that was decided deliberately and I lost the argument and the man who won it was right. Some of those names are a metallurgist and a pump operator and the whole of one night crew. Some of them are a driver, and a woman he has known since he was twenty-two. They are on the same page in the same ink, entered in the order he decided them, and there is nothing on that page that tells them apart. He can tell them apart. Nobody else ever will.

Our own actuary says that four thousand one hundred of the people going will die on the way. That figure is on page two hundred and nine of a document we printed four hundred copies of. That is publication. It is not telling you. I arranged it so that four thousand one hundred people would have to be looked for, and nobody looked for them, and I have listened to eleven people explain to me why that is not lying and none of them convinced me.

Nineteen years is the crossing. It is also how long we have been here. We built a town out of salt and a hospital that is better than anything within four hundred miles, and we are about to switch the town off. The hospital has eleven years of funding in it. Nothing else here has that long. I have said the first part in public a hundred times and I have never said the second part, and I have been made to stop and I have started again, four times.

In January the machines that built this rail stood across three miles of it and eleven thousand people stopped walking. Not one person was touched. There is no rule that permitted it and no rule that prohibited it, because there is no rule, and there was no rule the whole time, and the finding is not that somebody broke one.

Now the part that is actually the point.

Every person who made these decisions made them well.

That is what we cannot put on a hoarding, because it is the only frightening thing in it. There was no villain in any room I sat in for twelve years. Decent people wrote criteria to be defensible, and the criteria were defensible, and they stopped being able to tell human beings apart at ninety-nine per cent, and after that it was a man's handwriting. Nobody lied to you. We chose what to say out of a very large number of true things and you were left with a picture that is not true, and there is a word for that and it is not lying and I have spent twelve years failing to find it, and I do not think there is one, and I think that is why the job exists.

Here is what we are asking, and we have never once asked it out loud.

We are asking three hundred thousand people to stand on a flat white plain and watch a light leave, and then to go home, and to keep the water running for the eleven years it has left, and to raise children who will be told this story by people who were not chosen for it.

We have not said thank you in eleven years, and the reason is that thank you implies a transaction, and there was no transaction. You were not asked. There was never anything to ask you.

We would like to say it now. It is worth nothing. We are saying it.

If you remember one number from eleven years of us, do not make it six and a half billion.

Make it three hundred and ten thousand.

And then, set smaller, in the position where the campaign always put its call to action, the last four lines:

There is nothing to do about any of this.

That is not an oversight in the copy. Every advertisement ends with a thing you can do and this one does not, because there is not one, and if I invented one here I would be doing the same job I have done for eleven years with the lights off.

The ship goes on the twenty-second.

Come and watch it or do not.

Below that, the rule, and the standard footer, correctly set: the programme name, the site, and the line that had appeared on every piece of campaign material since the twenty-third year.

He was asked, for the rest of his life, why he did not run it, and he gave four reasons, and he said that three of them were true and one of them was the reason.

The three true ones: that it was not his to run, the campaign having been a consortium instrument and not a personal one; that the last fortnight of boarding was a bad fortnight to hand three hundred thousand people a document telling them how the sausage was made; and that a piece of copy which is itself very well written about the wrongness of writing copy well is a trap he could see and did not want to walk into in public.

The reason was the fourth one.

“It would not have worked. I do not mean it would have been refused — I never submitted it, so nobody refused anything. I mean that it would not have moved one person one inch. It is not persuasion. Every line in it is a man arranging his own conscience where other people can see it, and I have spent my whole working life being able to tell the difference between a text that does something and a text that makes the writer feel better, and I was not going to spend a hoarding on the second kind. I filed it, which is what you do with a thing that is true and useless.”

He was asked once whether he was proud of it.

“It is the only thing I ever wrote that I would put my name on. And the work I am judged by is the other eleven years, and that is correct, and I would not have it any other way, because the other eleven years is what got the ship off the ground and this is a page in a box.”

He went up last, on the twenty-second of January in the thirty-fifth year, on the four hundredth flight, and it has been read as courage for a hundred and seventy-one years and it was about two-thirds logistics.

The logistics are not complicated and he explained them at the time to anybody who asked. Somebody has to be the last person off a site. There are four hundred flights and one of them is last and the person on it is whoever has a reason to still be on the ground when the other three hundred and ninety-nine have gone. He had reasons: three signatures that could not be given from orbit, a handover of the site to a receiver who had to be met in person, and the fact that a launch programme running at five a day for eleven weeks needs somebody on the shore with authority to stop it, and that person cannot leave until there is nothing left to stop.

That is two-thirds of it and it is genuinely two-thirds. The other third is that he wanted to be last and would have arranged it if the logistics had not, and he admitted that once, in the ninth year of the crossing, to Halima Sow, who did not tell anybody until after he was gone from the record.

The last flight was not a ceremony. There were two hundred and forty seats and two hundred and thirty-nine of them were occupied by ordinary late-boarding personnel from three districts, which is to say by people who had been assigned to it by a scheduling clerk, and every account of the ascent that exists comes from them rather than from him, and they are the reason we know anything about it at all.

He was in a middle seat in the fourth row of the second bay, because the manifesting for the last flight had been done four months earlier by somebody who did not know who would be on it.

The three signatures were real and they are worth naming, because they are the last administrative acts of the shore period and two of them are not what people expect. One transferred the water plant, the hospital and eleven kilometres of the causeway to a municipal receiver, unencumbered, with the eleven years of hospital funding attached as a charge that could not be redirected — Nnamdi had drafted that clause four years earlier and had been made to defend it three times. One closed the project's accounts. And one was a personal instrument of about a page which released the site's plant, stock, vehicles and remaining fabric to the workforce as a body, to be disposed of as they saw fit, which is the reason the shore towns had four hundred trucks and a fabrication shed and why anything there ran at all for the next decade. He signed all three in the site office in the morning and was on the apron by two.

The ascent along the rail takes ninety-one seconds.

You do not see anything for the first sixty of them. There is nothing to see; you are inside a hull inside a shroud, accelerating along a track eleven metres off a salt flat, and the only thing available to the senses is a noise and a weight. What you get, after release, is a window — two of them per bay, small, badly placed, added late and over the objection of four engineers on mass grounds and retained because Nnamdi Okonjo had written a memo pointing out that a hundred thousand people were going to leave a planet and that the last thing you want to explain to a colony in eighty years is that nobody was allowed to look at it.

The window he was next to was on the left side, which on that bearing is the south, which is across the flat.

Here is what was under it.

Eleven thousand square kilometres of salt with two inches of water on it, at four in the afternoon, in the second month of the flood. From five kilometres it is not a landscape. It has no features and no shadows and no scale, and it does not read as ground; it reads as a hole in the planet with the sky at the bottom of it. Every account from that flight and from the three hundred and ninety-nine before it says some version of the same thing, and the phrase that recurs is that it looks like the world has been left unfinished in that spot.

And across it, from the shore to a point twenty miles out, one straight line.

The rail is four metres wide. From five kilometres it is a scratch. From twelve it is a hair. It is the only object on the salar and it is the only straight line in about four hundred kilometres in any direction, and it doubles, because everything doubles down there, so what he was actually looking at was two rails meeting at the piers, making a shape like a very long narrow eye.

Twenty miles of track with nothing on it, walked end to end nine hours before by twenty-seven people with lamps.

He had seen it before. He had flown that bearing eleven times in the previous fortnight on working trips and something over two hundred times in nineteen years, and he had looked out of a window at that flat in flood in every season it has, and this is a fact people leave out because it spoils the composition. The last ascent was not a revelation. It was a view he knew extremely well, from an angle he knew extremely well, for the last time, which is a different and smaller and more ordinary thing, and which is the thing that actually happens to people.

A woman named Consuelo Arze was in the seat across the aisle and one back, and she is the source for the part everybody quotes, and she was thirty-one and a dental technician and had no idea who he was until four years later.

“There was a man in the middle seat by the window and he had a notebook out and he was writing in it the whole way up, which I remember because writing during ascent is not comfortable. He looked out for a long time. Then he wrote something down and he did the thing you do when you are checking a sum — he went back over it with the pen without writing, twice. And then he put the notebook away and did not look out of the window again.”

The notebook is in the archive. There is one entry on that date and it is four lines, and it is the only place in the entire record where he ever wrote the thing down.

Wall, year 11: departure year 34. Slack 0.

Actual: year 35, month 1, day 22.

Over by thirteen months and nine days.

The wall was not corrected. I did not correct it. Somebody should have been told and the somebody was me.

That is the number he beat, and he did not beat it.

He had written on that wall, in marker, in the eleventh year, that the last line was a measurement and not a target and that if it ever read above zero somebody had made an error and he wanted to be told. It read zero for twenty-three years because nobody ever went and changed it, and the reason nobody ever went and changed it is that by the twenty-eighth year the wall had stopped being arithmetic and become the thing in the shed that people touched on the way past, and you do not correct that, and he knew perfectly well that you do not correct that, and he let it stand.

Thirteen months and nine days, against a twenty-three year run, is an overrun of about four and a half per cent. Every engineer who has ever looked at it has said the same thing, which is that four and a half per cent on a programme of that size is not a failure, it is an extraordinary result, and that the flood alone took six years and the schedule absorbed all six. Nnamdi said it to him directly and more than once. Teo Quispe said it to him on the causeway in the thirty-third year in language that was not respectful. He never disagreed with any of them, and he never once wrote the four and a half per cent down, and what he wrote down instead was thirteen months and nine days, which is the same fact stated in the units of the people who were on the ground when the months ran out.

The shed is still there. The wall is still there, under a roof that has held up better than anybody expected, and it still says zero.

He did not write anything about the flat, or the rail, or the three hundred thousand people who were at that moment standing in two inches of water watching him leave. There is no line in the notebook about any of it. Four lines of arithmetic and the last one is an admission of a procedural failure of his own, and then he put the pen away.

Consuelo Arze's account ends with the only other thing anybody observed about him on that flight.

“When the engines cut there is about a second where everybody makes a noise. Two hundred and forty people, all at once, some of them laughing. It happens on every flight, everyone says so. And I looked around because you do, and the man with the notebook had his eyes shut and his hands flat on his knees and he was not making a noise, and I thought: that one has been holding his breath for a long time.”

“And then a girl behind us was sick, and after that it was just a bus.”

Three hundred thousand people stood on the water and watched it go.

They had been coming out onto the flat every evening for six weeks by then, but the twenty-second was different in one respect only, which is that everybody came at once and everybody came early. By two in the afternoon there were people two miles out. By three the crowd had spread so far along the shore and so far into the flat that from the causeway you could not see either end of it, and there is no photograph that contains it, because there is no vantage point on the salar high enough to get above anything.

Nobody organised it. The safety office had moved the bowsers and the medical tents down to the shore edge four weeks earlier and that was the whole of the provision, and it was enough, and four people were treated for exposure and one child was lost for about ninety minutes and found.

Teo Quispe walked out at half past two with about twenty of her crew and they went a mile and a bit and stopped, which is nothing on that flat, and they stopped there because Julio Ari said it was far enough and nobody argued with him.

Efraín Mamani was there. His daughter Delia was there, who was twenty-six and a fitter and had failed the same criterion he had by a different margin. Julio Ari, who had written the last line in the log nine hours earlier and had slept for four of them. Twenty-odd of the twenty-seven who had walked the track that night, and most of their families, and Mamani's brother had come from Oruro to watch his own brother not leave.

They had brought chairs. Teo has always been careful to include the chairs.

“It was not a vigil,” she said. “People keep making it a vigil. We took chairs out and sat in two inches of water for an hour and a half and talked mostly about nothing, and somebody had brought a thermos and somebody else had brought the wrong kind of bread and this was commented on at length. There was a lot of laughing. If you want to know what three hundred thousand people sound like a mile apart on a mirror, it sounds like a swimming pool.”

The last flight went at four.

You see it before you hear it. That is the thing every account from the flat says first, and it is simply distance — the deck is on the shore and they were a mile out, and light is instant and sound takes five seconds, and for those five seconds the whole flat watched a stack go up the rail in complete silence.

Two of them. It always doubles. It went up along the track and its reflection went down along the reflected track and the two of them met at the release point at the western end and separated, one going up into the sky and one going down into the sky, and the water held it perfectly all the way because the water always holds everything perfectly, and then the sound came across the flat and hit three hundred thousand people at once.

“And that is when they made a noise,” Teo said. “Not at the launch. At the sound. Five seconds late, all together, the whole flat. I have never heard anything like it and I have thought about it for fifty years and the closest I can get is that it was the sound of everybody being surprised at the same time by a thing they had come specifically to see.”

You can follow a stack a long way from the altiplano. The air at three thousand six hundred metres is thin and dry and there is no light anywhere on that flat for a hundred kilometres, and in January the sun goes at about seven. They watched it for eleven minutes with the naked eye and Delia Mamani had a pair of surveyor's field glasses and got it for perhaps four minutes more, and then it was a moving point among fixed points and then it was not distinguishable from anything.

Nobody left.

“There was nothing to leave for,” she said. “That is the part that is not in any of the accounts. It went, and everybody had known it was going for nineteen years, and then we were three hundred thousand people standing in the dark in two inches of water a long way from the shore, and the stars came out, and they come out underneath you as well as above you, and there is no horizon, and you are standing in the middle of a sphere. We stayed out there until about one in the morning. Most people did. There was nowhere any of us had to be.”

The three hundred thousand is a real count and it is worth one paragraph on how it exists at all, because a crowd on a surface with no edges cannot be counted in the ordinary way. It was done from the launch telemetry aircraft on the same passes that recorded the ascent, by area sampling, twice, by two people who did not confer, and it came out at two hundred and ninety-four thousand and three hundred and eleven thousand. The site published three hundred thousand and Nnamdi Okonjo audited the rounding and cleared it, which is the last entry in the audit file that is not the advertisement.

Now the other thing, which she said once, at seventy-one, on a recording made on the shore road, and never again.

Teo Quispe was on the manifest.

She had been on it since the results went out in the thirty-second year. District one hundred and four. The letter came on the fourth day of the nine and she read it once at the kitchen table and put it in the drawer with the water accounts, and she went to work, and her crew read her face on the shore road that morning and every single one of them read it the same way and every single one of them read it wrong. They thought she had been refused. She let them think it, that day and for the two years and eleven months after it, and she never corrected one of them, and she has said that this was not a plan and that she simply did not correct it on the first day and then it was the second day.

She did not decline. There is no letter of withdrawal, no note, no entry anywhere. Ines Ferrer's procedure had a provision for withdrawal and it was used four thousand and eleven times over three years and Teo Quispe did not use it.

She did not go to the gate.

District one hundred and four boarded in the sixth week. Her berth was flagged not-presented at the end of the day and it went onto the held-open list and it stayed on the held-open list for five more weeks, and at two in the morning on the seventy-seventh day, in a portable building four hundred metres from where she was sleeping, a clerk who had never met her ruled off a page and closed it, and it flew empty with nineteen years of one person's water in it.

She was one of the four hundred and seventeen.

“I know exactly what it cost,” she said. “I stowed that ship. I know what nineteen years of one person's consumables weighs and I know what came off something else to make room for it, and I know that Ines held it open to the last minute because she is the most decent person I ever worked with and she did not know whose it was. I have never once said this was a fine thing to do. It was not a sacrifice. A sacrifice is a thing you announce. I did not announce anything, I did not go to a table, and the difference between me and the other four hundred and sixteen is that I have had fifty years to make mine sound like a decision and I am not going to.”

She was asked, on the same recording, whether she was ever going to give a reason.

“I was born on that flat. My mother's people worked the brine on it for four generations. I did my homework on it.”

“That is not a reason. That is where I am from. People leave where they are from every day of the week and most of them are right to. I do not have a reason. I have a drawer with a letter in it and a crew and a piece of ground, and every time I have tried to build those three things into an argument it has come out sounding like a poem, and it was not a poem, it was that on the day my district boarded I got up and went to work.”

Sonia Chura, who was one of the four who went and who had asked to stay on the roster to the day, found out sixty years later on the planet, from a transmitted recording, in a room with other people in it. She has described her reaction and it is not what people expect and she has refused to improve it. She said she was angry for about a week, and that the anger was not about the berth and not about the waste, and that it was about the fact that she had said goodbye to her foreman on a causeway believing they were two people in different situations, and had been the only one of the two who thought so.

Efraín Mamani found out on the flat, that night, in the dark, from Julio Ari, who had worked it out four years earlier from a roster and had said nothing to anybody.

He did not say anything for a while. Then he asked her one question, and it is the last thing anybody says in this book, and Teo has quoted it every single time she has been asked about that night.

“Does the track get walked in the morning?”

She said no.

He said right, and that was the end of it, and it was never mentioned between them again in the sixteen years they had left, and he went and sat back down on his chair in two inches of water next to his daughter.

The stars were under them and over them, the same number in both directions, and the rail ran twenty miles out into the middle of it with nothing on it and nothing coming, and three hundred thousand people sat on a flat white plain in the dark for another nine hours because there was nowhere any of them had to be.

In April the water went off.

A keel is the first member laid down and the member every other member is fastened to, and it is the only part of a ship that cannot be inspected once the ship exists, because the ship is on top of it.

Shipwrights used to say that you do not build a hull around a keel, you build it onto one, and that the difference matters when something goes wrong, because a hull can be opened and a keel cannot. If the keel is sound you may never think about it again for the whole working life of the vessel. If it is not, you will find out, and you will find out late, and there will be nothing to be done about it that does not involve taking the entire ship apart.

Nobody aboard called anything a keel. The Ark-Genesis has no keel and the word appears nowhere in nineteen years of engineering documentation, and Calder never used it in his life. It is the historians who use it, and they use it about the fourth day.

On the fourth day the ship was under way, the hull was complete, the Vault was sealed at frame nine, and ninety-five thousand five hundred and eighty-three people were aboard and awake.

That last part surprises people, because the crossing is remembered through the sleep tiers, and the sleep tiers were not loaded until the second year. The staged programme had always been a second-year programme; the first eighteen months were designed to be lived through by everybody, deliberately, so that a population arriving into a hull would learn the hull before most of them were put down in it. It is a good design and Halima Sow wrote it.

So on the fourth day there were ninety-five thousand people awake in a moving object, and the question of what governed them had not yet been asked by anybody, and it was answered for the first time at about eleven in the morning by a works supervisor named Bruno Adeyemi who did not know he was asking it.

Two districts wanted the same compartment.

That is the whole of it. Compartment nine-forty in the second ring had been fitted out as general workshop space and both district four hundred and six and district four hundred and eleven had scheduled it for the same eleven-week block, and the fitting-out schedule had them both down for it, and it was a clerical error four years old made on a different planet.

Adeyemi came to Calder because Calder was the person you came to, and put it to him in about ninety seconds, and Calder split it — four hundred and six for the first six weeks, four hundred and eleven for the second five and a fortnight of the following block by way of the difference — and Adeyemi wrote it down and went away and it was implemented that afternoon and nobody has ever disputed it.

It was a perfectly proper decision. It was made under a real instrument: compartment allocation is a fitting-out matter, fitting-out is construction, and the construction hierarchy on that ship was nineteen years old, thoroughly documented, universally understood and entirely legitimate. Adeyemi had gone to the right person and got the right answer through the right channel, and the channel had a paper trail, and there is a works order.

Calder walked back down the second ring afterwards and did the arithmetic, and this is the thing the historians mean.

He inventoried it as he walked, and it does not take long, because it is short.

The works order system, which governs the building of things. The two-signature rule above four hundred units, which governs a swarm sized to a task that no longer exists. The stowage and movement order system, which governs where objects are. The inspection regime. The consumables schedule. The reallocation rule, which is in the works log because he put it there himself, and which governs a berth. The Bound Ledger, which is a record and not an instrument and has never claimed to be one. Nnamdi's founding instrument, which governs the manifest, and the manifest is closed and in a sealed case four bays from a box he has never described.

Every one of them is a building instrument. Every single one. They are all excellent, they are all real, they were all written by careful people, and there is not one of them that says anything whatever about ninety-five thousand people other than as a load.

The consortium is on Earth. The manifest board dissolved itself in the thirty-fourth year, correctly, on schedule, because its work was complete. The Nine are a works committee and have never been anything else and their minutes say so on every page.

And the construction hierarchy is legitimate right up to the moment the building stops, and the building is going to stop, because a ship that has been built is a ship that has been built, and what is left after that is not a site. It is a population in a container for nineteen years.

What made him do the arithmetic was not the dispute. It was that the dispute had worked.

It had worked completely. A man had brought a real problem to the right authority and received a fair decision through a proper channel and everybody had been satisfied, and the only reason any of that had been true was that the problem had happened to be about a compartment. And sooner or later — and he put it at under a year, and he was out by four months in the wrong direction — Adeyemi or somebody like him was going to come down that ring with something that was not about a compartment, and there was going to be no instrument, and he was going to answer anyway, and it was going to be accepted, and that was going to be the founding.

Bruno Adeyemi lived to eighty-eight and was interviewed four times and never once understood why anybody wanted to talk to him about compartment nine-forty. He remembered the dispute clearly, because it had been his first week and he had been nervous about taking a scheduling error to the Founder over a workshop. He described the decision as fair and sensible and said it took about two minutes. Asked, in his last interview, whether Calder had seemed troubled, he said no; he said the man had listened, split it, told him to write it down, and asked after the fitting-out crew's water ration, which had been short for two days and which was fixed by the evening.

Nnamdi Okonjo was four decks up and about ninety metres aft.

He would have understood it in under a minute. He had spent his entire working life on precisely this and had said so out loud in a hot room eleven weeks earlier, in a portable building with a bad door, with a dead pipefitter's berth on the table between them; he had written the closure clause tight on purpose and taken the absurd result on purpose; he had put a two-signature rule into a swarm instrument because the machines were not the problem and the absence of an instrument was. Give him this and there would have been a draft inside a fortnight, and it would have been four pages, and it would have been airtight, and he would have been the one to say out loud that the Founder should not be the person who says what the answer is.

Calder went up.

He got as far as the cross-corridor outside Nnamdi's compartment, which is documented, because there is a passage log and he is on it at eleven forty-one and he is on it again, going the other way, at eleven forty-six.

Five minutes.

Nobody knows what happened in those five minutes and nobody ever will. The compartment was occupied; Nnamdi was in it; there is no record of a conversation and Nnamdi, who wrote down everything for another seventy years, never recorded one. Whatever it was, it happened in a corridor with nobody else in it, and it is the only five minutes in this entire book that has no witness, no document and no testimony from anybody.

He did not knock.

He went back down to the second ring and finished the walk he had been doing when Adeyemi found him, which was an ordinary inspection walk, and he did the rest of it properly, and there are four sign-offs from that afternoon in his hand.

This is where the historians put the keel, and it is the only reason the word is attached to this book at all.

Not the hull, which was fastened together on a salt flat over nineteen years by four hundred thousand people and is a matter of public record down to the last weld. The thing underneath it. Laid on the fourth day, in a cross-corridor, by one man in five minutes with no witness, and never afterwards inspectable, because everything that came later was built onto it and you cannot get at a keel without taking the ship apart.

Then he took the notebook out.

It is the same notebook. It is in the archive. The entry before it is four days old — the four lines of arithmetic written in a middle seat during a ninety-one second ascent, ending somebody should have been told and the somebody was me.

He did not write anything.

He put it away and went back to work.

The next entry in that notebook is dated eleven months later and it is about a water plant.