吹之不百年，当正也

Eighteenth of September, evening, halfway through drafting the 2033 exchange, and I stopped over a small thing. I was writing the line for house three's string, and remembered dividing it in two on 11 March. The half on the doorframe has hung six months. The half in the cupboard has been sewn shut and untouched. They came off the same string on the same day. They are no longer the same. One has been in the wind; the other has not. So I went looking for a figure: how much harder does a reference string get in a year on a doorframe? I do not have it. Four years, two hundred-odd findings, and that one is not among them — because nobody has ever measured the reference itself. Everybody measures other things with it.

Nineteenth: asked Alim to weigh both halves, and to re-sew the cupboard half immediately. Doorframe: 37.2 g. Cupboard: 39.1 g. And from him: when I divided them I weighed them, and the two halves were 0.3 g apart. Six months, 1.6 g of difference, on 39 g — about four per cent. Then Alim, unprompted: so a year is eight per cent. And: my father hung that string for thirty years — what was it by the end?

Yu Hao came on the twentieth. Forties, short working clothes, a cord round his waist with things hanging from it: a small rule, a plumb line, a thin worn slip of wood. Callus all over the hands, nails cut very short. He did not look at the papers. He looked up at my ceiling for about ten seconds and said: the north side of this room is two parts low.

I asked what he had built. The pagoda at the Kaibao monastery. Eleven storeys. Three hundred and sixty feet — the tallest thing in the city at the time. And: the day it was finished people stood round the bottom of it. Saying what? Saying it was crooked. Flatly. Leaning northwest. You could see it from below. Was it really crooked? It really was. I built it crooked on purpose. And the reason: the capital sits on flat ground with no hills, and the wind blows northwest all year. The wind comes from one direction. And: a hundred years of wind and the pagoda will be upright. I said that meant it would be crooked for a hundred years, and that everyone who looked at it in that time would think he had botched it. Yes. To both. And, and I picked up the pen: I will not be here for those hundred years.

So how did he dare? It isn't daring. He unhooked the plumb line and put it on the table. I measured. And: I measured the capital's wind for more than a year — which months, how hard, from where. And: and I went and looked at every tall thing in the city, one by one, to see which way it had gone over. More than ten of them. Every one leaning southeast. And, six times into the book: I did not guess which way the wind blows. I looked at more than ten things the wind had already bent.

Then the limit, which he volunteered. And if the wind has no settled direction? — Then you cannot build it leaning. Flatly. If the wind is disorderly, whichever way you lean is wrong. In a place like that I build it upright and then I brace it. And: leaning has exactly one precondition. And: you must first know which way it goes and how fast. And: missing either of those, build it straight and be honest about it.

I pushed the sheet over and told him about the two halves. He read the two weights. You have both of those now. Which two? Which way: drier. How fast: four per cent in half a year. And: so you may build it leaning. Next year, pick a reference one grade softer than what you want. Hang it a year and it walks to the grade you wanted on its own. Putting the plumb line back on the cord: but remember where those two numbers came from. You did not calculate them. And: you kept a second half in March and sewed it shut. And at the door, turning: the half you don't touch is worth more than the half you use. — M.

Memo to file. 18–25 September 2032. The reference drifts, and now I can measure it.

1. The gap I had never noticed in four years. Every household's reference string is a physical object hanging in moving air. It dries. Nobody has ever measured that, because a reference is the thing you measure with — it does not occur to anyone to measure it.

2. The numbers, and they exist only because of an unrelated decision in March. Split half onto the doorframe, half sewn into a bag as a backup in case he could no longer read the hanging half. That backup is now the only unmoving control I possess. Doorframe 37.2 g / cupboard 39.1 g, from a 0.3 g starting difference. ≈4% in six months, ≈8% a year. A control that was created for a completely different reason turned out to be the only instrument capable of measuring the instrument.

3. The design move, and the precedent for it. A pagoda built deliberately leaning into the prevailing wind, on the calculation that a century of wind would bring it upright. Wrong-looking for a hundred years, and the builder said plainly: I will not be here for those hundred years. Applied: pick next year's reference one grade softer than the target; a year of drying carries it to the target. Correction built into the design instead of applied afterwards.

4. And the precondition, which he stated harder than the trick itself, and which is the part I would have skipped. Leaning is permissible only if you know which way and how fast. Missing either — build it straight and brace it. Disorderly wind means every lean is wrong. And the source of his confidence was not theory: he looked at more than ten tall things the wind had already bent, and all of them leaned the same way. — He did not guess the wind. He read it off things that had already been deformed.

5. The rule issued 22 September. Every household, every closing, keeps three: one on the doorframe (working, dries); one sewn shut in the cupboard, not opened, not looked at; one posted to another household. The cupboard copy opens on exactly two occasions: (a) the doorframe copy is lost or spoiled; (b) at each closing, both are weighed together and the difference recorded — then re-sewn immediately. And: looking at it counts as touching it.

6. Chang's question, and I could not answer it. He divided his father's string three ways with the old man watching, who said nothing until: you say the cupboard one can't be looked at. Then what's it for? Chang Bin explained. A nod, and then: so in thirty-one years I never kept one. Then Chang Bin, on the phone: thirty-one years, replaced annually, each year's picked by comparison with the last. If it drifted a little drier each time — How big is that number? — I cannot compute it. I do not have his month of picking, his hanging duration, or his indoor humidity for any of those years. Nobody recorded them. So: whether those thirty-one years drifted, and by how much, is not recoverable. It is not a hard problem. It is a missing-data problem, and the data was never created.

7. What I told him, and I stand by the shape of it. (a) Your family may have been walking in one direction for thirty-one years. (b) Whether, and how far, cannot be known today. (c) From this year your family produces one new number annually — zero this year, one year's difference next year, and in thirty-one years somebody can read thirty-one of them backwards. He worked out that he will be sixty-four then — about his father's age now — and said: then I'd better sew this year's one tighter.

8. And from Kucha on the 25th, which reorganizes the whole thing. Ali went to Imin's twice; refused the first time, shown the notebook the second. Fifty-one cells, one small bag per cell, not one of them for sale. Nine cells blank in the middle. Why blank? — those nine years something happened to his houses and the goods weren't representative. And: he would rather leave them empty than fill them with another year. A man who cannot read worked out 阙 forty years ago on his own. And the last line, which Ali added himself: Grandpa Imin asked me whether my grandfather ever kept one. I said no. He said he knew. And: he said he'd been urging him for thirty years — and my grandfather said yes every time and never did it.

"The capital sits on flat ground with no hills, and there is a great deal of northwest wind. Within a hundred years the wind will blow it upright."
The pagoda was eleven storeys and the tallest thing in the city, and when the scaffolding came down it was visibly out of true. The crowd said what crowds say. The builder's answer is recorded in one sentence, and it is usually quoted as a piece of charming foresight.
What the anecdote does not dwell on, and what is actually remarkable, is the position he had put himself in. The correction would take a century. For that entire century the building would look like a mistake, and everybody able to judge his work by looking at it would judge it wrong. He was not going to be there for the verdict.
The second thing the anecdote does not dwell on is where his confidence came from, and it was not from theory about wind. He had measured the capital's wind for more than a year — months, strength, direction — and then he had gone round the city and examined every tall structure already standing, to see which way each of them had gone over. More than ten of them, all leaning the same way.
He did not guess the wind. He read it off things the wind had already bent.
Which sets up the limit, and the limit is the useful half of the lesson. Building with the correction pre-loaded requires two quantities: which way it drifts, and how fast. Where the wind is disorderly, every lean is wrong, and the honest construction is upright and braced.
That is a harder discipline than it sounds, because a designer who has correctly diagnosed a drift once tends to start pre-loading corrections for drifts he has not measured. The tower stands because the wind really did come from the northwest. It would have fallen just as surely if he had been clever about a wind that came from anywhere.