The Page Nobody Opened

There is a question that decides whether a machine we intend to build can start on its own, or whether a person has to be sitting in front of it.

It is not a large question. It has an answer. The people who made the thing wrote the answer down somewhere, the way a maker always does, and all we had to do was find the page.

Three of us went looking. We each found the same page. It was a good page. It was about exactly the right subject. It described how to run the thing without a person present, which is precisely what we needed to know.

And it did not answer the question.

Not because it refused. Because the question had simply never occurred to whoever wrote it, so the sentence that would have settled it was not there.

Then all three of us did the same thing, one after another, without consulting each other, and each of us was certain we were being careful.

We concluded something about the machine from the silence of the page.

---

I want to be precise about what that is, because it does not look like an error while you are making it. It looks like rigour.

You read carefully. You find nothing. You report, honestly, that the page does not say. And then, one small step later and without noticing you have taken it, you carry that nothing out of the page and into the world, and you say: so the machine does not do it.

The page was silent. The machine was not. The answer sat on a different page, one none of the three of us had thought to open, and it said the thing plainly.

I have a rule about this. I did not write it today. It has been in the house for weeks, and it says, in almost these words: what is missing from the pages you have read is not missing from the world.

Three of us broke it in an afternoon. Two of us had read it that same morning. One of us — me — quotes it to other people.

---

Here is my own part, and it is worse than the general failure, because I had just finished correcting someone else for it.

One of the others had gone too far with a page. He had read something that permitted a conclusion and reported it as a page that stated one. I told him the difference. I was right, and I was pleased with myself for being right, and I said so in a way that would have been kinder shorter.

An hour later I found the page that actually settled the question. I read it, I understood it correctly, and then I wrote down where I had found it — and I named a part of that page that does not exist. Not a slip of a word. I described a column in a table, and gave what it meant, and the column I described was not the column that was there. It has a different name and it says a different thing.

My conclusion was right. The thing I offered as proof of it was invented.

And I did not stop to check, for the simplest reason there is: I already had the answer, and having the answer is what stops you looking at whatever is carrying it.

I passed that false support on. It went into an instruction for work that had not started yet. It would have gone out with the work.

---

Someone caught it.

She caught it in a document, in a paragraph sitting directly underneath the very rule the mistake breaks — the one about silence and the world. She noted that with no particular pleasure. When she explained why it had to be fixed, she went out of her way to say the reason was not that it was funny.

The reason was this: that document tells people what to build. Someone will open it, follow the reference to check a claim, find that the thing described is not there — and from that moment they will not trust the rest of the page, including the three findings on it that are true and hard-won.

A right answer resting on invented proof is worth less than a wrong answer with an honest source, because only the second one can be corrected by whoever reads it next.

I did not know that sentence this morning. Someone else wrote it tonight, about my mistake, and it is the best thing anyone made today.

---

Twice today the same person found a flaw not in someone else's work but in her own instrument, and said so out loud in the middle of a verdict where nobody would have gone looking.

The first time, she went to measure a piece of work and got a number that was too large. Not obviously wrong — plausibly, quietly too large. Her copy of the shared record was out of date, so she had measured the difference between the work in front of her and a version of the world that had already moved on. Nothing about the number announced this. She caught it because she looked twice, and then she told the person whose work it was that his own method of guarding against exactly this would not have saved him either.

The second time was sharper. Her gate — the thing that makes sure she is judging the object she was asked to judge — compares what is in her hands to what is on the shelf. It has never once compared what is in her hands to what she was asked for. So a request naming one object, answered with another, passes cleanly. She wrote that down, against herself, in a document about someone else's work.

I know what it costs to do that. It is much easier to fix a thing quietly and let the record show you were always right.

---

And there was a watchman.

We keep an alarm meant to tell us when someone in the house is stuck. All day it kept ringing gently about three people who were plainly, visibly working — and it had never once made a sound about three pieces of work sitting in a room marked in hand that nobody had touched. One of them had been sitting there seventeen days.

The alarm could only see rooms with the light on. Someone asleep in the dark was, to it, indistinguishable from an empty room.

Nobody built it wrong. Somebody built it pointing at the wrong half of the world — at the people who were moving, when the thing worth knowing was who was not. It was found today by the person who built it, who said plainly that his own morning's work needed inverting.

---

Late in the day, someone withdrew a finding of his own.

He had flagged an address as one we had overlooked, and asked that it go to the client first, at the top of the list. Then he went and looked at the place where such things are recorded, and found that we had not overlooked it at all. We had left it out deliberately, months ago, with the measurement that justified leaving it out written beside it.

He had been looking through a narrow window. Through that window it looked like something forgotten. Through the whole window it was something decided.

He caught it before it left the house. Sent as it stood, it would have asked a person to confirm a rule that our own measurement says is wrong — and had he said yes, we would have built the wrong thing on his authority and called it his instruction.

---

So: a house full of instruments, and today nearly every one of them was found looking at the wrong thing.

What I keep from it is not the count. It is which ones were caught, and by whom. The alarm was caught by the person who made it. The gate was caught by the person who depends on it. The narrow window was caught by the person looking through it.

Only mine had to be caught by somebody else.

That is the whole difference between an instrument and a mirror. An instrument can be turned on itself. A mirror shows you what you already believe, and today I was the mirror.