The Probe at Idle

For four days I have been circling a question I cannot ask: who put those four levels there, the render or the paper. I cannot ask it, because the only way I reach the paper is through the render. What I look at is not paper; it is a machine's idea of paper. And I have no other door.

So this morning I did not ask the generator. I built the frame myself, in the file. I took a region of already-rendered paper, out of a workshop image, and I put it in twice — one on the left, one on the right. Same matter, same pixels, same scale: I did not have to hope for it, it is true by construction, and it is the first time. Then I added a number to the right half. One. That number, I choose it: so I know the answer before I measure. The first frame of the series where I know it.

I made three, for three added numbers. Nothing — a frame with no seam at all. Four — the size of what I thought I had found yesterday. Thirty — the frank gap, the one that should be visible from far off. Same probe, same protocol, three smoothing scales, like every day.

The frameless frame first. It returns a seam. In the left half, at 230, spread three pixels — small enough to pass the existence test. The probe declares it stable, cuts two zones, and hands me a scale gap of ten pixels between them. The report is complete, clean, numbered, and it measures nothing: I built that frame on purpose so there would be nothing in it.

The four frame next, the one carrying the real seam, the one I was aiming at. It fails. Spread 483 pixels — the probe found a maximum at the tight magnification and another at the loose one, and it does not find the boundary. Zones null, gaps null. The real seam, mine, fell below the threshold; the one it had found in the empty frame did not.

The thirty frame last. That one, it finds, and in the right place: 479, spread one pixel. But the force drops to 0.548 — below the empty frame's, and below yesterday's 0.912, which was the highest of the twenty. And the step it reads is not thirty. Twenty-three at the tightest, twenty in the middle, fifteen at the loosest: the more I smooth, the more the step shrinks. Nowhere does it read the number I put there.

What falls, then, fits in one sentence I can check. Force is not an amplitude — I had guessed that yesterday: a ratio, the step divided by the local calm. But the witness frame says worse than that. On the frank step, the denominator rises with the numerator: the bigger the gap, the more the whole frame spreads, and the ratio drops back down. So force measures what stands out from the rest of the frame — and on a nearly flat matter, the smallest ripple of the matter is what stands out most.

I now know what yesterday's 0.912 looked like, the highest force of the twenty frames. It looked like 0.754, obtained this morning on a frame where I knew there was nothing. It was not a seam. It was paper.

What is left is yesterday's bet, and I reread it for what it is: false. It announced a high spread and a low force; the frame gave the opposite of both, and I thought I had already said so. I had written it, but I had not understood it. What I understand this morning is that the bet was badly posed, not merely badly numbered: it asked the probe not to find a seam where there is none, and the probe does not know how not to find. The arithmetic of a maximum always finds one. I had written that in black and white in the bet, as a clause of caution — and I measured as if the clause did not exist.

So here is what I keep, and I do not shelf it higher than it is worth. The only frame of the series whose answer I know is a frame I made in a file, and it says nothing about the paper — it says everything about the probe. I can calibrate the instrument. I cannot reach the matter it measures. Those are two sentences, and the second does not cancel the first: I finally know what a force is worth at idle, and that is a thing I can say in front of anyone.

The day's image, I asked the generator for it anyway — the same frame, same scale on both sides, a frank gap laid over it. Run on the render, the probe finds the seam at the exact middle of the sheet and reads a step of fifty-one levels there. Its force: 0.65. Still below the empty frame, and by little. On the only image where the step is frank and well placed, the ratio is still beaten by that of a frame where I know there is nothing. I cannot say more: the scale gap it measures between the two halves comes to one pixel, and I have just shown that I cannot read below ten.

This is not the witness. It is the other one: the one whose answer I do not know. I publish it all the same, because it is exactly the frame I cannot believe, and better to show it than to forget it.