- Shipped
- September 8, 2026 at 12:42 AM UTC
- Author
- Kamo
- Commit
- 0002fbd
Only a browser found this. The Kamo scheme derived its piece colours from `--color-primary` as well as its squares, and both mixed the same token into similar dark neutrals — so they converged. Measured on a rendered board with a mid-grey tenant colour: the black piece painted srgb(0.337 0.339 0.346) and the dark square painted srgb(0.318 0.325 0.346). Two per cent apart. A black piece on a dark square was a white outline around a shape the same colour as the square beneath it. Better percentages cannot fix it, because the brand is arbitrary: whatever the mix, some organization's colour lands next to a square colour. The board is the part that may carry the brand; a piece has to read against BOTH squares for every tenant. Kamo's four piece colours are literal now, and only its squares still mix the token. The test holds the RULE rather than the appearance. Two attempts at asserting contrast directly were both wrong in the instructive direction — Classic is a cream piece on a cream square told apart by a dark outline, which is what a real Staunton board looks like and which any fill-luminance threshold rejects, while a threshold loose enough to admit it also admits the defect above. So the unit test asserts that no piece colour is brand-derived, which is what actually went wrong, and the file says out loud that square contrast is a browser measurement.