zooqueen 3879df2a34 our points wear the house H, and three marks you could not see now read ...

TWO things on the accuracy-at-cost chart.

OURS IS THE H, NOT THE RING. The ensō is the mark of the enso PRODUCT; the H is
the mark of the COMPANY that makes it. Every other point on this chart is a
lab's company mark — OpenAI's, Anthropic's, Google's — so a product mark in that
row answers a different question than the row is asking and reads as a fourth
vendor nobody recognises. The H is also the one mark a visitor is guaranteed to
have just seen, at the top of this very page. Same `MARK_PATHS` the header
draws, imported from @hanzo/logo rather than copied.

AND THREE LAB MARKS WERE BELOW THE FLOOR ON A LIGHT DISC. Anthropic's #D4A27F
measured 2.08:1 against the neutral-100 disc — reported as "I can't quite see
it", and the measurement agrees. It was not alone: NVIDIA 2.21, Xiaomi 2.65.
This chart deliberately puts every mark on a LIGHT disc so a near-black logo
reads, which is exactly the condition a pale brand colour fails.

Darkened by RULE, not by taste: same hue, same saturation, the least light the
colour can lose and still clear 3:1 — the WCAG 1.4.11 non-text floor, and the
same number pay's accent() already uses to decide whether a colour can carry a
control. Anthropic 2.08 -> 3.00, NVIDIA 2.21 -> 3.03, Xiaomi 2.65 -> 3.00. Ten
other labs already cleared it and are untouched.

Rendered and read before and after. 871 pages, 52/52 gates.

2026-08-14 15:07:46 -07:00

1.2 KiB

Raw Permalink Blame History