Self-audit
We pointed it at ourselves.
Every site in the showcase gets taken apart and published. It would be a strange kind of confidence to exempt the one doing the taking apart, so this page runs the same measurement on csscreme.com and publishes whatever comes back.
Two is a pairing. Three or more usually arrived by accident.
Most systems ship two to four. More than that and corners are being guessed per component.
Custom properties we actually declare. Higher is better here: it is the opposite of a sprawl signal.
What our own CSS declares
41 KB of CSS read across 3 stylesheets. Confidence: tokens, meaning named custom properties were found and the reading is precise rather than inferred from colour frequency.
Fonts in use
InterJetBrains MonoPixelify SansSoraSora
Radii in use
10px11px12px16px14px9px
Most-painted colours
#0a0a0f#4f46e5#f05a28#8a93a6#e5e7eb#5b6472
Our own DESIGN.md
The house palette is published on exactly the same terms as every other theme: a DESIGN.md with contrast computed from the tokens, and a shadcn install command.
#FBFBFDsurface#FFFFFFprimary#4F46E5accent#FF7A3Dtext#0E1116muted#5B6472 Why publish this
A measurement tool that has never been turned on its owner is a claim, not a tool. Publishing our own numbers costs us the option of quietly failing our own standard, which is the point: it is cheap to hold other people to a bar and expensive to be held to it.
There is no score on this page and there is none in the decoder either. A single number out of a hundred is unfalsifiable and tells you nothing about what to fix. Counts next to stated bands can be argued with, and a reader who disagrees with a band still has the count.
Point the decoder at your own site and you get this same read, plus the type scale, spacing rhythm, shadows and an accessibility pass.