Accessibility

Written by a person. Last read by a person on 2026-09-06, 2 days ago. Its facts were checked by the eval suite on 2026-09-07.

You need to know whether this site meets WCAG 2.1 AA before recommending it, you use a screen reader or keyboard and want to know what to expect, or you want to see which claims here are checked by a build rather than asserted by a person.

This site targets WCAG 2.1 Level AA. Most of that claim is checked on every commit rather than reviewed once, and this page says which parts are which.

What a build checks

Every page is complete with no JavaScript at all. The prose, the prerequisites and the next steps are all in the document as it leaves the build, and a check fails the build when any page stops being complete that way. Rubric.

Contrast. Every text and background pair is audited in both the light and dark grounds. The worst pair is 5.54 to 1, against the 4.5 that AA asks for at body size.

Keyboard operation. Every control the site adds is a native link or button, or carries a tab stop and a key handler. Each announces its role and its state. Tests assert this by operating the controls rather than by reading the markup.

Heading structure. Levels run in order, and every heading carries a stable id, so a reader can cite a section and return to it. A check reports when an id moves.

Text alternatives. Every diagram carries one, and the check fails a diagram without it.

Motion. Nothing on this site moves except scrolling, and the transitions that do exist are turned off entirely for a reader whose system asks for reduced motion.

What a person checks

Whether a stripped page reads as a document. A page can pass every check above and still be a list of fragments. An adversarial review runs against each declared reader, including one who uses a screen reader and cannot hover.

Whether the language works. No counter can tell a sentence that helps from one that patronizes. Words that call a task effortless are banned and checked; whether the replacement is clearer is a judgment.

What is not claimed

AAA. Some of it is met incidentally. None of it is targeted, and claiming a level you have not tested against is worse than claiming a lower one.

Testing with assistive technology in use. The keyboard and semantic work is verified by automated tests and by reading. It has not been through a session with a screen-reader user, which is a different kind of evidence and the kind that finds what the others miss.

The interactive layer that does not exist yet. The reading helper arrives in a later change. Announcing generated text to assistive technology as it arrives is specified and unbuilt.

Reporting something

If something here is unusable, that is a defect and worth an issue. Reports that name what you were doing and what your software did are the useful kind, and reports that only say a page was hard to use are still worth filing.