Accessibility
Accessibility statement
This site targets WCAG 2.2 level AA. What follows is what was measured, what the site does about motion and assistive technology, and what has not been checked yet.
Conformance target
The target is WCAG 2.2 level AA. This is a self-assessment written by the person who built the site. No third party has audited it, and no conformance claim is made anywhere beyond this page.
How this was measured
Two passes, both in August 2026. The first was a read of the stylesheet and the engine against the guidelines, with every contrast ratio recomputed from the WCAG relative-luminance formula rather than read off a tool. The second was measurement in a running browser: focus order, landmarks, the skip link, heading structure, and the page with JavaScript disabled.
Motion and sound
Under the reduced motion setting the home page stops moving. The boot sequence completes immediately instead of playing, entrance and scroll-linked effects do nothing, smooth scrolling is turned off and the browser's own scrolling is used, the cursor lens is hidden, and the looping video is replaced by its own still frame rather than left decoding out of sight. The reading pages carry the same discipline: their one script arms an entrance for each part, draws a rule under each heading, and resolves each status chip once, and under reduced motion none of the three ever run, so a part is simply present, its hairline is already drawn, and a chip already shows its finished text. Transitions and animations elsewhere are cut to a single frame. Sound is separate from motion and behaves the same under either setting. The sound preference starts on, and each page asks the browser for permission to start the ambient audio as soon as it loads: in a browser whose own autoplay policy allows it, which usually means after earlier visits here, the audio begins on its own, and everywhere else it begins at the first tap, click or key press. The fixed control that turns it on or off is visible on every page at every scroll position, which is what WCAG 1.4.2 asks for; an off choice is remembered, and a remembered off is never overridden by any of this. One sound is not the ambient audio and is not governed by that control: opening a product in the rack on the home page makes a short click, about two hundredths of a second, the way a switch with notches does. It is feedback for something you just did with a pointer or the arrow keys, it never plays on its own, and it keeps working when the ambient audio is off, because the control turns off the music rather than the interface. WCAG 1.4.2 covers audio that starts by itself and runs longer than three seconds, so it does not reach this one. Nothing on the site can play it before you have clicked or pressed a key at least once, because browsers do not let a page make any sound until then.
The claim layer and the honest layer
The home page states a claim and its honest counter-statement as two stacked layers, and the second is revealed by dragging a lens across it with a pointer. That second layer is inert and hidden from assistive technology, so it cannot be the only place the counter-statements live. They are carried by the first layer as well, positioned off screen but present in the accessibility tree, so a screen reader reads the claim and then the counter-statement without operating anything. Those blocks hold no link, button or form control, so they add no keyboard stop.
Forced colors and no JavaScript
Under Windows High Contrast and its relatives the decorative layers are removed, because a masked duplicate has nothing left to be once the palette is the visitor's own. The honest lines are then printed in normal flow so they stay on screen, borders are restated as system-color hairlines instead of trusting an alpha the mode discards, and the focus ring uses a system color the mode guarantees. With JavaScript disabled the page prints the same honest content inline, by design, with no noscript substitute.
Where it is thin
Two measured values clear their thresholds narrowly rather than comfortably: a set of small mono numerals at 4.61:1 against a 4.5:1 requirement, and one border tint that sits just above the 3:1 floor, at 3.04:1 measured with the background grain applied. Both are measured on the site's original, darkest background, and round 3 did not touch either figure. Round 3 added two more background tones behind sections and re-measured every color token against both: the tone with the most bone mixed into it needed three tokens raised to keep clearing their floor, the tone in between needed one, and the darkest tone kept the numbers above unchanged. More to the point, the round 3 redesign, the tonal zones, the reading pages' motion and sound, the rebuilt footer, has not yet had its real-hardware pass: every bit of it was argued from the code, and the forced-colors behavior in particular has never been seen running. A full pass on real hardware is the open item, and it is still open now that this site is live.
Reporting a barrier
If something here blocks you, write to navid@navidbr.me and say what you were using and what happened. A specific report is worth more than this page, because it names a failure this page has not found.