docs(status): the audit's outcome, and the four claims nobody has looked at - #82
Merged
Merged
Conversation
…ked at The section still described the audit as "written, nothing implemented" and claimed no application code had changed. That was true of PR #75 and has not been true since: seven PRs landed on 2026-09-27, four ADRs were accepted, three of five defects were fixed, and the suite went from 416 tests to 450. **The addition that matters is a list of four things believed on argument and a passing test, which nobody has looked at.** They were each recorded honestly in their own commit and ADR, and scattered across five documents that is not the same as being visible — the whole set reads as reassurance because the test count keeps rising, while the list of unconfirmed claims quietly grows beside it. - The landing page is not blank for a reduced-motion visitor. **Outstanding since 2026-09-25.** - `color-scheme` fixed dark-mode scrollbars and autofill. - Logical properties flip under `dir="rtl"`. - A learner hitting an error boundary gets a usable page. Three of the four are confirmations of fixes that shipped on specification rather than observation, and each row says why no agent can close it — a frozen `requestAnimationFrame`, a scrollbar whose colours are not exposed to `getComputedStyle`, a browser that does not re-resolve logical properties on a direction change, and a render error that can only be induced by breaking code. All four are minutes of work for a person with a browser. Putting them in one table is the point: it is the largest risk this workstream carries and it was previously only visible to someone who read five files. Also records the three tooling artefacts of the session — each of which first read as a bug in the code — because the lesson generalises past this work: isolate before concluding. In the `dir` case, not doing so would have cost a revert of a correct change. The defect list is now keyed by state rather than count: three fixed, one open (D3's pagination), one blocked on backend modelling. **Only D3 is open work**, and the 2.4.11 remedy cannot be scoped until A2 reproduces the fault. Docs only. 450 frontend tests unchanged; every ADR id and doc path resolves. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The section still said the audit was "written, nothing implemented" and that no application code had changed. True of #75; not true since — seven PRs landed on 2026-09-27, four ADRs accepted, three of five defects fixed, suite from 416 tests to 450.
Docs only, one file.
The addition that matters
Four claims are believed on argument and a passing test, and nobody has looked at any of them.
requestAnimationFrame— the pane cannot tell a working page from a broken one. Outstanding since 2026-09-25color-schemefixed dark-mode scrollbars and autofillgetComputedStyle; autofill needs a real interactiondir="rtl"Each was recorded honestly in its own commit and ADR. Scattered across five documents, that is not the same as being visible — the set reads as reassurance because the test count keeps rising, while the list of unconfirmed claims grows quietly beside it.
Three of the four are confirmations of fixes that shipped on specification rather than observation. All four are minutes of work for a person with a browser.
Also recorded: three tooling artefacts, each of which first read as a code bug
The frozen
requestAnimationFrame, a stylesheet that hadn't loaded reporting every border radius as0px, and the direction change above. Isolating before concluding is what separated them — and in thedircase, not doing so would have cost a revert of a correct change. That lesson generalises past this work, which is why it's in STATUS and not just in a commit message.The defect list is now keyed by state, not count
Three fixed, one open (D3's pagination), one blocked on backend modelling. Only D3 is open work, and the 2.4.11 remedy can't be scoped until A2 reproduces the fault.
450 frontend tests unchanged; every ADR id and doc path resolves.
🤖 Generated with Claude Code