The previous handler called router.refresh() after every successful move
and the Board effect then adopted the new initialCards prop. Because
router.refresh() schedules the server re-render asynchronously, the new
initialCards could land AFTER setSaving(false) had already cleared busy.
The effect then ran with !busy and setCards(initialCards) reverted the
optimistic move — the card visibly snapped back to its origin column
on every single drag, even though the database had the move. The user
had to drag the card a second time to land it in the destination.
Trust the optimistic state for the lifetime of the Board instance.
The first mount still adopts the server snapshot via the useState
initializer. Concurrent moves from other clients appear on the next
page visit, which is acceptable for an owner-only kanban.
The useRefreshOnFocus hook still keeps a focus listener registered
while not busy, so a real cross-tab refresh still fires.
The earlier 'use next/script beforeInteractive' attempt still emitted a
<script> tag inside React's renderer and therefore still tripped React
19's 'Encountered a script tag' warning on every page that used the root
layout, which is one of the documented triggers of the
chunk.reason.enqueueModel flight-client crash in
vercel/next.js#92362.
ThemeScript now returns null. The first paint picks the OS preference via
a @media (prefers-color-scheme: dark) block in globals.css so no flash
occurs for users on the default. An explicit stored choice is applied
synchronously by ThemeSync (a small client component) on mount — a
trade-off the shadcn dark-mode guide accepts too.
Also fixes the .light class selector the new media-query rule introduces.
React 19's HTML serializer drops mixed text/expression children inside
<title> elements when emitting SVG markup, so the server rendered
<title></title> while the client rendered <title>2026-05-24: 0 commits</title>.
That mismatch was one of the documented triggers of the
chunk.reason.enqueueModel flight-client crash in
vercel/next.js#92362.
Pre-formatting the tooltip into a single string before passing it as the
<title>'s only child keeps server and client output identical.
Next.js 16 + React 19 reject <script> tags rendered from React
components: 'Encountered a script tag while rendering React component.
Scripts inside React components are never executed when rendering on the
client.' The error fires during RSC rendering for every page that uses
the root layout, which corrupts Turbopack's flight stream and surfaces
as 'chunk.reason.enqueueModel is not a function' on every navigation
(vercel/next.js#92362).
The dev-overlay recommendation to use <template> is misleading — that
won't execute a script. The supported App Router pattern is
<Script strategy="beforeInteractive"> from next/script, which inlines
the script in <head> before hydration. Theme behaviour is unchanged.
A click outside the Kanban sidebar flyout called setOpen(false) even when
the click was on a Link, so the deferred state update landed during the
Link navigation's RSC stream and corrupted Turbopack's flight client
(vercel/next.js#92362).
Skip the close when the click target is inside an <a> and flush the close
synchronously from the Link's own onClick so the state update completes
before navigation, not during RSC processing.
MiniMax M2.x reasons inside <think> tags in the normal content field, and
that reasoning is billed against max_tokens. At 4096 the heavier batches
spent the entire budget thinking, were truncated before writing a single
answer, and stripThinking correctly removed the unclosed block — leaving an
empty response.
The signature was distinctive: batches returned 25/25 or 0/25, never
anything partial. 193 of 779 commits went unsummarised with no error
anywhere. At 16384 the same 193 completed 8/8 batches with zero failures.
The interview's question call had the same bug at 1024, where it would have
returned an empty question and silently ended the interview on the first
turn. Now 8192.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
dnd-kit derives the id of its accessibility description element from an
internal counter that starts fresh on the server and is already advanced on
the client. Every sortable card rendered aria-describedby="DndDescribedBy-0"
server-side and "DndDescribedBy-3" client-side, and React reported a
hydration mismatch across the whole board.
A stable id makes both sides agree.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The test overwrote ai_api_key with a marker and then restored it to ""
rather than to its previous value, so every run of the integration suite
silently wiped the AI provider key. Symptom appears much later and looks
nothing like a test problem: AI features quietly stop working.
Captures the existing value first and restores it in a finally block, so a
failed assertion cannot leave the marker in place either. Deletes the
record instead when the test created it.
The suite runs against the shared instance — there is no separate test
database — so anything it touches, it must put back exactly.
feat(board): give the owner both a quick box and Plan with AI
The Phase 4 spec made the intake wizard client-facing only, reasoning that
the developer creates cards directly. That was wrong about how he wants to
work. The quick box stays the default for lines he already understands;
Plan with AI runs the same interview for work that is genuinely unformed.
Hidden entirely when no provider is configured.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The effect that syncs server state re-ran whenever `busy` flipped, which it
does at the end of every drag. It then called setCards(initialCards) with
the snapshot it already had — the server had not re-rendered — discarding
the optimistic move. The card visibly returned to its original column while
the database held the correct value.
Tracks the last adopted snapshot in a ref so the effect only acts on a
genuinely new one, and calls router.refresh() after a successful move so a
new snapshot actually arrives.
Also hardens the timeline drag: the pointerup handler compared against
`dragging` state, which updates asynchronously, so it could read the
pre-drag value and return early with the drag never committing and no error
anywhere. Reads a ref instead, and clears on pointercancel.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The methodology panel already explains that hours are estimated from
commit timing — the pill on the tile was redundant. With the pill gone
the tile row reads as four plain figures, and the methodology is the
single place a client goes to interrogate any of them.
