The window-title widget arrives as a leading titlebar accessory — a
quiet chevron on board windows only, installed and removed by the
window controller's own attach lifecycle — anchoring the one
board-level surface as a transient popover (the board window
deliberately grows no toolbar item for it). Inside: board rename
editing frontmatter title only (the folder is never renamed; an empty
commit removes the key and the window title falls back to the folder
name), the embedded shared style editor permanently targeting the
board, and the labeled Git section that this milestone only reserves
— a mode-none explanation and a disabled stub where m7's add-git,
branch, remote, and authentication controls land. Cmd-I (File >
Board Info) toggles it per window through a focused scene value,
kept apart from board-scoped transient state since a titlebar
popover belongs to one window, not to the board. Escape reverts a
dirty rename field and falls through to dismiss otherwise; a foreign
rename resyncs the field only while unfocused. 12 new tests.
Claude-Session: https://claude.ai/code/session_01SR4XGjmBE16ZUYWpfFHXwY
One style-editor component, anchor-agnostic: a background grid (None
well plus the 12 palette colors) and a curated symbol grid (the
pathfinder's five-dozen set, leading well removing the icon key for
the level default), selection-aware across cards, lanes, and the
board itself. Batch edits compute per-dimension state — uniform,
mixed (no well selected), or an off-palette value labeled verbatim
outside the grids — and choosing a well applies to the whole target
set as one write bracket, skipping no-ops per field. The popover
tracks its target set live per the freshly ratified rule: targets
re-resolve by UUID on every reload, a vanished target leaves the set,
an emptied set dismisses the editor, and nothing ever silently
retargets to the board. Anchors landing now: Board > Style
(Opt-Cmd-S) and the card/lane context menus, which also carry the
quick-style recents row (app-wide, persisted, capped at six, None
never recorded) and the lane's width control twinning the menu
chords. The styling system's other two renders arrive with it: a
lane's background paints the C7 top-edge band, the board's paints
the window content background — malformed values paint nothing and
stay byte-identical on disk. 31 new tests.
Claude-Session: https://claude.ai/code/session_01SR4XGjmBE16ZUYWpfFHXwY
The lane title bar becomes real: leading SF Symbol (hand-written names
render leniently, unknown ones fall back to the level default), title
or secondary untitled placeholder, a quiet count badge that counts
exactly the cards the body renders (so the m5 search filter is
followed by construction), and a new-card button. The whole bar is
the reorder drag surface — no grip — with click-vs-movement splitting
select from drag; a pure proposal function maps the drag to an
insertion index and release commits through the Writer's same-parent
degenerate reorder, compacting and retrying when midpoint precision
runs out. Clicking never edits: inline rename is Return on the sole
selected card or Board > Rename for either kind, a third transient
editor beside the placeholder that tracks its target by UUID, commits
on focus loss, discards silently when the target vanishes, and
removes the title key on an empty commit. The new-card placeholder
renders at last — the settled Cmd-N target rule (pure, tested) files
it after the anchor card, at a selected lane's bottom, or into the
last-active lane; Return commits and re-selects the lane, Cmd-Return
also opens the card window, and a failed create discards the overlay.
New Card / New Lane / Rename land in the menus with focused-editor
and read-only validation; rename gets its own WriteOperation case in
the banner vocabulary. 59 new tests.
Claude-Session: https://claude.ai/code/session_01SR4XGjmBE16ZUYWpfFHXwY
The lane strip replaces the placeholder board: window width divides
across the lanes' width units (no horizontal scroll, no minimum width,
degenerate compression accepted), a lane of n units flowing its cards
into n round-robin masonry columns via a measurement-cached Layout.
Width has two deliberately opposite controls, both landing here: the
right-edge drag (ported verbatim from the pathfinder's ColumnResize)
freezes the 1x standard at drag start, snaps between integer widths
with the asymmetric shadow-leads tick and 10pt re-entry, grows the
window one standard width per snap so siblings keep their exact
pixels, and rubber-bands at the screen's visible frame — uncapped
otherwise; the Increase/Decrease Lane Width items (new Board menu,
Cmd-Opt-arrows) are the stepper's keyboard face and re-divide the
existing window width instead, never touching the window. Width
changes write through the new .resize WriteOperation ("Couldn't
resize…" in the banner vocabulary, which grows with the surfaces by
design); malformed width values render as one unit and stay untouched
on disk. 27 new tests port the pathfinder's resize-math suite onto
the uncapped range and pin the write path's fidelity.
Claude-Session: https://claude.ai/code/session_01SR4XGjmBE16ZUYWpfFHXwY
Four scenes (welcome, restore bootstrap, board group, card group) with
system restoration disabled in favor of the registry's open-now flags:
set when a window actually opens, cleared only on user close, so quit —
and crash — leave exactly the restoration set behind. AppModel joins
windows to sessions (shared store, registry record, card refs, held
security scope); CloseFlushCoordinator pins 02's strict close order as
a seam-injected machine (card sessions end, windows drain, store
flushes, record stamps, teardown) with named slots where m6/m7 flushes
land. HostedWindowController proxies — never replaces — SwiftUI's
window delegate to intercept windowShouldClose for the flush, report
frames, and place saved frames onto live screens. Card windows are
(board path, case-folded card id) values: reopen focuses, and a
snapshot-pure fate function dismisses on delete, tombstone, tombstoned
lane, or cross-board move.
Claude-Session: https://claude.ai/code/session_01SR4XGjmBE16ZUYWpfFHXwY
TransientBoardState, one per store, holds state by how a reload treats
it: item-referencing sets (selection, drag membership, pending cut)
share one shape and one constraint rule — members must exist in the
current universe — applied in two directions by one primitive, so the
search filter's hidden-cards rule and reload survival are one rule
expressed once; derived state is stored as its inputs only (the query,
never its result set); and the new-card placeholder is a lane-anchored
overlay with no UUID until commit, discarded when its lane vanishes or
tombstones, handed off when the created card's UUID appears. Trash
visibility rides along per-open, never persisted. The decision is
written back into DESIGN/02 § Changes from Kanban — the TBD is closed.
Full suite 352 tests in 64 suites green. Two findings filed.
Claude-Session: https://claude.ai/code/session_018BjQRYBR6jQja3jCRi5S3A
The banner surface as one vocabulary (Kanban/LiveStore/BannerCenter,
Kanban/UI/BannerStripView): a pure precedence rule — in-progress pinned
above the collapse (ratified mid-build), lock > breakage > one-shot
write failures > commit+attachment, signposts last — with all
user-facing phrasing owned here via exhaustive switches over the closed
WriteOperation enum; free-form English survives only in diagnostics.
performWrite posts its failures before rethrowing, so no one-shot can
bypass the strip; refusals under lock post nothing.
The lock vocabulary completes: vanishedRoot and unwritableLocation join
bracketedReloadFailed, each with its own clearing rule (unwritable
clears only on a reconciling reload's writability re-probe). The
registry now owns root recovery: bookmark re-resolution absorbs renames
transparently, a dead root locks read-only and re-arms FSEvents on the
gone path so the root's return round-trips back through rootChanged,
re-minting and re-keying on the way. DirtyBufferGuard is the one modal
moment, retry / save a copy / discard, no fourth button.
36 new tests; full suite 333 tests in 62 suites green. Five findings
filed on the Redesign board.
Claude-Session: https://claude.ai/code/session_018BjQRYBR6jQja3jCRi5S3A
The parallel design session settled the tombstoned-ancestor question:
a card counts as trashed if its own flag or its lane's says so, so
tombstoning a lane ejects its cards from a live selection — they render
nowhere once the lane collapses to a single trash entry, and nothing
invisible may stay selected, drag-included, or pending-cut. One-line
predicate change in Selection.resolved(against:), plus the test.
Full BoardStore suite green (15 tests).
Claude-Session: https://claude.ai/code/session_018BjQRYBR6jQja3jCRi5S3A
Two registries (Kanban/LiveStore/): BoardStoreRegistry shares one live
store and one started, fully wired watcher per open board across its
windows — keyed by file identity (fileResourceIdentifier), never path,
refcounted to order teardown; release matches by store identity so a
board renamed while open can't leak its watcher. BoardRegistry persists
app-private per-board state in Application Support as diff-stable JSON:
records anchored by security-scoped bookmarks, recents = the registry
sorted by lastOpened (counts registry-cached, never scanned), graceful
orphaning with Forget, corrupt files quarantined aside, and files-first
verified — the board tree is untouched byte-for-byte. Timestamps use
ISO8601DateFormatter with fractional seconds: the FormatStyle variant
truncates-then-rounds and drifts a millisecond per round trip.
16 registry tests; full suite 297 tests in 56 suites green. Four
findings filed on the Redesign board.
Claude-Session: https://claude.ai/code/session_018BjQRYBR6jQja3jCRi5S3A
The parallel design session resolved the m3 watcher findings; two
changed behavior: .git internals are now filtered at any depth (a
nested clone or submodule is a stray whose internals never render),
and a root change landing mid-bracket is owned by the root-change
path — endBracket() skips its mandatory reload once a .rootChanged
tore the stream down, cleared when a fresh stream attaches. The skip
is scoped by an explicit flag, not by stream absence: a watcher whose
stream never came up still delivers post-bracket reloads, since
distrust of FSEvents delivery is the mandatory reload's whole reason.
Two new watcher tests; full suite 281 tests in 54 suites green.
Claude-Session: https://claude.ai/code/session_018BjQRYBR6jQja3jCRi5S3A
Per-board @Observable MainActor hub (Kanban/LiveStore/): watcher signals
drive off-main tree walks with a generation guard, single-flight
coalescing (strongest pending origin, watcher's merge rule), and the
resilience contract — a failed reload never replaces a good snapshot,
per-file breakage never locks editing, and a wholesale operation
(performWholesale) arms a reload-must-succeed-or-lock floor so a failed
post-bracket reload flips the board read-only until a good reload heals
it. Selection is a pure UUID-set value re-resolved on every swap;
liveness flips eject. performWrite brackets the watcher so Writer
round-trips come back app-mediated.
14 store tests; full suite 279 tests in 54 suites green. Three findings
filed on the Redesign board.
Claude-Session: https://claude.ai/code/session_018BjQRYBR6jQja3jCRi5S3A
Debounced FSEvents watcher over the board tree (Kanban/LiveStore/):
trailing 200ms debounce coalesces bursts; nestable brackets suppress
delivery and close with exactly one app-mediated reload; origins merge
reconciling > appMediated > foreign; missed-events flags degrade to a
reconciling reload; WatchRoot vanish tears the stream down (streams die
and are recreated — reattach() covers rename re-resolution); board-root
.git churn is filtered against the symlink-resolved root. The FSEvents
release callback must be top-level and non-isolated — a MainActor
closure traps in dispatch_assert_queue during stream deallocation.
15 watcher tests (4 consecutive green runs); full suite 265 tests green.
Five design findings filed on the Redesign board.
Claude-Session: https://claude.ai/code/session_018BjQRYBR6jQja3jCRi5S3A
rzen