Loomworks — feat-in-engagement-dual-pane parked-state resolution — v0.1

Version. 0.1 Date. 2026-08-25 Author. Claude Code on DUNIN7-M4. Operator: Marvin Percival. Status. RULED — PARK. Disposition decided by the Operator 2026-08-25 on the evidence below. This note exists so the question is not re-opened; it is not a recommendation awaiting a call. Subject. Operator-Layer branch feat-in-engagement-dual-pane, tip 9141447, 7 commits, last touched 2026-06-26. Sibling. investigations/loomworks-b0dc30b-parked-state-resolution-v0_1 — same shape, different branch.

The ruling

Park it. Do not merge, do not rebase, do not mine it for Phase 1.

The branch is superseded by an independently-built descendant — superseded in substance, not in git history. Nothing on it is worth recovering except one trivial refactor named in §4, which carries no urgency.

1. What the branch is

Seven commits building an in-engagement surface: a responsive dual/single-pane shell, the Memory room wired to real engine data, the Rendering room wired to real outputs, held-notes confirm (the Operator's reserved act), and mockup-fidelity input-zone and header work. Forked from main at e2e1887 (2026-06-26).

Its own directory shape — CenterPane.tsx, LeftRail.tsx, RightRail.tsx — is a three-column layout that does not exist on main in any form. That architecture was not merged and not continued.

2. Why "superseded, not by history"

None of the branch's 7 commits is reachable from main. Verified per-SHA with merge-base --is-ancestor: zero hits. Git-wise the two lines never touched after the fork.

But main independently built the same capabilities, further. main's MemoryRoom.tsx history begins at db5da27 ("Companion surface step 2: in-engagement surface (responsive) to mockups 1 & 2") and continues through the type-scale fix, list-loading CR-2's keyset spine, and the add-side held-card Dismiss — a separate lineage reaching the same destination by a different road.

The decisive evidence: useIsWide.ts is byte-identical between the branch and main — implementation and docstring, including the "The switch is on WIDTH, not device type" comment and the SSR-safety note — with no shared commit. A file does not arrive at byte-identity twice by chance. Whoever built the Companion-surface arc on main had this branch's work in front of them and carried the good parts across by hand.

Size confirms the direction of travel:

| File | Branch | main | |---|---|---| | MemoryRoom.tsx | 286 lines | 684 | | RenderingRoom.tsx | 162 lines | 315 |

3. What main has that the branch does not

The rest of the surface, none of it on the branch: ManifestationRoom, ShapingRoom, MobileSurface, SeedViewer, CredentialPanel, CandidateCommitBanner, CandidateFindingsPanel, WalkPanel, WorkCard, and the RoomView dispatcher itself. Plus a persisted conversation-order setting (conversation-order.ts — a /me/settings key with a stored preference, where the branch has only a UI toggle), keyset pagination from list-loading CR-2, and add-side held-card dismiss.

At the fork point the branch was 7 commits ahead of a main that has since taken 152 commits on that surface.

4. The one non-superseded item

MessageOrderIndicator extraction. The branch pulled the conversation-order toggle out of TopBar.tsx into its own component at src/app/operator/engagement-navigation/MessageOrderIndicator.tsx; on main it remains an inline function MessageOrderIndicator inside TopBar.tsx (line 233).

Both bodies were diffed: same data-testid, same data-message-order / data-icon attributes, same Tabler IconSortDescending/IconSortAscending selection, same styling tokens. Functionally identical — the difference is file placement only. The branch version adds a standalone prop to drop a pill-specific negative margin.

This is a trivial refactor with no urgency. It would matter only if the toggle were needed in a second location; it currently is not. Recorded so the observation is not re-derived, not because it should be actioned.

5. What this note forecloses

• Do not re-open "is the branch ahead?" It is not. §2's byte-identity finding is the answer and it will not change.
• Do not merge or rebase it. The three-column layout it carries would conflict with RoomView at every point and would be reverting main, not advancing it.
• Do not treat it as a source for Phase 1. Everything of substance already shipped, more completely, by the other path. The risk here is the inverse of the usual one: not that work is lost, but that someone spends a day recovering work that is already surpassed.

If the branch is ever swept, sweep it. Its 7 commits are single-copy (on no remote), so a sweep is a real deletion — but the deletion costs nothing except §4's refactor, which is a fifteen-minute redo whenever a second call site appears.

DUNIN7 — Done In Seven LLC — Miami, Florida Loomworks — feat-in-engagement-dual-pane parked-state resolution — v0.1 — 2026-08-25