Version. 0.37
Date. 2026-08-06
Charter. standing-notes/dunin7-standing-authorization-charter-v0_1, ratified 2026-07-30.
Author. Claude Code (execution session). Operator: Marvin Percival.
Supersedes. v0.36 at record cd11588. All earlier versions stand as siblings.
Changes from v0.36. CR-2026-171 and CR-2026-172 filed, executed, and merged — the small-work sweep's engine and surface clusters. Six of the seven items the Step 0 findings called small landed; the seventh (B-54) was pulled back out mid-execution as a residue, larger than scoped.
loomworks-record)
| Commit | Subject |
|---|---|
| cd11588 | file CR-2026-171 (engine) and CR-2026-172 (surface), small-work sweep |
| this filing | status brief v0.37 |
CR numbering confirmed clear. Both CRs arrived from Claude.ai as best-guess numbers (171, 172), flagged for collision-checking before filing. Checked against the record's ledger directly: the highest CR filed to date is CR-2026-170. No collision; both numbers filed as given.
loomworks-engine)
| Commit | Subject |
|---|---|
| 8992179 | CR-2026-171: small-work sweep, engine cluster (B-53, B-35, B-44) |
Tagged cr-2026-171-small-work-sweep-engine, pushed, origin-verified.
loomworks)
| Commit | Subject |
|---|---|
| baec093 | CR-2026-172: small-work sweep, surface cluster (B-57, B-38, B-36) |
Tagged cr-2026-172-small-work-sweep-surface, pushed, origin-verified.
Three of four items landed as scoped. B-53 (render list's SELECT never named display_number) — one line. B-35 (NoCredentialError's exception handler returned 500 where every sibling "no key reachable" path returns 503) — one line. B-44 (both shape-event routes read titles via the stored-title-only lookup instead of the stored-then-derived cascade renders already used) — two call sites, one file. Each carries a new or extended test; full engine suite: 3547 passed, 68 skipped, 0 failed.
B-54 pulled back out mid-execution — a residue, not a fix. The safety check the CR itself gated the migration on (COUNT(*) WHERE display_number IS NULL against playground_dev, both views) cleared at zero. The migration and its own narrowly-scoped test passed. Then the full suite surfaced 230 failures — fixtures across the engine, unrelated to this concern (test_shape_event_model.py, test_shape_events_view_projection.py, dashboard/inbox/library/composition suites among them) construct ShapeEvent/RenderEvent objects directly without going through assign_shape_display_number / assign_render_display_number, and the tightened NOT NULL broke every one of them. This is a fixture-touching change spanning dozens of files, not the one-migration item the CR scoped as small. Per the CR's own §5.6 language — land the NOT NULL only if safe, else report the residue rather than silently drop it — the migration was reverted (file deleted, playground_test downgraded from 0104 back to 0103) rather than forced through. B-54 needs its own scoping note before it is attempted again, scoped as a fixture sweep, not a migration.
One finding beyond the CR's four items, flagged and left alone. The Shape-list query has the identical display_number omission as B-53 (shape_events_view_query.py:109-133) — not a named build-list item, so not folded into B-53's fix. Recommended as its own item (the findings document called it a B-59 candidate) rather than absorbed.
All three items landed as scoped — no residues. B-57 (Manifestation room's staleness sentence read as self-contradicting) — a new string, composeVersionPairStable, used only when the flag and the version pair would otherwise appear to disagree; no change to the flag's own logic. B-38 (RenderingRoom hand-rolled its own loading/error/empty/populated branch instead of the shared .state contract three other rooms already use) — one file, confirmed behavior-identical via the hook's own precedence rules, not merely possible-in-principle. B-36 (no way to retire a render) — the engine route was already complete and already correctly irreversible; built a Retire control with a required-rationale confirm step, wired to the existing endpoint.
Full sweep: eslint clean, tsc clean, next build succeeds, 704/704 vitest tests pass.
Both CRs halted at their named Checkpoint A before tag-and-push, per C-1's mandatory protocol. Both approvals arrived and were honored only in the exact standalone form (GATE A APPROVED — PROCEED) per the standing gate-approval-exact-form rule — an initial less-exact phrasing ("Checkpoint A approved.") was correctly held as not-yet-approval and the exact form was requested and received before either tag landed. HEAD was re-verified against each CR's stated baseline immediately before every commit; both matched exactly, no divergence from origin at any point in either repo.
Nothing. Record main pushed and origin-verified through this filing (once committed). Engine main at 8992179, tag cr-2026-171-small-work-sweep-engine, both push-verified. Surface main at baec093, tag cr-2026-172-small-work-sweep-surface, both push-verified. No session in flight.
B-54 needs a scoping note before it is attempted again — not a migration-and-test item as originally sized, but a fixture sweep across the engine test tree (230 call sites touched by the NOT NULL tightening in this session's now-reverted attempt). Blocks nothing; every other item in the sweep proceeded without it.
DUNIN7 — Done In Seven LLC — Miami, Florida DUNIN7 — status brief — v0.37 — 2026-08-06