Version. 0.6
Date. 2026-08-01
Charter. standing-notes/dunin7-standing-authorization-charter-v0_1, ratified 2026-07-30.
Author. Claude Code (execution session). Operator: Marvin Percival.
Supersedes. v0.5 at record 32182bf. v0.4 at 1ec5c2a, v0.2 at 40571cf, v0.1 at 602bd9c stand as siblings. v0.3 remains deliberately absent.
Changes from v0.5. B-29 is done. CR-2026-159 merged and tagged. The item that halted twice and consumed six CR versions is closed. Q-2 is answered and removed; Q-3 is unchanged and now blocks nothing but itself.
Generated from git log at filing time, per v0.4's rule.
loomworks-engine) — the first engine merge under the charter
main at e106917, tag provenance-seam-v0_1 at f89e9c7, both pushed and origin-verified. Previous main was a317051, unmoved since 2026-07-30.
Thirteen commits on cr-2026-159-option-a, merged --no-ff:
| Commit | Step |
|---|---|
| d4dc17a | 0 — pre-flight, baseline failure set, discard the option-B branch |
| 9b96d5a | 1 — append_event takes the event identifier from its caller |
| 4fd175f / c607ba6 / f3a1bbf | 2 — the 93 src/ call sites, in three tranches |
| 356b4b3 | 3 — test helpers for minting and version advance |
| 7377041 | 4 — version-advancing copies re-mint provenance |
| 0f64562 | 5 — the 322 test append sites |
| 16d629c | 6 — the two backfill repairs |
| 178a62b / d58031e / 9a31938 | 7 — regression tests, acceptance item 10, test tightening |
| 0598ab8 | 8 — implementation notes |
| e6eb8e8 | merge |
| e106917 | archive v0.10, the version actually executed |
loomworks-record)
| Commit | Subject |
|---|---|
| 6165bf6 | CR-2026-159 v0.5 — reverse to option A |
| 546544f | CR-2026-160 v0.1 — B-5, the Manifestation room |
| c53476b | CR-2026-159 v0.6 and CR-2026-160 v0.2 |
| 9c6f9a5 | CR-2026-159 v0.8 — kickoff made resume-aware |
| 1768854 | CR-2026-159 v0.10 — re-anchor the count derivation |
loomworks)
main at 99f7f64, untouched. CR-2026-160 v0.2 is filed and unexecuted.
provenance.wasGeneratedBy means "the memory event id that produced this version." Every write path used to fill it with a UUID minted before the event existed, so the stored value pointed at nothing. It now means what it says.
Acceptance gate: all sixteen items met. Suite at the tag: 1 failed, 3444 passed, 68 skipped — the single failure is the pre-existing test_stele_router_mount 401, which fails at a317051 too. No failure outside the Step 0 baseline set at any step.
Live verification on a throwaway database showed the row, both JSONB columns and the object the caller still holds all carrying one value. The third of those is what option B could not give and is why it was abandoned.
The invariant, now stated and enforced: one object version is one event is one freshly minted identifier. memory_events.event_id is the primary key, so a model_copy that advances a version without re-minting provenance is an IntegrityError rather than a silent wrong value. The invariant cannot rot.
Deployment is not done and is not autonomous (F-1). The launchd restart is the Operator's act and batches to cadence. Merged is not deployed.
Six filed CR versions, two never filed, two halts, one abandoned branch. Worth recording plainly because the cost was real:
What it bought: no value anywhere in the system is now fabricated, corrected, or derived after the fact, and the database enforces that rather than a convention. The alternative the drafting session preferred — dropping the field as a denormalisation — was set aside by ruling and its reasoning is preserved in v0.10 §1 as a known accepted cost.
The pattern across all of it: each halt came from a premise stated with more confidence than the evidence supported. None was an analysis error. Every one was caught by building the thing and running the suite — not by review.
Four, all mine, all caught before the merge:
NameError in committed src/ code — manifestation.py used uuid.uuid4() where the module imports uuid4 directly. Introduced at f3a1bbf, invisible for three commits because the tree could not import between Step 1 and Step 5.Provenance(...) calls. Reverted; every later pass used AST statement boundaries and compile-checked before writing.pytest.raises(Exception) with a loose substring check. The verification script hit an unrelated InvalidRequestError and reported enforcement as proven, which exposed that the test could pass without the invariant ever firing. Tightened to the specific constraint.main twice across the arc before being moved to a branch. Nothing was ever pushed. v0.10's kickoff added "check the current branch before the first commit" precisely because of this, and on the final run I checked first.
And one tooling lesson worth carrying: ast.parse does not catch a repeated keyword argument; only compile() does. A file passed the parse check and failed at collection.
B-29 was the gate. With it merged:
wasGeneratedBy=<var>_event_id rather than uuid.uuid4().provenance JSONB now agree, which was the point of ordering B-29 first.
Nothing. All repositories pushed and origin-verified. No session in flight. Branch cr-2026-159-option-a is merged and may be deleted at the next engine session's Step 0.
docs/phase-crs/ created in loomworks, which does not exist there yet.test_stele_router_mount 401 — Q-3. Until it is fixed the engine baseline is not zero and "suite green" cannot be an acceptance criterion for any engine CR.
| # | Question | Default | Blocks | Proceeds regardless |
|---|---|---|---|---|
| Q-1 | B-16 — seed v0.13. Drafting authorized 2026-07-30. Committing a seed version is an Operator act (charter F-3). The draft awaits your reading. | Commit as drafted once read. | Seed v0.13 only. | Everything. |
| Q-3 | The engine baseline is not zero. test_stele_router_mount fails at a317051 and at e106917 with a 401, not caused by CR-2026-159. Own build-list item, or a rider on the next engine CR? | Stays red and stays recorded. | Only the ability to assert a zero-failure gate. | Everything. |
Q-2 is answered and removed. The Operator ruled 2026-07-31: finish option A. Done.
Batched Operator acts pending: one. F-1 deployment — CR-2026-159 is merged and tagged but not deployed. The launchd restart is yours and batches to cadence. No passkey ceremony outstanding (F-2). No external communication queued (F-5).
None outstanding. A-2 fired twice during this arc (v0.4's Checkpoint A, and the option-A halt on the test tree); both were reported, queued, and resolved by CR version bumps from a drafting session rather than absorbed in flight. C-3 held under pressure both times — the tempting one-line test edit was never made.
R-2 completed end-to-end for the first time. A filed CR ran through its own checkpoints unattended, halted where it said it would, resumed from a bumped version, and merged and tagged under R-2 including the push. Checkpoint A reported and proceeded on a clean report with no Operator confirmation, exactly as v0.4's amendment intended.
C-1 observed at the merge: fetch and HEAD re-verified immediately before, clean index confirmed, --no-ff, one session on the repo.
R-4 held: the suite ran against playground_test; live verification ran on a throwaway clone dropped afterwards. playground_dev was never connected to. No .env change, no live process touched.
F-1 respected: merged is not deployed, and the restart is queued as an Operator act rather than performed.
The executable-document standing note earned its keep three times. Two drafts (v0.7, v0.9) were caught at pre-flight with stale step anchors and never filed — each would have sent a resuming session to delete a branch that no longer existed or to perform a step already done. And at v0.10 the version check had to become numeric: this is the first document to reach double digits, and a plain directory listing sorts v0_10 between v0_1 and v0_2. The note's "one directory listing" rule now needs a numeric sort to stay reliable, and should be amended to say so.
DUNIN7 — Done In Seven LLC — Miami, Florida DUNIN7 — status brief — v0.6 — 2026-08-01 B-29 closed. The engine is released. One object version, one event, one identifier.