Version. 0.11
Date. 2026-08-01
Status of this document. The single guide we work to: everything we are building, in order, with its current status. This is the one page the Operator reads at each sitting. Charter v0.1 ratified 2026-07-30; autonomous regime in effect. Day-to-day status also lives in current-status/dunin7-status-brief, which carries the decision queue.
Author. Claude.ai; v0.3–v0.4 by Claude Code; v0.5–v0.9 by Claude.ai; v0.10 by Claude Code; v0.11 by Claude.ai.
Changes from v0.10. Two items closed: B-29 and B-5. Both are built, merged and tagged. B-29's entry was materially wrong in v0.10 and is corrected here — it described a fix we tried, proved unsound, and abandoned. Four new items, all small, all from things the two runs surfaced. One parked item removed because the approach that created it was abandoned. And one thing now waits on you that did not before: the restart that puts both merges live.
It is written for you first. Every item has a plain-English line saying what it is and why it matters. Technical cross-references sit at the end of each line for the sessions; you can ignore them.
It updates as we go. Whenever an item's status changes, the responsible session saves the next version of this file with the change made and the date on it — old versions stay as history, per our filing rules. The highest-numbered version is always the truth — and once a series passes nine, sort the versions as numbers, not as text, or v0.10 hides between v0.1 and v0.2.
The statuses, in plain terms: READY — a session can start it today · IN PROGRESS — a session is on it now · NEEDS YOU: decision — waiting on an answer (it will be in the status brief's queue) · NEEDS YOU: act — waiting on something only you can physically do · WAITING — queued behind an earlier item · DONE — finished, filed, verified · PARKED — deliberately set aside.
The finish line, in one sentence: the full live demonstration works with no excuses — a firm's own document becomes a working engagement, information flows in, the summary view appears, a finished report comes out, a question gets a truthful answer with its sources shown, and an outsider can contribute without seeing anything else.
B-1. Fix the memory plumbing. DONE (2026-07-30). (CR-2026-158 v0.2.)
B-2. Put the fix live. DONE (2026-07-30).
B-3. Your browser walk-through. DONE (2026-07-30).
B-4. The completion plan for the missing screens. DONE (2026-07-31).
B-5. The Manifestation screen. DONE (2026-08-01). The "organize what we know" room was an empty placeholder that told you "nothing has been organized yet" on engagements that had been organized that morning. It is now a real screen, and the rule it carries is the one that matters: a room cannot claim the record is empty unless a completed read says so.
How that rule is enforced, in plain terms. The two states that make a claim about your record — empty and has content — cannot be created in the code without the timestamp of the read that produced them. A room can still choose a poor sentence. It can no longer invent one out of nothing. (The first merge on the surface side under the charter. CR-2026-160 v0.2; loomworks main 7e81e30, tag manifestation-room-v0_1.)
B-6. The Shaping screen. READY — unblocked by B-5. (Change request C, with B-7.)
B-7. The Rendering flow, end to end. READY — unblocked by B-5. Two things missing — no way to download, and no sign of which document is current when an older and a newer sit side by side. (Change request C.)
B-8. Teach the Companion the last three rooms. WAITING on B-6 and B-7. (Change request D. First Step 0 item: does a vaguely-worded request skip the authority check entirely?)
B-9. The "show me where this came from" walk. WAITING — depends on B-25. Click any statement in a finished report and walk backward to the person who said it. Both of its original blockers are now gone — the version-pinned lookup it was said to need existed all along, and the fabricated provenance field is fixed. What remains is B-25, because a walk over false authorship demonstrates a lie convincingly. (Change request E.)
B-10. Outside contributors, safely. READY (can run now). (CR-2026-157, filed, confirmed not executed.)
B-11. Make answers truthful. READY. The answer isn't wrong, it's unstable — the same question three times gave correct, then a false confession of fabrication, then correct with its source cited.
B-12. Make the record searchable. WAITING on B-11. The biggest single piece of engineering on this list. May merge with B-11.
B-13. Ask your engagement. WAITING on B-12.
B-14. Read spreadsheets and slide decks. READY (independent).
B-15. Catch the record up on June–July. DONE (2026-07-30).
B-16. Update the foundation document. READY to draft · NEEDS YOU: act (read it first). Drafting is authorized; committing a seed version is yours and cannot be delegated. (Seed v0.13, three riders. Queue entry Q-1.)
B-17. The portfolio filter. READY. One parameter.
B-18. File the four protocol requirements. READY.
B-19. Keep this list alive. IN PROGRESS (this document).
B-25. The record writes that aren't true. READY — unblocked, and next on the engine. Eleven places where the system writes something into the permanent record that is plausible and wrong. Nine of the eleven are invisible from any screen. B-29 no longer blocks it — that work is merged, so the collision it would have caused is gone. (Two open questions and one unresolvable line reference carried to its Step 0 — see the corrections note.)
B-26. Correct the record about door 3. READY (small).
B-27. Fix the door that can't be closed. DONE — it was already fixed. (W-2, resolved at a425ecd.)
B-28. Stop losing contributed documents. READY. The system advertises that it accepts Markdown documents, then refuses them on the way back out, and the contribution is lost rather than merely unreported. (W-3. Step 0 of change request C.)
B-29. The provenance field that was fabricated everywhere. DONE (2026-08-01).
Every record we write carries a field meaning "the event that produced this version." It was filled with an invented value at eighty-three places across thirty-six files, and correctly at exactly three. All eighty-three now carry the true value.
What v0.10 of this list said, and why it was wrong. v0.10 described the fix as "one small change in one file", with the eighty-three sites left untouched. We tried that and it failed. The check we built for exactly this purpose caught it before anything merged: because of how the code protects records from being altered, that approach corrected a copy and left the original — so the record would have been right while the system's own working copy stayed wrong, in the same breath. We took the expensive route instead, and changed all eighty-three places so the value is correct from the moment it is made. (The prior position is kept here rather than deleted, per our corrections discipline.)
What it cost and what it bought. Ten versions of the change request, two abandoned designs, and a run of about ninety minutes to build. What it bought is that the record, the stored copy, and the value the system is holding all agree — which the cheap fix could not have delivered at any price. (CR-2026-159 v0.10; engine main e106917, tag provenance-seam-v0_1.)
B-30. The trap that caused B-27. READY — small. The thing that broke the door in B-27 was never repaired — only the broken door was removed. A shared piece of plumbing demands a value that only some addresses can supply, and any new address built without it will break the same way, silently. The test suite cannot see this, because the tests replace that piece of plumbing with a stand-in.
B-31. The post-admission lifecycle investigation. READY to open — B-29 and B-5 have both closed. An investigation arc, not a build. The parent asked on what basis an agent is admitted at all; this asks what happens after — an admitted agent accumulates authority, calls other agents, may hand authority onward, goes dormant, is revoked, and leaves records that outlive it. Eight questions. No build items come out of it until the parent's decisions settle.
B-32. The re-derive button on the Manifestation screen. READY — new, small. The new screen tells you when your record has moved on since the summary was built. It does not yet give you a way to rebuild it. That was an oversight in my own change request: the design note said the screen should offer the action, and then the build instructions never listed the button. It stays a button you press rather than something that happens by itself, because rebuilding costs credits and changes the record — so it is yours to trigger. (Queue entry Q-4, resolved into this item rather than left waiting on you.)
B-33. The engine's one broken test. READY — new, small. One test has been failing for some time, unrelated to anything we have built. It is harmless in itself, but it means nobody can ever say "the tests pass" — they have to say "the tests pass except that one," which is exactly the kind of standing exception that hides a second failure when it arrives.
B-34. The surface's twelve warnings. READY — new, small. The same problem on the other side. The surface's tests report no failures, but the run still ends in an error state because twelve background requests are left unhandled. Same consequence: a clean run cannot be recognised as clean.
B-20. The marketing website. READY. All five answers confirmed.
B-21. Reply to Aldous. DONE (2026-07-30).
B-22. Package Stele for the outside world. WAITING (steady background track). (689de41, tag v0.4.0.)
B-23. The security story, told right. READY. All four answers confirmed.
B-24. Investor-visible FORAY. READY — unblocked. B-29 was the gate: the tamper-evidence layer must not be switched on over fabricated values, and that is now fixed. (Still sequenced behind B-10 for content.)
Restricted-visibility slices 2–3 · target-hosted visibility · the protocol wire-format changes · OVA's remaining standalone role · cross-fund questions · the stray zip file in the record · how many live records already carry a fabricated author (needs a production database query; surfaced, not folded in).
Removed from this list: "whether to sweep the eighty-three now-dead lines." That item only existed under the cheap fix, which left eighty-three lines writing a value nobody used. The route we took visits all eighty-three, so there is nothing left over to sweep.
Two things now, where yesterday there was one.
The restart. Two merges are sitting built and not yet live — the provenance fix on the engine and the Manifestation screen on the surface. Merged is not deployed. The restart is yours by the charter and cannot be delegated to a session. Both are ready and batch together into a single act.
B-16. Read seed v0.13 before it commits. Drafting was authorized on 2026-07-30; committing a seed version is yours.
Everything else on this list proceeds without you. Decisions inside the authorized build items are taken by the sessions and recorded with their reasoning in the status brief, where you can overturn any of them at cadence in one word.
DUNIN7 — Done In Seven LLC — Miami, Florida DUNIN7 — the build list — v0.11 — 2026-08-01