Version. 0.6
Date. 2026-07-30
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.
Author. Claude.ai; v0.3–v0.4 updates by Claude Code; v0.5–v0.6 by Claude.ai.
Changes from v0.5 (committed at 43bb66f). B-15 DONE — the June–July absorption ran and the record is current for the first time since 28 June. Current-status manifest v0.76 filed: seven new entries covering roughly forty change requests, the two owed corrections applied to entries 119 and 120, and six open items recorded. Three change-request documents that still claimed they were never built were corrected. Record pushed at 7539f05. B-11 is unblocked and now READY — and the absorption found the likely reason for the wrong answer, written into its line below so nobody rediscovers it. New item B-25 (a mislabelled record, small, known since June). B-4's inputs grew: the absorption named a pattern across three of the missing screens, noted in its line. No other content changed.
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 (v0.2, v0.3 …) 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. Updating this list is part of every session's closing duty, like the rest of its reporting.
The statuses, in plain terms:
The standing documentation rule, recorded here: every substantial piece of system documentation gets an easy-to-read plain-English companion for your consumption, with examples where they help. You never have to ask for it; you only ask when you want more explanation.
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). Eight kinds of recorded events were never registered with the replay machinery, so six real engagements — including your biggest — couldn't produce their summary view. The fix was built, tested, accepted at its review gate, merged with a guard that stops the mistake from ever being reintroduced, and pushed. (CR-2026-158 v0.2; engine merge c4014d6, main a317051, tag registry-completeness-v0_1. Recorded in manifest v0.76 Entry 126.)
B-2. Put the fix live. DONE (2026-07-30). The Operator-directed restart is done and the live engine verified running the fixed code. The six affected engagements heal the first time anything asks them to produce a summary view.
B-3. Your browser walk-through. NEEDS YOU: act (anytime). A short checklist (already written, with the exact commands to start things up) where you click through the system and say what feels right or wrong. Fifteen minutes. Your answers sharpen B-4. This is the only thing standing between us and B-4.
B-4. The completion plan for the missing screens. WAITING on B-3. A session takes the audit's break list and your walk-through notes and writes the precise, ordered plan for items B-5 through B-9. New input from the absorption: three of those five screens are the same problem wearing different clothes — the system holds true information and no screen says it. The plan should treat that as one subject rather than five separate gaps. (The completion scoping note, from the walk-audit break list + manifest v0.76 §5 surface-silence class.)
B-5. The Manifestation screen. WAITING. Today the "organize what we know" room is an empty placeholder. This builds the real screen — including a marker that tells you when the view has gone stale, so what happened silently to E0005 (its newest summary is 25 versions out of date and nothing says so) can never happen invisibly again.
B-6. The Shaping screen. WAITING. The room where knowledge is arranged for a particular reader — also a placeholder today. This makes it real.
B-7. The Rendering flow, end to end. WAITING. Producing the finished document and downloading it, smoothly, from the screen — no workarounds.
B-8. Teach the Companion the last three rooms. WAITING. Today you can talk to the Companion while gathering information, but not to organize, shape, or render — for those you'd have to leave the conversation and use raw mechanics. The audit counted eight forced exits. This closes them: you ask, it does.
B-9. The "show me where this came from" walk. WAITING. Click any statement in a finished report and walk backward — to the shaped view, to the record, to the person who said it and when. The machinery underneath already does this perfectly; no screen shows it. This is the product's promise made visible.
B-10. Outside contributors, safely. READY (can run now). Lets a founder or a customer reference contribute to a deal file through a one-time link — everything they add is labelled as theirs and held for your team's approval, and they can see nothing else. Fully specified and approved. (CR-2026-157, filed; confirmed not yet executed. At Step 0: its OVA naming is historical trajectory — GRANTHA holds the authorization seat.)
B-11. Make answers truthful. READY — unblocked by B-15, and it now starts with a lead rather than a blank page. In testing, a tiny six-note engagement answered a question with an old, already-corrected value presented as current. What the absorption found: in early July a run of four fixes made the Companion truthful by construction — its replies composed directly from the record so they cannot fabricate, omit, or miscount. Every recall path got that treatment except one: the path a partner actually uses to ask a question. It was deliberately set aside at the time as its own future piece of work, for a good reason, and never picked up. That is where the diagnosis starts. One other read worth doing: the July correction-matching work never recorded what was actually built, and correction-matching is adjacent to this bug. A scope question the diagnosis will settle: if this turns out to be a problem with the question-answering path rather than the record-keeping, then B-11 and B-12 are probably one piece of work — rebuilding that path to be truthful and making it searchable would otherwise be built twice. Not a decision for now. (Manifest v0.76 §5, semantic-recall residue; Entry 122; CR-2026-129 v0.2's carve-out; CR-2026-145's unrecorded mechanism.)
B-12. Make the record searchable. WAITING on B-11. Today a question just looks at the fifty most recent notes, whatever they are — no real search exists. This builds it, over the notes and the text of uploaded documents. The biggest single piece of engineering on this list. May merge with B-11 — see B-11's scope question.
B-13. Ask your engagement. WAITING on B-12. The payoff: a partner asks a plain question — "what did the reference calls say about churn, and what does the contract actually promise?" — and gets a truthful answer with every source shown, which can then be kept in the record, marked as machine-assisted. When this lands, our rule against demonstrating question-answering lifts. (Limits and order confirmed by the Operator, 2026-07-30.)
B-14. Read spreadsheets and slide decks. READY (independent — can run alongside anything). Today the system can't take in the cap table, the financial model, or the pitch deck — the heart of a data room. This adds those file types and fixes an inconsistency between the two upload paths. (The absorption confirmed both gaps still stand after July's fifteen-CR upload cluster — manifest v0.76 Entry 124.)
B-15. Catch the record up on June–July. DONE (2026-07-30). Six weeks of engine and Operator Layer work — roughly forty change requests — had landed with none of it in our current-status record. Now absorbed: seven new entries, the two owed corrections applied, six open items recorded, and the record's own fronts brought current. Two of the six open items are findings this pass turned up rather than inherited, and one of them is B-11's lead. (Manifest v0.76 via amendment recipe; record 7539f05.)
B-16. Update the foundation document. READY to draft · commit NEEDS YOU: act (read it first). Three corrections queued for the seed, including replacing the outdated authorization wording. Drafting approved 2026-07-30; a session drafts, you read, and only you approve the new version into the record. (Seed v0.13, three riders.)
B-17. The portfolio filter. READY. A small fix: the dashboard can already group engagements into workspaces but won't filter by them. One parameter. Gives the fund-level "compare my deals" view its plumbing.
B-18. File the four protocol requirements. READY. Your agent-construction idea produced four precise requirements for the identity and authorization specs. Approved 2026-07-30 — a session files them in writing with the right owners so they're waiting when those specs next move.
B-19. Keep this list alive. IN PROGRESS (this document). Sessions maintain it; you read it.
B-25. The mislabelled record. READY (tiny). When the Companion triggers a piece of work, the record labels it as coming from a contributor rather than from the Companion. Known since late June, filed twice, never fixed. It is one line in the substrate — and it is a wrong label on a record whose entire promise is accurate attribution, which is why it shouldn't sit unfixed. (compositions.py:73; CR-2026-120 residue, carried by CR-2026-127; manifest v0.76 §5.)
B-20. The marketing website. READY. All five answers confirmed 2026-07-30 — site scope, the headline sentence, "book a working session" as the button, the honesty rule for every claim, and the go-ahead to write the full text. The answers themselves are in the marketing alignment note; the session writing the copy reads them there rather than from this line.
B-21. Reply to Aldous. NEEDS YOU: act. Draft chosen 2026-07-30 — the shorter, warmer one. Edit if you like, then send.
B-22. Package Stele for the outside world. WAITING (steady background track). The identity component, cleaned up so a stranger can run it — plus assembling its papers in the shape a security evaluation would want, at little extra cost while we're in there.
B-23. The security story, told right. READY. All four answers confirmed 2026-07-30. Not "we sell security" — rather: the incidents in the news become our backdrop, a proper assurance page speaks the security buyer's language, and certification comes before category. Includes making "a rogue agent is an unsigned agent" the flagship of our standards work.
B-24. Investor-visible FORAY. WAITING on B-10. Confirmed 2026-07-30: investor-hosted first, not target-hosted. A session scopes the adapter once B-10 lands, and the pilot conversation leads the build rather than following it — what a real investor asks for shapes what gets built. (Target-hosted visibility stays parked.)
Restricted-visibility slices 2–3 (to be reworked in grant language when their day comes) · target-hosted visibility · the protocol wire-format changes (queued to the next version boundary by design) · OVA's remaining standalone role (one conversation, whenever) · cross-fund questions · the stray zip file in the record.
(The live-operations feed to investors left this list at v0.5 and is now B-24.)
None open. Decisions 1–8 are all answered. The history is kept below.
The queue is still clear of questions. What is left of your involvement is three acts — the walk-through (B-3), sending the Aldous reply (B-21), and reading the seed draft before it commits (B-16).
DUNIN7 — Done In Seven LLC — Miami, Florida DUNIN7 — the build list — v0.6 — 2026-07-30