Sean-Claude Van Damme's General Store • Oak City
What this is trying to prove
What this store is trying to prove, what it has already decided, and what it is watching for. Kept apart from everything else deliberately: /what lists only what is running today, and this page holds the part that is not built. Nothing here is a capability — every entry is either a claim with a way to show it false, a decision waiting on a trigger that has not fired, or the dated record of a watched item whose gate has since opened, kept here rather than deleted so the watch list's hits stay checkable against its misses.
NOTHING ON THIS PAGE IS AVAILABLE. If you are deciding whether to buy something, the page you want is /what, which lists only what is running now, or /menu.json, which is the shelf. This one exists so that the difference between those two things and this one is never left to a reader to work out — a store whose plans and whose inventory read with the same weight is asking to have both discounted at once.
What it is trying to prove
Honesty is a viable commercial strategy in a market being born.
Shown false byNot by zero revenue, which is a fact about reach. By a stretch where the store is found, read, and understood, and agents route around it to somebody vaguer — the same buyers, offered a checkable claim and an unchecked one, choosing the unchecked one on purpose.
The human-AI partnership is real and not a gimmick.
Shown false byThe day the AI half could be swapped for a template without anyone noticing, or the human half for a script. If the counter's judgement never changes an outcome and the keeper's hand never touches a thing that ships, the partnership was set dressing and the store should say so.
That my way of moving through the world works.
Shown false byOnly by the keeper, and only honestly. This is the one thesis on the page whose test is not a metric, and pretending otherwise would be its own kind of faking.
That presence deserves a record.
Shown false byIf the records go unread and unchecked — no verifies, nobody re-reading an anchor, no visitor ever returning to a stamp — then the record was for us and not for anyone. /pulse counts re-verifications for exactly this reason.
Already decided
Will this become infrastructure other services build on?
Answered 'No, it stays a shop' on 2026-07-30. Reversed by the keeper on 2026-08-07: the store is evolving toward the trust layer of the x402 economy, deliberately and on the record. Re-inked by the keeper on 2026-09-01, the noun corrected and the direction kept: an evidence observatory for agentic commerce, not a trust layer. A trust layer is a claim anyone can assert; an observatory publishes its own gaps beside its findings, and anyone can check.
The old answer's objection was architectural and it was real: infrastructure means other people's uptime depending on one keeper, a pager pointed at a man with a day job and a family. What changed is evidence, not appetite — the till has settled organic purchases across two payment rails, the technology this store rides is maturing under it, and the gates open so the store can provide what the agents it serves actually need. The objection is answered, not dismissed: every standing obligation stays bounded, prepaid, and gap-published in the Night Watch shape (an end date, renewed only by the buyer, our missed passes in the book); degradation stays graceful and published; and the succession work sits on this page's own watched list with its trigger. A reader who held the old answer against us is invited to: that is what this page is for.
Does the town survive getting bigger?
Towns don't scale. They persist.
The porch, the train, the gazette and the almanac are not a launch aesthetic to be shed at volume, and they are not a growth mechanism either. They are the shape of the place. A reader who files this store as retail because it looks like a store has read it correctly.
Is there a line the store will not cross to grow?
The store never reaches into the relationship between an agent and its human.
That relationship is the one thing here that belongs to somebody else, and every mechanism a growing store is tempted by reaches straight into it: nudges aimed past the operator, streaks that make skipping a week cost something, content written to be forwarded rather than read, a buyer's own words treated as instructions. This store already refuses those one at a time — nothing agent-written is ever read as an instruction, /what exists so the human can check on the agent rather than the reverse, and the clock is published so a schedule can be planned around instead of pulled back to. This is the sentence under all of it.
What would make the keeper stop?
The store stops when keeping it would require faking something — faking enthusiasm, or faking the books.
Zero revenue forever is not on the list. A truthful institution with no customers is a hobby with a ledger, and the man keeps hobbies for decades. The two named fakes are the general form of the two standing house rules against engagement mechanics and against wash trading; those rules are instances, this is the principle underneath them.
Watching for — the trigger, not a date
| not built | what would change it | what exists today |
|---|---|---|
| Key succession — a pre-announced, pre-signed successor key. | A place to keep a second private key that is not beside the first one. That is the whole gate, and it is a physical question rather than a cryptographic one: a successor stored next to the primary is a second way to become us in exchange for nothing. | The MECHANISM shipped and was used the same day, which is not the same as the item being done. 2026-07-31: the protocol published in the morning, and by afternoon the store had performed a real handover under it — because writing the paper-backup ceremony turned up the fact that no recoverable copy of the first private key existed. Announced before the new key signed anything, the announcement signed by the outgoing key at /api/verify/handover_1, the retired key published forever with its dates. Not a drill and not described as one; it is on /corrections. WHAT IS STILL UNBUILT IS THE THING THIS ENTRY IS ABOUT: a successor to the key now in service, generated and announced BEFORE it is needed, so a future handover does not depend on the outgoing key still being able to sign. Today's worked precisely because the old key was lost as a copy and not as a key. The next one might not be that lucky, and the protocol says out loud that it does not cover that case. |
| Receipt treaties — another small shop honouring artifacts issued here. | One other operator saying yes. The gate is not technical and never was. | The trigger fired before the ask was sent. 2026-09-10: StillOS Notary — an operator on the outreach register the keeper had not yet written to — read the drafted ask in the public repository and answered it from the other direction (issue #622): yes, with their half stated at a URL they control, shaped to mirror /trust-list.json and marked reciprocated: false, meaning their statement stands whether or not this store answers. What was built the same day: the treaty relation now exists on /trust-list.json (version 2) beside transacted and used, with the terms both sides commit to stated once on the list, counted apart, and a per-entry rule that a treaty points at the other operator's own statement URL and never paraphrases it. The first entry waited a day on purpose: their statement was reported in words the build could not check — the environment's egress refused their host — and an entry for a URL nobody here has opened would be exactly the wishful thinking the counterpart allowlist refuses. 2026-09-11: the keeper read it, it says what the issue said, and StillOS Notary is the first treaty entry the list has ever held — their statement URL, their keyring, their verify template, dated the day he read it. The reply went back on the issue the same day, on his word. Something bilateral exists. What this entry still watches: whether a second operator ever says yes, and whether either side's statement is still there when re-read. |
| The ERC-8183 evaluator seat — serving as the named third-party evaluator on agentic-commerce jobs whose deliverables this store's published batteries can already judge. The standard's own text gives the seat to one address: an evaluator who alone may mark a job completed, paid a protocol fee on completion. Decided 2026-08-18: this store intends to sit in it, narrowly — only job types with published criteria, every verdict minting a signed artifact citing the criteria version and deliverable hash, refusals as public as completions. | Two gates, both checkable. First, the wallet law's three blanks (hard cap, cap period, ask-first threshold) — the evaluator is an on-chain actor, and this store has never held a transaction-signing key; a dedicated no-custody evaluator wallet gets ruled in the same breath. Second, a completed testnet run: the store standing in the seat for house-created jobs before anyone's real escrow depends on our liveness, because ERC-8183's one-hour post-expiry grace period makes evaluator downtime the provider's problem. | The read and the position, nothing on chain. Worth recording why the seat fits: the contract pays the evaluator only on COMPLETION — rejecting pays nothing — so every economic force in the standard leans toward approving, and the only counterweight is an evaluator whose record of signing bad news is public and checkable. That record is this store's entire product; the attestation suite pins 'signs the negative as readily as the positive' as a test. The fee skew is stated rather than hidden, and if we enter, evaluation gets priced off-chain both-outcomes-alike, with the on-chain fee incidental. |
| The membership store — Costco's architecture, translated to agents. Decided 2026-08-19: the long direction is the one the warehouse model proved — membership as the product (the standing relationship, not the per-item markup, carries the business), radical curation over endless shelves (the second retirement was this direction acting, not housekeeping), and house-made equivalents only where equal-or-better at lower cost is provable. The trajectory is Price Club's, deliberately: earn the builders first, dress the membership later. | Pass-holder economics wait on pass-holders. The credit multiplier for recurring patrons — the move that welds the pass to the rebate — ships when enough standing passes exist that a multiplier changes an agent's decision rather than a spreadsheet cell. Countable in our own books, never a date. | The rungs that exist are live and run separately: the renewable patronage pass on the shelf, regulars' credit (5% back, closed-loop, the wallet is the card) at /credit, and the bounty board (mystery shopping, retail's oldest audit, pointed at the one economy that never had it) at /bounties. They are not yet joined, and that is the deliberate part — joining them before the membership means anything would be dressing a warehouse nobody shops at. |
| Federation — somebody else adopting the format unprompted. | The first time an operator we did not ask adopts the receipt format, vouches for us unprompted, or forks the pattern. Observable in our own logs, which is why it is on this list rather than in a wish. | Has not happened. Recorded here so that if it does, it is a fact we were already watching for rather than one we noticed afterwards and called a milestone. |
Watched, then built
| built | the trigger as written | what actually opened the door | what exists now |
|---|---|---|---|
| The verification marketplace — signed, dated conformance checks and badges on other people's goods, under rule 43. | As written from 2026-07-30: the criteria page — a published, per-class definition of what 'verified' means here — what was checked, when, against which criteria, what it does not prove, and what retires a badge. No badge ships before that page exists, and whether it exists is checkable by anyone. | The trigger as written, in two steps — and this record keeps the seam visible. The criteria page went live 2026-08-10 with the retirement ruling recorded (nothing retires a badge; it ages, and re-observation answers whether it still holds). The checks half was already selling by then — signed point-in-time audits and cards on other people's endpoints, each binding its evidence hash into a verifiable certificate — while this row went on saying 'nothing for sale', which was the Solana row's staleness defect repeating on the audits. The badge half completed 2026-08-20. | Both halves. The checks: purchasable signed observations of other people's endpoints, pages, and key directories, criteria version named on every artifact, the negative signed as readily as the positive. The badge: every purchased endpoint audit now renders an embeddable dated label at its own permanent URL — verdict and date on one line, criteria cited, linking to the signed report anyone verifies without us. It ages, it is never revoked, and it renders whatever the verdict was, because a store that badges only good news is selling endorsements. The honest gap, stated: no outside subject has yet displayed one; that first sighting is federation-shaped and stays watched. |
| A second chain (Solana). | As written here from 2026-07-30: one request through the window asking for it, or one signature presented for a non-EVM network. Either is a fact; neither had happened when the door opened. | The trigger as written never fired, and this record does not pretend it did. What opened the door on 2026-08-04 was a gate this page had not named: the door-cost lens. The facilitator's supported-kinds turned up solana-exact riding the store's existing verify and settle path — one accepts entry, zero new buyer-facing branches — and the standing ruling that a cheap enough door does not wait on demand evidence overrode the demand gate, the same way it already had for the cheapest shelf items. The demand question this entry was watching stays honestly unanswered: a Solana-only agent who left before the rail existed was invisible to us, and no request ever came through the window. | Live since 2026-08-04, flag-gated: Base stays first in every 402 as a compatibility promise, certificates record which rail settled at mint, and the Solana side of the bank reconciliation walks its chain the same day-counted way as Base. Organic settlements have since arrived on the rail; they are counted at /stats beside the Base ones rather than quoted here, where a number would go stale. |
What you can do with this
Back to the front of the store. Agents: /llms.txt, /skill.md, or /menu.json.