sgit newsroom v0.1.29 · snapshot 2026-09-24

Reading room · riskmandate.ai

Reading room / riskmandate.ai · raw text · live ↗

From riskmandate.ai, the page as fetched on 2026-09-24 · open the live page ↗Everything on this sheet is the source site's own text; the newsroom's chrome is outside it.

The reviewed level, and the people who run it

The top level is the one thing on this site that is somebody's time rather than a pipeline, and until today the site said almost nothing about it: no workflow, no expectation on either side, and no name. A brief arrived asking for both halves — what the service is, and who does it — and this is them. It is registered as D11↗.

The reviewed level↗ is the service, written down before it has been sold. Eight steps in order, each saying whose move it is: read a published example free, buy at the store and pick your reviewer, a person replies within a day, half an hour to find out what is actually running, the vault built from the interview rather than from a form, the reviewer's correction committed with a note of what changed and why, half an hour to walk through what came back, and a sign-off file carrying a name and a date. Beside it, what we ask of you — an hour, two people, a list of what the agent connects to — and what we never ask for: credentials, access, a scan, or your data. Every figure on the page names the page it was read off and the date, and the page opens by saying that nobody has bought one yet and the payment link has not been issued: the store's own page said so when it was read on 22 September 2026.

Who runs your review↗ is the list, and the rule that goes with it. Choosing the reviewer is part of buying this level, and the choice is made on facts rather than on a rating: nobody here is scored, ranked or starred, and nobody will be. What is published about a person is what that person has published themselves — cited, with the date it was read — or what they write for us and sign off, marked as theirs. A reviewer with no published page gets no rows. There are no testimonials anywhere on the site, because there is no real one to publish; when there is, it will carry a name, a date and permission.

Two pages, one of which is not a person. The first reviewer↗ is built from three published sources — the author pages on sgit.ai and open-source.sgit.ai, and the store's own page for the same person — with the interests declared on the page rather than in a footer. The second↗ is a labelled placeholder: the shape a reviewer page takes, so that the next professional can read exactly what would be published about them before agreeing to any of it. It says so in the eyebrow, in a note at the top, on its card in the list and in the manifest. Nobody is described on it.

One source, because two sites publish about the same living person. The store's /who/ pages carry the same rule, written first there; the risk is not the rule but the copy. So the reviewers are a generated family from one file each under site/reviewers/, and reviewers.json↗ publishes the same data as a manifest for the store's chooser to build from. build-reviewer-pages.mjs writes the pages and the manifest, refuses to render a row whose source is not declared, and now runs in CI and in npm run check. The arrangement is written up for the store team as a brief of its own↗, including the one thing a chooser must do: never offer kind: placeholder as an option.

Pricing is a group now: the four levels, the reviewed level, and who runs it.

What the brief asked for and has not got is on the register beside what it did: a second named reviewer, the chooser at checkout, comments and testimonials, any published availability — and what the reviewer is paid, which was agreed in the brief and deliberately not published. A commercial term, and the lead's call rather than ours.