Module 10 — The claim map
Every card of the project's knowledge base, grouped into its areas, with what it says, what its status is, and what evidence it rests on. Nothing is hidden from this map: a card that has never been through a referee pass is shown and marked as such, and a piece of evidence that lives in the private companion is shown and marked as held privately, rather than linked to a page that would not resolve.
proved and refereed are complete structured proofs, the second having survived an independent pass. certified is a bound by construction (a spectral-window lower bound, or an enclosure of the compressed problem); the defect-matrix quadrature behind it is not certified. numerical is a measurement with an error bar — never a theorem, and never styled like one. conditional holds only under the hypotheses named on the card. conjectural and open are unsettled. refuted is kept because the record of what failed is part of the result. verified, active, done and superseded mark lessons, decisions and tools rather than mathematical claims.
Three further marks are about the record, not the mathematics: never refereed on a card with no verdict anywhere in its record, awaiting a referee pass on one whose review line says pending — which is not the same thing, since its history may already hold several verdicts — and last verdict failed where the most recent pass failed and the repair is not yet recorded.
The knowledge base itself is not published: it lives beside the repository, and
this page is generated from it by
tools/build_site_data_extra.py
into docs/data/extra-claim-map.json. Whether a piece of evidence is public is decided
at build time by asking both repositories over this one work tree what they track:
git ls-files for the public one, and the same question of the private companion's
.git-private. Tracked publicly gives a link; tracked privately gives
held privately and no link; tracked by neither gives
not committed yet. A file being present on the machine that
built this page says nothing about whether it ships, which is why the question is asked of git and
never of a list written by hand.
The same claims, with the same statuses and one page each, are also generated
as documentation pages: every claim by area ·
what is open · the documentation index.
Each is rendered from a Markdown file of the same name under docs/, which is what GitHub
shows when you browse the repository.
Clicking a dot opens its card. Every card is also in the list beside the graph, which is reachable with the keyboard.
Select a claim.
What this map is for
Modules 1 to 9 each carry a handful of claims. This one carries all of them, including the ones no module displays: the open questions, the lessons learned, the decisions, and the cards that record something that turned out to be false. A reader who wants to know what this project believes, and how firmly, should be able to find every one of them in a single place, with the same status vocabulary throughout and no card quietly left out.
Which cards those are is decided by one rule, the same rule the generated Markdown pages use: every
card of the knowledge base's claims/ folder, and every card of its evidence/
folder that has reached a displayed status — a finding promoted to a result. The findings that
have not are audit records of what an adversarial pass found in a document, not claims about the
mathematics; their number is in the count above, and the claims they support carry them as evidence.
Holding both surfaces to one rule is the point: a card on one and not the other would be a card
quietly left out.