API

The machine-readable layer is first-class here, not an afterthought: every page has a JSON twin and a Markdown twin, rendered from the same query layer as the HTML, so they cannot drift apart.

URL grammar

Every node has exactly one canonical page at /<kind>/<slug>--<id8>. The slug is cosmetic — lookup is by the eight-character id suffix only, so a label change never breaks a link (a stale slug 301s to the current one). Documents live at /source/<id8> with no slug.

Three query parameters work on every content page and are honoured server-side:

Twins

Append .md to any content page for the same data as Markdown — a Question, for instance, is also /question/<…>.md. The JSON equivalents are below.

Endpoints

EndpointReturnsParameters
/api/v1/question/{id8} A Question: every branch, its share, its evidence, the residual. lens, asof
/api/v1/claim/{id8} A claim: standing with the full term breakdown, tree, voices, sources. lens
/api/v1/entity/{id8} An entity: involvement, statements made, resolution doubt. lens
/api/v1/event/{id8} An event: event-time with its basis, participants, claims.
/api/v1/source/{id8} A document: what it claims (verbatim spans), syndication, runs.
/api/v1/story/{id8} A Question's neighbourhood: questions, cast, clock, ranked worlds. lens
/api/v1/world/{id8} An assembled possible world, including what it rules out. lens, w
/api/v1/changes The what-changed log, filterable. kind, question, page
/api/v1/search Search across every kind, with each hit’s relationships. q, kind
/api/v1/suggest Type-ahead suggestions. q
/api/v1/peek/{id8} A one-line preview of any node. lens

Machine-readable description: openapi.yaml. Every endpoint is read-only; there is no write API, and there will not be one before the moderation policy exists.

Feeds

Lenses

log-odds@1 — Evidence-weighted
Belief computed from the support/challenge graph in log-odds: independent legs add, challenges subtract, restatements add nothing.
structural@1 — Structural
No probabilities at all — only how well tested each branch is: evidentiary legs, independent sources, challenges. Untested and contested are different shapes, not different numbers.

If you are quoting this site

Quote the residual with the share. "34% under the evidence-weighted lens, with 49% unknown" is the claim this site makes; "34% likely" is not. Name the lens, do not turn leading into established, and remember that repetition moves nothing here. llms.txt says the same thing to agents.