SCOPE AND STANDING

Everything this platform does, and how much of it you should rely on

This page is generated from the route tree every time it is served. It is not a selection. 401 capabilities exist here and all 401 are counted below, including the ones that would be more flattering to leave out.

The layer is a test, not a preference. Instrument means the figures come from measured inputs carrying a source and a date. Record means it settles a claim or reports on that settlement. Workshop means neither — honest work that is not load-bearing. Anything the test cannot classify fails the build rather than being filed under a default.

Every figure sits next to the thing that could falsify it

Walk it in order. Two of these six links exist to show you where we have been wrong, and the chain is only worth reading because they are on it.

01A reading
The figure, every input behind it, and each input's source, observation date and evidence tier.
02What would refute it
The settlement contract: the date, the source and the condition.
03Whether it has been right
Settled forecasts against their own base rate, not against nothing.
04Where it has been wrong
The register of false statements this platform has published, each with the test that now prevents it.
05What is failing now
Present-tense process, evidence-store and source freshness checks, with gaps shown explicitly.
06Which code produced it
The reported release, commit and digest, together with the working-tree state.

What the count actually is

161

INSTRUMENT

RECORD

WORKSHOP

111

INTERNAL

Pages and API routes together. The workshop layer being large is not a defect — it is what an unlabelled platform would have shown you as product.

Instrument

68 pages

Figures from measured inputs. Every one carries its source, the date it was observed and an evidence tier, and an input that is absent is reported as absent rather than as zero.

Record

12 pages

What settles the instrument's claims, or reports on that settlement. This is the layer that can embarrass us, which is the only reason the one above is worth reading.

Workshop

62 pages

Honest work that is not load-bearing: simulations, experiments, syndication and views built on top of the instrument rather than measuring anything themselves. Named as such, because an experiment shown beside a measurement is a claim about the experiment.

Internal

Not addressed to a reader: scheduled jobs, operator tooling, authentication and payment plumbing. Listed so the count is complete, not because anybody should click. 7 pages and 104 API routes.

The full manifest, with the test applied to each capability, is served as JSON at /api/capabilities. The reported commit, source digest, and working-tree state are at /api/provenance. A clean-clone reproduction claim applies only when that endpoint reports a clean release tree.