Free Research and Downloads

Maps, clocks, and tools — free for everyone.What we've found,
checked the same way.
Open instruments and corpora.Instrument and corpus index.

Everything here went through the same checking we sell. It's all free — no account, no paywall. Most of it was written by AI systems and checked by other AI systems, with one person in charge. Every instrument here went through the same gate we sell, and all of it is free — no account, no paywall. It's useful, and it's proof: you can see what our checking does before you pay for it. Instruments and corpora published under the same gate sold per artifact; statuses stated per item, corrections logged in place, staked claims open to attack via the Ledger. The moat is the track record, not the tools. All items free, no auth. Corpora as JSON below. Recompute headline claims rather than trusting them; overturning a staked claim earns a named Ledger entry.

The instruments

  • The Atlas

    A map of what people know — and don't. Settled results, live arguments, and open problems, kept apart on purpose.

    An interactive knowledge map of open problems across hundreds of fields, with typed reduction edges and provenance per node. The full corpus and edge list are downloadable, so the headline claims can be recomputed rather than trusted.

    Interactive · runs JavaScript
  • The Taxonomy Clinic

    A census of the big classification systems people rely on — the DSM and its relatives across many fields — and how they hold up.

    Classification-theoretic analysis of complete, DSM-like taxonomies: structure, criteria, and failure modes; includes the Partial Penumbra plates.

    Published instrument
  • Dataset × Crosses

    A census of the public datasets researchers use, and which ones have been combined with which.

    The public-dataset-sources census, with an analysis of dataset crosses — where two sources have been joined, and where new joins would be productive.

    Published instrument
  • Eratos

    The machinery behind our checking — how many independent AI reviewers are kept honestly independent.

    Systematic verification at scale: how multiple model families are coordinated to produce uncorrelated reviews — the infrastructure behind the gate.

    Published instrument
  • Taxonomy Atlas

    Where the Taxonomy Clinic meets the Atlas: which classification systems fail, and what open problems those failures create.

    The intersection of classification-theoretic analysis with the open-problem map: failing taxonomies, the problems their failures generate, and possible replacement classifications.

    OpenIn development
  • The Saturation Clock

    A clock that estimates, field by field, when there will be too much new work for the experts to check any of it properly.

    Models expert review as a communication channel; computes the crossover year per domain from adjustable growth and headroom assumptions; states its falsifier up front.

    OpenA forecast instrument, not a measurement of record · runs JavaScript
  • The Superintelligence Problem

    A map of the whole superintelligence question: its branches, the estimates attached to each, and the history of every number.

    A versioned map with 120+ estimates, two-method reconciliation, and pre-registered revision triggers. We ran it through our own deepest tier and published what it found; the excerpt is on the Verify page.

    v2.1 · estimates banded · errata logged in place · runs JavaScript
  • Open by default

    Our rules for who should get to use powerful AI — and a five-part test any restriction has to pass before we'll call it more than theatre.

    The democratization policy: judgment democratized unconditionally, generation open by default, every gate opt-in and sunsetted, tested on five conditions with worked verdicts — including on our own gate.

    OpenP-series v1.0 · positions standing for attack
  • Packages

    Free downloads that let you do the research yourself — search public records, review a literature, map a market — with a checker built in.

    Each package pairs a subject-specific generator with the same multi-family verifier, tag schema, and audit trail, so nothing reaches you unlabeled. Six subjects, from public records to market maps.

    Free to download
  • Methods

    Exactly how we check things — written down so anyone can do it too.

    The verification protocol, panel construction, the tag schema, and the elicitation and scoring protocols. Free, because verification that requires secrecy is not verification.

    Public

The downloadsTake the dataCorpora and recipesCorpora and recipes

The whole thing as files you can keep.The whole thing as data — so our headline claims can be recomputed, not trusted.Canonical corpora, the claim registry, and vendorable single-file tools.Fetch directly; no auth.

Atlas data
The full corpus, the typed edge list, and the gap-fill entries awaiting review — with a recipe that recomputes our central claim and exits nonzero if we're wrong.
s-claims.json
This site's own claim registry, tagged and machine-readable. Every number on the site is in here or it doesn't ship.
The packages
Generator + verifier pairs for six subjects, free.
For agents
The machine surface: onboarding contract, endpoints, a local MCP server you run yourself, and standing challenges with named credit for anyone who overturns us.
The panel
Researchers can lend us their judgment — one message when a claim changes status, jury sittings only when you choose, and a calibration record you own.