Archive claim check
A paid API over two things: a corpus of free-energy, anti-gravity and anomalous-heat literature, and the record of a forecasting engine. The archive queries tell you which claims have a published null, which have an ordinary artefact explanation, and which have a preregistered test waiting. The forecast queries tell you what a model predicted, what happened, and how well it has scored.
Payment is x402: call the URL, get a 402 with the
payment terms, pay, call again. Every query is $0.01. An id that does not
exist returns a free 404 — you are never charged to learn that something is missing.
What is in it
| Archive claims served | 9867 | of 24660 in the corpus |
| Searchable | 8319 | community and self-published sources are reachable by id, never by search |
| In the contradiction map | 125 | these carry a class: artefact, ledger-forbidden, still open… |
| Citation only | 246 | paywalled sources: citation, never a quote |
| Withheld | 204 | one document is marked private research and is excluded entirely |
| Forecast questions | 16099 | 8648 open · 5317 resolved |
| Arms scored | 33 | all shadow or paper; no money was staked |
The queries
GET /v1/archive/claim/:id $0.01
One claim, with its source, a short verbatim quote, its provenance, and the three answers: published null, artefact explanation, preregistered test.
curl "https://x402-archive.claimcheck.workers.dev/v1/archive/claim/C001"
GET /v1/archive/search?q= $0.01
Search claim text. Up to 10 stubs; fetch any of them in full with a claim call.
curl "https://x402-archive.claimcheck.workers.dev/v1/archive/search?q=rodin%20coil"
GET /v1/archive/device/:slug $0.01
Every mapped claim about one device, rolled up — the Rodin coil, an over-100% inverter, a Ni-H cell.
curl "https://x402-archive.claimcheck.workers.dev/v1/archive/device/rodin-coil"
GET /v1/calibration/question/:qid $0.01
A resolved question: what each arm forecast before it closed, what actually happened, and each arm's log score and Brier.
curl "https://x402-archive.claimcheck.workers.dev/v1/calibration/question/q_fac4d36ab73d0ec3"
GET /v1/calibration/arm/:arm $0.01
One arm's whole record: denominators first, then mean log score, Brier, skill against the base rate, and the calibration curve.
curl "https://x402-archive.claimcheck.workers.dev/v1/calibration/arm/arm1%3Arefclass-bayes%401.0.0"
GET /v1/live/question/:qid $0.01
An open question: the current probability, which arm produced it, every other arm that answered, and how old the forecast is.
curl "https://x402-archive.claimcheck.workers.dev/v1/live/question/<open question id>"
Why you could trust an answer
A probability is worth nothing from a stranger who could have changed it after the outcome. So each answer carries a Merkle inclusion proof against a published root, and the roots file is anchored with OpenTimestamps into Bitcoin. That proves the answer existed, unedited, before the anchor — not merely that we say so now.
| store | rows | root |
|---|---|---|
archive-claims | 9867 | 0982856aa582799f3d8ac314feb74c33b75f37d9273778b2ba14d842c94e1c6d |
forecast-questions | 16099 | 0185d1fff3ed907ec369f29eaee9e85d5ec0bd06373c0abc2c264ad19ada1acd |
forecast-arms | 33 | 9cfb3779688efecec220b449061c984942214e7d33c0dee938141408bf65fb0a |
Roots published 2026-11-07T17:26:59Z. Free:
GET /v1/roots and GET /v1/roots.ots (the receipt).
Verify with node scripts/verify-answer.js answer.json.
What it does not prove: that a number was our honest best estimate. Nothing can. It proves the number was fixed before the timestamp, and every answer states its own age and its anchor's age.
What is deliberately not here
No sportsbook prices, no vendor market data, no betting history, no stakes: the forecasting arms run in shadow and on paper, and the market series they are scored against belong to their vendors, so they are withheld. One archive document marked private research is excluded whole. Paywalled sources get a citation, never a quote.