Method & Proofed API

Architecture

1. 01Signal Ingestion Layer
2. 02AI Event Clustering
3. 03Atomic Claim Extraction
4. 04Source Provenance & Lineage Engine
5. 05Research Agent
6. 06Evidence Database
7. 07Sceptic Agent
8. 08Proof Scoring Engine
9. 09Risk Classification
10. 10Human Review Gateway
11. 11Article Generation
12. 12Publication
13. 13Continuous Re-Verification
14. 14Immutable Revision Ledger

Verification states

⚪ UNPROOFED

Detected but insufficient evidence exists.

🟡 DEVELOPING

Independent evidence is accumulating.

🔵 CORROBORATED

Multiple genuinely independent sources support the claim.

🟢 PROOFED ✓

Strong primary or independently corroborated evidence, and adversarial review found no material contradiction.

🟠 DISPUTED

Significant credible evidence conflicts with the claim.

🔴 DEBUNKED

Strong evidence demonstrates the claim is materially false.

States are never silently altered. PROOFED → DISPUTED is always possible when new evidence appears. Accuracy outranks embarrassment.

Source hierarchy

There is no universal rule that government or newspapers are true. Sources are weighted for the specific claim: rail infrastructure data for a cancellation, meteorological observations for wind speed, Companies House filings for company finances. A police statement is authoritative for what the police officially state — allegations remain allegations until established.

Human escalation

Future Proofed API

The verification system is architected so third parties can eventually submit text, a URL, a social post, an image, a video or a public statement for verification.

POST /verify
{
  "input": "text | url | social_post | image | video | statement"
}

200 OK
{
  "claims": [
    {
      "id": "CLAIM-001",
      "text": "...",
      "status": "PROOFED",
      "proof_score": 94,
      "supporting_evidence": [...],
      "contradictory_evidence": [...],
      "source_lineage": { "independent_lineages": 8, "tree": [...] },
      "last_checked": "2026-08-08T09:51:00Z"
    }
  ]
}