General Media Company

Labs Pilot

Human-grounded factuality infrastructure for frontier AI

General Media converts journalist verification into structured eval, RLHF, and retrieval grounding data. Stringers turn raw signals into events, decompose them into typed facts, and complete verification tasks that produce confidence-scored claims and TruthBench-ready benchmark items.

Pipeline

The system is already a two-sided marketplace: stringers submit structured news events and complete labeling tasks; labs consume the artifacts through /api/v1. This page reframes that flow as factuality infrastructure: from raw signal to confidence-scored facts to TruthBench eval items and deliverable data-assets.

End-to-end factuality flow
Mini Map
Where TruthBench & knowledge-gaps fit

TruthBench converts verified events into eval items and measures hallucination rates over time. Knowledge gaps (collection: knowledge-gaps) capture systemic failure modes, then commission targeted stringer tasks and new data-assets.

90-Day Pilot

Factuality pilot (enterprise walkthrough)

Verifier network
2,000 journalist-verifiers
Claim throughput
100,000 verified claims
Domains
Politics · Local · Public health · Finance · Regulation
Outputs
  • Factuality labels (factType + confidence) at claim granularity.
  • Citation-validity labels (primary-source present, supports claim, timestamped).
  • Verifier agreement scores (inter-verifier consistency + editor adjudication).
  • Temporal freshness scores (how quickly facts decay vs. update cadence).
  • Benchmark items (TruthBench-ready eval items and slices).
  • Preference pairs (model outputs ranked vs. grounded reference).
  • Evaluator traces (evidence chain + decision rationale + error taxonomy).
  • Retrieval-grounding datasets (queries ↔ citations ↔ verified facts).
Delivery surface

Labs receive structured artifacts as an API feed or packaged assets. Internally, the workflow uses collections like events, tasks, truth-bench-runs, knowledge-gaps, data-assets, and licensing-offers.

API surface
/api/v1/*
{
  "events": "GET /api/v1/events/query?status=verified&minConfidence=0.85&limit=50",
  "firehose": "GET /api/v1/firehose  (SSE stream of updated events)",
  "truthbench": "GET /api/v1/truth-bench-runs",
  "knowledgeGaps": "GET /api/v1/knowledge-gaps",
  "dataAssets": "GET /api/v1/data-assets",
  "licensing": "GET /api/v1/licensing-offers"
}
Quality gates

Pilot deliveries gate on agreement, citation validity, and temporal freshness. Low-confidence or disputed facts remain in factType states that preserve uncertainty rather than forcing false certainty.

Live Product Mock

Mocked but schema-faithful artifacts. Shows a single signal promoted from queue into an event with facts[], journalist verification via tasks, and final delivery as a data-asset.

Raw queue item
queue
{
  "id": "queue_6f3c1f",
  "snippet": "New York DFS posted draft guidance on AI use in insurance underwriting and opened a public comment period.",
  "sourceOrigin": "gov_filing",
  "sourceUrl": "https://dfs.ny.gov/publications/ai-underwriting-guidance-draft.pdf",
  "priority": "high",
  "queueStatus": "converted",
  "suggestedCategory": "economy",
  "suggestedRegion": "northeast_us",
  "enriched": true,
  "extractedClaims": [
    "The New York Department of Financial Services released draft guidance on AI underwriting on 2026-05-12.",
    "Public comments are open for 30 days from publication.",
    "The guidance requires documented governance, audit trails, and fairness testing for models."
  ],
  "extractedEntities": [
    {
      "name": "New York Department of Financial Services",
      "type": "organization"
    },
    {
      "name": "Insurance underwriting",
      "type": "domain"
    },
    {
      "name": "Model governance",
      "type": "concept"
    }
  ],
  "aiSummary": "NY DFS released draft AI underwriting guidance and opened a 30-day public comment period for insurers and vendors.",
  "createdAt": "2026-05-12T18:22:00Z"
}
Structured event
events
{
  "id": "event_1029",
  "headline": "New York DFS releases draft guidance on AI underwriting and opens 30-day comment period.",
  "eventType": "policy",
  "category": "economy",
  "impact": "high",
  "eventStatus": "verified",
  "occurredAt": "2026-05-12T14:10:00Z",
  "location": {
    "placeName": "New York, NY",
    "city": "New York",
    "stateProvince": "NY",
    "country": "US",
    "region": "northeast_us"
  },
  "facts": [
    {
      "statement": "The New York Department of Financial Services released draft AI underwriting guidance on May 12, 2026.",
      "factType": "verified_fact",
      "source": "NY DFS draft guidance PDF (publication notice)",
      "sourceType": "official_document",
      "sourceUrl": "https://dfs.ny.gov/publications/ai-underwriting-guidance-draft.pdf",
      "confidence": 0.93
    },
    {
      "statement": "The draft guidance specifies a 30-day public comment period from publication.",
      "factType": "verified_fact",
      "source": "NY DFS draft guidance PDF (comment period section)",
      "sourceType": "official_document",
      "sourceUrl": "https://dfs.ny.gov/publications/ai-underwriting-guidance-draft.pdf",
      "confidence": 0.93
    },
    {
      "statement": "The guidance requires documented governance, audit trails, and fairness testing for model use in underwriting.",
      "factType": "verified_fact",
      "source": "NY DFS draft guidance PDF (requirements section)",
      "sourceType": "official_document",
      "sourceUrl": "https://dfs.ny.gov/publications/ai-underwriting-guidance-draft.pdf",
      "confidence": 0.93
    },
    {
      "statement": "A senior regulator said final guidance may arrive in Q3 2026 after stakeholder input.",
      "factType": "attributed_claim",
      "source": "Recorded press call with DFS spokesperson (verbatim quote)",
      "sourceType": "named_source",
      "confidence": 0.8
    }
  ]
}
facts[] decomposition
facts[]
[
  {
    "statement": "The New York Department of Financial Services released draft AI underwriting guidance on May 12, 2026.",
    "factType": "verified_fact",
    "source": "NY DFS draft guidance PDF (publication notice)",
    "sourceType": "official_document",
    "sourceUrl": "https://dfs.ny.gov/publications/ai-underwriting-guidance-draft.pdf",
    "confidence": 0.93
  },
  {
    "statement": "The draft guidance specifies a 30-day public comment period from publication.",
    "factType": "verified_fact",
    "source": "NY DFS draft guidance PDF (comment period section)",
    "sourceType": "official_document",
    "sourceUrl": "https://dfs.ny.gov/publications/ai-underwriting-guidance-draft.pdf",
    "confidence": 0.93
  },
  {
    "statement": "The guidance requires documented governance, audit trails, and fairness testing for model use in underwriting.",
    "factType": "verified_fact",
    "source": "NY DFS draft guidance PDF (requirements section)",
    "sourceType": "official_document",
    "sourceUrl": "https://dfs.ny.gov/publications/ai-underwriting-guidance-draft.pdf",
    "confidence": 0.93
  },
  {
    "statement": "A senior regulator said final guidance may arrive in Q3 2026 after stakeholder input.",
    "factType": "attributed_claim",
    "source": "Recorded press call with DFS spokesperson (verbatim quote)",
    "sourceType": "named_source",
    "confidence": 0.8
  }
]
Confidence calculation
confidence
{
  "factType": "verified_fact",
  "sourceType": "official_document",
  "inputs": {
    "sourceWeight": 0.95,
    "factModifier": 1,
    "stringerTrustScore": 0.82,
    "weights": {
      "sourceType": 0.4,
      "factType": 0.3,
      "stringerTrust": 0.3
    }
  },
  "formula": "(sourceWeight*0.4) + (factModifier*0.3) + (stringerTrustScore*0.3)",
  "confidence": 0.93
}
Assigned journalist task
tasks
{
  "id": "task_18b2",
  "title": "Verify DFS AI underwriting guidance: comment period + requirements",
  "taskType": "source_verify",
  "assignmentClass": "verification",
  "beat": "economy",
  "region": "northeast_us",
  "priority": "high",
  "status": "in_progress",
  "slaMinutes": 180,
  "pay": 45,
  "assignedTo": {
    "id": "stringer_44d1",
    "displayName": "Senior verifier (NY/finance)",
    "tier": "senior",
    "reputation": {
      "trustScore": 0.82
    }
  }
}
Delivered data asset
data-assets
{
  "id": "data_asset_501",
  "title": "Verified Claims Feed (Pilot) — Regulation / Finance — May 2026",
  "assetType": "verification_dataset",
  "sellerOrg": "General Media Company",
  "rightsSummary": "Pilot license for internal model evaluation + training; includes citations and agreement scores.",
  "coverageRegion": "US (multi-region)",
  "structuredness": 0.92,
  "availabilityStatus": "negotiating"
}