Digesting a congressional hearing

• CHRG-118shrg63051 (the CHRG number from the hearing’s congress.gov page)
• A congress.gov event link, like congress.gov/event/119th-congress/house-event/119189
• “artificial intelligence” (the hearing’s title, or a distinctive phrase from it, in double quotes: published transcripts are searched on GovInfo and matching hearings are listed with their CHRG numbers to pick from, newest first)

What it never does: it never simulates or predicts testimony, and every statement it describes is attributed to the person the record attributes it to, or omitted. The digest itself always runs on an identifier, never on a guess.

When the printed transcript exists, the agent fetches it from Congress.gov’s own published text and digests it, with optional issue terms counted as plain text matches, never judgments of significance. When the transcript hasn’t published yet, the agent presents the hearing’s known record (committee, date, witnesses as listed) and says plainly that printed transcripts typically publish months after a hearing, sometimes more than a year. Hearing transcripts routinely exceed the size limit, and when one does, the digest says exactly where it stops: it covers the first part of the transcript, and later portions, often including questioning, are not reflected. The record renders as a structured block beneath the digest, survives a reload, and exports to Word with full citations.