Epstein Files Engine: Agentic Search for Investigative Journalism
4.60T1 sourcearXiv cs.HC
Source record
Published by arXiv cs.HC (T1 source). The original is at https://arxiv.org/abs/2609.30611.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryThe New York Times built the Epstein Files Engine, an AI agent that translates reporter questions into BigQuery SQL across DOJ Epstein releases, the Times archive, and external news headlines. Over 100 journalists used it, contributing to at least 20 published stories. The paper details query patterns, a Diff text-and-visual duplicate-matching method, and argues newsroom agents work best as interfaces to source material rather than autonomous writers.
Why it mattersFirsthand account of a high-stakes newsroom agent deployment with concrete architecture, query logs, and a principled stance against autonomous AI writing. Useful template for investigative and document-heavy workflows.
Cited by
No citations on record.
