We built pursueindex.com to make the DOW PURSUE corpus actually searchable. The official portal is a DataTables widget. You can browse one card at a time. You cannot search the text of the documents.
Our pipeline OCR'd all 4,153 pages, pinned the manifest to a SHA-256 hash, and serves full-text search with mandatory per-page citations. Every claim traces to a specific page of a specific war.gov PDF.
The Muroc entry is the one that stopped us: on July 8, 1947, while Roswell was publicly announcing and retracting a "flying disc" recovery, five Air Force personnel at Muroc gave sworn statements about two separate disc sightings hours apart. The FBI investigation file sat in 62-HQ-83894 from 1947 until declassification in 2007. Reading guide with quoted excerpts and per-page citations: https://pursueindex.com/finds/muroc-1947
Limitations up front: OCR on mid-1950s FBI carbon copy is imperfect even with the LLM fallback. Methodology and known failure modes are documented: https://pursueindex.com/methodology
This is Release 01, 161 cards. More tranches expected. Corrections welcome via GitHub issues.
by Narrow_Market45