Build cited ancestor records from scanned family papers
At a glance
Alex Jeannopoulos uses Claude to turn multilingual family scans into reviewed records with explicit unresolved questions. This workflow is listed under Family History, Document Organization, Research.
- Agents / tools
- Claude, Local files, Markdown
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Give a file-capable Claude workspace access to a batch of scanned family PDFs and a durable archive folder. 2) Have it rotate sideways pages, copy the scans into the archive, and create INDEX.md with one row per document. Process PDFs in batches of four. 3) Maintain one Markdown file per ancestor with documented facts, source citations, spelling variants, open questions, and a last_distilled date. Cross-reference conflicting dates and record which document takes precedence and why. 4) Read every generated person page yourself. When a record changes materially, generate a public version containing only reviewed, double-sourced facts and retain uncertainties in the working files.
How it went
The author reports about 25 person pages across five generations and a naturalization date established from three independent papers.
Worth knowing
The author's /archive-import and /distill-to-site are custom routines. Their behavior is described above; the article does not supply their implementation.