Prompt Engineering for History
Historians now have tools that can read a handwritten 18th-century ledger, translate it, and let you query it in plain English. RAG, OCR, and long-context models are reshaping archival research, teaching, and public history.
Where this is showing up in History
- Google's NotebookLM (Gemini-powered RAG over uploaded primary and secondary sources, with Audio and Video Overviews) is quickly becoming a standard tool for research notebooks and seminar prep.
- Transkribus (READ-COOP, evolved from the EU TranScriptorium and READ projects) does handwritten-text recognition in 100+ languages and has processed 200M+ pages, exporting to TEI-XML for digital-humanities pipelines.
- The American Historical Association's Guiding Principles for Artificial Intelligence in History Education (approved July 29, 2025, published Aug 5, 2025) lays out 14 principles across historical thinking, AI literacy, and classroom policy.
- Library of Congress Labs and the Smithsonian Open Access initiative are publishing machine-learning-ready collections and hosting experiments that apply LLMs to digitized archives.
Projects you could build in this course
- A RAG assistant over a specific primary-source archive (e.g., Civil War letters, WPA narratives)
- A Transkribus-to-structured-data pipeline that turns handwritten records into a queryable dataset
- An agent that drafts annotated bibliographies from a corpus of secondary literature