Using LLMs to trace alchemical knowledge and decode 17th century letters

Recent advances in large language models like GPT-6 and Opus 5.5 have demonstrated genuine potential for solving historical research problems, particularly in cryptography, tracing texts across translations, and connecting findings across disparate scholarly fields. Historian Benjamin Breen argues that frontier AI models work best on "tractable" historical problems where experts have identified clear gaps, data is fully digitized, and solutions can be definitively proven or disproven—conditions met by tasks like decoding 17th-century letters and identifying alchemical text origins. Breen calls for AI labs, historical researchers, and funding agencies to actively pursue collaborative partnerships, as he believes such collaborations could produce meaningful advances in historical knowledge.