'Computational Historiography' — using AI as a digital co-historian.
The engine describes it as a framework for machine-reading history at massive scale.
It combines Schmidt's AI2050 effort and the Seshat historical databank with pipelines built from OCR, the LLaMA 3 70B model, and named-entity recognition to automatically extract patterns from 10,000 years of data.
It matters as a tool that lets machines find historical patterns across a sweep of time far beyond human reach.