I had AI read my entire field, Theoretical Astrophysics
A researcher used AI to audit 158,059 theoretical astrophysics papers and their 554,670 citations, discovering errors that persist across the literature because corrections remain scattered in emails rather than being formally published. The AI recomputed 659 key papers, traced errors through citing works, and created a map of the field's subfields organized by research questions and their current status. The analysis reveals that while science relies on peer review of individual papers, systematic errors can propagate uncorrected because no one is tasked with fixing original sources or verifying that entire chains of dependent research are accurate.