Casey Handmer audits 150,000 astrophysics papers with Opus 5.5, finds ~1,000 cited errors

TheLivingstoneBIG286 · reddit · 2026-11-17

Casey Handmer used Opus 5.5 to audit theoretical astrophysics: indexing 150,000 papers, deep-auditing 50,000, and finding nearly 1,000 major errors that were cited many times and mostly remain uncorrected. He estimates peer review is poor at catching errors, and compiled dozens of open questions anyone with spare Claude Max tokens could answer (context .md files included).

His takeaway: without any further model progress, AI can speed up quantitative science by >1000x — but universities show no sign of pushing this, while founder types are "speedrunning" entire disciplines. He argues academic publishing is now effectively dead as a unit of scientific record.

Related event: AI Audit of Astrophysics Papers Finds Nearly 1,000 Repeated Citation Errors(2 posts)→

Original post →

More from Models

Models channel →