Claude Opus 5.5 audits 150,000 astrophysics papers, finds ~1,000 major errors
zakkohane · x · 2026-11-18
CJHandmer used Opus 5.5 to audit his old field, theoretical astrophysics: indexing 150,000 papers, auditing 50,000, and deeper-checking thousands more. The sweep found nearly 1,000 major errors that were cited many times, most still uncorrected — and let him estimate peer review's error-correction efficacy ("not very good"). He also compiled dozens of open questions with .md context files anyone with Claude Max tokens could tackle. Bottom line: AI can speed up quantitative science >1000x with no further model improvement, but he sees no evidence academia is acting on it.
More from AGI Musings
- Chris Albon: Offload to AI what makes you better — never what you want to become — chrisalbon · 2026-11-18
- VC distills African AI startup playbook: narrow wedge, labor replacement, 10-100x savings — prasanna_says · 2026-11-18
- Speculation Swirls That OpenAI's 6.2 Astra Leans on Looping Transformers, Raising Monitorability Concerns — scaling01 · 2026-11-18
- If autonomous AI can own capital, humans may no longer own the productive economy — PierceLilholt · 2026-11-18
- alth0u: a huge capabilities-to-utility overhang separates model power from user needs — yacineMTB · 2026-11-18
- Podcast: What if Machiavelli time-travelled to today, with Ada Palmer — Ben_Reinhardt · 2026-11-18