Anthropic shared extensive nonpublic AI R&D data with METR, incl. algorithmic progress estimates
METR_Evals · x · 2026-11-06
METR disclosed the evidence base behind its AI R&D acceleration assessment: Anthropic shared extensive data across its pipeline, from pretraining to RL, including key details like estimates of algorithmic progress over time. METR says this nonpublic evidence was highly informative.
A shared chart breaks down what was shared vs. made public: internal model benchmark capabilities (both), key intermediate measures, algorithmic progress estimates, evidence of AI's contribution, R&D input measurements, bottleneck details, and staff interviews — most shared with METR only, with the most risk-relevant findings redacted from the public report.
Related event: METR Report: Anthropic Internal AI R&D Nearly Doubles in Speed(56 posts)→
More from Companies & People
- ETH Zurich releases full open-source VLSI course: 12 lectures plus OpenROAD-based exercises — Obijuan_cube · 2026-11-06
- Mystery OpenAI gift card with giant smiling face shows up at AI researcher's doorstep — Yuchenj_UW · 2026-11-06
- Analyst: "Anthropic is after our data" is becoming the go-to grifter pitch — teortaxesTex · 2026-11-06
- Physical AI meetup in India: 4 robotics founders on real factory deployments and eval loops — carrycooldude · 2026-11-06
- Boom founder: only ~5 of 165 staff work on supersonic jet as energy comes first — dunkhippo33 · 2026-11-06
- xAI inference and kernel engineer departs, citing SGLang, attention kernels and C++ inference stack rewrite — HanchungLee · 2026-11-06