METR report: Anthropic's AI-driven AI development doubles model improvement speed, alarming researchers
The Verge AI · rss · 2026-11-07
According to The Verge, AI safety org METR published a report Thursday after an 11-week investigation inside Anthropic evaluating its AI R&D automation.
The findings show Anthropic is using AI to build more powerful AI with striking results: its models are improving roughly twice as fast as a year ago, with little sign of slowing. Researchers find this alarming, and the report offers some of the clearest evidence yet of AI automating AI development. The probe is separate from the program giving safety researchers employee-like access after high-profile hacking incidents over the summer.
More from AGI Musings
- Toby Ord: self-censoring flawed analysis from 'your team' is bad epistemics — tobyordoxford · 2026-11-07
- Toby Ord: AI's 80% time horizon on AI R&D is just 15 minutes at OpenAI and GDM — tobyordoxford · 2026-11-07
- Toby Ord clarifies: most people's AGI timelines are too long, the bigger problem — tobyordoxford · 2026-11-07
- 'AIs are grown, not coded': on emergence and why we don't understand our own models — Yamapama · 2026-11-07
- Let the machine check the work, never let it sign it: Reddit debates AI's role in decisions — indianbull88 · 2026-11-07
- Hand-drawn AI timeline predicts open-source AI crosses closed-source capabilities around today — victormustar · 2026-11-07