METR: Anthropic now improving models twice as fast, driven by AI-assisted R&D
Recoil42 · reddit · 2026-11-06
METR published an independent assessment of Anthropic's AI R&D acceleration as of October 2026:
- Anthropic is likely improving model capabilities on well-specified tasks at roughly 2x its historical 2024–2025 rate (uncertainty range 1.3x–3x).
- Most of the acceleration is likely attributable to Anthropic's extensive use of AI to do R&D itself, though other causes likely contributed in late 2025.
- No observed bottlenecks or strong diminishing returns, implying further progress could accelerate capabilities even faster.
- Since October 2025, AECI benchmark scores have grown at 1.9x the prior rate (95% CI 1.5–2.8x), meaning roughly two years of historical progress were packed into the last year.
Related event: METR Finds AI Has Roughly Doubled Anthropic's AI R&D Speed(11 posts)→
More from AGI Musings
- US Census Bureau starts surveying whether employees use generative AI at work, every two weeks — robseamans · 2026-11-06
- People worry agents are non-deterministic as if their old software was more predictable — zanbrum · 2026-11-06
- Ex-OpenAI VP Brundage: AI 2027 Is Unrealistic — Things Will Move Faster — Miles_Brundage · 2026-11-06
- Redditor rails against 'zero-effort AI novel in hours' viral-get-rich fantasies — Recent-Day3062 · 2026-11-06
- Opinion: AI industry needs soft power — explain itself to the public like crypto did — RichardsonDx · 2026-11-06
- Altman teases "AGI is gonna be wild"; Gary Marcus replies 2027 will be wild — GaryMarcus · 2026-11-06