METR data implies a second doubling in the rate of AI capability progress
peterwildeford · x · 2026-11-06
Peter Wildeford highlights an analysis of METR task-length data suggesting the rate of AI capability progress has sped up a second time in three years — the first doubling occurred in mid-2024, and the pace of that acceleration may itself be accelerating. If true, extrapolations of capability timelines should shift earlier.
More from AGI Musings
- Rethinking predoc interviews in the AI era: agent transcripts and blindfolded video tests — aniketapanjwani · 2026-11-06
- MOSAIC agent swarm discovers fractional Chern insulator gaps 10x beyond ideal quantum geometry — BenBlaiszik · 2026-11-06
- Paul Graham: schools reward bloated essays, which is why good writing is rare — paulnovosad · 2026-11-06
- Coding 'coming to a close'? X debate says AI engineering era begins — TheTuringPost · 2026-11-06
- Claude now leads 26% of Anthropic's model R&D tasks; simple extrapolation puts it near 68% now — peterwildeford · 2026-11-06
- Five fundamental relationships — money, time, health, people, knowledge — are where AI labs will build products — signulll · 2026-11-06