Anthropic's capability gains per $1B of R&D keep falling, ramez pushes back on 2x AI acceleration
ramez · x · 2026-11-11
After digesting the METR report and the Mythos/Fable 5.2 model card, ramez argues the data doesn't support anything close to a 2x acceleration from AI, though some acceleration is real. The accompanying chart shows Anthropic's AECI points gained per $1B of R&D sliding from 6.58 in 2024 to 0.76 by Q3 2026, with no clear bend in the trendline.
austinc3301 counters that R&D dollars aren't exogenous: a growing share of Anthropic's compute spend goes to AI agents doing internal R&D, so acceleration shows up as spending. The real question is whether Anthropic could have absorbed 5x/year spending growth at constant returns without AI labor. ramez concludes that since every model has had misses, the right stance is uncertainty and hedging across scenarios rather than ruling RSI in or out.
Related event: METR's 2x AI R&D Acceleration Report Sparks Takeoff Debate(19 posts)→
More from AGI Musings
- OpenAI researcher tells healthtech founders: bet on the models getting better — madhavsinghal_ · 2026-11-11
- Anthropic: Claude now leads 26% of its model R&D, up from under 1% in February — RealGeneKim · 2026-11-11
- Living with AI doom: save for the future or spend it all on psychedelics? — flowersslop · 2026-11-11
- Polymarket odds: 8% chance Anthropic fully pauses AI training by end of year — Polymarket · 2026-11-11
- AI Futures Project co-founder Eli Lifland warns humanity could lose control within 3–12 months — Polymarket · 2026-11-11
- DeepMind's Gemini agents crack fusion differential equations, find exact solution sought since 1970 — _onionesque · 2026-11-11