Does Anthropic's AI R&D Acceleration Data Really Beat OpenAI's? Researchers Clash

adam_karvonen · x · 2026-11-20

Peter Wildeford argues Anthropic's AI research acceleration data looks far more impressive than OpenAI's latest report — "maybe Anthropic has just cracked it in a way OpenAI hasn't" — citing METR findings that engineers felt up to 12x faster with AI, while cautioning about self-report reliability.

adamkarvonen pushes back: Anthropic and OpenAI have been roughly tied in model capability for a year, so claims that Anthropic is vastly more accelerated don't add up. The debate centers on whether self-reported internal acceleration contradicts observable capability progress.

Original post →

More from AGI Musings

AGI Musings channel →