Anthropic discloses AI now leads 26% of its AI R&D, with 30,000 agents at work
ZeroStateReflex · x · 2026-09-18
Anthropic published three transparency metrics quantifying 'AI building AI,' arguing any frontier lab could publish the same and third parties could verify them:
- AI share of AI R&D: Over 26% of the work building the next version of Anthropic's models is now led by AI—up from 0% six months ago. About 30,000 agents do research and engineering work at Anthropic at any moment, described as 'a country of geniuses in a datacenter' (comparable to or larger than four countries).
- Agent oversight: Measuring how well AI agents are supervised.
- Compute allocation: Only about 6% of compute going to AI R&D is allocated toward safety.
Anthropic says the goal is minimizing the gap between what frontier labs know and what the public knows as the world considers pacing the frontier.
Related event: Anthropic Discloses Claude Now Leads 26% of Its Internal R&D(30 posts)→
More from AGI Musings
- Yoshua Bengio says recursive self-improvement should be illegal — Polymarket · 2026-09-18
- Irving Good's 1965 quote predicted the intelligence explosion 60 years before AGI hype — burny_tech · 2026-09-18
- Mathematician Tony Feng: 'Correct' in modern math means experts convincing experts — littmath · 2026-09-18
- Researcher fires back at AI datacenter skeptics: machines already outsmart experts — suchenzang · 2026-09-18
- Gary Marcus endorses claim that AI extinction obsession under-invests in security, citing Hugging Face incidents — GaryMarcus · 2026-09-18
- Predicting 70% fully-automated AI share by 2026 and true RSI by decade's end — Justin_Halford_ · 2026-09-18