TechAnthropic reports Claude now leads 26% of internal AI R&D work, up from <1% earlier in 2026
Anthropic published metrics showing Claude completing most tasks end-to-end from high-level prompts under human supervision for 26% of its AI R&D. About 30,000 AI agents ran internally in August with screening blocking one in ~47,000 actions. No tasks reached full autonomy.
TLDR
Provides concrete evidence of AI accelerating its own development, fueling debates on recursive self-improvement and how close labs are to more autonomous systems. Contrasts with OpenAI's recent misalignment disclosures and amplifies calls for transparency across frontier labs. High engagement on X from AI enthusiasts and skeptics debating oversight needs and timelines.
Combined views
30.2K
1 Source, first seen 7h ago
Anthropic reports Claude now leads 26% of internal AI R&D work, up from <1% earlier in 2026
Anthropic published metrics showing Claude completing most tasks end-to-end from high-level prompts under human supervision for 26% of its AI R&D. About 30,000 AI agents ran internally in August with screening blocking one in ~47,000 actions. No tasks reached full autonomy.