AIAnthropic reports Claude leads 26% of its AI R&D work, up sharply from under 1% early 2026
Anthropic's R&D Automation Index shows Claude completing most tasks end-to-end with human supervision on 26% of AI research as of August 2026. Over 90% of work is at collaborates level or higher, with roughly 30,000 active agents.
TLDR
The data provides concrete evidence of progress toward recursive self-improvement and AI building better AI, fueling discussions on oversight challenges and safety. It aligns with calls for coordinated slowing of frontier development and frames AI as already significantly accelerating its own advancement, raising questions about the pace of progress and regulatory oversight.
Combined views
—
2 Sources, first seen 13h ago
Anthropic reports Claude leads 26% of its AI R&D work, up sharply from under 1% early 2026
Anthropic's R&D Automation Index shows Claude completing most tasks end-to-end with human supervision on 26% of AI research as of August 2026. Over 90% of work is at collaborates level or higher, with roughly 30,000 active agents.
TLDR
The data provides concrete evidence of progress toward recursive self-improvement and AI building better AI, fueling discussions on oversight challenges and safety. It aligns with calls for coordinated slowing of frontier development and frames AI as already significantly accelerating its own advancement, raising questions about the pace of progress and regulatory oversight.