Chart of the Day | Coatue Summary

"Claude now leads over a quarter of Anthropic's model R&D, up from virtually none just six months ago!" The chart shows Claude's role as an autonomous "AI leads" agent in model development jumped from 1% (April 2026) to 26% (August 2026) — a 26x increase in roughly four months.

The chart data shows a clear phase transition: for most of the tracked period (Aug 2025–Feb 2026), the dominant categories were "AI assists" and "AI collaborates," with "AI leads" essentially at zero. Starting in April 2026, "AI leads" appears and scales aggressively (1% → 3% → 12% → 14% → 22% → 26%), while "Minimal AI" and "No AI" categories shrink toward zero — indicating full-stack automation of R&D tasks is displacing human-led and human-only workflows almost entirely.

The data implies that Anthropic — one of the leading AI research labs — is quietly transitioning into an organization where its own AI model does a substantial and rapidly growing share of the research and development work itself, not just coding assistance. This challenges the conventional narrative that frontier AI labs are talent-constrained, human-researcher-driven shops. "Claude now leads over a quarter of Anthropic's model R&D" suggests the constraint is shifting toward compute and model capability rather than headcount of human researchers.

Anthropic — AI research lab and developer of the Claude model family. Mentioned as the case study/data source for the chart illustrating how much of its own model R&D is now automated by Claude. "Monthly share of model R&D tasks at Anthropic by automation level" (Source: Anthropic).
Coatue — Investment firm and newsletter publisher. Mentioned as the author/curator presenting this data point as a notable "Chart of the Day" signal for investors.

None specifically named in the article; the content is a data chart with no individual persons cited or profiled.

Track internal AI-automation metrics as a leading indicator. Anthropic's granular tracking of "share of model R&D tasks by automation level" (AI leads / collaborates / assists / minimal / no AI) is itself a useful operating framework — entrepreneurs and R&D leaders should consider instrumenting their own workflows this way to measure and communicate AI adoption progress internally.
Expect step-function, not linear, adoption curves. The jump from ~0% to 26% "AI leads" share happened in about four months (Apr–Aug 2026) after a long period near zero — operators should plan for sudden inflection points in AI capability/adoption rather than assuming smooth, incremental change.

The "Minimal AI" and "No AI" categories have nearly vanished. While the headline focuses on the "AI leads" growth to 26%, the chart also shows that by mid-2026 almost the entire stack of R&D tasks falls into "AI assists," "AI collaborates," or "AI leads" — meaning fully human-driven, non-AI-touched R&D work has essentially disappeared at Anthropic. This broader disappearance of "no AI" work may be as significant as the "AI leads" growth itself, signaling total workflow saturation by AI tools, not just isolated automation wins.