Future of work

Anthropic says AI now leads 26% of its own R&D

Published
Source
Anthropic Institute

Summary

Anthropic has published three measurement tools for the first time that track how much of its AI research and development is actually done by AI. As of August 2026, Claude leads 26% of Anthropic's AI R&D work, meaning it completes most tasks end-to-end under human supervision, and more than 90% of measured work is at or above the collaboration level. No measured area runs fully autonomously, and roughly 30,000 research and engineering agents run at any one time on its most-used internal platform, watched by online and offline monitors. Because these are self-reported figures, Anthropic says it plans to embed independent third-party evaluators to verify them.

Why it matters for our work

The speed of humans and AI working together is now being quantified inside an AI lab's own research operations. Other organizations can build similar metrics to decide which tasks to hand to AI and where to place oversight.

Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.

Read original (opens in a new tab)