Anthropic has released new metrics revealing that its AI system, Claude, now leads over a quarter of the company's internal research and development work. The data provides fresh insight into how much of an AI lab's own technical development is being conducted by artificial intelligence.
What Happened
According to new measurements from Anthropic, its AI assistant Claude now leads 26% of the company's AI research and development work. This represents a significant increase from February, when Claude led less than 1% of such tasks.
Anthropic defines the term "leads" as the AI completing the majority of a task end-to-end based on a high-level prompt, while operating under human supervision. In addition to leading more than a quarter of R&D, more than 90% of the company's AI research and development is now performed at or above the level where AI collaborates directly with humans.
Key Details
- Claude leads 26% of Anthropic's AI R&D work, up from under 1% in February.
- "Leading" is defined as completing most of a task end-to-end from a high-level prompt with human supervision.
- More than 90% of AI R&D is conducted at or above the level of human-AI collaboration.
- Anthropic had roughly 30,000 AI agents simultaneously running research and engineering work on its primary internal platform in August.
Why It Matters
The new figures speak directly to a central question in ongoing discussions surrounding AI development: whether increasingly capable artificial intelligence systems could eventually accelerate the creation of even more advanced systems. By demonstrating that Claude is independently driving a growing share of its own creator's R&D process, the data offers concrete metrics on the automation of AI engineering.
