Anthropic: AI Writes Over 80% of Its Product Code, Signals Self-Improvement
Anthropic's internal data shows that Claude, its AI model, now writes over 80% of the code for current products, with code volume per engineer growing eightfold in Q2 2025. The company sees signs of approaching recursive self-improvement, though it says the point of no return has not yet been reached.
According to internal data from Anthropic, more than 80% of the code for the company's current products was written by its AI model Claude. In the second quarter, the volume of code generated per engineer increased eight times compared to 2024, as team members increasingly delegate model development to AI systems.
Anthropic Institute head Marina Favaro and co-founder Jack Clark wrote that, given sufficient compute, this trend could lead to a system capable of "fully autonomously designing and developing its successor." They emphasized that the company has not yet crossed a point of no return, and recursive self-improvement is not inevitable, but it may arrive sooner than most institutions are prepared for.
Source: ForkLog