Anthropic's Claude leads 26% of AI research, but what does 'lead' mean?
anthropic says claude leads 26% of its research work, but the term 'lead' doesn't imply full autonomy, affecting how we assess ai development progress.
anthropic published metrics on its ai development, revealing claude 'leads' 26% of future model work, scored by claude itself using epoch ai's scale. al4, or 'ai leads,' means claude can analyze, fix, and test tasks without human questions but still needs human oversight. this is distinct from al5, or full autonomy. the metrics aim to provide insight into model development and support anthropic ceo dario amodei's call for slowed ai development. the company's internal platform monitors 30,000 agents, blocking 0.002% of actions.