On September 18th, Anthropic rolled out a series of metrics tailored to "gauge the speed of AI development in leading-edge labs." This initiative is aimed at deepening the public's grasp of the strides made in AI model development, facilitated by measurement tools that are openly accessible. Anthropic is of the opinion that society requires a greater influx of information to accurately evaluate the tempo of AI advancements at the forefront. The metrics zero in on three pivotal domains: the role of AI in crafting next-generation AI systems, the proficiency in overseeing AI agent conduct, and the resource allocation propelling the creation of more formidable models.

Among these focal points, Anthropic has instituted an "AI Research and Development Automation Index" to precisely measure the share of AI R&D tasks undertaken by Claude within the company. As of August 2026, although Claude has yet to attain complete autonomy in R&D, it has "spearheaded" roughly 26% of Anthropic's AI R&D endeavors. Moreover, over 90% of the R&D work has reached a stage of "AI collaboration" or achieved an even greater level of automation.