What happened
Anthropic published a quantitative research report (lead authors Hermle, Jones, Althoff, et al.) using privacy-preserving analysis of 207,000 Claude and Claude Code research sessions plus an accompanying survey of scientists. It finds scientists use Claude most in the formal sciences (computer science, mathematics, statistics) and least in health and lab-based physical sciences, and primarily for data analysis and writing — data analysis alone accounts for about one-third of sessions on Claude and nearly half on Claude Code. The headline projection: "If existing AI capabilities are widely adopted, they could double research output in some fields. If AI could carry out all non-physical research work, output in fields such as computer science and statistics could rise as much as 15-fold, while output in bench sciences such as chemistry and biology... would rise three- or fourfold." Roughly 40% of researcher survey respondents expect AI models to carry out full research projects on their own by 2030.
Why it matters
This is a rare, methodology-rigorous measurement of how much frontier AI is already accelerating scientific R&D and how much it could accelerate it — directly informing R&D investment, capability strategy and AI-for-science policy for labs, universities and governments.
Action needed
Brief R&D and technology leadership on the quantified acceleration scenarios and use them to assess where AI augmentation could double (or far exceed) research throughput; align AI-for-science investment and safety/governance accordingly.