宝玉 (@dotey)T2·Industry
72 pts
Anthropic analyzed roughly 207,000 research-related Claude sessions and found data analysis accounts for about a third of usage and writing another 30%, while hypothesis generation—ranked most important by scientists—makes up only 4.2%. The report pairs an AI capability score with a usage-multiplier metric, estimating that full automation of non-manual work could speed computer science and statistics research by about 15x, versus 3-4x in chemistry and biology where lab work remains a bottleneck.