Citrini Research, in partnership with Anthropic, used the Claude Science model to design drug candidates against scorpion toxins. From a 35-page prompt, the AI generated approximately 180 molecular candidates in a few days, work that would normally take a year using traditional methods. Twist Bioscience synthesized and tested 211 designs, of which 17 successfully bound to target toxins, exceeding Citrini’s expectations. In a report published on October 20, 2026, Citrini recommends investing in experimental validation companies like Charles River Laboratories and Revitty rather than in AI model developers. This experiment suggests that candidate creation is no longer the bottleneck, but physical lab validation has become the constraint.
Source: Read the original article
