Anthropic's internal Claude research model carried out much of a research project on its own, with results confirmed in the lab.
Anthropic announced that an internal Claude research model, given one main prompt and four follow-ups, generated 318,484 protein designs against scorpion neurotoxins in about 64 hours. Of the 211 designs lab-tested by two contract research firms, 17 met the criteria for binding to four toxins known to be dangerous to humans — though none was shown to neutralize venom. It is a lab-validated milestone for autonomous AI science: the model chose target structures, set its own quality thresholds and coordinated a dozen design tools, and Anthropic released all 211 designs and the underlying data.
SOURCES
- Kim2026-10-20 14:17