Anthropic published on October 27 an update on drug-development work with six organizations, BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group, plus its own discovery programs, including rare childhood forms of muscular dystrophy. The post reports wet-lab and workflow outcomes rather than generated candidates alone.
At Generate Biomedicines, six scientists used Claude Code to propose redesigns of part of a CAR-T construct's genetic instructions. One design was about 70% shorter and, in lab tests, let immune cells make roughly 10 times as many receptors, killed tumor cells more efficiently and stayed active longer. Divergence Labs had Claude turn its goals into a design plan and generate roughly 35,000 antibody designs in about a month. Early tests found three designs that bound IgE and 20 that dislodged IgE from immune cells, and three of those 20 performed better than the standard medication.
At BigHat Biosciences, Claude was connected to the lab's data system and proposed changes to two antibody fragments. About 1,010 variants were made and tested in 10 weeks, with the best versions binding more than 10x tighter, one lead staying in solution 18°C higher and another clumping more than 100x more slowly. At Enveda, Claude reviewed 1,061 pages of manufacturing documentation, catching all 11 issues human reviewers found and flagging seven they missed, and cut review time from over 25 hours to under three. It also drafted two regulatory documents, one in four hours against a roughly three-month manual baseline.
Our takeWhat stands out is the mix of wins and near-misses: 35,000 designs produced 23 early hits, and one regulatory draft still needed a human correction. Those ratios are the useful data. They suggest the model compresses search and paperwork rather than substituting for validation. The reports are partner-authored and not yet peer reviewed, and the strongest claims are wet-lab results whose protocols sit in separate documents, so treat them as promising case studies rather than a general discovery rate.