Anthropic's first drug development results with Claude

Anthropic published early results from its life sciences team, which spent several months working with six drug developers and started its own discovery programs. Most of the work is preclinical, so nothing has been tested in humans yet, but the specific numbers are worth reading. At Generate Biomedicines, six scientists used Claude Code to redesign a genetic component for CAR T-cell therapy aimed at solid tumors. One design came out about 70% shorter than the standard version and worked better in the lab, with immune cells producing roughly ten times as many cancer-spotting receptors and staying active longer.

The pattern across partners is that Claude did not just generate candidates, it questioned the measurements. At BigHat Biosciences, Claude proposed changes to two antibody fragments and the lab built and tested about 1,010 variants over ten weeks. Partway through, Claude and an Anthropic researcher reanalyzed the raw data, decided the standard heat-stability metrics were misleading, and proposed new ones. The best versions bound more than ten times more tightly and stayed in solution at temperatures 18°C higher. At Enveda, Claude reviewed 1,061 pages of manufacturing documents, caught all 11 issues human reviewers had found plus seven they had missed, and cut review time from over 25 hours to under three.

Anthropic also acquired Mithril Medicine, a one-person company whose founder used Claude to coordinate experiments across 11 external labs and took a muscular dystrophy treatment from idea to promising mouse results in under 18 months. The full write-up is on Anthropic's site.

Why it matters

If you work in biotech, the useful signal here is not drug design but process: Claude caught flawed lab metrics and cut documentation time by 89%. Those are tasks you could test now, without betting a program on model-generated molecules.