Claude cut drug-batch record review by 89% and produced 35,000 antibody designs in a month

What the agent did

• At Divergence Labs, turned the team's goals into a design plan and, largely on its own, produced about 35,000 antibody designs in about a month; early tests found 3 that bind IgE and 20 that dislodge it
• At BigHat, connected to the lab's data system, proposed changes that became about 1,010 tested variants in about 10 weeks; the best bound more than 10x more tightly
• At Enveda, checked 1,061 pages of manufacturing records, caught all 11 issues human reviewers had found plus seven they missed, and cut review from over 25 hours to under three
• Also at Enveda, drafted a regulatory document in four hours that takes about three months by hand

What it means for you

• Connect the agent to the system of record, as BigHat did with its lab data system and Enveda with all its manufacturing documents, instead of pasting data into chat.
• Before trusting an agent with filings, score it against human reviewers on the same documents, as Enveda did with its 11 known issues.
• Each partner has a technical report linked from Anthropic's post; read the methods before copying a setup.

Why it matters

Most talk about AI in pharma is about inventing molecules; these results show agents also speeding up the slow loops around them, like lab iteration and quality paperwork, where Enveda's chemists spent up to 75% of their time. Anthropic says the work is mostly preclinical, so none of it has reached human trials yet.

Results as reported by Anthropic about its own model and its partners; mostly preclinical and not independently verified.