The most telling number in Anthropic’s new drug development update isn’t about molecules. It’s that manufacturing chemists at Enveda spent up to 75% of their time on paperwork. Drug development has plenty of bottlenecks, and many of them are dull.
Anthropic’s life sciences team has spent several months working with six organizations: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. It has also started its own discovery programs, including work on rare childhood forms of muscular dystrophy. The company’s premise is that no single step limits progress. Target selection, molecule design, manufacturing, regulatory filings and trial operations all slow things down, so AI needs to touch all of them.
The company says the work serves three purposes: learning where its models fall short, producing case studies for other biotech firms, and, if the trajectories hold, delivering real medicines. Results so far are early and come from the partners’ own lab tests, so they should be read as promising rather than proven.
Designing better cell therapies and antibodies
Generate Biomedicines wants CAR T-cell therapy, which has mainly treated blood cancers, to work against solid tumors. That requires shortening one part of the therapy’s genetic instructions to make room for more. Six Generate scientists used Claude Code to propose new designs. One was about 70% shorter than the standard version. In lab tests it let immune cells make roughly 10 times as many receptors, killed tumor cells more efficiently and stayed active longer.
Divergence Labs is pursuing a different idea for food allergies: pulling IgE off immune cells entirely, rather than only neutralizing the IgE floating in blood. That needs a two-armed drug, one arm to bind IgE and one to pry it loose. Claude turned the goals into a design plan, then largely on its own generated about 35,000 antibody designs in roughly a month. Early tests found three designs that bound IgE and 20 that dislodged it. Three of those 20 appeared to beat the standard medication. The next step is combining both arms in one drug.
BigHat Biosciences tested about 1,010 variants of two antibody fragments in about 10 weeks, with Claude connected to the company’s lab data system. Midway, the team suspected the standard heat-stability and clumping metrics were misleading. Claude and an Anthropic researcher reanalyzed the raw data and proposed new metrics and a protocol change. The main predictions held. The best versions bound more than 10 times more tightly, one stayed in solution 18°C higher, and another clumped more than 100 times more slowly.
Cutting the paperwork
Enveda connected its manufacturing documents to Claude and asked it to flag possible issues. Across 1,061 pages, Claude caught all 11 problems human reviewers had found, plus seven they had missed. Review time fell 89%, from more than 25 hours to under three. Claude also drafted two regulatory documents. One took four hours instead of about three months. Enveda’s standard human review of the other found one minor error.
Changing which ideas get funded
The Column Group is setting up Pleiades, a company meant to run many more, and more complex, drug programs with a small team using Claude. The logic is economic. Because discovery is slow, costly and failure-prone, investors tend to back previously validated ideas. Cheaper, faster discovery could make bolder bets easier to fund.
What it means
The results span design, optimization, operations and finance, which supports Anthropic’s argument that value lies across the pipeline, not only in molecule generation. Still, none of these candidates has reached patients, and clinical trials are where most drugs fail. The real test is whether faster early work produces more approved medicines. Anthropic says it plans to scale up substantially, and the partner reports offer clinicians and founders a concrete look at where AI is already saving time.