Anthropic Bought a One-Person Drug Company
Anthropic Bought a One-Person Drug Company
Cody Tranbarger ran a whole biotech on Claude, coordinating 11 outside labs by himself. In under 18 months he had a muscular dystrophy treatment showing promise in mice — and now Anthropic owns the company.
Anthropic disclosed on October 27 that it has acquired Mithril Medicine, a biotech with exactly one full-time employee: its founder, Cody Tranbarger. The deal itself closed back in July 2026, and Anthropic didn't disclose the price.
What makes it worth your attention isn't the acquisition. It's what Tranbarger did before it. Using Claude — Anthropic's AI assistant — he organized contracts, logistics, and experiments across 11 external labs, and took a possible treatment from an idea to promising results in mice in less than 18 months.
How one person ran eleven labs
Mithril was working on two rare forms of muscular dystrophy: LAMA2-related congenital muscular dystrophy and alpha-dystroglycanopathy. Both are severe genetic disorders that damage the connections holding muscle fibers to the tissue around them, so muscles weaken and waste away over time.
Tranbarger's approach was an experimental antibody — a lab-made protein designed to act like a bridge — meant to restore that broken link between muscle fibers and surrounding tissue. The labs did the actual bench work. Claude did the coordination that would normally take a whole operations team.
According to TechStartups' account of the project, Claude once reconstructed the status of experiments from email threads and 14 months of presentation files, flagged what was still outstanding, and drafted handover notes for lab partners. In another case it consolidated hundreds of antibody-screening files and judged which candidates were worth pursuing.
Tranbarger estimates that drafting an initial research strategy for a new program, a job that used to take one to two months, now takes two to three days with Claude. Anthropic calls the model "Pipeline-in-a-Person": one scientist overseeing several drug programs with AI support.
That framing needs a caveat. Humans still picked the targets and made the final calls, and the lab work still happened in the real world. Claude didn't discover a drug on its own. It removed the coordination bottleneck that usually stops a tiny team from trying.
What the mice actually showed
In a six-week study in mice with a severe form of LAMA2-related disease, animals on the highest antibody dose showed roughly twice the grip strength and muscle force of untreated mice, according to TechStartups. Researchers also found the antibodies in the muscle tissue where they were supposed to act, which is early evidence the treatment reached its target.
But let's be clear about what this isn't. The treated mice were still weaker than healthy mice, so this is not a cure. The treatment hasn't been tested in humans at all, and all of this work is still preclinical — lab and animal studies, not clinical trials.
Anthropic says that if the candidates keep holding up, it hopes to pick formal drug candidates in 2027, with human trials for the LAMA2 treatment starting as soon as late 2028 and the alpha-dystroglycanopathy program following in 2029. Those dates depend on more safety testing and regulatory approval, and plenty of drugs that work in mice fail in people.
It wasn't just Mithril
The acquisition came packaged with early results from six outside partners, and together they show where Claude actually helps in drug development. It's not one magic step. It's a lot of unglamorous ones.
At Generate Biomedicines, six scientists used Claude Code to redesign part of an experimental CAR T-cell cancer therapy — a treatment that re-engineers a patient's own immune cells to attack cancer. One design came out about 70% shorter than standard, and in lab tests it let immune cells make roughly 10 times as many cancer-spotting receptors and stay active longer.
Divergence Labs, working on a food allergy treatment, had Claude generate roughly 35,000 antibody designs in about a month. Early tests found hits on both halves of the design: three that could bind to IgE, the molecule behind allergic reactions, and 20 that could strip it off immune cells, with three beating the standard medication.
BigHat Biosciences tested about 1,010 variants of two antibody fragments over roughly 10 weeks, based on changes Claude proposed. The best versions bound their targets more than 10 times as tightly, one stayed stable in solution at 18°C higher, and the other clumped more than 100 times more slowly.
The least flashy result might be the most useful. At Enveda, manufacturing chemists were spending up to 75% of their time on paperwork. Claude reviewed 1,061 pages, caught all 11 issues humans had found plus seven they'd missed, and cut review time by 89%, from over 25 hours to under three. It also drafted a regulatory document in four hours that would have taken about three months by hand.
Bristol Myers Squibb is using Claude to assess early-stage clinical trial plans and estimate their odds of success. One of its executives said a single prompt produced an initial analysis that would otherwise have taken days, and the two companies plan to expand the work in 2027.
Why Anthropic wants to own drugs, not just sell tools
Until now, AI labs have mostly sold tools to drugmakers. Buying Mithril puts Anthropic in the business of making medicines itself, one day after it filed for its stock market listing. That's a real shift, and the reason is economics as much as science.
Rare childhood diseases often don't get treatments because the patient numbers are small and a full biotech team is expensive. If one scientist with AI support can run programs that used to need a whole company, diseases that were ignored for financial reasons suddenly become viable to pursue.
Anthropic is also opening the door wider for other researchers. It has rolled out Claude Science, an AI workbench for researchers, plus compute credits for life scientists, a verification program for trusted organizations, and an open call for guest investigators in the life sciences.
None of this makes the mouse results a cure, and none of the partner results are proven medicines yet. But the operating model is the story to watch: if a one-person company can reach animal trials in 18 months, the question is which ignored diseases get a shot next. Watch the 2027 candidate picks, not the hype.
