Anthropic Buys One-Person Biotech Mithril Medicine

Anthropic acquires Mithril Medicine, a one-person biotech

Anthropic said on October 27, 2026, that it has acquired Mithril Medicine, a drug discovery company whose only employee was its founder, Cody Tranbarger. The news came in a post on Anthropic's work with six drug developers and its own drug programs. The deal closed in July 2026, according to Anthropic's technical report, and the price was not disclosed.

Mithril worked on two rare forms of muscular dystrophy, and Anthropic calls it "a one-person company." Tranbarger relied on Claude to set up contracts, shipping, and experiments across 11 outside labs, and let Claude handle much of the day-to-day running of the business.

One person did not do all of the work, though. The technical report says members of the Mithril team were paid consultants or advisers, and that the outside labs were paid for their work.

Founder Cody Tranbarger continues the research at Anthropic

Tranbarger now works at Anthropic, where he is continuing Mithril's research and plans to start several new programs. The technical report's competing interests statement says Anthropic owns the antibody program that Mithril developed.

Mithril targets a rare childhood muscular dystrophy

Mithril's main target is LAMA2-related congenital muscular dystrophy, a childhood disease that has no treatment able to slow or change its course. It is caused by the loss of a protein called laminin-α2, which holds each muscle fiber to the membrane around it. Children with the disease are weak and have low muscle tone from birth or early infancy.

Muscle without laminin-α2 makes more of a related protein, laminin-α4, but this substitute binds only weakly and cannot do the same job. Mithril's idea is an antibody (a protein that binds to a specific target) built to grab two targets at once.

One end holds laminin-α4 and the other holds a protein on the muscle fiber's membrane, so the fiber is anchored again. Mithril plans to use the same approach for its second target, α-dystroglycanopathy.

From idea to promising mouse results in under 18 months

Going from an idea to encouraging data in mice took Mithril under 18 months, Anthropic says. In a pilot study in the technical report, mice with the disease were given a tetravalent version of the antibody, one with four binding sites. They developed about twice the peak force and grip strength of mice given an inactive solution.

Human clinical trials as soon as late 2028

If the drug candidates keep holding up, Anthropic expects a drug to be ready for human clinical trials as soon as late 2028. A clinical trial tests a drug's safety and effect in people. Mithril's results so far are preclinical, meaning they come from lab and animal work before any human testing, and the technical report has not been peer reviewed.

One researcher and Claude plan tests of 113 antibodies

The technical report logs one week in which a single researcher and Claude put together a plan to test 113 antibodies at five contract research organizations (CROs). CROs are outside labs that drug companies pay to run experiments.

On day four, Claude read a 341-page manufacturing report on the 113 antibodies. It gave a bottom line in 2 minutes 40 seconds and a full summary of all of them in 7 minutes 53 seconds. It then proposed testing 82 as they were, purifying 7 further, and holding 24 back.

Claude was also much quicker to respond than the labs. Its median time to a first answer was 96 seconds, while the CROs' first replies took a median of 8.5 hours.

The researcher overrules Claude and cuts the list to 89

On day seven, the researcher overruled Claude's recommendation and cut the plan from 113 antibodies to 89. Four minutes later, Claude had revised its memo, the run order, and the unsent email drafts to match. The researcher, not Claude, made the final call on what to test.

Claude drafts 17 emails, and the researchers send 12

Over the week, Claude drafted 17 emails but did not send a single one. The researchers edited the drafts and sent 12 of them.

Claude catches five of the six errors examined in detail

The researchers examined six of Claude's errors in detail, and Claude found five of them itself. Three of the six still reached a CRO in emails the researchers had edited and sent.

How Anthropic's six partners use Claude in drug development

In the same post, Anthropic described how its six partners use Claude, from designing drugs to checking manufacturing records. The main examples are below.

The other two partners are BigHat Biosciences and The Column Group. Anthropic presents BigHat as its example of improving drug leads, and The Column Group as its example of starting new drug companies.

Most of the work has not yet reached human trials

Anthropic says the research, both at its partners and at Mithril, is still mostly preclinical. For the most part, the drugs have not yet been tested in people, so no one knows whether they work in humans.

For researchers, Anthropic also offers Claude Science, its AI workbench. (Related: Claude Science: An AI Workbench for Researchers)