Published: 28 October 2026
Anthropic has acquired Mithril Medicine, a biotech startup with one full-time employee that used Claude to take an experimental muscular dystrophy treatment from idea to mouse results in under 18 months.
The deal closed in July 2026 but was only disclosed on Tuesday 27 October, in a research update from Anthropic on its drug development work. Financial terms were not released.
Mithril's founder, Cody Tranbarger, has joined Anthropic and will carry on the research there. Anthropic says a treatment could enter human trials as early as late 2028 if the current candidates hold up.
Who is behind Mithril Medicine?
Mithril Medicine was founded roughly 18 months ago with backing from Atlas Venture, the US biotech investor, according to Endpoints News. It ran as a virtual company: one founder, a handful of part-time consultants and no lab of its own.
Tranbarger had spent six years posting about biotech and venture capital on X under the pseudonym Professor Oak. He revealed his identity on the day of the announcement, and says he will keep posting under the same handle.
Before Mithril, he worked in venture capital and crossover investing. He says conversations with researchers at Coefficient Bio, which Anthropic bought earlier this year in a deal TechCrunch reported at around $400m, helped him turn his approach into a repeatable operating model.
What did one person and Claude actually do?
Mithril's work targets two rare genetic conditions: LAMA2-related congenital muscular dystrophy and alpha-dystroglycanopathy. Both break the links that hold muscle fibres to the surrounding tissue, which leads to progressive weakness in children.
Tranbarger designed bispecific antibodies meant to restore those links. Claude handled the coordination: paperwork with partner labs, logistics, scheduling and data analysis across 11 external laboratories.
In one example Anthropic describes, Claude rebuilt the status of every live experiment from email threads and 14 months of slide decks, flagged the open problems and drafted handover notes for lab partners. In another, it combined hundreds of antibody screening files and ranked which candidates were worth taking forward.
Tranbarger estimates that drawing up an initial research strategy for a new programme, which used to take one to two months, now takes two to three days. He calls the model "Pipeline-in-a-Person": one scientist running several drug programmes at once, with AI absorbing the overhead.
How strong are the early results?
In a six-week study of mice with a severe form of LAMA2-related muscular dystrophy, animals on the highest antibody dose showed roughly twice the grip strength and muscle force of untreated mice.
Four of eight untreated mice died between two and six weeks of age. Only one of eight treated mice did. The antibodies were also found in the muscle tissue where they were expected to act.
Anthropic is careful about what this means. The treatment has not been tested in humans, the treated mice were still weaker than healthy ones, and the company says Claude made mistakes along the way. This is early preclinical work, not a cure.
Mithril has now produced more than 100 new antibodies. Anthropic hopes to pick formal drug candidates in 2027, with possible human trials for LAMA2-related muscular dystrophy in late 2028 and for alpha-dystroglycanopathy in 2029.
What else did Anthropic announce?
The acquisition came alongside early results from six partner organisations: Bristol Myers Squibb, BigHat Biosciences, Generate Biomedicines, Divergence Labs, Enveda and The Column Group.
Generate Biomedicines used Claude Code to redesign part of a CAR T-cell cancer therapy. One version was around 70% shorter than the standard design and let immune cells produce about ten times as many cancer-detecting receptors in lab tests.
Divergence Labs had Claude generate around 35,000 antibody designs in a month for a potential food allergy treatment. BigHat tested about 1,010 antibody variants, with the best binding more than ten times more tightly than the starting point.
Enveda reported an 89% cut in human review time for manufacturing documents, with Claude catching issues manual review had missed. Bristol Myers Squibb is using Claude to weigh up clinical trial plans for an early-stage drug and estimate their chances of success.
Anthropic first said it would run its own drug programmes in June, when it launched Claude Science and told STAT News it would focus on neglected diseases the commercial market overlooks. It has since set up a Bay Area wet lab, and this week filed for a Nasdaq listing that disclosed $16.3bn of revenue in the first half of 2026.
What this means for UK founders
The headline is not the mouse data. It is that a venture-backed biotech reached animal results with a headcount of one, and a frontier AI lab thought that was worth buying.
For UK founders in life sciences, the cost structure is the story. A rare childhood disease programme that could never justify a 20-person team might justify one scientist, a model and a budget for outsourced lab work. That changes which diseases get worked on at all.
It also raises the bar on what "lean" means. Bizztor has already covered how Anthropic's own agents run six hours for every human hour inside the company, and UK healthtech startups such as adsilico, which tests medical devices on virtual patients, are already selling a version of that pitch to investors: fewer people, more done in software before anything touches a patient.
The UK has the pieces for this. The government's £100m AI biosecurity lab with OpenAI and Anthropic shows Whitehall wants frontier models close to biology research. The gap is whether UK seed investors will back a one-person biotech the way Atlas Venture did.
One caution. Mithril's founder had years inside venture and biotech before he started. The model is one expert plus AI, not AI instead of an expert. Human scientists still chose the targets, judged the evidence and made the final calls.
