Anthropic Buys Mithril, One-Person Biotech Using Claude for Rare-Disease Drugs

Anthropic disclosed on Oct. 27 the early results of six pharmaceutical collaborations using its Claude model and confirmed a more significant move: its July acquisition of Mithril Medicine, a one-employee rare disease biotech based in Cambridge, Massachusetts. The transaction adds two rare genetic disease drug programs to Anthropic’s internal research portfolio and gives Anthropic a live experiment in applying AI to end-to-end drug development. The disclosed collaborations cover antibody design, candidate optimization, production documentation and clinical planning. With Generate Biomedicines, Claude Code helped shorten a signaling domain in a CAR-T construct by about 70% while producing an experimental structure with roughly 10 times higher receptor expression, pointing toward possible improvements for solid tumors. Divergence Labs used Claude to generate 35,000 antibody designs in about one month while pursuing treatments that may disrupt IgE binding in food allergies, although the candidates remain early-stage. Other partnerships emphasized workflow efficiency. BigHat Biosciences paired Claude with automated labs to test 1,010 antibody variants in about 10 weeks, identifying molecules with more than 10 times better binding. Enveda used Claude to review a 1,061-page manufacturing file, finding all 11 human-identified issues and seven additional concerns while cutting review time by 89%, and drafted a regulatory document in hours that could otherwise take months. Bristol-Myers Squibb is applying Claude to early clinical strategy and success assessment, while The Column Group is using the tools to evaluate disease-target combinations for a new venture, Pleiades. The centerpiece is Mithril Medicine, which before acquisition had a single full-time employee: founder Cody Tranbarger. Operating through a virtual model with part-time advisors and 11 external laboratories, Mithril used Claude to coordinate contracts, experiments, data and communications. Anthropic called this model Pipeline-in-a-Person. It helped move a potential therapy for LAMA2-related congenital muscular dystrophy to mouse proof-of-concept in less than 18 months. In short-term animal tests, treated mice showed roughly double grip strength and some muscle force, along with better survival. Mithril has also generated more than 100 bispecific antibody combinations and plans to nominate candidates in 2027, with human trials targeted for 2028 and 2029 for LAMA2-CMD and alpha-dystroglycanopathy, respectively. Tranbarger, revealed in an X post as the biotech blogger Professor Oak, joined Anthropic’s life sciences team and has a background spanning Prime Medicine, Casdin Capital, Atlas Venture and Revision Bio. He said AI’s near-term value in drug development may lie less in replacing scientific judgment than in coordinating parallel workstreams and reducing administrative burden. Anthropic has not disclosed the acquisition price. Its early drug programs remain preclinical, and the company says broader clinical trials are needed to prove whether Claude can materially accelerate real-world drug development.