Dario Amodei, Co-Founder and CEO of Anthropic
Across the eleven external laboratories that carried out the work of Mithril Medicine, nobody was walking the corridors with a clipboard. Coordination ran through one person, Cody Tranbarger, and a set of Claude models that scheduled, tracked and reconciled the research. His company targets two rare forms of muscular dystrophy, conditions that large pharmaceutical groups have found hard to justify commercially. Anthropic has now acquired the venture, and Tranbarger has joined the firm. The financial terms were not disclosed. A drug based on his research could enter human trials as early as late 2028.
Rare disease development has long suffered from arithmetic. Small patient populations rarely repay the overheads of a conventional programme, and a large share of those overheads is coordination: contracting, scheduling, reconciling results from different sites. If one founder with capable models can direct work across eleven laboratories, the threshold at which a condition becomes worth pursuing falls. That matters in Britain, where roughly one in seventeen people will live with a rare condition at some point, and where families often wait years for a diagnosis and longer for a treatment option. Anything that makes small-population research viable deserves a welcome.
Anthropic's wider activity gives the purchase weight. The company is working with Bristol Myers Squibb, BigHat Biosciences and four other partners on tasks ranging from antibody design to clinical trial planning. Its filing for a public listing places the discovery of new cures first among the applications it expects to benefit humanity most. A commitment made to investors is a firmer basis for UK life sciences planning than a conference promise, and trial planning in particular touches parts of the system that Britain controls.
The bottleneck moves downstream. Faster discovery produces more candidates that need testing in people, and testing needs sites, patient registries, regulators and clinicians with time to run studies. Britain has real assets here. Specialist neuromuscular centres hold deep clinical expertise, and their patients are concentrated enough to recruit small trials efficiently. The clinical trial regulations have been reformed, and ministers have committed to shortening the time taken to set up studies. Yet research delivery in the NHS still depends on staff who also cover waiting lists, and contracting delays at trust level remain a long-standing complaint from industry sponsors. A sponsor choosing where to run a first-in-human study for an ultra-rare condition will compare countries on speed of set-up. Operational pressure therefore bears directly on whether the UK is chosen.
Data forms a further link. The ten year health plan sets out an ambition to sequence the genomes of newborns, and the Generation Study already sequences babies at scale. Well-characterised patients supply the raw material for AI-assisted development and make recruitment faster. They also raise the stakes on public trust. Earlier data-sharing controversies in the NHS left a long memory, and any arrangement that gives commercial AI developers access to patient information will be judged on transparency and visible public benefit. Setting governance early costs less than recovering confidence later.
Payment follows. If AI-coordinated programmes do lower the cost of reaching the clinic, NICE and the specialised commissioners of NHS England will face a steadier flow of therapies for very small populations. The highly specialised technologies route was built for such cases, but its budget is finite and the health service has limited tolerance for unplanned spend. Horizon scanning that treats 2028 as a planning date, however uncertain, would cost little and could prevent a scramble.
For NHS leaders the practical steps are modest. Trusts with neuromuscular services can audit their research readiness, from contract templates to pharmacy capacity and clinic time protected for trial visits. Integrated care boards can ask what commissioning for ultra-rare therapies would demand of them. Policymakers can protect the research workforce inside operational planning, where it is often the first casualty of winter pressure. UK health-tech and life sciences firms have an opening to supply what AI-led developers lack, including established trial site networks, curated datasets and regulatory experience. Patients and the public gain most if these preparations happen before a candidate is ready, since families living with these conditions have little time to spare.
Mithril is one small company, and its drug has yet to enter a single patient. The significance lies in the working method it demonstrates, a lean team directing distributed science with the help of AI. Britain can benefit from that capability without building it, provided its health service can say yes quickly when a promising candidate arrives, with sites, data and public consent already in place. The first patients will be treated in the countries that are ready for them.
