Anthropic buys a one-person biotech to develop its own drugs

Developing a new drug from scratch used to require dozens of specialists and millions in capital just to get past early screening. Anthropic is showing that a single founder with an AI copilot can now challenge that playbook.

The AI lab just revealed that it is running its own internal drug discovery programs and has acquired Mithril Medicine, a one-person biotech startup whose Claude-assisted candidate for muscular dystrophy went from idea to promising mouse data in under 18 months. The drug could reach human trials as early as late 2028.

The one-person biotech pipeline

Mithril was founded by Cody Tranbarger, who ran a pseudonymous biotech account on X named Professor Oak for six years. Tranbarger describes his setup as a pipeline-in-a-person.

Instead of hiring an army of scientists and ops staff, Tranbarger used Claude to manage contracts and coordinate scientific testing across 11 external contract research labs. Claude essentially acted as his scientific coordinator and regulatory assistant.

Anthropic also released fresh data from commercial partners using Claude in production. At biotech company Enveda, Claude generated a first draft of a critical regulatory filing in just four hours, a task that typically took human researchers three months.

The system slashed human review time on drug manufacturing paperwork by 89%, while catching seven batch-record errors that human reviewers completely overlooked.

Designing 35,000 antibodies in a month

Claude is also moving into molecular design. Divergence Labs used the model to design roughly 35,000 antibodies in about a month.

Lab tests showed that three of those designs outperformed today's leading commercial drug at stripping allergy triggers from cells. In other laboratory experiments, Claude helped achieve 10 times more receptors on cancer-fighting immune cells.

Anthropic's ambition goes beyond pure computation. According to its IPO filing this week, Anthropic runs its own wet lab and Claude can now operate physical lab equipment, including microscopes.

The global pharmaceutical market is expected to top $2 trillion this year. Still, Anthropic tempered expectations in its filing, explicitly warning investors that its proprietary drug discovery bets may never generate revenue.

Why it matters

For independent scientists and boutique biotechs, this is a watershed proof of concept. The regulatory filings, vendor negotiations, and screening workflows that once required deep venture backing can now be orchestrated by a single researcher. If you are developing preclinical compounds, using Claude to streamline paperwork and run virtual biotech operations is already worth doing right now.

For the broader AI race, Anthropic is pivoting away from just selling tokens toward directly capturing value in physical, high-margin industries. Dario Amodei teased back in August that early glimmers in biology were coming, and running in-house wet labs proves Anthropic wants to be a direct pharma player rather than a passive software vendor.

We think integrating digital reasoning with physical lab equipment like microscopes is a brilliant strategic move to build proprietary scientific datasets. The catch, however, is that biology does not follow Silicon Valley timelines. Anthropic's candid warning that these drug bets might never yield revenue is a sharp reminder that preclinical success in mice rarely guarantees clinical success in humans.