Anthropic’s first drug pipeline came from a one-person company. On October 27 it disclosed buying Mithril Medicine, a Cambridge, Massachusetts outfit, back in July.
Founder Cody Tranbarger was that person. Eight paid consultants and 11 outside laboratories sat around him, coordinated largely through Claude, a model he calls pipeline-in-a-person. Two rare muscular dystrophy efforts are now the first drugs Anthropic has run internally.
The mice, and the caveats
The lead program targets LAMA2-related congenital muscular dystrophy, which weakens the anchors holding muscle fibers to surrounding tissue. Mithril built bispecific antibodies meant to bridge that gap.
Less than 18 months, by Anthropic’s count, separated the starting point from mouse proof of principle. Detail sits in a technical report awaiting peer review. Grip strength and peak triceps force both roughly doubled in seven treated mice, each about six weeks old, compared with four given a control solution.
Then come the caveats. Nothing randomized dosing, since litters served as the unit, and the treated cohort was read out 126 to 161 days after the control cohort with the colony moving midway. The report asks for blinded replication.
Six partners, and some honest error counts
Alongside the deal, Anthropic published early results from BigHat, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and Column Group.
At Enveda, Claude read 1,061 pages of manufacturing records, matched all 11 issues human reviewers found, spotted seven more, and cut review time 89%. At Divergence Labs it generated roughly 35,000 antibody designs within a month.
Failures were disclosed too. Three of the six Claude errors it examined left the building in emails staff had adjusted and sent. Two others slipped past scrutiny for more than four days.
Mithril aims to name candidates in 2027, with LAMA2 human trials possibly by late 2028.
