Microsoft Research Launches Project Quine. Five Pharma Rivals Pool Their Data to Beat Every Public Model.
Last issue frontier AI labs went deeper into life sciences, with Novo Nordisk signing Claude across R&D, Anthropic opening its own wet biology lab, and ARPA-H funding the first autonomous cardiology AI programmes. During the last two weeks the pattern spread: Microsoft Research introduced Quine, a multimodal world model of biology built with the Broad Institute, and used it to rank thousands of pancreatic cancer compounds in a weekend. Five top-twenty pharma companies confirmed that their proprietary structural data, pooled through federated learning, trains a protein co-folding model that beats everything built on public structures alone. Genentech committed over $1.5 billion to Earendil Labs' AI antibody platform, AbbVie signed Iambic days before Iambic filed to go public, and the US Department of Veterans Affairs put a $775.72 million ceiling on ambient AI across its hospital system. Three frontier labs now have their own biology research programmes in the open, pharma is choosing which AI partners its R&D runs through, and federal healthcare has priced ambient AI as production infrastructure.
