Stanford Publishes Biomni in Science: The First AI Agent That Autonomously Conducts Biomedical Research

My take: What Stanford published this week in Science is not a medical chatbot or a literature search tool: it is an agent that reads research, forms hypotheses, selects datasets, writes code, interprets results, and proposes next experiments, all within the same workflow. Layered on top: 150 specialized biomedical tools, 105 software packages, and 59 databases covering all 25 subdisciplines defined by bioRxiv. More than 10,000 labs are already using it in production.

The data point that best illustrates what this means in practice: a data analysis task that normally takes a researcher 60+ hours, Biomni completed in 40 minutes. With full citations and step-by-step traceability, making the science more rigorous, not less. That is real efficiency, not marketing.

For anyone working in biomedicine, bioinformatics, or health sciences, this changes the scale of what is possible without a team of ten. Whoever learns to integrate agents like Biomni into their workflow will produce at a speed that was previously impossible for a single researcher or a small team.

The question I keep asking: if agents like this already exist for biomedical research, which scientific or professional field will be next to get its equivalent this year?