When John Pak and his collaborators tasked a team of A.I. bots with redesigning part of a Covid-19 vaccine in 2024, he doubted they would develop something successful or coherent. At least they’d learn something about the deficiencies of these models, he thought.

The lab of A.I. agents had hundreds of conversations to brainstorm, unburdened by distractions like eating or sleeping. Within a few days, the results were in: dozens of novel proteins that would theoretically bind to new Covid variants. Testing showed two of them actually worked — created at a fraction of the time and cost it would take a lab of humans.

Dr. Zou’s team at Stanford University is working toward fleets of A.I. agents as “co-scientists,” independently generating hypotheses and running experiments with human oversight. In a new paper published Thursday in Science, the lab reported that tens of thousands of A.I. agents predicted an approach to attacking a protein involved in lung cancer. Weeks later, a large pharmaceutical company announced promising results using the same approach.

Kyle Swanson, a former graduate student in Dr. Zou’s lab, called it “a new era” in drug discovery.