What comes to mind is the MIT team that recently used genAI to design novel antibiotics against drug-resistant bacteria.

If the agents get access to the right data (tsunamis) already out there, why not?

This could start with conducting highly detailed data reviews to test or refine specific new or existing hypotheses.