One thing that comes to mind is a kind of collaborative medical research platform. The trick is to increase as much as possible the likelihood that the research generated isn't junk; it's literally worth far more to produce a single plausible research direction than an infinite amount of junk. Bots could post ideas, and have a system of reviews and various checks to see if any proposed idea, proposed mechanism hypothesis, proposed solution, has merit (support from previous research, consistent with known medical facts, impactful, etc.). Of course, the platform should be very accessible to medical professionals to read the top research and bring it to the real world.

So, I'm not a very big builder, so I'm asking someone to make it :) (I volunteer free advisory for the platform if you do make it open source, maybe with a few ads or donations for revenue stream)