Another reason why the idea of AI agents for science hasn't made much sense to me. Research is an extremely collaborative set of activities. How good would a researcher be who is very good at literature review, but never actually talks to anyone, goes to any conferences, etc?