I mean, to be fair, "knowing all of the relevant literature" is a great first step for solving a problem! And a LLM can probably be better about not forgetting

I mean, to be fair, "knowing all of the relevant literature" is a great first step for solving a problem! And a LLM can probably be better about not forgetting than humans are (though I'm guessing they do much worse on the hallucination front than humans do- humans tend to know when they are playing a hunch). But "This paper from a lower-tier journal two years ago suggests you should look at X" is a very valuable thing to have!