A caveat that "reputation", like competence, is more variagated and localized than is often appreciated. As with someone who is highly competent and well regarded in their own subfield, while simultaneously rather nutter about some nearby subfield where they don't actually work.
One can have a reputation like "good, but beware they have a thing for <mechanism X>". Or "ignore their results using <technique> - they see what they want to see". Subtext that gets passed among professors chatting at conferences, and to some extent to their students, but otherwise isn't highly accessible.
When people speak of creating research AI's using just papers... that's missing seemingly important backchannels. And corresponding with authors. Attempting research AI as developing-world professionally-isolated professor.
A caveat that "reputation", like competence, is more variagated and localized than is often appreciated. As with someone who is highly competent and well regarded in their own subfield, while simultaneously rather nutter about some nearby subfield where they don't actually work.
One can have a reputation like "good, but beware they have a thing for <mechanism X>". Or "ignore their results using <technique> - they see what they want to see". Subtext that gets passed among professors chatting at conferences, and to some extent to their students, but otherwise isn't highly accessible.
When people speak of creating research AI's using just papers... that's missing seemingly important backchannels. And corresponding with authors. Attempting research AI as developing-world professionally-isolated professor.