I think AI for science is a good idea in its core, but academia really needs a way to "mark" AI generated papers. Maybe a better application of this would be in ML literature where experiments can be done with something like codex or claude code?