“The only thing that can stop a bad guy with an LLM…”

Yes that's a law! But perhaps it can be specified further by using any drop in kids-to-caregiver ratio? 🤔

I'd say we have to get used to AI all over the place fast, that's not all bad but proper documentation should become code if conduct. More and more urgently needed guidance is coming: doi.org/10.1186/s410...
And teachers, reviewers and editors will need to sharpen their bullshit detectors...

Fair point, totally agree. Yet it's often an unfortunate situation for researchers, and trying to get such a paper peer-reviewed can be very bothersome. I think, once more, this makes the case that registered reports should become the normal way to do empirical research.

What if authors realize they didn't sufficiently prespecify quality checks or positive controls and without a positive finding the results are not interpretable? (One could argue the study itself the QRP in that case and you're doing the world a favor not publishing it)

Is that Darwin, the guy who got famous for removing essentialism from species in biology, and who then thought "Hey, let's reintroduce essentialism for emotions. They're inherited and fixed. Flies and frogs get angry and sad just like humans. Quote me on this forever, don't ask any questions."?