How I read papers now, with the help of a large language model and a lot of suspicion

I read more papers than I used to, and I trust them less. Not because the papers are worse but because I’ve gotten better at noticing when I’m being told a story.

A workflow piece — what I actually paste into a model, what I never paste, the questions that consistently surface the weak parts of a paper, and the embarrassing number of times I’ve been wrong because I trusted the summary instead of the methods section.