To be fair LLMs can be used in an empowering way to learn, etc. But at that point it feels disingenuous - why not just one shot the problem and get the result more efficiently than just babying it around?

And yeah, fair enough. I guess my bio sample size is limited, but many of the PhD students I've seen in the experimental sciences seem to operate this way a majority of the time. I remember my brain being absolutely fried after 4 hours of thinking about math, only to hear these folks say they pulled a 12 hour shift in the lab. And I'm just like - how? In any case it's all anecdotes from me so feel free to ignore them.