Question for my science peeps. I’ve been resisting using AI coding tools for almost anything beyond documentation looking while building my data analysis pipeline for a few reasons. From most practical to most personal: (1/n) 🧵
1) I feel strongly that I should know and understand what every single line in my code is doing. A mathematical error/hallucination/magic number in my data analysis code would be an absolute disaster. (2/n)
2) Even if it’s 100% accurate, I feel like it takes away my opportunity to learn and integrate that knowledge into my bones.
3) My experience as a writer tells me the writing it produces is bland and mediocre… I’m a less experienced coder but why would I assume its code is any better? (3/n)
Jun 18, 2025 14:534) Copious ethical concerns (stolen work of artists and stack exchange contribs alike, horrifying climate impact, etc.)
5) Personal bias against techno utopian overhype. The gap between the usefulness of the tech and the amount of investment is a canyon, and that canyon is filled with hype (4/n)
So my question(s): am I being unreasonable? Don’t all tools prevent you from learning the old way of doing things? Is it producing excellent code I can’t hope to match? And is there ANY justification for the ethical concerns re:coding specifically? Would love to hear people’s thoughts… (fin)
Documentation lookup* … dammit autocorrect. (Actual fin)
