I do use LLMs for some specific tasks, they can be quite good at some stuff but the general hype of it by non-technical folks trying to fit it into every single use-case under the sun is absolutely tiring... Having to explain for the n-th time why what we are trying to do is not a good fit for AI™ is exhausting, not because I have to explain it again but because I know I will have to do it again next week, at least another couple of times.
AI is being viewed in this hype as almost literal magic, it can do anything, we just have to wish for AI to do it (whatever the fuck AI means by now, it's just an umbrella for magical thinking).
I'm tired, and definitely bored.
I don’t use it because our products have the potential to harm other people and I’m not personally comfortable assuming that risk. Nobody else seems particularly moved by that argument, however.
That sounds... Good?
Lots of people run into using whatever is vogue without even thinking twice about it. See the massive move to using cloud for absolutely everything, money be damned, and you'll see what I mean. Cargo culting is a huge issue in the industry.
At least they're actually considering if it's making them more or less productive, compared to the vast majority of the ecosystem.
Just writing everything down isn't data. You have to have a standardized methodology and low-ambiguity signals. You should have systems of blinding.
What's your proposed metric for "productivity" for software developers? You could probably get a management prize for it.
At 1,000 hours of "debate", it is unlikely that anyone is considering if it is more or less productive, but rather are using that time to convince themselves that their position is the right one, even when it isn't.
I mean, if you take that not as an exaggeration but at face value, it's 200 days of 5 hours of discussions per day. I'm fairly sure the specific number is an exaggeration, but if it isn't, I'd probably agree with you :)