As the years go by, I'm finding myself being able to rely on those less and less, because every time I do, I eventually get disappointed by them working against their user base.
For me this been a pretty fundamental shift, where before I either had to figure out another way so I can move on, or had to spend weeks writing one function after learning the needed math, and now it can take me 10-30 minutes to nail perfectly.
Just think about the type of code these things are trained on and the fact you’re clearly some random non-specialist.
That's just a misunderstanding, I'm not "vibing" anything. The tests are written by me, the API interfaces are written by me, the usages are written by me, and the implementation of these functions are written by an LLM, but reviewed to be up to code standards/quality by me.
If a function gives me the right output for the inputs I have in mind, does anything beyond that really matter?