Ask HN: GPT-4 showing bad performance in coding tasks (02 Dec 2023)?

Yes, now it's giving me sub-par results for the same tasks.

I tried generating some Angular Jasmine-Karma tests, and none of them worked. (It used to be fine back in August and September.)

GPT-4 Turbo seems to be fast but not so good at coding.

It's been pretty decent for Python-related tasks that I've been asking it, mostly involving TensorFlow construction, pre/post-processing tasks, and unit tests for them, though still need to question it about certain decisions it made, etc.

It's never worked well for me. Lots of results just flatout don't/won't work, not even in a nuanced way.