I just shared my link in this post: Major Issues in new GPT 3.5 : DO NOT DEPRECATE OLD ONE - #20 by SomebodySysop
This isn’t a prompt issue or “hoping for a better response” issue. It’s a model degradation issue.
Now, I’m not complaining too loudly because the GPT-4 codex model, despite a hiccup there and there, is doing an exemplary job as my coding assistant. I have built a whole working ingestion/query system, from scratch, in php, with it’s assistance. Very happy with the results.
But, GPT-3.5 and the GPT-4 Browsing models? I don’t know. Something’s not right.