I have following bug.
- embedd a GPT (A) of mine with @ in another GPT (B) of mine
- did not like the answer click on regenerate
- GPT looses (A) embedded and switch to the princiap GPT (B) and regenerate nonsense
I have following bug.
I’m surprised the OpenAI team missed it. There should be a unit test.
Anyways… it seems like a bug to me, maybe you can report somehow…
It is getting even worse, the prompt inside an other GPT and outside and even in the training window deliver all completly different outputs.
It is getting more than annoying that there is no choerence anymore.