The GPT Seems to lose its aces as it keeps performing less and less accurately to the user’s requests while also not following instructions at all 
!
Even given a clear long instruction and request, it will keep doing it by itself without following the instructions and meeting the request.
I found other threads that explore the same issue. Experiencing poor performance as well as of lately. It needs to be addressed or OpenAI will lose lots of subscribers soon.
I agree, performance and reliability have deteriorated to the point where clients are declaring that this tech is not ready for prime time.
Accuracy seems random at best, and lately it is bad most of the time.
And here we are pulling our hair trying to come up with prompts that will make the GPT answer more reliably, but I think the problem is really lack of compute power in the back-end to pay sufficient attention. No amount of Prompt Engineering can make up for that.
I hope this is fixed quickly, otherwise GenAI will become a joke and will set back adoption of this tech by at least two years or more.
A victim of its own success.