Anybody else notice that the latest version refuses to remember and use the context in the chat thread?

Until the last few days, I could start a chat thread, and it was reasonably good (not perfect by any means) at remembering and using the thread history. It no longer will do that. Each new message seems stand alone.

For example, here is a typical use case for me:

  • New thread
  • “Here is a framework that shows you my style of python coding. Use it for the remainder of this chat thread. [and then I paste in a python script]”
  • “Modify this script to add a CLI argument -s/–something and when provided, it calls a function that does [XYZ]”

And then chatgpt will basically do what I ask – give me a new version of the script with additional functionality that I’ve asked for that matches the style provided.

Another use case:

  • New Thread
  • “I will provide 3 chapters of a book I’m writing. Wait until I’ve give you all three chapters, and then give me a summary of the characters.”
  • And then I would proceed to paste each chapter into separate message windows.
  • Then chaptgpt would do as told and give a summary of the characters from all 3 chapters.

This just plain isn’t working for me right now.

Does anybody have the same experience?

1 Like

I think it’s working as designed if you really are opening a new chat for every message. GPT only remembers a few dozen prompts in one chat thread and other threads are completely isolated.

You can add generic personalized startup data for GPT if you wish it always knows something about you when you open a new thread.