This is a rant and I hope it will get the attention of OpenAI.
I am a power user if OpenAI's GPTs for a long time, and I am using them because, as it was the clear motivation from OpenAI as well, that I don't want to append a context message to my chats every few messages as a reminder of the context.
By the way by context I mean the 8000 characters of information that GPT builders are providing to their GPTs. I noticed that GPTs remember their contexts in the beginning and then they forgot about their context as the time progresses.
This means that whoever developed the code for GPTs, just made the context to be appended in the beginning of the conversation, and leave it this way with the hope that the discussion will not be too long, more than the 8k or 16k tokens that now are provided, which means a few pages of text.
If we forget for a moment that the context window in GPT's at least is probably far less than 32k tokens, (probably 16k tokens?), the real problem here is that this situation is the result of lazy programming.
Just appending the context as a first message when a "chat", a discussion is initiated it sounds like a "smart hack", but it is actually a very lazy and very sloppy programming. The right way to do it is by making sure to indicate to the GPT in each query what its context is and then proceed with the rest of the chat history.
If you want to make GPTs useful agents, then please fix that. It is just embarrassing, and if you are just too cool to get bothered with things like that, then just give us a decent context window, to at least have fighting change to see our work to getting done.
GPTs are not search engines with cool names, they supposed to have an active context.