I hope this makes sense! Consider this scenario: you’re chatting, it suggests a code change. You go to the editor making the change. You run your code, get a new error for example. You select the error from the terminal and click ”Add to chat”.
Since I ran the code based on its suggestion the circumstances changed. The error message may be the same, but the context is different since the code is different.
So I expect to attach the new error to my new message so that GPT knows the context of how we got to this error, based on its previous suggestion. My expectation is not to edit my previous query.
However what happens now is that when you focus on the editor and attach something from the Terminal or the Editor, it attaches the code to the old message.
Always consider the new message textarea a priority, not the previous one.
I mean I don’t know maybe a lot of people polish the query instead of having a conversation (wonder what’s the point, you still spend your tokens, maybe just keeping the convo clean?). But for me, if I wanted to edit the initial request, I’d explicitly go there.