Every time a dialog with LLM is cleared via (+), you can make a dialog summary in the background, asking LLM if it can extract the user’s wishes or preferences from this dialog to form new rules. And if it can recognize them, it can record them as rules in a separate automatically generated file.
2 Likes
Interesting…I’d like to learn how to do this. My simple application of this would be debugging. As others (at least one) have commented, it’s quite tedious to implement a change, switch to other IDE (xcode), clean, build, and identify issues. Then copy the issues into the agent for it to fix.
It took me a day of this tedium, note I’m not a Dev or Coder, I figured I could have a simple command in a text file I copy into the agent telling it to loop until it cleans and builds properly. While this is a bit better, it’s not great.
Also, I have noticed I can allow execution of commands automatically where Cursor runs various commands without manual impact. I’m not opposed to using this, but I’d be interested in others thoughts if this works/helps or is a risky approach.
Thanks,
Jeff
This would be so helpful. I just don’t want to be interrupted during my coding flow just to formulate as rule what I told cursor already many times.