Cursor struggling with simple edits

Thanks for the detailed report, @Margarida_Carvalhosa. I can see how frustrating this is - having AI make incomplete changes and ignore explicit instructions really disrupts workflow.

Others have reported similar issues with AI reliability on larger codebases. A few things that might help:

Project Rules:

  • If your rules are set to “Apply Intelligently,” they may not attach reliably. As Dan from Cursor explained here that models can fail to pick rules and too many instructions cause them to miss things.
  • Try: Set key rules (like “no print statements”) to “Always Apply” and keep them short and precise.

Also try:

  • Create new focused chats for each task
  • You can also try other models which might do better.
  • Add the 34k reference folder to .cursorignore if AI doesn’t need to edit it

Let me know if any of these help!