I’ve found that I need to switch back to the standard GPT-4 model to successfully apply the changes to my code. gpt-4-turbo-2024-04-09 almost always has issues when I try to apply changes. It will miss an edit, or leave out an addition, fail to remove something, etc.
I don’t know if this is an issue with the model itself or with Cursor’s implementation, but I wanted to leave this feedback and see if anyone else has observed the same thing.
I find that gpt-4-turbo-2024-04-09 can generate code much quicker in the Chat sidebar, but to apply successfully in a single pass I always need to switch back to GPT-4 before pressing Apply.