I wish Cursor would help me to keep track of these daily use limits per model in the interface. I could try to remember it, but in the end I would probably have to write it down per model. Having a counter next to the model selected would be more convenient.
Does that really work for you? I expected the codebase indexing to have an effect on being aware of the project structure, but it doesn’t. The only way to make the queries aware of it is to feed it via the .cursorrules file.
I recently upgraded to the Pro plan, but I’m doubtful about renewing. The issue for me isn’t the 500-request limit itself—I think that’s a reasonable amount for $20. My main concern is the repeated instances of poor-quality code generated by the AI, which causes me to burn through those 500 requests far too quickly.
For example, if I make 500 requests, the AI claims it has fixed the issue, but I often end up needing to request additional fixes for the same problem multiple times—sometimes up to 400 attempts—just to get it working. Why should those repeated attempts count against my 500-request limit? This model doesn’t benefit anyone and feels unfair.
Is that still the case for you or is it working better for you now?