Hi Cursor team and community,
I’ve experienced a troubling pattern with Auto mode that seems tied to billing cycles (or usage), and I need clarification from Cursor.
The Issue
Before My new billing reset:
- Auto mode refused to modify my ~3,000-line codebase
- Claimed code was “too long and complex” to understand context
- Forced manual coding for around 2 weeks, expanding code to ~5,000 lines
After July 23 billing reset:
- Same Auto mode now handles the 5,000-line code perfectly
- No complaints about length or complexity
- Works flawlessly on even more complex sections
I’ve seen the thread about “slow requests” when 50% usage is consumed, but that’s not the issue.
The problem wasn’t delayed responses – it was “outright refusal to make even simple edits,” only to work perfectly after the billing (or the usage) reset.
Questions:
- Has anyone else experienced Auto mode becoming uncooperative on medium-large codebases specifically near or after reaching the end of their monthly limit?
- Cursor team: Is there intentional performance throttling or model downgrading based on usage percentage?
This inconsistency makes it impossible to rely on Cursor for consistent development work. Given the recent pricing changes and user complaints, we deserve transparency about how usage limits affect Auto model performance.
If Cursor claims “Auto usage does not count towards your usage limits,” what’s the point when actual performance is clearly tied to the usage?
Please explain what’s happening.