Cursor with Claude accuracy significantly degraded

Bug Reports
1

Thanks for reporting a bug you have found in Cursor!
Please add the following info to help us diagnose your issue:

I am currently on the latest cursor on an M2 Macbook Pro.

The last few weeks Cursor has basically been unusable, specifically with Claude 4. The agent is not performing nearly as well as it used to. The big problem is that it’s not listening to feedback. It’s not debugging problems it’s facing. When the change it makes does not work the agent does not correct it’s mistakes. It just keeps digging in.

I can’t even get the bot to write tests correctly any more, it completely ignores conventions it used to previously follow perfectly.

2

I came to a similar conclusion. While working on a project for my tech startup using Claude 3.7 Sonnet-thinking, everything was running smoothly—fixing bugs, resolving errors, and even backing up the entire project. I was able to access it and launch the app without any issues. But now, trying to access anything has become unnecessarily complicated and inefficient.