Critical Feedback on Cursor 2.0 Update (Version: 2.0.43 – System Setup)

Feedback
1

Subject: Critical Feedback on Cursor 2.0 Update (Version: 2.0.43 – System Setup)

I’m experiencing a significant drop in performance with the Cursor 2.0 update compared to v1.7.

I’m having a very poor experience with Cursor 2.0, especially compared to v1.7.

I’ve tested both Claude 4.5 (Max) and Opus 4.1, and both deliver a noticeably worse experience than before. Key regressions include:

  • Less intelligent responses: Answers lack depth and coherence.
  • Frequent mid-response cutoffs: The model often stops abruptly, even when a single task is incomplete.
  • Loss of multi-step execution: In v1.7, the AI could generate long-form plans, execute multiple sequential tasks, or complete an entire todo list in one go without interruption. Now, it gives very short replies and halts prematurely—even if the task isn’t finished.

Since upgrading to 2.0, I’ve had to temporarily switch to Claude Code Max. For comparison, Claude Code CLI handles problems coherently: it outlines a clear todo list and completes the entire list before stopping—something Cursor 2.0 currently fails to do (not even considering speed differences yet).

Critical Issue: Plan Mode Regression

  • In v1.7, Plan Mode worked well and was beginning to match Claude Code’s capabilities.
  • In 2.0, Plan Mode has regressed significantly:
    • It fails to update an existing plan properly.
    • When asked to continue, it loses almost all context and responds with extremely short outputs—sometimes shorter than auto-free mode.

This bug is confirmed by other users:
:link: 2.0 - Plan mode not update plan.md

“Same here, it will not update the plan, then when I ask again, it will overwrite the existing plan with the new changes. There is no way to save the plan.”

Limited Positives in 2.0

  • Multi-agent: An interesting concept, but doesn’t yet improve code quality or productivity in practice.
  • Browser interaction: This is the only reliably good feature in 2.0 so far.

Final Thoughts

Overall, Cursor 2.0 feels like a step backward. This sentiment is echoed by other users:
:link: Cursor 2.0 is not a good update!

“I had the same experience. I’m not paying for it… I cancelled my sub. It’s a bummer cuz I’ve had it less than a month.”

I hope the team prioritizes fixing these critical regressions in upcoming patches.

1 Like
2

…and QUICKLY…

Cursor has become in integral part of my daily workflow as a DevOps and Software engineer for a startup.

The Cursor agent, though far from perfect, has greatly increased my ability to deliver code.

However, in Cursor 2, the Agent is slower, less functional, more inconsistent in it’s behavior (between Agent instances) and far more error-prone.

This is a HUGE step backwards.