Cursor 2.x (auto) seems have degraded AI

Ideas Feedback
1

I lately noticed that Cursor since v2.x and now v3.x (auto mode) is getting worse than its v1.x in that it behaves more like robotic than 1.x. By ‘robotic’, I mean it has less logic reasoning capability. It has increased my code review efforts to ensure software engineering quality and thus slowed down my code review cycle a lot. Is this a result of the new versions training on cheaper model from a ‘competitive country’, as hearsay?

I hope Cursor sticks to its quality and improves the situation. Or, you will soon see that people leave for ClaudCode or Antigravity.

BTW, here is the version I am using:

Version: 3.5.38 (system setup)
VSCode Version: 1.105.1
Commit: 009bb5a3600dd98fe1c1f25798f767f686e14750
Date: 2026-05-26T21:32:06.537Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Windows_NT x64 10.0.26200

5

Hey @Chris_Shen

Auto is ultimately a model router not tied to a specific Cursor version, and our goal is that the models behind Auto have only gotten better, not worse! If you have any specific examples where reasoning went south (great if you can share a Request ID), happy to look into it!