0.45 vs 0.46 Testing

Cursor 0.45 vs 0.46: Experiences and Feedback Request

Hi everyone,

I’m sharing this post to compare versions 0.45 and 0.46 of the Cursor tool, and to invite you to log your specific use cases and experiences here.


My Testing Process

Setup:

I downloaded both Cursor 0.45 and Cursor 0.46 as separate versions, renaming them accordingly. After installation, I opened the same projects in both versions and executed the identical command within the same composer environment.

Test Execution:

Cursor 0.46:

• I ran the same command 10 times using the same model. Unfortunately, in about 5 to 6 of those runs, I encountered connection issues with Anthropic.

• For the remaining runs, I experimented with different models (including 3.7 Thinking, 3.7, and 3.5) in an attempt to resolve the error. Despite this, I repeatedly received nonsensical responses.

• These responses not only attempted to fix the script but also, inexplicably, created new scripts in entirely different directories instead of correcting the existing one. The error persisted throughout all trials.

Cursor 0.45:

• When I performed the same tests using Cursor 0.45, both the 3.5 and 3.7 models in one shot were able to resolve the error I encountered. This clearly indicates that something in the handling between the versions has been compromised in 0.46.


Concerns and Call for Feedback

I conducted these tests in a controlled manner and observed that version 0.46 is not working as expected. This leads me to wonder whether the Cursor team is fully aware of these issues or prioritizing their resolution. Moreover, it seems that new versions are being released without addressing these critical problems. Frankly, I find the current explanations provided by the Cursor team to be quite insufficient, and it appears that they might not fully grasp the severity of the situation.

I kindly request that you share your own experiences with both Cursor 0.45 and 0.46. If possible, please test the same use case across both versions and log your results here. Your input will be valuable in providing better feedback to the team.

My previous topic was flagged as inappropriate. As a regular customer, I’m asking the team not to do that again. I pay for this service and believe I have the right to share my ideas freely, as long as I’m respectful.

4 Likes

I have not had issues where it messes things up as badly as you described here, however, my main issue is that I either get “Connection failed” or “cannot reach Anthropic”. This happens anywhere from 50-80% of the time on average. Sometimes it will not work after 10+ retry attempts and I have to give up and paste my context into chat gpt to get help.

how big is your codebase? (number of scripts/length of scripts)

what version are you using?

Current codebase is about 223MBs including the venv and caches. Its a full FastAPI backend. Not huge but not super small.

Version: 0.46.7 (Universal)
VSCode Version: 1.96.2
Commit: 3611c5390c448b242ab97e328493bb8ef7241e60
Date: 2025-02-27T07:15:37.748Z
Electron: 32.2.6
Chromium: 128.0.6613.186
Node.js: 20.18.1
V8: 12.8.374.38-electron.0
OS: Darwin arm64 23.6.0

which 0.45 version are you using? 0.45.14? I am trying to see which 0.45 version to download - GitHub - oslook/cursor-ai-downloads: All Cursor AI's official download links for both the latest and older versions, making it easy for you to update, downgrade, and choose any version. 🚀

Version: 0.45.17 (Universal)

I’m using this one. Working fine for me