Feedback on Using Cursor for 3 Months

Feedback
1

As a serial founder with no prior coding experience, Cursor has been a transformative tool for me. It’s the first time I’ve been able to step into the role of a technical founder, enabling me to validate ideas and execute every part of a startup independently. I cannot code without Cursor, and I deeply appreciate the efforts behind building such a complex yet essential solution. In my opinion, it embodies the right approach to integrating AI into coding, and I am genuinely thankful for this innovation.

My Usage Patterns and Challenges

How I Work:

  • I guide code changes primarily using business language.
  • I rely heavily on AI’s suggestions as I can barely judge whether the code makes sense.
  • I frequently tag entire folders and depend on AI’s code navigation.

Pain Points:

  1. AI Struggles with Multiple Files:
  • The AI often fails to read multiple files thoroughly, leading to inefficient suggestions, duplicate functions, or missing crucial pieces of code. It sometimes focuses on code it can interpret rather than addressing the real issue.
  1. Long Chat Threads Diminish Efficiency:
  • Extended conversations reduce clarity. The AI tends to lose track of the core problem, occasionally hallucinating irrelevant issues that are hard to identify and resolve.
  1. Inconsistent Adherence to Personal Instructions:
  • Instructions like introducing itself are consistently followed, but others—such as avoiding hardcoding languages—are frequently ignored unless explicitly added to each prompt.

Suggestions for Improvement

  1. Project Status Storage:
  • Implement a project memory feature to record development progress and key decisions. This could improve AI navigation across large or complex projects and help it identify and address issues more effectively.
  1. Incorporated Code Versioning:
  • Adding integrated version control could maintain alignment with project records, aiding AI in handling long-term projects. Allowing users to manage backups directly within Cursor would also enhance this functionality.
  1. Improved Side Chat Issue Management:
  • Create a task list feature where users can specify the exact issues they want to solve. AI should focus exclusively on these tasks to avoid confusion when unrelated problems arise during implementation.
  1. Enhanced User Rules for AI:
  • Develop a condensed, sharp version of user instructions (processed with AI) and integrate it into every user message at a low-temperature parameter. This could enforce adherence to user preferences more effectively.

Observations on Cursor 0.43.6 vs. Previous Versions

  • Improved Code Composer: The performance has noticeably improved.
  • Interface Issue: The side chat occasionally fails to display buttons to apply code, which remains frustrating.
  • Increased Instruction Inconsistency: AI appears to ignore user-defined instructions more often than in prior versions (based on subjective experience).

Cursor has been a game-changer for me. I see tremendous potential in the platform, but refining the consistency and functionality of existing features should be a priority. I hope these suggestions can help drive the development of an even more robust solution. Thank you again for building such a valuable tool—fingers crossed for further improvements!

1 Like
2

Hey, thanks for the feedback, we’ll try to consider your suggestions.

3

Hi, I was using Cursor also already around 3 months.
Generally I love it, you did an amazing job! Continue the same way.

But with the last update it just became a headache for me…Code implementation applyig is just became terrible. After last update I’ve noticed that in one reply some changes overwrites other, sometime changes in code not applying, sometimes it just dont let me accept the changes that brings mess. And I need to check each time how the code applied, if it applied and other. So at some point now its easier to code it manually.
But again this is only a negative exp with lates update.

Generally the project is very cool.