After developing 10 projects with Cursor, here are 7 real experiences I’ve summarized!
1. Give AI a “memory memo”.
AI has a poor memory. When reconstructing large projects, it’s easy to forget which files have been modified. You can create a “memo” for it. For example, create a “to be processed.txt” file and list the files to be processed and their processing status. This can greatly improve efficiency.
2. Put “instructions” on the code.
Programmers usually hate writing comments. But after using Cursor, adding clear comments in the prompts not only makes AI understand the requirements instantly but also earns praise from colleagues during code review. At the same time, you can ask Cursor to add specific and detailed explanatory comments when completing the code to improve code readability.
3. Let AI be your “test assistant”.
Writing tests used to be a headache. Now you can let AI automatically generate test cases and analyze test results. Writing unit tests also has the advantage that AI can judge the execution result and then make the next modification according to the result, allowing AI to complete the entire project process independently.
4. Modular development for twice the result with half the effort.
For a project, trying to implement the entire large function at the beginning may cause卡顿. You can first let AI design the framework, then split it into small modules, and finally assemble and integrate them. This can improve efficiency.
5. Don’t panic when encountering new technologies. Let AI help you find information.
When you receive a WebGL project and don’t know how to do it, you can throw the online documentation to Cursor and let it search for the latest documentation and look at code examples to quickly get started. This is a boon for technical novices.
6. There are tricks for reconstructing old projects.
When taking over “legacy code”, such as a project using Koa 1.x, you can create a new project and let AI help with reconstruction.
7. Multimodal development makes products more outstanding.
When working on a web project, I found that Cursor can recognize pictures and generate interface code from design pictures. The operation is very impressive.
After using Cursor for so long, the biggest feeling is that AI is not omnipotent, but it can really save a lot of trouble. The key is to learn to “communicate” with it and make it a powerful assistant. For example, the Koa upgrade project that has been delayed for several years is thanks to this little AI assistant. If you are also using Cursor, welcome to share your usage skills in the comments section and let’s make progress together.
