Vast majority of work we do while building something, is to read existing code, understand the context and then add more to it.
In cursor, we select a piece of code and just press CMD+L to chat and get clarity on the context. Rather than jumping across and reading files manually to understand everything.
Other majority of work goes into planning out new work in terms of how to build something, referring to standard library and other tools.
In cursor again, just command + L and ask on standard library functions/methods one at a time, in a single chat thread. And potentially ask it to generate first piece of code right here in your editor, to get you started with rest of the implementation. Without Cursor, you will have to google it, go to the docs, read on about the function, repeat the same until we get fully clarity, come back to the editor and start working.
Later, when the code you wrote doesn’t work as needed, and you don’t seem to get why, just ask. Command + L and you will get possible mistakes and options to fix them.
The time saved on context switch alone in the entire process is huge!
From our usage, I see that the accuracy of all this response Cursor gives is usually very very good. We never switch the default model set in our editor - Claude 3.5 sonnet. I’m not sure why others would switch to a specific other model. Have you found a specific model useful for a specific usecase?
What are your thoughts on CoPilot and Cursor? Are you using one of these? I’d love to hear more about your experience.
