For me the main advantages of Cursor are GPT-4 and a much better codebase context.
Copilot doesn’t use GPT-4, so the answers will generally be worse.
Copilot also only sees the code that’s currently visible inside your window. In Cursor, you can quickly throw a bunch of full files into your chat or automatically retrieve the most relevant code snippets from your entire codebase. This gives you many more chat possibilities and makes the answers even better on top of having the GPT-4 advantage.
However, I still like to use the normal Copilot autocomplete inside Cursor. These very quick and short auto completions that run at all times are something that Cursor currently doesn’t have.
copilot chat also has this annoying bug where many queries are ‘filtered’ and error out - Oops, your response got filtered · community · Discussion #56134 · GitHub - it’s been like this since May this year and renders the chat function more or less useless, unreal they haven’t looked at it. I don’t know how a trillion dollar company lets one of their flagship products sit in such a state for so long
once i got the hang of adding custom references in Cursor, then i fully switched. ( @ and then add the docs link to something if not already ingested.)
the other thing i’ve been playing with is running CoPilot and Cursor together. CoPilot can do relatively quick code completion, while Cursor w/ GPT-4 can handle the questions that need the full repo context (or my specific @ context).
i’m currently all-in on Cursor, but it is a rapidly evolving landscape so i hope these young chaps are ready to run
Actually GitHub copilot chat can use the whole workspace as context and use gpt4 if needed (You can see that in the “output” window of vscode)
Also answers quality are good (with almost no hallucinations)
Inline chat on the other hands is really bad