How Is Cursor Different Than GitHub Copilot or Sourcegraph Cody?

(assuming default configurations, and this was my personal experience in a 4 year old software project)

Copilot felt like an incompetent colleague you needed to slap to get anything coherent out of it.

Cursor feels like an overly eager junior colleague that wants to refactor everything. It works, sometimes it makes sense. Your project overall increases in velocity anyway.

Cursor include an agent that can write code for you, fix errors on its own. Copilot is mostly autocomplete.

Cursor is far superior at indexing files, i.e. knowing which part of the code does what. This is probably their special sauce. There's some codebases that it can't index and Copilot works better there.

Like IDE vs text editor, some people have different comfort levels on how much code is written by the AI. Copilot is lighter, Cursor is in the middle, and on the other extreme, you have Devin which you chat with like it's a person.

I personally think it is very much worth it. But they have a liberal free trial, so you can find out for yourself if it is worth paying for your particular situation.

I use it regularly, and used to use GitHub Copilot.

I think Cursor is probably overhyped and not worth the money, and I want to explore switching back to Copilot or trying Cody, or other alternatives. However, it's hard for me to test against GitHub Copilot because the latter's tab-completion conflicts with Vim keybindings in VSCode. If the AI-autocomplete from GitHub Copilot can get the job done, I would not keep using Cursor, I would pay the smaller amount for Copilot instead.

I don't use the Cursor chat much, because it's worse than just copying and pasting code back-and-forth between IDE and "native" ChatGPT. IME, it is better than Copilot's though.