Copilot versus Cursor: (This is so embarrassing for Copilot) I opened the same project in Visual Studio Code and Cursor. I asked both agents to complete the same task: • Cursor: 34 seconds • Visual Studio Code: 122 seconds (4x slower!) At this point, Copilot is so behind Cursor... that it's sad to even remember they were once ahead.show more

This was using the latest version of Visual Studio Code. Changing the model doesn't help at all. It's not that Claude 3.5 is slow; the issue here is with Copilot. This experiment wasn't even that bad. At least Copilot completed what I asked it to do! If you try, you'll see that it fails many times or just spins forever in the middle of whatever you are doing. The Copilot agent is just a pre-alpha product. This is not working software.

You should compare @windsurf_ai and @cursor_ai .

I tried the same by generating requirements through and then using them in Copilot and Cursor. Still, copilot losses, but not this bad.

earth really changed in 88 secs extra taken by Copilot, so damn it and make everything superfast

I can't read whether you are trying to be sarcastic. If you are, this is a very low IQ take. If you are not, ha ha ha.

It's truly a generational fumble

Have you tried roo code?

It’s amazing how much CoPilot fucked up. Multi-billion dollar fuck up. Don’t see that every day.

A lot of it depends on the model you or the IDE selects.

Copilot is bad regardless of the model in everything I've tried.