Cursor at $20: My Honest Review After a Month

Agreed—these days it's definitely not like it used to be lol. I was also thrown off recently because code suggestions kept popping up in the wrong places.

Got a source? Is the claim that it got slower based on feel or actual benchmarks? I actually feel like it's gotten faster since the recent update—guess it varies from person to person.

Well, that's a bit... $20 is reasonable enough, but "worth it" is kind of vague, isn't it? If you break it down per hour, it's really cheap, but the problem is that time isn't always productive. I'm ultimately using it alongside Copilot. Cursor is strong for refactoring, while Copilot is more stable for plain autocomplete. Using both costs over $30 a month, and for that money... hmm.

Same here. I admit I can't go back to stock lol. Once you get a taste, there's no going back.

I miss the days when one tap would type it all out for me... Now I have to tap three times before it finally gets it right ㅠ

I don't agree. I used it for a month and ended up canceling. Once autocomplete starts guessing wrong often enough, you hit a point where productivity goes negative. Especially in projects with a mix of legacy code, I really couldn't make it work. I'm curious what stack the author is using. It seems like that would completely change how you evaluate it.

Oh, I didn't know this. There are a lot of comments saying it's gotten slower lately.

Have you tried switching models? When I switched from the default to a different one, it missed the mark way less often. Try checking the model selection in settings. Of course, even accounting for that, it does seem like it used to be smarter...

Isn't that old news? I tried it again recently, and compared with six months ago, it actually got better. It could be about the timing of swapping in paid models and region issues. Generalizing it as "it got slower" no matter what seems a bit hasty.