Post

I had that issue with Github Copilot, never with Cursor and I thought it was because I was using Claude with Cursor. This issue seems to be worse if you check for helping them improve their AI with your data. As if you’re being used for A/B testing with excess recommendations.

Further on, I would say you’re doing AI programming wrong if you’re still typing manually. Stop that, send the file to AI and explain the change you want. Every time, even for small 2-3 line changes.

I agree on Frontend. For backend my experience is that there are fewer errors when I start programming and then let AI autocomplete as it "understands where I am going". Just recently it doesn't understand me very well

Take a look at Windsurf. The pricing is a bit off since the latest update, but I prefer it over Cursor. Or you can just use Cline with your own API Key. (already works with Googles latest Model for freeish).

Yes, I'm seeing people raving about Windsurf - that's what triggered my tweet. I don't wanna switch yet, as switching costs are just too high (relearning) - I'd rather want to get Cursor working

Not much of a switching cost. Everything looks the same. I switched fro cursor to Windsurf too (though time to time I switch back to check)

Never happened to me, but then I only use the chat/apply/composer feature at most a few times in a day. File a bug report maybe?