[ko_pivot]

> In fact, Cursor’s code completion isn’t much better than GitHub Copilot’s. They both use the same underlying models

Not sure that this is true. cursor agent mode is different from cursor code completion and cursor code completion is a legitimately novel model I believe.

[alook]

They definitely do train their own models, the founders have described this in several interviews.

I was surprised to learn this, but they made some interesting choices (like using sparse mixture-of-experts models for their tab completion model, to get high throughput/low latency).

Originally i think they used frontier models for their chat feature, but I believe theyve recently replaced that with something custom for their agent feature.

[tom_m]

You can have Copilot use Claude now but I'm not sure it's the default. I found the latest Gemini Pro 2.5 to be much better than Claude Sonnet anyway...but yes, these are all more or less the same at this point.

One thing people don't realize is there's also randomness in the answers. Also, some of the editors allow you to tweak temperature and others do not. This is why I found Roo Code extension to be better.

[tough]

According to this article, Cursor's tab completion is SOTA, and was built by a genius that also created some other important foundational libs on AI and worked briefly at OpenAI

https://www.coplay.dev/blog/a-brief-history-of-cursor-s-tab-...