I was very sold into Claude Code. Tried Windsurf, tried GitHub Copilot, tried TRAE, and none of them seemed to compete to Claude Code. The simplicity and effectiveness of Claude Code was better for me.
Then, Cursor 2.0 was released and it has been slowly gaining my attention. Right now I am barely using Claude Code, which is surprising even for me.
The very first feature was not actually in the Code Editor, it was the Composer1 model. It’s fast, VERY FAST. The other models are turtles next to it. Even Haiku which as a considerable improvement in speed compared to Sonnet feels slow next to Composer1. Obviously it’s not as smart as Sonnet but for the speed it provides, it is really good. Probably better than Haiku. It has become my go-to model for simple tasks.
Then when we go to the specific features of the Code Editor, here are the ones I’m regularly using:
- Integrated web browser - this is a killer feature. Being able to inspect components and pass them as context is a game changer.
- Planning mode - it is getting better, I still think that Claude Code has way better planning but this one is convenient since you can plan with
Sonnetthen execute withComposer1. - Agent Mode - I’m not using it as often, but it’s another convenient feature. Sometimes, I appreciate the freedom of switching to a Conversation mode rather than having a code editor in front of me.
The only thing that I still get frustrated about is how to organize the steering files for the AI. In cursor, there’s the rules, in Claude Code, there is the CLAUDE.md files. I guess that I need to find a way to standardize these files in the docs/ folder and have these read from there so there is a single source of truth.