The landscape of AI coding agents is moving at a breakneck pace. Over the last year, I’ve done what many developers have: hopped from tool to tool in search of the “perfect” AI coding tool. I started with GitHub Copilot, hopped over to Cursor, migrated to the terminal-heavy power of Claude Code, and experimented with various CLI agents.

Full disclosure: I still actually maintain active subscriptions to all three tools. I like to keep my options open, but for my day-to-day heavy lifting, my preference has shifted dramatically.

I also have dedicated udemy courses on GitHub Copilot, Cursor, and Claude Code.

The Journey: Copilot → Cursor → Claude Code → Copilot

Why I Dislike Cursor Now

Don’t get me wrong, Cursor used to be amazing. They were the first to really show us what an AI-native IDE could look like. But lately? The experience has soured. It has started to become buggy (responses seem to hang or cut out fairly often), and they are constantly making updates to the UI that actually make it worse and more cluttered.

The biggest thing is that the “Cursor Gap” (that massive lead they had in features) has COMPLETELY closed. Being locked into their specific fork of VS Code feels restrictive when the core editor features start lagging or breaking.

Why I’m Moving Away from Claude Code

Claude Code is a beast. The reasoning power is intoxicating. However, the reason I don’t use it as much anymore is that I LOVE to experiment with models outside of Anthropic. Claude Code is obviously very tied to their own ecosystem.

With the way GitHub Copilot Chat has evolved, experimenting with different models (Anthropic, OpenAI, Google, etc.) is SO MUCH EASIER. I want that flexibility without having to jump through hoops.

Why GitHub Copilot is Killing It Right Now

GitHub didn’t just catch up; they refined the experience in a way that is SO MUCH MORE stable than the competition. Here is why I’m back:

1. This UI is SO Good!

   The interface has improved MASSIVELY. It no longer feels like a tacked-on sidebar. The Agent HQ is a total game-changer; it makes it REALLY easy to view previous and ongoing tasks. When you’re managing complex refactors across multiple files, having that high-level bird’s-eye view is EVERYTHING.

2. Multi-IDE Freedom

   One of the biggest downsides with Cursor is that you HAVE to use Cursor. GitHub Copilot Chat is now robust across a whole load of popular IDEs.

3. Feature Parity is FINALLY Here

   For a long time, Copilot felt like it was lagging in features compared to Cursor. Not anymore.

4. Model Switching is a JOY

   As I mentioned, I like to swap models depending on the task. Copilot Chat makes this SO EASY. I can jump from GPT to Gemini to Claude within the same interface. You can even run the Claude agent directly within GitHub Copilot now.

Final Thoughts

The “AI Editor Wars” are far from over, but GitHub has proven that you can be the big player and still innovate FAST. By integrating the best features of specialized agents into the tools we already love, they’ve created a package that is simply BETTER than the rest at the moment.

Despite still paying for all three, GitHub Copilot is the one I find myself opening every single morning. It’s just a joy to work with.