Why does it look like everyone is abandoning GitHub Copilot?
GitHub Copilot is a powerful tool in my opinion. Most people I know use Cursor, Codex or Claude Code instead of Copilot, since Copilot is mainly an inline completion tool.
But doesn’t the "new" agent mode do the same things other coding agents can do? Yes, it has a much smaller context window, but it still feels very good to me. For the current subscriptions you also get far more than almost anywhere else: for 10 dollars a month you can send around 100 Claude Opus messages, without a limit on how long they run (because of the 300 premium request credits).
Right now Claude Code and other agents are hyped, but what about GitHub Copilot? do you all still use it?
I still use it as my main tool. I find it great for an existing code base because of the VS code integration by default. I can see how AI changes code I'm familiar. Plus, I think it's a great value. $10 for so much usage and access to Opus 4.6 and GPT 5.4.
It doesn’t do the same thing. Each has its own system prompts, tooling etc.
GitHub copilot sometimes doesn’t give you the model’s full context or something else.
The issue is also what does the model mean? Is Opus on Copilot the same as in Claude direct? There’s been lots of hosting providers trying to save money and cut corners. Some employ compression techniques.