Been a few months since I used co-pilot, but they use a model that's worse than GPT-4/4o which is a big step down from the reasoning models.

Try out Cline, aider, or one of the tools devs actually use with the latest models from Anthropic/Google/OpenAI.

https://aider.chat/docs/leaderboards/

Didn't look through all the issues but there were things like

The agent was blocked by configuration issues from accessing the necessary dependencies to successfully build and test. Those are being fixed and we'll continue experimenting.

Been out less than a week, let's see how it's doing in a year.