Trust and Verify Your AI Coding Agent

I was working on a project recently using VS Code with GitHub Copilot. I used the agent mode with Claude Sonnet 4 model and I had a few imperfect experiences.

In one instance, the agent mentioned that it reduced the time complexity from O(n²) to O(n), but it wasn’t O(n²) to begin with.

In another instance, the agent created code that is not being used at all:

In summary, we should trust and verify the output of our coding agent, at least for now.