Small systems.
Real problems.

AI can produce code faster than most of us can read it. That changes the pace of building, but it does not remove the work of understanding. Every generated line still enters a real system, carries assumptions, and creates consequences.

The valuable interaction is not a single prompt followed by acceptance. It is a loop: explain the intent, expose the constraints, inspect the result, test the edges, and ask why. A person must remain close enough to the system to recognize when it is elegant, when it is merely plausible, and when it is wrong.

I use AI as a collaborator in that loop. The goal is not to type less. The goal is to think more clearly—and to remain able to explain, maintain, and take responsibility for what gets built.