We start by reading. The code, the need, the constraints. Then we propose. Not the other way around.

We prefer a small tool that holds to a large system nobody dares to touch.

When we get it wrong, we say so early. It costs less than hiding it.