Instant feedback makes coding more tangible but can be harmful. One may get caught in an edit-eval cycle and spend time solving the wrong problem.

Regularly stepping back and reevaluating assumptions/goals helps. (And write supporting tests, if you swing that way.)

For me this advice is about developing your spec to an appropriate level of detail before writing any significant amount of code. I'm still in the phase where writing code is fun, and it takes a little discipline to think my spec through enough to make my code-writing time efficient.