“I’ve noticed what might be called a “law of subverting successful approaches,” by which I mean once you’ve hit on something that works, don’t expect it to work again,”