The interesting problems don't have a closed feedback loop. You don't know what the requirements are until you try something yourself and decide that isn't it.