Reviewing plans is a waste of time.
It was useful in the past when planning cognitive load <<< implementation cost. But when implementation is free, just ask the model to bring you 3 complete, end-to-end tested solutions and iterate from there.
And if all of them suck, most likely the model was missing some crucial piece of context.
๋ ๋ณด๊ธฐ
