[vales]

you have to argue with everything ai says in order for the end result to be close to human reasoning (including arguing in favor of its initial position)