> If it has no concrete reasoning, how is this looped revision intended to be more-accurate It can work if the reasoning (the logic) is part of the world model

> If it has no concrete reasoning, how is this looped revision intended to be more-accurate It can work if the reasoning (the logic) is part of the world model and is similarly refined. Then, like in the normal case, you gather a good amount of information than draw your conclusions based on you logic and the "wisdom" built. Were this automated, you would have an automated champion in judgement - doing its best with the imperfect data and function available. Like us, just better.