And yet, if models could indeed reason, that would not be an issue. You're arguing my point.