SoerenMind comments on Technical AGI safety research outside AI