adamShimi comments on AGI safety from first principles: Superintelligence