Neat! Did you measure whether the surgical runs still produced correct results or only the token savings?