Yitz comments on The case for becoming a black-box investigator of language models