From reading the proof, I think he meant “n composite => 2^n - 1 composite” and it’s just a typo