Randomized, Controlled comments on Simple distribution approximation: When sampled 100 times, can language models yield 80% A and 20% B?