Entropy of a Large Language Model output | Hacker News