For anyone who doesn’t read the article, the biases shown in the thumbnail are not the final result. After doing a million runs, every digit had close to the same probability of appearing.

Love it! He had an assumption, created a method of testing that assumption, tried it multiple times, was proven wrong and accepted it.

Now, if more people could operate this way…

Anti Commercial-AI license

Pretty cool to show that sample size matters a lot during testing…

Sample size = 10: “There’s 20% 8! WTF, should be 10%”
Sample size = 10k+: “Oh wait nevermind”