Buck comments on K-complexity is silly; use cross-entropy instead

[Buck]

Note that Nate’s proposal is not original; this is him rederiving and popularizing points that are known, rather than learning new ones.