Thanks for great article. I’ve been exploring the same topic and feel the same way about limitations of statistical methods and entropy. General purpose procedural compressions algorithms ought to be explored more thoroughly.