I think that's the first time I've heard someone specifically use the concept of Huffman coding in evaluating language design. It makes a ton of sense, and maybe it's a more well known viewpoint than I realize.
Larry Wall has been using it for quite a while. His essay about the problems with regular expressions as they’re known and used today is insightful and carefully argued: http://dev.perl.org/perl6/doc/design/apo/A05.html (from 2002)