It has nothing to do with English being terse, a claim that probably isn't even true for code when measuring character or rune or glyph length. It has everything to do with the proportional frequency of a language in the training corpus.