I would challenge that "languages are more stable when they're only read by a select educated few" -- what's more stable is the written version of the language (as, for instance, written Chinese). The spoken language kept on changing and splitting, but the written records that we look at appear to show less change. Now, writing technology is easier to access and share, so we're seeing a broader swath of people using the language . . . personally, I think we have to think a lot more about sampling before we can compare the evidence we have for different periods.