“David Friedman states that tokenizing Mandarin requires 2 to 3 times more tokens than English, and source code requires 1.5 to 2 times more, which would lead to unfair taxation under a flat token tax.”