I think tlrobinson's point is that number-of-tokens renders a trivial objection to measuring code size such as "what if we rewrite everything to have one-letter

I think tlrobinson's point is that number-of-tokens renders a trivial objection to measuring code size such as "what if we rewrite everything to have one-letter variables" irrelevant, since that doesn't change the number of tokens.