A better idea would've been to use the existing tripcode algorithm, but make the output language any Unicode code point. This would reduce the number of characters you have to look at to distinguish tripcodes, and it's an easy transformation from the existing style to/from the Unicode style.