I've been reading up on different text-based encoding schemes yesterday. Anything above base64 is really getting esotheric. There is base85 which is apparently used by adobe PDF. It maps 4 bytes of binary data into 5 ASCII characters which is a few percentage points more efficiency than base64.

Then there is Base2048 and Base65536 where it gets really weird because they use a much larger part (or the entirity!) of the unicode character space (all international characters, all the strange symbols etc). I don't know what these tokens would look like but Base65536 can encode 2 bytes of data into a single character. That means a 100 byte Cashu token could be 50 chars (that's insane). I don't know if all compures would be able to render these characters but if so, you could basically fit ~2.85 Cashu tokens into a single 140 char SMS.

Wonder why I haven't seen people abuse this to store large chunks of data on Twitter or other text-based media. I've seen projects that abuse Twitter as cloud storage (by storing data in forms of tweets) but they usually use base64.