Earlier quoted context omitted.
I'm interested too. I'm guessing the data is actually encoded into the character as they do look ever so slightly different with different content. This would suggest this means that you can create a dynamic font/character in unicode. I had no idea this was possible, can someone confirm/slap me down for saying silly things?
I'd like to see the algorithm used to do this sort of compression.
It's not actually compression, according to the author, the 'single character' is 6x larger than the original UTF8 bytes.