Tried a little experiment building an alphabet out of just six bits. Came out sort of interesting, but the s, t, x, z are highly questionable.
Meanwhile, I'm excited about this "structure of information networks" class Jon Kleinberg is teaching this semester. It's only been two actual classes so far, but it's just neat little clean discrete math proofs and stuff so far. Very refreshing compared to the muddy empiricality of machine learning last semester.