But you're wrong about this:
> if each particle, at each step, made a two-way choice, that single particle could run through that 10^140 potential states in no time flat
No, each particle can compute 10^43 states per second, or 10^60 states in 10^17 seconds (10 billion years). I think that's where I got the 20 byte number. It's 20 bytes per particle, 60 for a universe of 10^80 particles computing in parallel.
So if you find you need more than 60 bytes to do anything, that just shows that you haven't properly optimized your representation ;-)
To make matters even worse, I just realize that the custom to write the logarithm in base 2 of x as ²log x (superscript digit two, 'log', 'x') may not be used world-wide and may even be specific to the Netherlands (http://nl.wikipedia.org/wiki/Logaritme#Natuurlijke_of_neperi...) That may have made this even more incomprehensible.
Also, the number of states that the universe has time for to compute may be a lot smaller than the number of possible states. For example, if each particle, at each step, made a two-way choice, that single particle could run through that 10^140 potential states in no time flat. Because of that, you cannot say "64 bytes ought to be enough for anybody"