I like your comment, but would it really be so difficult to for the next generation of Glass to tether to a smart phone in your pocket that did all the computing and only used Google's cloud services for supplemental input when available? I.e. the voice recognition gets better when you've got a good data connection. The phone could be larger than average because you'd rarely need to pull it out of your pocket.

I don't see why you think there's such a large difference between local horsepower and cloud computing. They compliment each other, and the ultimate technology will be a mixture of both. User interaction is the much more interesting and difficult problem.

(And your battery life comment is misplaced.)