Judging from the technical description [1] and the most recent version of the draft white paper [2] alone, I think the project would benefit greatly from more substantial input and expertise from the academic cryptography community. I'm afraid they are spending a lot of effort cargo-culting on the wrong things, while missing where the real challenge lies: How to deal with the inevitable orders-of-magnitude increase in transaction size/complexity while preserving consensus-based distributed verification.
To be more precise: The metered computation mechanism is a very clever solution for dealing with unrestricted computation and unbounded data storage, but their proposed solution does not convincingly address the inevitable increase in space required. Hand-waving about greedy subtree verification without actual numbers/scenarios that shows this could work at all (which would be very surprising to say the least) is not convincing.
In all a great idea, though. It would be great if it works out.
[1] http://gavwood.com/Paper.pdf
[2] https://github.com/ethereum/wiki/wiki/Whitepaper-2-Draft