My initial read of their paper was interesting, but it was two to three orders of magnitude more resource intensive than would be required to make it actually viable.  ... This is still impressive since 1000x too big/slow is still way better than infinite, which was the best alternative I had for something that was actually decentralized.

(The lay explanation of Bitcoin was _meh_ as it glosses over the blockchain which is the only really novel and somewhat non-obvious part of the system at large)

My greatest concerns were: 50Kbyte transactions with 0.5 second validation time, stored in a step-2-then-a-miracle-occurs (DHT, presumably an attack resistant one created by unicorns), with a cryptographic accumulator which grows without bound and can't be pruned like the block-chain or compactly zero trust queried like the UTXO can if we add a commuted UTXO tree.

Something like this could be used in an external system and tied in via N of M multisig, and the authors acknowledge that but if you're going to take accept a (distributed) point of trust for that, you can use a chaum token like service can be constructed less computationally and bandwidth intensive than this.

On the plus side— approaches can only get better.