Great interview! The whole proof-of-work approach is fascinating, and reminds me of a very old email concept he mentions in passing, where an email server would only accept a msg if the sender agreed to pay like a dollar. Then the user would accept the msg, which would refund the dollar. This would require spammers to front way too much money to make email span affordable. In his version the sender must do a processor-intensive computation, which is fine at the volume legitimate senders use but prohibitive for spammers.