I suppose it's difficult to say without knowing anything about his implementation. He's saying he's doing 500 million * 60000 address comparisons per day. If his implementation is naive (checking each generated key against each address in the "database") that works out to ~350 million key comparisons per second. If his implementation is less naive (using a hash table for the 60000 "known" addresses) then that works out to nearly 6000 keys generated and compared per second.
If he's doing it the first (naive) way, then I think he's probably made a mistake somewhere without realizing it, because that number (350 million comparisons per second) seems obviously out-of-range for an i5 (to me, anyway).
If he's doing it the second (less naive) way, then assuming it's "homebrew" that's been quickly slapped together in 3 days, it's still suspiciously fast, IMHO.
Maybe I've misread/misunderstood something or calculated incorrectly. I'm always happy to be corrected