Alan, what do you think of "Hash Hyena" claims?
He basically says that he is generating trillions of addresses non stop and getting a few collissions that will grow over time. His logic is that the entropy of Armory, Bitcoin Core, Multibut, Electrum etc. (and generally any other wallet that uses a RNG based on software) is flawed and results in a highly reduced keyspace, which will result in collissions with enought computing power and space devoted to private keys bruteforcing.
See:
https://bitcointalk.org/index.php?action=profile;u=380718;sa=showPosts...
...
On that note, Armory uses Crypto++ was is considered a cryptographically-secure RNG (X9.17 with OS-provided seeding). On top of that, Armory pulls in system files, mouse clicks, keypresses, and a desktop screenshot, to add to the Crypto++ RNG entropy pool. I made sure when selecting these sources that it would guarantee at least 256 bits of entropy to be added to the pool even if Crypto++ was really weak.
...
It's rather telling that he claims a "paperwallet" (I assume he means bitaddress.org) has a safer RNG for key generation than Armory (which, among others, has undergone "heavy testing" by his team) because "paperwallet" uses mouse input....