mmmmmmm imagine the Ghash's, but there more likely to use it for cracking pgp and truecrypt
Since they will just have a lot of parrallel processing power, and have not discovered any real vulnerability in those protocols, it seems it will be enough for truecrypt and gpg users to double they maximum key size and be done with it. As I understand it, the difficulty to crack them increases exponentially with longer keys.
I use 4096 bits for everything, so I'm fine.
Using larger key sizes won't really do any difference unless your password also has double the entropy. We're slowly reaching a point where humans are having trouble remembering pass phrases with sufficient entropy (we're not there yet though!).
Once you get to that point, you can just hash the part you remember and use the hash as the password. The part you hash would still need to be strong, and it would force the cracker to either, use hashed values, the actual values, or both. We could also go the route of using patterns instead of characters.
Well that's true. You could, for example as Armory does, use a really slow key derivation function, and apply it so many times that it takes, for example, one second to get the key from the password. That would limit the cracker's attempts to one key per second as far as I can tell (unless the key derivation function is compromised).
Interesting idea about using shapes to gain entropy. Though I have a hard time figuring out how much entropy this actually provides.