Brainwallets just don't produce enough entropy for the seed of your private key. It's known fact that there are several groups with GPU farms creating giant rainbow tables for these purposes. It's generally considered that every passphrase that can be Googled is not secure enough for the brainwallet. There's a guy who reported that passphrase created from the entire poem written in some obscure language (Afrikaans) has been bruteforced. For instance, I've took the four words from your sentence above in random order "capable dispute creating everyone" and it returns 0 hits on Google (until I post this, at least). This may be a good brainwallet by your criteria since if I haven't post it it would be probably safe against attackers for a very, very long time, but it has nowhere close enough entropy compared to any decent PRNG. The question is this: If it's inferior from the security standpoint then any address generated by the Bitcoin-QT wallet, why don't you let the Bitcoin-QT generate the address and after that passphrase protect the wallet with the same "capable dispute creating everyone" passphrase, making it infinitely harder for the attacker since he has to hack your machine first + hack the passphrase, instead of only hacking the passphrase?