- double SHA-256 proof-of-work isn't computationally useful for anything (GridCoin is still affected by this criticism to a certain extent);
Even this is arguably a feature. A "computationally useful" proof-of-work algorithm wouldn't be any more efficient, it would simply result in a higher difficulty.
GridCoin appears to depend on at least one third party, Berkeley (the University of California). The purpose of a blockchain is to enable consensus without a special trusted entity.