RE: Red Alert! supercomputing Is Back!
This is the quote from @arhag's post about the new algorithm.
So the real solution is to simply get rid of the degree of freedom provided when the system allows the user to arbitrarily choose the private key d (or at least the system should allow no more of a degree of freedom than that allowed by iterating a nonce or account name which changes the pseudorandom cryptographic hash output that is used as the private key). So the private key d should instead be deterministically determined by a cryptographic hash dependent on the latest_block_id, miner_account_name (this is to necessary to prevent the PoW from being stolen by another account within the same block), and the nonce. input should also depend on those three values, so it can simply be defined as the SHA256 cryptographic hash of the private key d. The rest of the algorithm is more or less kept the same.