Actually, no, it doesn't involve finding new primes. In fact it states clearly that it is not interested in finding large primes numbers (which is a requirement for finding new primes).
This is at least more useful than running a double sha on some input to produce some output that is hopefully less than a target t. But a altcoin that does something useful with its proof of work system is still not available.
Someone should make a coin whose proof of work involves solving difficult (but verifiable) problems that are in NP, that are frequently used in commerce.
This would provide a more direct incentive for researchers to make these algorithms faster, and would therefore provide a tangible benefit to society.
(I don't know why this is useful, as I am not a mathematician.)