But does it have any actual use? I'd love your take on it.
Does it add value? (Primes cannot - by definition - be factorized, so no one could ever arrive at your private key by factorization and multiplication [if that's even a thing].)
If you confuse the concepts between the ECDLP and the RSA Problem, then you should not be generating your own private keys.
Are primes handled "better" (whatever that means) when for example signing transactions, resulting in better byte economy for the blockchain?
No.
Is this unsafe? (I think not, according to the
prime number theorem (PNT) there should be about 6.54*10^74 primes - a huge number - in the range as defined by the secp256k1 with the ECDSA algorithm. In other words, of all possible private keys, approximately 0.056% are prime, but only a few of these correspond to public addresses whose hash160 happen to be prime too.)
It is slightly less safe since the search space is being reduced.
That's around 2^248.53,
You are reducing the number of possible private keys.
He is also changing the distribution of the keyspace. Private keys are supposed to be selected from a discrete uniform random distribution within the keyspace (and those who don’t know what that means
should NOT be messing with this stuff...). The subset from which he is selecting keys does not have that distribution—unlike keys that are simply restricted to a contiguous part of the keyspace.
I have
no idea if or how this difference of distribution could be exploited—or even whether or not that would require solving one of the biggest, oldest problems in mathematics... However, I would not bet that a cryptographer could not think of something clever!
What do you think?
The NSA loves you.
You have not even adequately answered the question implied in the topic title:
Why?