I like this way of using numpy. This resembles the sieve of eratosthenes, so you'd think it was fast. However `n%p[-1]` isn't fast to do, your inner loop ends

I like this way of using numpy. This resembles the sieve of eratosthenes, so you'd think it was fast. However `n%p[-1]` isn't fast to do, your inner loop ends up taking nearly linear time. At least linear in the final number of primes, which is larger than `N/logN`. Hence your algorithm runs in `N^(3/2)` whereas eratosthenes is `NloglogN`. Funny how many versions of this algorithm there are out there.