If true, that would be a colossally cool theorem! Dan W. Edwin Clark wrote: ----- If my program is correct, starting with the n-th prime (n<=10^5) the procedure always ends with {3, 5, 7, 11, 13, 19, 23, 47} which is sequence A020575: (In particular the number of primes obtained will always be finite.) -----