Provably Correct Crypto?

Hmm. Instead of "vetting" the PGP PRNG code, as Dr. Cohen has been trying to tell me in private email is several programmer- years worth of work perhaps I should just write my own version of the PRNG using the algorithm as defined in whitepapernumberwhatever. If there are artificial "bugs" to throw off the PRNG currently in the PGP MIT code, they ought to simply not work right with the code I write independently and therefore cause a crash or something. Or perhaps just rewrite the program to use plain arrays instead of pointers. ;-) Phil