You seem to want pseudo-random numbers that aren't pseudo-random, with a higher probability of consecutive numbers being 'significantly' different than pseudo-randomness requires. I doubt that any common prng will do this, whatever your seeding strategy.