This is an educational-, research-, and CTF-oriented implementation of the Mersenne Twister PRNG, providing features such as recovering internal state from outputs, rewinding the internal state, or recovering the seed from a freshly seeded internal state.

There are may implementations of Mersenne Twister on the Internet, and even may implementations that allow cloning internal state. This one is better because:

• Both 32-bit MT19937 and 64-bit MT19937-64 are implemented
• A rewind feature is provided to "turn back time" on the PRNG
• The value of the seed can be recovered from a freshly-seeded state
• The code is written in a conceptually simple manner, making it easier to reason about. Performance does take a hit, though.

An example that demonstrates cloning a standard library random.Random is provided in the code, and also below:

def test_with_stdlib_random():
    import random
    stdlib_random = random.Random()
    mt = MT19937()
    output_for_cloning = [stdlib_random.getrandbits(mt.w) for _ in range(mt.n)]
    mt.clone_state_from_output(output_for_cloning)
    for _ in range(10000):
        assert stdlib_random.getrandbits(mt.w) == mt.get_next_random()
    print('Test successful')

Have fun!