My new conjecture in P vs. NP

If you set up the rules so that a self-reproducing machine can create an exponentially large number of copies, then indeed you can solve NP problems in polynomial time.

But you should realize that an exponentially large number of machines also requires an exponentially large amount of space to fit them, so at the very least it takes an exponentially long time at the speed of light for the winning machine to report its result back to you.