> is it possible to design a large scale self replicating machine that runs more efficiently than evolution can give us?

Transistors miniaturization may fit under that category. We created something that every year doubled down the scale, augmented performance and reduced energy costs; and for that we needed those previous transistors. I don't think natural evolution has done any other process that fast.

I may be wrong...