It seems like whatever breakthrough occurs with self-replication could have been long ago demonstrated with software. You could have some piece of software that you could hit a button and it would copy all its files and installation routine to wherever you designated free of charge. Of course most people try to prevent unauthorized (i.e. uncompensated) copying of their software. Furthermore all software has the ability to fabricate other types of bits strings in memory or external storage.

Not to mention, the little matter of self-directed, autonomous self-assembly. [Cf the living cell . . .]

Here’s good old Evo Mat-leaning prof wiki:

A self-replicating machine is an artificial construct that is capable of autonomously manufacturing a copy of itself using simpler components or raw materials taken from its environment . . . . A self-replicating machine would need to have the capacity to gather energy and raw materials, process the raw materials into finished components, and then assemble them into a copy of itself. Further, for a complete self-replication it must, from scratch produce its smallest parts, such as bearings, connectors and delicate and intricate electronic components. It is unlikely that this would all be contained within a single structure, but would rather be a group of cooperating machines or an automated factory that is capable of manufacturing all of the machines that make it up.

The factory could produce mining robots to collect raw materials, construction robots to put new machines together, and repair robots to maintain itself against wear and tear, all without human intervention or direction. The advantage of such a system lies in its ability to expand its own capacity rapidly and without additional human effort; in essence, the initial investment required to construct the first self-replicating device would have an infinitely large payoff with no additional labor cost.

Such a machine violates no physical laws, and the basic technologies necessary for some of the more detailed proposals and designs already exist.

The same article mentions:

In 2005, Adrian Bowyer of the University of Bath started the RepRap Project to develop a rapid prototyping machine which would be able to manufacture most of its own components, making such machines cheap enough for people to buy and use in their homes. The project is releasing its designs and control programs under the GNU GPL.[27] The RepRap approach uses fused deposition modeling to manufacture plastic components, possibly incorporating conductive pathways for circuitry. Other components, such as motors and discrete electronic components, would be supplied externally. As of 2006 the project has produced a basic functional prototype.

A long way to go yet, I’d say.

Not to mention, as you underscored, the machine was . . . er, ah, ahem . . . DESIGNED.

GEM of TKI

It also makes no mention of the energy source or of processing the material needed for it to replicate the parts.

By contrast, what a marvel is the simplest self reproducing cell.

As such, there is certainly no evolution going on with it just yet.

I do expect at some point (probably within 5-10 years) someone will create a truly self replicating maching — either on the microscopic scale or the macroscopic. As you very rightly allude to, it will be interesting to see what its design tolerances are, error correction, and self repair, if any.