Really not a mystery. This is not that complex for people with low level programming experience (embedded, (older) games, demoscene, performance systems, etc.).
Knowing these tricks and knowing how to apply them is basically a requirement on these areas.
I think this is an excellent skills showcase, even if I don't think it's magical.
* More specifically, miniaturized robots consisting of one or more robot arms with multiple degrees of freedom, a CPU, some kind of storage (perhaps something like wound-up DNA serving as tape on reels), executing programs written by humans.
On a macro scale, life tends to be bilaterally or radially symmetrical, but our robots are not necessarily like that, just considering even factory robots which are often an arm and little else. So, at the micro scale, I don't think they have to resemble "life" there, either. I'm hardly suggesting some kind of tinkertoy with one atom here and another atom there and the strut being, instead of wood, a covalent bond. No, I think we would need more atoms than that.
Frankly, we haven't much tried to scale down our robotics. Oh, you'll find someone who will produce the flagellar motor (a mere forty-five nanometers, compared to the ten thousand nanometers of a human cell) but not much else. I wouldn't worry about the thermodynamics and quantum effects until we're down to that motor level.
By just taking standard equations found in textbooks, plugging in some numbers, and realizing that things such as Ideal Gas Law don't hold at such scales.
You can even derive the equations for yourself to confirm your not being misled by the textbooks, but that takes more skill and knowledge then is likely feasible to acquire without advanced study.
We will face different challenges at different scales, of course, but even as recently as 2016, researchers using a scanning tunneling microscope crafted a 8,192 bit message using less than ten thousand atoms. So, no, I see absolutely no reason why we couldn't have robots within an order of magnitude of the human cell -- we have bacteria which are much much smaller capable of a variety of functions as well as reproducing themselves
You might look into Feynman's "There's Plenty of Room at the Bottom."
Clearly cell-like robots are 100% possible at very small scales, there’s no one debating you on this.
I feel like this kind of knowledge is becoming a lost art these days despite having lost no relevance.
I had a question recently: would an extra 65th bit in some registers have a good use case with stuff with SWAR?
The much-hyped RISC-V is one of the few that doesn't have a flag register, because it's been claimed to be a performance bottleneck - in my somewhat heretical opinion, the actual reason is that its designers are deluded into thinking that C and UNIX are as fundamental to software as transistors are to hardware, and thus anything beyond what's needed to support that environment is not worth implementing.
But an extra bit per register would be like having a separate carry flag for each, potentially solving some problems with parallelization and also allowing checks for integer overflow at the point a value is stored into memory or used in a subsequent operation.
Having some kind of carry flag enables various programming tricks that can also be useful for SWAR, for example subtract-with-carry of a register with itself will either set or clear all bits.
There is nothing magical in computers once you can do that deconstruction in your head.