John B posted some good technical questions in discussion following an earlier post.
However, let's assume that the only material that can be made into a programmable atomically-precise self-replicating assembler is diamondoid. .... How will you generate and conduct power with just diamondoid? OK, you could conduct via mechanical means, but how do you GENERATE power via pure diamondoid? Or are you going to limit yourself to non-selfreplicating power sources for your devices?
Well, if we can build diamondoid, then we probably have graphite and/or buckytubes, which gives us conductors, and buckytubes can be semiconductors as well. With a conductor and diamondoid, you can generate electrical power from heat (such as concentrated sunlight) via thermionics--CVD diamond thermionics have been demonstrated--and of course you can use the electricity in electrostatic motors.
Even if we have only mechanical devices, we can build macro-scale Stirling engines.
How will you fulfill an exponential need for feedstock with just diamondoid devices? Given a reasonably pure source of atomicly rigid materials, you might be able to pull off sorting rotors, but - how do you generate that reasonably pure source in the first place, and how do you scale that source up exponentially along with the need?
Bootstrapping won't need much material. Once we have a few sorting rotors built, we can use them to purify more feedstock. Microfluidics will probably be usable to generate the feedstock. If you don't have to pay extra for fabricating more features, then you can scale up a microfluidic design simply by making more of them.
Don't you have to make a functioning [nanofactory] first before you start commercializing it? And isn't the whole product rather more than just the self-replicating device in a closed system? Does it not include the controller for the devices (unless all you want is self-replicating devices, you'll have to control them somehow, not to mention error-checking and -correction routines), most likely some sort of clean chamber for atomic precision work, power supply, coolant (unless things go real slow), feedstock transport(s)?
A nanofactory could include simple computers, able to handle errors at a local level, so you'd just have to feed it the blueprint stream. (Mechanical computers work just fine at that scale.) And of course the nanofactory could also build the general-purpose computer with large memory to feed the blueprint to the factory. My nanofactory paper does include both computers and internal clean working chambers. Power, whether fed in electrically or mechanically, would not require much volume to distribute. Likewise for feedstock. My nanofactory design (which by the way is obsolete--there are much better architectures now) does not include the feedstock tank, refrigeration engine, or external heat exchanger, but those are all simple mechanical things that would not require much mass to build with diamond.
By now, some readers will have clicked away; some will be nodding in agreement; and some will be saying, "But that doesn't answer ______. There are so many issues to be solved!"
Yes, there are a lot of practical issues to be solved. To discuss all the issues would require a book; even to list them would require several pages. No one has pulled together this discussion, and anyone looking at only part of it will be able to find lots of [apparent] holes. But I have spent sixteen years thinking about this, and several years studying it intensively, and I think I have looked at most of the issues. Every one appears solvable. For most of them, the solution appears straightforward and does not impose much constraint on the rest of the system.
I am not saying that developing a nanofactory infrastructure will be easy. I'm saying that it could be done fairly quickly, given a well-funded well-focused effort--and that effort will be very worthwhile.