It occurs to me that the "mathematical approach to produce desired configurations of nanoparticles by manipulating the manner in which the particles interact with one another" mentioned in the "Utility Fog" post might be a great way to produce carbon nanotubes.

If you could essentially put carbon on a conveyer belt and move it through a system that would dictate how the carbon interacts with other carbon atoms, then why wouldn't you be able to produce carbon nanotubes in the strengths and lengths necessary to do all sorts of interesting things?