Richard Feynman once challenged the world to produce print the information in the encyclopedia britannica in a space no larger than that of a head of a pin. Science has since developed techniques that could potentially be used to accomplish this goal, but may be "impractical" for a variety of reasons - too expensive, too difficult of a fabrication process, etc. This talk isn't about printing the encyclopedia, but simple shapes.
This is a method of constructing nanometer scale shapes with DNA based on a long single-stranded "scaffold" strand and many little "staple" strands that crosslink portions of the scaffold, essentially weaving it into a particular secondary structure. The success of this method and the final shape of the depends sequence of the scaffold.
Practically speaking, take M12mp18 viral genome (~7kb), use a computer program to analyze its sequence and generate short staple sequences that will bend the scaffold sequence into the desired shape. Almost any arbitrary shape can be built; the staples are synthesized and added to a solution with the scaffold sequence and about 2 hours later many of the desired shape should be floating around, waiting to be set down on a sheet of mica and examined with an electron microscope.
This approach can generate nanoscale structures of equal or better compexity to those possible with more "conventional" methods, but whereas each of the latter must be produces one at a time, the origami method generates billions of the structure at once.
Future directions: "Nanobreadboard" - a way of organizing any nanoscale size componenents (quantum dots, fluorophores, gold balls, molecular switches) in a generalized, arbitrary way.
Perhaps using this technique could be applied towards engineering a synthetic cytoskeleton in vivo.