Suppose you wanted to direct the evolution of molecular systems that behaved according to parameters you set. By whatever means you implement that, you must have a method for directing evolution of individual molecules toward desired functional properties, since the system must have the appropriate parts to generate its functionality. In that case, you must have a molecule generating factory for the component parts. Starting with enormous numbers of generators of different efficiency/capacity, and limiting the substrate precursors of the desired molecules, iteratively in rapid succession, and imposing the parameters defining the desired functional molecule, natural selection will favor preservation of the high-capacity, fast-reacting, efficient generators that yield the targeted molecule.
That process admits of automation on a chip, as demonstrated by Brian M. Paegel and Gerald F. Joyce, in the Departments of Chemistry and Molecular Biology, of the Scripps Research Institute, La Jolla, California, and in the Skaggs Institute for Chemical Biology, of the Scripps Research Institute, La Jolla, California, both in the U.S.
The procedure produces a continuous stream of real-time data, giving the experimenter a record of the evolutionary course in terms of population size and heterogeneity, and growth conditions, including availability of limiting resources. Each microchip contains multiple microfluidic circuits independently addressable. The method costs modestly, so the process makes Darwinian evolution readily accessable. As easy almost as implementing the evolution of the system on a computer.
Darwinian Evolution on a Chip. Paegel BM, Joyce GF. PLoS Biology Vol. 6, No. 4, e85 doi:10.1371/journal.pbio.0060085