A short comment on Drexler’s paper Biological and Nanomechanical Systems: Contrasts in Evolutionary Capacity: He distinguishes two types of design, O-style (like organic) and M-style (like mechanical) systems. He points out that O-style systems are much more robust to incremental design modification, where M-style systems require coordinated changes that are much, much less likely to happen all at once in the course of natural variation.
Note, however, that it’s not really necessary that the physical instantiation of the system not be M-style. Evolution happens in the genome, not the physical body. The condition for evolution to work is that workable designs be close together in the design space where variation is occurring. Thus we might have a machine whose genome is a highly compressed bit-string, such that changing one bit causes a vast, coordinated change in the ultimate physical embodiment. Then, rather than requiring that ten separate mutations specify the change from a quarter-inch pipe, fittings, valves, supports, and so forth to half-inch ones, only one change need happen and the mechanical process of ontogeny would carry out the coordination.