Darwinian evolution is not restricted to biological organisms as we know them. In fact, selection will occur provided there is (1) competition for limiting resources, (2) faithful replication of entities with occasional copying changes (mutations) that are themselves replicated, and (3) variation in competitive ability associated with those heritable differences. In fact, Darwinian evolution can even occur within an electronic microcosm. One such example is Tom Ray's Tierra system, wherein self-replicating computer programs compete with one another for the limiting resource of central processing unit (CPU) time. Tierra differs from the standard computer simulations used by evolutionary geneticists, and from most genetic algorithms used by computer scientists, in that the fitness of the Tierra programs is not determined by an experimenter-manipulated measure, but solely by their ability to compete for CPU time. In Tierra, parasitism and hyperparasitism can evolve if programs have access to the instructions of other programs.

A recent paper by Yedid and Bell [1] explores the properties of microevolution within Tierra and compares them with observations from experiments of microbes evolving in chemostats. To focus on microevolutionary aspects, the authors eliminated the access of programs to other programs' files and thus prevented the evolution of parasitism. As a correlated response to selection for faster replication, the mean size of genomes in the Tierra simulations generally decreased. However, smaller genomes did not always outcompete larger ones because some genotypes were more efficient than slightly smaller ones.

When mutation rates were low (on the order of 0.01 per genome per generation), Tierran populations followed patterns similar to the classic periodic selection often seen when microorganisms adapt to a chemostat environment. Under periodic selection, the population consists mainly of a single dominant variant with all other variants having very low frequencies. Periodically, a mutation derived from that dominant variant would appear, quickly increase in frequency, and then become the next dominant variant. With higher mutation rates (∼0.1-0.2 mutations per genome per generation), Tierran populations exhibited a different pattern: no single genotype dominated, and the most common variant was generally less than 40 percent of the population. Here, the most abundant genotypes usually did not arise from genotypes that had been the most abundant, and instead arose from intermediate-frequency or even rare genotypes.

There are profound differences between Tierran programs and biological organisms in how information is transmitted and expressed. Tierran programs lack a clear distinction between genotype and phenotype and between transcription and translation. The code of the Tierra system is also more brittle than the genes of biological systems. That is, mutations in Tierra are more likely to be notably deleterious (and thus less likely to be near neutral) than mutations in most organisms. Yet despite these profound differences, Tierran evolution appears to share many properties with the evolution of biological microbes. Perhaps, as Yedid and Bell note, the Tierra system can be a useful "halfway house" between classical theory and chemostat experiments.