Adaptive optimization of cellular automata is explored as a minimal model of biological evolution. The rules for the cellular automata correspond to genotypes; the patterns generated to phenotypes. Successive point mutations are applied to the rules, being kept when they do not decrease fitness (such as lifetime of finite patterns). Complex phenotypes are produced, and many phenomena familiar from biological evolution are seen. The models are minimal enough to allow complete studies of the space of possible evolution paths.

A fundamental explanation for the Second Law is given in terms of the interplay between the phenomenon of computational irreducibility in underlying processes and the computational boundedness of us as observers. This explanation appears to significantly clarify the character and origin of the Second Law and to resolve core aspects of it that were previously mysterious. The approach used allows the Second Law to be broadly generalized beyond its traditional framing in terms of the dynamics of heat-though the approach also suggests limitations to the Second Law, notably in connection with what is termed here the mechanoidal phase. Drawing on some of our other recent work, it is suggested that all three major theories of twentieth-century physics- general relativity, quantum mechanics and statistical mechanics-are ultimately analogous in their foundations and are all derivable from the interplay between computational irreducibility and our characteristics as observers.