How much information is required to program multicellular development? Motile animals must exert exquisite control of their ontogeny to place organs, muscles, their support structures and enervation systems into proper alignment and connectivity, as illustrated by the architecture of human bones. Plants, and sessile animals such as corals, have greater flexibility but still require a high degree of functional organization. The nematode worm has 1,000 leaves in its developmental tree that are genetically hard-wired. Similarly, humans must make trillions of divide or differentiate cell fate decisions with high accuracy, also hard-wired as evidenced by the phenotypic congruency of identical twins. It had been proposed that Boolean combinatorics of transcription and other regulatory factors acting on cis-acting regulatory DNA sequences would suffice to direct developmental ontogeny, but this proposition has never been justified theoretically, mathematically, or mechanistically. By contrast, a decisional tree with N leaves requires NlogN regulatory decisions, which is consistent with the quasi-quadratic increase in the number of regulatory genes with total gene number in bacteria. Consequently, the proportion of regulatory information increases with complexity, imposing a constraint on genetic programming and providing a theoretical explanation for the expansion of non-coding DNA with increased developmental and cognitive sophistication. The fact that there are millions of different epigenetic marks imposed at specific locations in different cells at different developmental stages and that the genomes of plant and animals are transcribed in dynamic patterns during development suggests that RNA regulation has been enlisted as the most flexible and information efficient solution to the challenge of orchestrating multicellular ontogeny.