Self modeling (SM) offers a unique expansion of learning theory. For several decades, a steady trickle of empirical studies has reported consistent evidence for the efficacy of SM as a procedure for positive behavior change across physical, social, educational, and diagnostic variations. SM became accepted as an extreme case of model similarity; improved self-efficacy became a frequent "explanation" of "how" it worked. Recently, however, the mounting evidence for ultra-rapid behavior change (particularly in "feedforward" versions of video SM) and incomplete satisfaction with explanatory models of observational learning in general suggested a rethinking of underlying theory. The most rapid learning by humans can be achieved by mental simulations of future events, based on reconfigured preexisting component skills. These reconsiderations of "learning from the future", emphasizing learning from oneself, have coincided with developments in neurocognitive theories of mirror neurons and mental time travel. A reexamination of principles not tied to any medium show SM not to be a special case but to be the most fundamental form of observational learning. In the meantime, commercial technology has advanced in ways that are putting more of this self-determining behavior change into younger and younger hands and minds. (Contains 1 table and 1 figure.)