Much has been made of both human and machine superperformance, typically as a means of achieving enhanced levels of performance. We introduce a theoretical approach for understanding superperformance. Our approach requires a form of human-machine symbiosis, where human-machine interactions are mediated by both sensory information and planning capacity. The joint optimization (maximization) of sensory and planning capacity results in superperformance. Superperformance is dependent on defining intelligence as the difference between generativity (the results of sampling), and selectivity (the outcomes of planning). An intelligent agent is defined using three constructs: Intelligence (involving both between generativity and selectivity), the sensorium (detecting relative motion from sensory information), and modality-specific sensory gradients. The relationship between the world model and an intelligent agent can be described as a closed-loop regulatory system where each system must match each other’s state space. This is a form of Intelligence Augmentation (IA), where the human and machine acquire a symbiotic form of superperformance. Successful superperformance results from a mutual sampling rate, particularly one that does not exceed that of the world model. We conclude by considering more sophisticated forms of optimized superperformance, more directly relevant to human cognition.