An energy-gradient theory of embedded predictions in the brain.

Activity gradients measured with neuroimaging play a fundamental role in brain function, yet their relationship to the brain's internal predictive models during rest remains poorly understood. Clarifying this relationship can reveal how the brain processes information efficiently and adapts to a changing environment. Here, I discussed how energy flows may give rise to gradients of activity between voxels and to spatial coding in fMRI, and I described computational algorithms that can be applied directly to brain-activity images. I proposed that the brain continuously maintains and updates predictions about its environment through internal models, and that these models are embedded in the dynamic activity flows observed at rest and during cognitive tasks. This perspective emphasizes the role of energy turnover in supporting cognition: Neural circuits reflect predictions and adjustments shaped by past experience. I argue that the interplay between predictive processing and energy dynamics offers a richer account of cognitive mechanisms and points to new research directions in psychology and physiology. Together, these insights underscore the importance of energetic principles in brain physiology.