The article discusses the concept of interoceptive predictions in the brain, proposing that these predictions are generated by agranular visceromotor cortices and are constrained by ascending visceral sensations. It introduces the Embodied Predictive Interoception Coding (EPIC) model, which integrates anatomical models of corticocortical connections with Bayesian active inference principles. The model suggests that interoceptive predictions are essential for maintaining homeostasis and are involved in various mental and physical illnesses. The brain is viewed as an active inference generator that minimizes prediction errors by adjusting predictions, modifying body movements, or altering sensory input processing. The EPIC model also highlights the role of agranular visceromotor cortices in interoception, which are less sensitive to prediction errors than more granular regions. The article discusses the implications of these findings for understanding mental and physical illnesses, including depression, and suggests that interoceptive predictions may be a key factor in these conditions. The model also proposes that interoceptive predictions are fundamental to conscious experience and that abnormalities in these predictions may contribute to various illnesses. The article concludes by emphasizing the importance of further research to validate these hypotheses and to explore the neural mechanisms underlying interoceptive predictions.The article discusses the concept of interoceptive predictions in the brain, proposing that these predictions are generated by agranular visceromotor cortices and are constrained by ascending visceral sensations. It introduces the Embodied Predictive Interoception Coding (EPIC) model, which integrates anatomical models of corticocortical connections with Bayesian active inference principles. The model suggests that interoceptive predictions are essential for maintaining homeostasis and are involved in various mental and physical illnesses. The brain is viewed as an active inference generator that minimizes prediction errors by adjusting predictions, modifying body movements, or altering sensory input processing. The EPIC model also highlights the role of agranular visceromotor cortices in interoception, which are less sensitive to prediction errors than more granular regions. The article discusses the implications of these findings for understanding mental and physical illnesses, including depression, and suggests that interoceptive predictions may be a key factor in these conditions. The model also proposes that interoceptive predictions are fundamental to conscious experience and that abnormalities in these predictions may contribute to various illnesses. The article concludes by emphasizing the importance of further research to validate these hypotheses and to explore the neural mechanisms underlying interoceptive predictions.