World models are fundamental to how organisms interpret raw sensory inputs, enabling structured and purposeful behaviors such as navigation, decision-making, and object manipulation. Despite their critical role, the processes through which these models are learned, represented, and maintained across distributed neural networks remain largely elusive. This research seeks to address these questions by investigating how active engagement with the environment dynamically shapes neural circuits to support goal-directed behavior, offering profound insights into the interplay between sensory and motor systems.