This paper discusses predictive processing, focusing on the unpredictable results of ritual magic in the interplay between training, material culture, didactics, social support and cognitive processes. According to the predictive coding framework, perception can deliberately be “hacked”, as it is shaped by habitual estimations at the neurological level that can be altered through training. “Top-down processing” can override the bottom-up sensory input and make us sense what is not physically there. As seeing the unseen is a frequent goal with esoteric practices, this theory becomes highly relevant for the study of esotericism. In this paper, I apply it to my learning during a three-year fieldwork in an initiatory society. I experienced the effects of re-wiring my top-down processing, leading to better kataphatic abilities. Repetitive training in ritual and meditation sharpened my sensations and led to alterations in consciousness, shaped by anticipation and cumulative experience of magical practice. Predictive coding insights suggest that perception involves continuous adaptive learning, influenced by surroundings and dedicated training. I discuss how my wardrobe became a “designer environment” that facilitated concentration in conjunction with ritual tools and clothing, yet highlighting how, with training, the need for material anchors subsided. Ritual practice became increasingly portable, entangled with everyday life, as I learnt to practice magic in cafés, during short coffee breaks, on the commute, and while queuing in the supermarket. 

However, my learning outcomes, such as the frequency and intensity of astral visions, remained unpredictable. Despite diligent training, my kataphatic abilities did not increase in a clear-cut progression, as one might expect from daily, systematic training. This randomness complicates the predictive coding theory. It raises questions about whether my measurements, such as the magical diary and self-assessments, fail to capture the process fully or if predictive coding theory itself needs supplementation.