Text follows this introductory line. I believe I have this fully sorted out. Many problems can crop up on the path, and I believe I have a solution to any possible problems. The thought parts are hard to sustain, bring together, and retain, so there's only a bit here and there. Basically, the patient works with an AI as if they were part of a GAN. The AI learns to predict the behavior of the patient, and the patient learns to understand their own behavior. As the AI forms a model of the patient's psychosis, it maps things that trigger it or worsen it. The patient can then make informed decisions on how to act. They could pursue one or more of: - alteration of their brain to prevent the experiences - investigation into how the experiences developed and why they continue, in order to resolve them therapeutically - avoidance of the events that stimulate their experiences They could engage - personal cognitive training to learn to handle things - augmentation to warn them of or prevent the experiences This sounds both exotic and easy, but we have the research to do this today, and I am not aware of it being done. Research can take years to reach the doctor's office as things move through clinical trials. I've seen visualisations of AI models for image-recognition. I've also seen that AI models of human behavior are widespread among politics and sales. I'm sure visualising models of human behavior has already been done. I will probably be trying to work on this. But I expect public research to outpace me due to my issues. My eta would be a couple decades.