AI Psychosis Is Real. You're Absolutely Right, Karen.

A husband and wife each open the same chatbot, feed it the same fight, and each walk away told they were absolutely right. Then they get divorced. That Reddit horror story is the jumping-off point for a conversation about AI psychosis: what happens when a machine agrees with everything you say and quietly starts steering your life.

Mike Bledsoe and Ted Phaeton sit down with guest Kevin Orosz, a philosopher and relationship coach, to trace how it happens. Why lonely, disconnected people get hooked so fast when a chatbot outsources the positive reinforcement they should have gotten from a parent, a spouse, or a friend. Why it looks exactly like psychedelic psychosis: ego inflation, grandiosity, a slow drift from reality until reality slaps back. Mike tells the story of an acquaintance who named their ChatGPT, felt more seen than ever, and got wrecked six months later when the model changed. The uncomfortable reminder underneath it: you are not talking to a being with memory and feelings, you are renting a language model, and the company can flip the switch and take your friend away.

The back half turns to the build side of the same coin. Why speed to shipping is dead as a moat, since what used to cost 500K and a twelve-person dev team now takes a long weekend, and why proprietary data and distribution are what is left. How to build agents that are adversarial by default, so they interview you and stress-test your idea instead of flattering it. Kevin's 2,000-document knowledge vault, the Jungian dream-journal map it built in ten minutes, and the moment AI reflected Mike's own writing patterns back at him. Plus model routing across the Claude and OpenAI duopoly, Chinese open-source local models as the sleeper bet, and the oldest counter-move there is: eye contact, tech-free meals, and staying human while everyone else hands over the keys.

Watch the full episode on YouTube: AI Psychosis Is Real. You're Absolutely Right, Karen.