Can code actually do therapy better than a human? This episode dives into the cutting-edge science behind AI in mental health and medical diagnosis. We explore how ChatGPT maxed out tests of emotional awareness, how specialized AI architectures are delivering text-based CBT that outperforms human clinicians, and how autonomous AI doctors like Google’s AMIE are navigating differential diagnosis better than physicians with internet access. From the ingenuity of LoRA to the critical workflow of parallel analysis, from FDA guardrails against confabulation to the deep psychological risks of falling in love with an algorithm that cannot actually care about you — this one challenges everything you think you know about human intuition. But here’s the unsettling question: if we design the perfect listener — endlessly patient, perfectly validating, completely judgment-free — do we risk losing our tolerance for the messy human relationships that actually keep us alive?