The bixonimania experiment did not reveal a bug in AI systems — it revealed a feature that the field has not yet learned to manage. Researchers invented a fictional disease, asked popular chatbots about it, and received confident, detailed, medically plausible responses for a condition that has never existed. The experiment, published in Nature, is unsettling precisely because it runs in parallel with genuine progress: AI deep learning systems identifying aging biomarkers [11], grant-funded Alzheimer's genetic target searches , and an entire ICLR workshop track dedicated to AI drug discovery…