A brother, and a bill nobody could pay

What Can AI Actually Do for Us in Healthcare? A three-part series Sep 27, 2026 · @joy People keep telling us that AI will cure disease and take us to 150 years old. I want to make a narrower argument that I think is more useful: the science is arriving faster than the price is falling, and the price is the part we can engineer. This is a short series in three parts.

1. Why I care about this question — a family story, and why cost is the real bottleneck
2. The 5P framework — the five principles precision medicine actually runs on, and where AI fits
3. Baby KJ and the million-dollar problem — what 2025 proved, and what still isn't solved No background in genomics needed. Part 1 — A brother, and a bill nobody could pay If you could not afford an expensive treatment and had to lose someone in your family, how would you feel? If you lost that person to a disease that could have been prevented, how deep would that sorrow be? My father and mother were born in what is now North Korea. The question that always follows is whether I am North Korean too. I'm not. My parents fled south during the Korean War. My father lost his entire family in that flight and arrived in the South with only his older brother. Not long after, while both were still children, he lost that brother to peritonitis. There was no money for the surgery, so he had to let his brother go. He carried that for the rest of his life. Today almost nobody dies of peritonitis alone. What changed was not compassion. My father had plenty of that, and it saved no one. What changed was that the treatment became cheap, standardized, and reachable before the damage was irreversible. That is the whole argument of this series, in its oldest form. The burden is not where most people think Two numbers frame the problem. Noncommunicable diseases — heart disease, cancer, chronic respiratory disease, diabetes — killed more than 43 million people in 2021, about three-quarters of all non-pandemic deaths worldwide. Around 18 million died before the age of 70, and 82% of those early deaths happened in low- and middle-income countries. The WHO's own estimate is that 62% of premature deaths from these diseases are preventable. At the other end of the distribution: roughly 300 million people live with one of more than 7,000 rare diseases. About 80% are genetic, nearly 70% appear in childhood, diagnosis takes 4.8 years on average, and about 95% have no approved treatment at all. Those two groups look unrelated — the enormous and the vanishingly rare. They fail for the same reason. In both cases the decisive information exists in the patient's biology long before it exists in the clinic, and the system is built to act only once symptoms make the problem undeniable. That is what precision medicine is trying to invert. Part 2 is about