The Failure of Ultra-Stability: Robert Pearl on Why American Healthcare is Quietly Rationing Us to Death

“It’s ultra stable. Health care doesn’t move. If you biopsied American health care in 2010 and again in 2026, no one could figure out which slide was which.” — Robert Pearl, MD

Bad news. The patient, I’m afraid, is ultra-stable. Robert Pearl, former CEO of Kaiser Permanente for eighteen years and author of ChatGPT MD, returns with the bleakest diagnosis we’ve heard all month. American healthcare, Dr Pearl says, is “ultra stable.” That might sound good. But it’s actually very very bad.

If you biopsied American healthcare in 2010 and again in 2026, Pearl says, no clinician could tell the slides apart. Both were and are overpriced. Both underperforming. Hospitals still represent between 30-35% of expenses. Costs continue to rise at between 7-9% a year. There remain four hundred thousand misdiagnosis deaths annually. Burnout is stuck at 50%. The numbers haven’t moved in fifteen years.

Meanwhile, a stealth revolution is already underway. 40% of Americans use generative AI every month for medical questions. 70-80% of physicians use it weekly. While the patients and doctors have moved, the system hasn’t. It remains ultra-stable. It’s a Kodak moment — healthcare’s business model, Pearl suggests, is selling sickness. So, for example, the new new medical thing is GLP-1 drugs that cost $5 to manufacture and sell for $400.

So will the system collapse? No, Pearl insists. It has too much strength for that kind of drama. Instead, it will quietly ration us to death — more chronic disease, earlier deaths, more people making a major sacrifice to pay their healthcare bills. Ultra-stability, then, is what is killing the American healthcare system. It will, quite literally, ration us to death.

Five Takeaways

•       Ultra Stable: Pearl’s diagnosis of American healthcare in one phrase. Hospitals stay at thirty to thirty-five per cent of total expenses. Costs rise at seven to nine per cent annually. Life expectancy hasn’t budged. Four hundred thousand misdiagnosis deaths a year. Burnout at fifty per cent. Biopsy 2010 and 2026 — no one could tell the slides apart. Both overpriced. Both underperforming.

•       The Stealth Revolution Has Already Happened: Forty per cent of Americans use generative AI every month for medical questions. Seventy to eighty per cent of physicians use it weekly. The patients and doctors have moved. The system hasn’t. It’s a Kodak moment — they had the first filmless camera and let it die because their business model was selling film. Healthcare’s business model is selling sickness.

•       Quietly Rationed to Death: There will be no dramatic collapse. The system has too much strength for that. Instead: rationing, more chronic disease, earlier deaths. Like airlines moving everyone into first class while the rest drive. Twenty-five per cent of Americans already made a major sacrifice to pay healthcare bills last year. When it hits fifty per cent, maybe the polling places will notice. Pearl is doubtful.

•       GLP-1s Cost $5 to Make and $400 to Buy: Yale’s analysis: the manufacturing cost of a GLP-1 drug is $5 a month. They sell at a discounted price of $400. That’s eighty times markup. Pearl’s math: to make GLP-1s cost-neutral against the medical savings, the price has to be under $200. Trump Rx won’t help most people because you can’t use insurance there and $400 cash is still impossible on $60,000 a year.

•       Vibe Coding Is the Prescription: One year old. Lets clinicians build software in plain English without code. Pearl’s example: a heart failure patient at home, weighed daily on a Bluetooth scale, with an electronic stethoscope, ankle video, blood oxygen, exercise tolerance — all in an app a doctor could build in a weekend. Three days of fluid retention caught bef