This week Bill Gates published the most important thing a technology optimist can write: a warning. His essay on the turbulent AI era argues that the transition ahead will be among the most disruptive in human history, that its benefits and harms are arriving simultaneously, and — the sentence that should be read twice — that there is “no plan to ease the entry into the AI era.”

He is right about the diagnosis. I am writing because, for health, he is no longer right about the plan.

The lag is the enemy, and the lag is a choice.

Gates observes that past transitions produced winners and losers. True — but incomplete in a way that matters. Every general-purpose technology has been both equalizer and injustice at once; which one dominated for whom was decided not by the technology but by the institutions that received it. Industrialization eventually raised living standards for nearly everyone — but only after two generations of stagnant wages that economic historians politely call a “pause,” borne by people who died before the gains arrived. The pattern repeats because the institutional response has always lagged the technology by decades. That lag is not physics. It is a decision civilizations keep making by default — and the entire question of the AI era is whether, for once, we build the receiving institutions at the same time as the technology.

Gates calls for exactly this: new coordinating institutions, national and international. His pessimism is that they require political consensus nowhere in sight, while competitive incentives push everyone to full speed. He names the trap honestly: whoever is cautious loses to whoever is not.

That trap is real — at the frontier, where capabilities are built. But it inverts at the layer where AI meets patients. In frontier development, caution is a tax. In clinical deployment, caution is the moat: regulatory clearance, liability, clinician acceptance, and above all patient trust mean the reckless deployer of medical AI doesn’t win the market — he produces the scandal that closes it, while the deployer who published its safety cases and answers to something other than shareholders becomes the one health systems dare to adopt. In medicine, being trustworthy is the competitive strategy. Which means health is the one arena where Gates’s institutions can be built now, by willing parties, without waiting for the consensus he doubts will come. That is what we have done.

What makes health different — and dangerous.

For all of history, the scarce input in medicine has been expert time. The physician-hour is why care costs what it costs here and why it barely exists in much of the world. Clinical AI converts that scarce input into software: for the first time, medical cognition can be copied at near-zero marginal cost. Universal medicine becomes economically possible in a way it never was — not as charity, but as arithmetic.

But zero marginal cost has never once guaranteed distribution in health. Oral rehydration therapy costs pennies and children still die for lack of it. Insulin is a century old and gets rationed in the richest country on earth. When a health good’s natural scarcity disappears, artificial scarcity gets constructed on top of it — patents, pricing, licensing — because someone’s business model requires the price and the cost to diverge. A shareholder-owned medical AI will follow that script within a decade of its founding, whatever its founders intend. And as this year’s dismantling of America’s vaccine infrastructure demonstrates, even a health commons with fifty years of proven universal distribution can be undone in a few political seasons if its governance can be captured. The lesson is not despair; it is a design specification.

The plan.

Over the past year I have published, open-access and DOI-registered, an architecture built to that specification — The Trajectory Engineering Papers and the companion book Doctor AI. In brief: health treated as an engineerable trajectory rather than a series of billable events; a system of care in which the AI is an accountable analytic layer and a human remains the moral agent (Gates proposes “Human Reserved” roles and offers a health example — delivering the news of an incurable disease; our architecture reserved that ground in print months ago, and made AI deployment conditional on safety and bias audits in every population panel it serves, before deployment, not after the damage); a financing architecture that ends the zero-sum logic in which prevented disease is lost revenue; and — the load-bearing choice — commons ownership. The clinical AI at the center of this architecture will be built and owned by a public-purpose trust with a non-amendable charter: no shareholder exists whose returns require artificial scarcity, no donor can buy governance, and the founder — me — holds no financial stake by design, with the disclosures published in every paper.

The plan is also deliberately refutable. It specifies three curves — cost, outcome, and trust — with dated falsification gates. If the curves do not cross their status-quo counterfactuals on schedule, the architecture fails and says so. A plan you cannot fail is a slogan.

Region Zero, not region only.

And this is a global health proposal, not an American one. The architecture designates the United States as Region Zero — the hardest test case, with its employer-based financing, legal fragmentation, and deep mistrust — on the logic that an architecture which works here has proven itself against maximum resistance. But the papers specify translation frameworks for every health system family on earth: Beveridge systems like the UK’s, Bismarck systems like Germany’s and Japan’s, mandatory-savings models like Singapore’s, and greenfield settings across the Global South. The greenfield case may matter most: where AI-era medicine arrives without an entrenched fee-for-service system to dismantle, the trajectory architecture can be built clean — which is why the field’s first international teaching site is planned for Nairobi, not New York. The moral arithmetic is simple: medical cognition at near-zero marginal cost matters most exactly where the doctors aren’t. Four and a half billion people lack access to essential health services. They are not an afterthought of this plan. They are its point.

The invitation.

None of this diminishes Gates’s essay; it answers its bleakest paragraph. He wishes for institutions and doubts the world will build them in time. In health — the sector where his own foundation has done its most consequential work, and the one where trustworthy deployment is the winning strategy rather than the losing one — the institutions are being built now: an academic home in formation, a governance trust already chartered, the scholarship public and open for any reader’s red pen, including his.

I present the framework at the University of Oxford on September 16. The papers are at doi.org/10.5281/zenodo.19791977 and on SSRN. Mr. Gates: the health chapter of the plan you called for has a draft. Come mark it up.

Robin L.P. Blackstone, MD, FACS is a surgeon and health systems architect; founder of Blackstone Health, the H4 Alliance, and the H4 Alliance Trust; and author of “Doctor AI: Reimagining Healthcare, Rebuilding Trust, Delivering Health 4.0” (2026). She holds no financial interest in the Trust or its assets; full disclosures appear in every paper of the series.

Also published in my LinkedIn newsletter, Architecting Future Health.

https://lnkd.in/p/gaAjhieV

At 7:14 on a Sunday evening, Daniel Harper died at the table his mother had carried through three apartments.

Ninety-two seconds later, strangers began bringing him back.

He had been slicing roast beef when the pressure in his chest became impossible to ignore. He put down the knife, said the room felt hot, and lowered himself into the chair beside his mother. Then his eyes changed.

His daughter Lena called 911. A dispatcher told her to pull him to the floor and press hard in the center of his chest. A neighbor ran in. Sirens turned onto the street. Paramedics cut open Daniel’s shirt, placed pads against his skin, and saw the lethal rhythm moving across the monitor.

The first shock lifted his body from the floor.

The second restored a pulse.

From that moment forward, an astonishing system assembled around one failing human heart.

The ambulance transmitted an electrocardiogram before it reached the hospital. The emergency team was waiting. Blood was drawn, medicines were given, and a cardiologist met Daniel in a room built for precisely this emergency. Dye moved through his coronary arteries on a screen. One vessel narrowed to a thread and then disappeared.

A wire crossed the obstruction. A balloon opened it. A stent held it open.

By midnight, blood was flowing again.

Everything medicine did that night was extraordinary. Knowledge accumulated over generations had been converted into protocols, teams, instruments, drugs, electricity, sterile rooms, and trained judgment. Hundreds of people who would never know Daniel had contributed to the chain that saved him.

Health 2.0 could do what the system of Clara’s childhood could not: recognize a biological catastrophe in one specific person and mobilize immense capability around him.

But the catastrophe had not begun that evening.

Daniel was fifty-eight. He had worked rotating shifts at the same manufacturing plant for thirty-two years. The cafeteria served fried food through the night because that was what stayed hot. Cigarette breaks were where supervisors heard what was really happening on the line. Men who complained about pain were teased until they stopped complaining.

Daniel’s father had died young. His blood pressure had been high at two physicals. He had filled one prescription, stopped it when it made him feel tired, and intended to ask about another when work slowed down.

Work never slowed down.

Culture taught him to endure symptoms quietly and measure worth through reliability. Environment made the easiest choices the most damaging ones. Biology supplied vulnerability. His decisions were real, but they were not made in open space. They were made inside those forces, day after day, year after year.

Medicine saw almost none of that when he arrived.

It saw the artery.

The next morning, the cardiologist showed Daniel the image. On one side of the screen, the vessel ended abruptly. On the other, after treatment, it opened like a road cleared after a landslide.

‘That was the blockage,’ the physician said.

Daniel stared at the dark interruption on the first image.

‘How long was that happening?’ he asked.

‘Years,’ the physician said.

The answer unsettled him more than the procedure.

For years, his body had been recording a trajectory that no one was holding as a continuous story. The hospital could identify the precise location at which the story became an emergency. It could not reconstruct every force that had made that emergency increasingly likely.

Daniel went home with six medications, a cardiac-rehabilitation referral, dietary instructions, warning signs, appointments, and a folder thick enough to feel like a plan.

He followed much of it. He stopped smoking. He attended rehabilitation until his shift changed. He learned to read labels but still ate whatever was available at two in the morning. He became more attentive to his body, though he never stopped being the person everyone else called first when something broke.

Clara sat beside him at the old table during his first week home. She was smaller now, her hands folded around a cup of tea.

‘When I was a baby,’ she told him, ‘they changed the water.’

She touched the center of his chest, just above the bruise left by the defibrillator pad.

‘For you, they changed the artery.’

Daniel smiled. He knew she meant it as wonder, and it was wonder.

Modern medicine had returned him to his family after his heart had stopped. Few achievements in human history are more remarkable.

But his coronary artery was where the trajectory became visible.

It was not where the trajectory began.

THIS WAS HEALTH 2.0.

TREAT THE EVENT.

It brings extraordinary expertise to the individual when biological failure becomes visible—often after culture, environment, and accumulated decisions have shaped the trajectory for years.

Share

Author’s note: The Harper family is a fictional composite created to illustrate the four eras of health. The circumstances are drawn from common patterns in public health and care delivery.

The person who saved Clara Harper’s life never entered her house. He never touched her forehead, never heard her cry, never knew that her mother had already chosen the dress in which she would be buried if the fever came.

Two summers before Clara was born, Anna Harper had lost a son.

Thomas was eleven months old when the heat settled over the city and sickness began moving from one apartment to the next. First came the vomiting. Then the fever. Then the terrible stillness of a child too weak to cry.

The physician arrived with a leather case and a grave face. He could name what was happening. He could sit beside the bed. He could not reverse the course.

By morning, Thomas was gone.

On their block, three other infants died that week. One mother blamed the milk. Another blamed the heat. Anna blamed herself. The city recorded four deaths on four separate lines.

No one yet held the whole pattern.

Then strangers began appearing near the neighborhood pump. They collected water in glass bottles. They asked where waste was emptied, where milk was stored, and which houses had lost children. An engineer placed marks on a map. A public-health nurse carried names from kitchen tables into a municipal office.

What people had been taught to endure as fate began to look like infrastructure.

The work was slow, disruptive, and largely invisible. A new main was laid beneath the street. Sewage was separated from drinking water. The supply was filtered and disinfected. Milk sellers were inspected. Refuse was collected before it could rot in the summer heat.

No one called it personalized medicine. No one asked about Clara’s genome. They changed the world immediately around her.

When Clara was born, Anna trusted almost nothing. She boiled water until the room filled with steam. She covered the milk. She washed the feeding bottle twice. Every time the baby slept too deeply, Anna placed two fingers beneath her nose.

Then the first summer passed.

Clara remained well. So did the baby upstairs, and the twins across the courtyard. The small white coffins appeared less often.

There was no dramatic rescue in Clara’s story. No ambulance came. No surgeon opened her body. No medicine pulled her back from the edge.

The edge had been moved farther away.

Clara’s susceptibility as an infant was no different from Thomas’s. What changed was the environmental force pressing upon it. Clean water made thousands of decisions before any mother had to make them.

Culture changed more slowly. Anna still believed a good mother should be able to protect her children through vigilance alone. It took years for her to understand that Thomas had not died because she failed him. The conditions surrounding them had failed them both.

Clara grew up at the same oak table where Anna had cooled boiled water in a blue enamel bowl. She did her schoolwork there. Years later, she fed her own son there. When the family moved, the table went with them, its legs shortened once, its surface scarred by hot pans and children’s pencils.

Clara lived long enough to hold a granddaughter — one who would someday ask why the family had kept such an ordinary table for so many years.

Clara never knew the name of the engineer who mapped the deaths, the nurse who carried the pattern from house to house, the workers who lowered pipe into the earth, or the official who finally agreed to fund it.

The person who saved Clara was not one person after all.

It was a public capacity: the ability to see a shared danger, change the environment, and protect people before they became patients.

A pipe beneath the street had given an entire family a future.

**THIS WAS HEALTH 1.0. PROTECT THE POPULATION.**

It acts on shared conditions—especially the environment—before the individual is known and before biology becomes disease.

*Author’s note: The Harper family is a fictional composite created to illustrate the four eras of health. The circumstances are drawn from common patterns in public health and care delivery.*

We talk about the heavy lifting culture and environment play in health outcomes. Today we go deep on that sliver of biology and something you may never have heard of: Endotyping, the idea that will decide whether precision medicine becomes real or stays a brochure.

I want to start with a drug that failed.

In the late 1990s, researchers built an antibody against interleukin-5. IL-5 is the signal that recruits and sustains eosinophils — the white blood cells found in enormous numbers in the airways of people with asthma. The pharmacology was elegant. Block the signal, lose the cells, relieve the disease.

They gave it to patients with asthma. Circulating eosinophil counts collapsed, exactly as engineered. And the patients were, clinically, about the same.

When a drug does precisely what it was designed to do and the patient does not improve, there are two possible conclusions. The target was wrong. Or the patient was wrong.

The field went with the first. Careful, serious people concluded that eosinophils were passengers rather than drivers — along for the ride, not steering. The program nearly died.

It was the second answer that mattered. The target, eosinophils, was right.

We had given it to the wrong patients.

The way that mistake got corrected is the most useful thing I know about chronic disease. And the fact that we are, right now, making the identical mistake with the most consequential drugs of my career is why I am writing this.

The word that does the work

Two terms, the whole argument lives in the space between them.

A phenotype is what a disease looks like from outside. Age of onset. What triggers it. How severe. Whether you also have eczema, or reflux, or allergies. Phenotypes are observational, and I do not want to be dismissive about them — a clinician who has seen ten thousand patients has real knowledge encoded in her pattern recognition.

An endotype is a subtype of disease defined by the mechanism actually driving it. Not what it looks like. What it is, at the level of a molecular pathway.

A phenotype tells you what to call something. An endotype tells you what mechanism to aim at if you want to impact it.

The truth is apparent. We “treated” phenotypes for much of the history of medicine. Almost every diagnosis I write down in a day was originally defined as a phenotype, because it was named before we had any way to see mechanism. Habits die hard.

Asthma is named after a sound, from the Greek for panting. Obesity is defined by a ratio of weight to height, using a formula developed in the 1830s by a Belgian mathematician who was studying populations rather than patients. Fatty liver was a shadow on an ultrasound. Heart failure is a description of what a person can no longer do.

We named diseases after their surfaces, because for most of medical history the surface was all we could reach. That part was honest. It was the limit of the available instruments.

The failure — and I include myself in it — is what we built on top of those names. We wrote guidelines for them. We created billing codes, and the codes became the operative reality, because the codes are what pays. We designed clinical trials around them, which means we tested drugs in populations defined by their outward appearance. We trained physicians to recognize them and treat them. And we told patients: this is what you have.

We enshrined the phenotype, me taught it to our patients and our colleagues. We billed by it.

Then, when the treatment for that name failed in a patient who carried that name, we did not question the name.

We questioned the patient.

Asthma: the proof

The people who did not give up on IL-5 asked a different question. Not who has asthma, but whose asthma is being driven by this pathway right now?

That turns out to be measurable, and cheaply. Eosinophils in the blood. Eosinophils in sputum. Nitric oxide in exhaled breath. IgE. A history of frequent exacerbations.

So the trials were run again, restricted to patients with evidence of active eosinophilic inflammation. The same class of molecule cut exacerbations by roughly half. It let people taper off oral steroids they had been taking for years — with everything chronic glucocorticoids do to bone, glucose, mood, and face.

Nothing about the drug changed. The definition of the patient changed.

That experience reorganized severe asthma into something closer to actual disease biology. Today the field splits it, roughly, into two. Type-2 high is the IL-4, IL-5 and IL-13 axis — eosinophilic, allergic-inflammatory, measurable in blood and breath. It is somewhere around half of severe asthma, and it now has a whole shelf of targeted biologics aimed at different points in that pathway, including one that works upstream at the alarmin level and reaches somewhat further than the rest.

The other half is type-2 low: neutrophilic, or paucigranulocytic, or obesity-associated, or smoking-related, and some of it not primarily inflammatory at all. It remains comparatively underserved, and the honest reason is that we understand its mechanisms less well. There is no clean biomarker and no clean target yet.

But consider what happened to those type-2 low patients in all the years before we knew to draw this line.

She arrives on maximal inhaled therapy, still symptomatic. So we escalate. We add oral steroids and hand her the metabolic consequences. We review her inhaler technique. Again. We note in the chart that adherence may be contributing. We may use the word compliance, which has an entire negative worldview inside it.

She was on the wrong drug. Not the wrong dose of the right drug — the wrong drug, for a mechanism it does not touch. Her body was behaving exactly as its biology dictated. The failure was ours: a mechanistic hypothesis held with unearned confidence.

And our explanation for our own failure was her character.

That is not a story about asthma. Asthma is simply where we caught the pattern first, because the biomarkers were cheap and the pathway was clean. That pattern is the default operating mode of twentieth-century chronic disease care, and it is still running in most clinics today.

GLP-1: the drug without the endotype

I have spent my career in metabolic and bariatric medicine, and I did not expect to see anything like the last five years.

The incretin-based therapies are the most consequential pharmacology of my professional life. Semaglutide produces roughly 15% total body weight reduction. Tirzepatide, in its pivotal obesity trial, roughly 21% — territory we previously reached only with surgery. And in people with obesity and established cardiovascular disease but without diabetes, semaglutide reduced major adverse cardiovascular events by about 20%. That last result should have ended an argument I have been having for thirty years. You cannot look at obesity and still call this a lifestyle problem. We are preventing deaths.

Here is how we prescribe them.

Start low. Titrate up. Push to the highest tolerated dose. If it isn’t working, switch to the other one. If that fails too, the patient is a non-responder. And if it does work — once she reaches her “goal” weight, her insurance often stops paying for the drug that got her there, as though the disease were cured rather than treated.

That is the algorithm. There is no mechanism in it anywhere.

Fifteen percent is an average, and an average is a fact about a population, not a prediction about a person. Inside those trial distributions are people who lost a third of their body weight and people who lost almost nothing at full therapeutic dose. Something real distinguishes them. We cannot currently tell them apart in advance. We find out by trying, over months, at extraordinary cost.

And the heterogeneity is not hiding. It is visible to me in clinic every week.

There is the patient whose satiety signal never seems to arrive — not psychological hunger, neurological; the message that she has eaten is not being delivered. There is the patient with abnormally rapid gastric emptying, full for twenty minutes. There is the patient who eats to regulate emotion, whose appetite biology may be close to normal and whose problem is not appetite at all. There is the patient with a genuinely low resting energy expenditure, eating no more than her neighbor and gaining anyway.

Serious work, much of it out of Mayo, has begun formalizing patterns like these into measurable categories with differential drug response. That is exactly the right direction of travel. But notice what it still is: phenotyping. Observable patterns. We are in obesity medicine roughly where asthma was in the 1990s. We have the drugs before we have the endotypes.

Now add the economics, because this is where it stops being academic.

These drugs are expensive and supply has been constrained, so access is being rationed with the crudest instrument available: BMI thresholds, prior authorization, step therapy, employers dropping coverage mid-year.

A BMI threshold is a phenotype, and a bad one. It is a ratio of mass to height. It says nothing about mechanism, nothing about who will respond, nothing about whose organs are actually failing.

So look at the position we have engineered. A class of drugs that transforms some lives and barely moves others. No way to know which in advance. A price that forces rationing. And a rationing method that is biologically illiterate. Then we act surprised when patients conclude the whole enterprise is a racket, and when payers conclude the same.

Here is the way out, and it is the pivot of this entire essay.

If we could identify in advance the woman who will lose 25% and the woman who will lose 4%, we would not be rationing anymore. We would be targeting.

Those are categorically different acts. Rationing means withholding something that works from someone who needs it, on the basis of cost. Targeting means directing a mechanism-specific therapy at the mechanism it addresses, and offering the other patient something that actually fits her biology.

To be plain, because this argument gets borrowed by people who want the opposite: targeting is not a case for spending less on obesity. It is a case for spending the same money accurately instead of blindly.

The first destroys trust and generates litigation and legislation. The second is simply good medicine. And it happens to cost less.

MASH: what it looks like when they arrive together

For most of my career, “fatty liver” was a shrug on a radiology report. An incidental note on an abdominal ultrasound. Lose some weight, follow up in a year.

Here is what that shrug was hiding. Under those three words sat two completely different fates. For most people, liver fat that would never harm them. For a minority, a progressive fibrotic disease heading toward cirrhosis and hepatocellular carcinoma. One label. Two prognoses. No way to sort them — and therefore, conveniently, no urgency about sorting.

Then the field did three disciplined things in quick succession.

First, it renamed the disease around its mechanism. Nonalcoholic fatty liver disease became metabolic dysfunction-associated steatotic liver disease. Nonalcoholic steatohepatitis became MASH.

That looks like nomenclature politics. It is not. Look at the old name: nonalcoholic. The disease was defined by what it was not. That is a diagnosis of exclusion wearing the costume of an entity. The new name defines it by the metabolic dysfunction actually driving it — and you cannot endotype a disease that is named after an absence.

Second, it identified the population that actually progresses. MASH with moderate-to-advanced fibrosis, stage F2 to F3. Not the amount of liver fat. Not BMI. Not transaminases. Scar tissue. Fibrosis stage turned out to be the dominant predictor of liver-related death.

Third — and this is what made it real — that group became findable without a biopsy. Staging fibrosis used to require putting a needle into someone’s liver, which is precisely why almost nobody was staged. Now FIB-4 calculates from labs most patients already have, and elastography measures stiffness in minutes. An endotype that nobody could identify at scale became one you can find in an ordinary clinic.

And then the drugs arrived, approved for that group specifically.

Resmetirom, in March 2024: a thyroid hormone receptor-beta agonist acting inside the hepatocyte, indicated for noncirrhotic MASH with moderate-to-advanced fibrosis. In its pivotal trial roughly a quarter to a third of treated patients achieved MASH resolution, and a comparable share achieved fibrosis improvement, against markedly lower rates on placebo.

Then semaglutide, in August 2025: the same fibrosis stages, an entirely different mechanism, coming at the liver through systemic metabolic disease. Europe is moving in the same direction, with a positive regulatory opinion in early 2026.

Let me be honest about the effect sizes. They are not miraculous. I would rather say that than oversell.

The point is not the effect size. The point is the architecture. Developed for an endotype, tested in an endotype, approved for an endotype — and the endotype identifiable non-invasively. A treatable population carved out of an untreatable label.

Which produces the question that will define the next decade of chronic disease practice: two mechanisms, one endotype — which patient gets which, or both, and in what order?

Notice that this question could not have been asked five years ago. You cannot ask it about “fatty liver.” There is no version of that conversation that begins with a shadow on an ultrasound. You can only ask it once the label has been cut into mechanisms.

One more thing before I move on.

Everything I have said so far treats mechanism as though it were intrinsic — as though an endotype were something a patient simply has, like eye colour. Most of it is not. The T2-low asthma endotype, the one with no targeted biologic, is also the one most associated with traffic pollution and housing stock. Which means the patients we cannot treat and the patients we have failed to protect are substantially the same people.

There is a name for the other half of this. The exposome — the sum of what a person has been exposed to across a lifetime, proposed as the complement to the genome. A genome is a fixed input. An exposome is a trajectory. And almost none of it appears anywhere in the medical record.

Why we didn’t do this sooner

The comfortable explanation is that the science was not ready. That is partly true and mostly an alibi.

The harder explanation is that endotyping exceeds the cognitive capacity of the clinical encounter we built the system around.

Consider what defining and applying an endotype actually requires. Multiple biomarkers interpreted jointly rather than sequentially. Imaging quantified rather than described in prose. Genetic, and increasingly proteomic and metabolomic, data. The longitudinal trajectory — not this eosinophil count, but the pattern over three years. Response to prior therapy, fed back as evidence about mechanism. And then all of it integrated into a probabilistic judgment about which mechanism is driving this person’s disease, at the moment a decision must be made.

No physician can do that in fifteen minutes. I cannot do it, and I have spent decades in a subspecialty built on precisely this kind of metabolic reasoning. The human mind holds a handful of variables in genuine simultaneous consideration. Endotyping routinely requires dozens, weighted, with interactions.

So we did what any overloaded system does. We simplified until the task fit the resource. Diseases became labels. Labels became guidelines. Guidelines became algorithms. Algorithms became defaults. We built quality metrics that reward applying the default to the label, called the result standard of care, and told ourselves the variation left over was noise, or non-adherence, or social determinants.

This is why I resist framing artificial intelligence in medicine as a productivity story. Ambient documentation is useful. Inbox drafting is useful. Neither is the point.

The point is that computational integration of high-dimensional patient data is the only known mechanism by which endotype-level precision becomes deliverable to a population rather than available to a research cohort. Precision medicine has been promised for twenty-five years and remains, for most patients, a brochure — not because we lacked conviction, but because the cognitive infrastructure to deliver it did not exist.

It exists now. And that changes the moral situation. When precision was impossible, treating everyone with the same label the same way was reasonable practice.

Now it is a choice.

The trust argument

I care about this because of what imprecision has done to the relationship between people and medicine.

Ask why trust has collapsed and the answers reach for something big and recent. A scandal. A pandemic. Politics. Social media. I think it is smaller than that, and much more our own doing.

Consider what the non-responder experiences. She is told she has a disease. She is told there is a treatment. She takes it exactly as prescribed, often at real cost and real side-effect burden. It does not work.

And then the system’s explanation for its own failure is her.

She must not be taking it correctly. She must be eating more than she reports. She must not want it enough.

We have said versions of this to asthma patients for decades and to patients with obesity for a century — and we said it while being wrong about the mechanism of their disease.

Multiply that by millions of encounters, across generations, and you do not have a communications problem. You have several million people with direct personal evidence that an institution claiming to understand their bodies did not.

Precision is the repair. Not because accuracy is elegant, but because “that treatment didn’t work because your disease runs on a different mechanism, and here is the one that targets yours” is a fundamentally different human event than “you must not be trying.”

The first is a physician using knowledge on a patient’s behalf. The second is an institution protecting itself.

Trust is not rebuilt with messaging strategy. It is rebuilt by being right more often, and by having an honest account of the times we are not.

Unthinkable — Debatable — Obvious

I think in terms of the Overton Window — the range of positions a society treats as reasonable to hold in public. Policy does not change because an argument is proven. It changes when a position moves from unthinkable to debatable to obvious.

So the practical question for anyone trying to change a health system is: which sentences need to move? Here are five. Every one is defensible today with evidence already in hand. Not one is consensus.

A diagnosis is a hypothesis, not a fact. Today this sounds like an attack on medical authority. It is the opposite. A physician who treats a diagnosis as a hypothesis to be tested is doing more rigorous work than one who treats it as an answer.

Coverage decisions made on labels rather than mechanisms are medically indefensible. Right now a BMI cutoff or a diagnostic code determining access to therapy is treated as ordinary administration. It should be as unacceptable as prescribing a biologic without checking the biomarker. Payers are not the villains here; they are using the only instrument the clinical world handed them. When we produce endotypes, we remove that excuse — from them and from ourselves.

Treatment failure is evidence about the disease, not about the patient. This is the highest-yield, most important sentence on the list, because it converts a moral judgment into a scientific observation. Non-response is data. It tells you your mechanistic hypothesis was wrong. In a properly designed system, every failed treatment would update the model rather than sully the patient’s reputation.

Delivering precision at population scale requires computational capacity beyond human cognition — and using it is a professional obligation, not a threat to the profession. This is the hardest one, because it asks physicians to accept there is a limitation on our own ability to reason. But no one thinks less of a cardiologist for using an echocardiogram instead of an ear. The instrument extended the sense. Intelligence is an instrument that extends judgment.

Care, payment and research should be organized around mechanisms rather than organ-based specialties and diagnostic labels. The eosinophilic endotype crosses asthma, chronic rhinosinusitis with nasal polyps, eosinophilic esophagitis and atopic dermatitis. Four specialties, four clinics, four prior authorizations, one mechanism, sometimes one drug. Our organizational chart is an artifact of nineteenth-century anatomy. Biology does not respect it.

Health 4.0

If Health 1.0 was the physician’s senses, 2.0 the laboratory and the hospital, and 3.0 the digitized, protocol-driven, industrialized system we are currently drowning in — then 4.0 is not more digitization. It is the point at which the system’s unit of analysis stops being the label and becomes the mechanism, and the person.

Practically, that system looks like this. Diagnosis becomes probabilistic and provisional: a stated hypothesis, with a confidence, and a plan to test it. Biomarkers, imaging and molecular data are integrated computationally at the point of decision rather than filed as documents. Therapy is selected against mechanism, with predicted response stated up front so that both physician and patient know what is being tested. Response is measured and fed back, so non-response revises the model. Payment follows mechanism-matched appropriateness rather than label-based eligibility. And research runs continuously inside care, because every treated patient with a characterized endotype is an observation.

That system is cheaper than the one we have. It is also kinder — which is not a coincidence, because most of what is expensive in health care is the cost of being wrong repeatedly and slowly.

None of this is speculative. Every component exists somewhere today. Severe asthma clinics run on this logic. Oncology has run on it for two decades and nobody calls it futuristic; they call it standard. MASH just did it, in public, in under two years.

What does not exist is the decision to build the whole system this way — for chronic disease, for everyone — and the honesty to admit that until we do, we are practicing a medicine that systematically mistakes its own limitations for the patient’s failings.

That interleukin-5 antibody was never a failed drug. It was a working drug, given to a population defined by a sound the lungs make.

We have thousands of drugs like that. And millions of patients being told, right now, that a treatment should have worked for them.

The instruments finally exist. What is missing is the willingness to say plainly that a diagnosis was never the answer — only the beginning of the question.

Next week: the other half of the mechanism. Environment does not sit beside biology — it becomes biology, and it stays. Three patients, three exposures, three endotypes. Including a question I have been avoiding: whether the MASH endotype I just finished celebrating is itself two different diseases.

The full architecture — evidence, governance, and what it would take to build — is in Doctor AI: Reimagining Healthcare, Rebuilding Trust, Delivering Health 4.0.

robinblackstone.com

Notes and sources

Anti-IL-5 in unselected versus eosinophilic asthma. Early negative trials in unselected populations (Flood-Page et al., AJRCCM, 2007) versus subsequent biomarker-selected trials: DREAM (Lancet, 2012) and MENSA (NEJM, 2014).

Tezepelumab and the type-2 low population. NAVIGATOR (NEJM, 2021); approved December 2021.

Semaglutide 2.4 mg in obesity. STEP 1 (NEJM, 2021): −14.9% mean body weight at 68 weeks.

Tirzepatide in obesity. SURMOUNT-1 (NEJM, 2022): −20.9% at 72 weeks on 15 mg.

Cardiovascular outcomes. SELECT (NEJM, 2023): 20% relative reduction in MACE (HR 0.80) in patients with overweight or obesity and established cardiovascular disease, without diabetes.

Obesity phenotypes. Acosta and colleagues, Mayo Clinic — phenotype-guided pharmacotherapy in obesity.

MASLD/MASH nomenclature. Multi-society Delphi consensus, June 2023.

Resmetirom. FDA accelerated approval, March 2024, for noncirrhotic MASH with moderate-to-advanced fibrosis (F2–F3). MAESTRO-NASH (NEJM, 2024).

Semaglutide in MASH. FDA approval, 15 August 2025, for noncirrhotic MASH with moderate-to-advanced fibrosis (F2–F3). CHMP positive opinion for conditional marketing authorisation in the EU, January 2026.

A note on numbers: effect sizes are stated as ranges or approximations throughout, because the precise figure depends on dose, duration and endpoint definition. Where I have rounded, I have rounded conservatively.

There is no end of commentary. A take on drug prices. A thread about prior authorization. A panel on AI. What we rarely get is someone holding onto a single claim and following it down — through the biology, then the economics, and then the part where it implicates the person making the argument.

That is what we are doing in August. One argument, taken as far as I can take it. Long enough to be uncomfortable, sourced well enough to be argued with, and written as much for people outside medicine as for those inside it. Actually, maybe even more relevant to people struggling to find an answer to a difficult health problem.

I am a surgeon. I spent my career in metabolic disease, which means I spent it watching people be blamed for their own biology, for their obesity. I am also the author of Doctor AI Reimagining Healthcare, which is a book about rebuilding trust in a system that has spent most of it. Those turn out to be the same subject.

Some weeks the argument will be clinical. Some weeks it will be about money, or law, or the sentences we are not yet willing to say out loud in public. What they will have in common is that I will be frank, I will show my sources, and I will tell you where I think I might be wrong.

The first essay lands Friday.

It is about a drug that failed for a decade — but in reality had not failed at all. We gave it to the wrong patients, and then we blamed the patients. I think we are doing precisely the same thing right now with the most consequential medications of my career.

If that sounds like an argument worth following, subscribe. One email a week. Nothing else.

From metabolic surgery to histotripsy, medicine is moving from managing consequences to making precise, mechanism-level interventions. Our health system has not caught up.

By Robin P. Blackstone, MD, FACS

The new pattern of medicine: precision at the mechanism, not endless rescue downstream.

Recently, a patient with a liver tumor lay still on a table while a focused beam of sound destroyed the tumor from outside the body.

No incision. No treatment needle. No radiation.

The technology is called histotripsy. It uses focused ultrasound to create microscopic bubble clouds that mechanically break down targeted tissue. One recent patient was home and back to normal activities the same evening. The FDA authorized the Edison system in 2023 for the noninvasive destruction of liver tumors; more than a thousand people have since undergone liver-tumor procedures with the system, and in May its manufacturer submitted a request to expand the indication to kidney tumors.

Long-term cancer outcomes are still being studied. That matters. But so does the shape of the advance: a treatment that once required entering the body can now, for selected patients, be delivered from outside it.

I have spent my career as a metabolic and bariatric surgeon, and I want to tell you that this story is not really about a machine.

It is about a pattern.

Once you see it, you cannot unsee it.

The knot

In 2019, in the Bulletin of the American College of Surgeons article “Surgery and the Gordian Knot of Metabolic Syndrome”, my coauthors and I described metabolic syndrome using an image from antiquity: the Gordian knot.

Obesity, type 2 diabetes, hypertension, dyslipidemia, fatty liver disease, polycystic ovary syndrome and some cancers have been divided into separate specialties, separate appointments, separate drugs and separate payment streams. Medicine spent a century trying to untie the knot one strand at a time—managing each consequence after it appeared, often for the remainder of a patient’s life.

Our argument was that these were not simply independent failures occurring in the same body. They were often connected manifestations of a shared metabolic pathology. A decisive intervention at the mechanism could change several strands at once.

Metabolic surgery had already demonstrated the principle. In one long-term prospective study, Roux-en-Y gastric bypass produced durable weight loss and type 2 diabetes remission in approximately half of affected patients at twelve years. Surgery did not merely make the stomach smaller. It changed metabolic signaling, appetite regulation and the body’s defense of weight.

But the country was not ready to hear what the biology was saying.

Surgery was called extreme. Obesity was still treated as a failure of discipline. Patients were told, in one form or another, that the knot was their fault.

Then the GLP-1 drugs arrived, and something changed that had never changed before.

The number that had never moved

In July 2026, Gallup reported that its measure of U.S. adult obesity had fallen from a peak of 39.9% in 2022 to 36.4% in 2026, while the share of adults reporting current GLP-1 use for weight loss rose from 3% in 2024 to 11%.

These are self-reported survey data, not measured national examination data, and parallel trends alone cannot prove that the drugs caused the entire decline. But Gallup has used the same method over time, and the direction is statistically meaningful. After decades in which the curve seemed capable of moving in only one direction, it moved the other way.

No lecture about personal responsibility accomplished that.

A class of molecules helped make it possible.

The drugs did not solve the biology. They solved the politics.

Metabolic surgery had demonstrated the biology decades earlier. The drugs made that biology culturally legible. A weekly injection made it finally sayable, at scale, that obesity is a chronic biological disease—not a referendum on character.

That was a therapeutic breakthrough. It was also a political one.

The pattern is bigger than obesity

The obesity story is the loudest instance of something happening across medicine.

Look at the shape of it.

In liver tumors, focused sound can destroy targeted tissue without an incision. Kidney tumors may be next.

Histotripsy status as of July 2026. A human study does not mean that a treatment is approved, proven or appropriate for every patient. Prostate work shown here is for benign prostatic hyperplasia—not prostate cancer.

In severe aortic stenosis, a new valve can be delivered through a catheter rather than through an open chest. In selected patients, randomized trials have found comparable major outcomes between transcatheter and surgical valve replacement through five and six years, even as questions of lifetime durability and reintervention remain.

In some advanced cancers, immunotherapy has produced durable survival—and, for a subset of patients, something approaching long-term disease control—where survival was once measured in months. For some patients, managing cancer as a chronic disease is now current therapy. Preventing more cancers from developing at all—and intercepting others before they become clinically apparent—is the horizon we can already see.

In hepatitis C, eight to twelve weeks of oral medication now cures more than 95% of patients, where medicine once waited for cirrhosis, liver cancer or transplantation.

Different organs. Different specialties. Different technologies.

The technologies differ. The structural move is the same: pull disease upstream of rescue.

The same move: pull disease upstream of rescue. Stop paying forever for the consequences when a precise intervention can alter the mechanism.

Stop untangling the knot.

Cut it.

The honest part

None of this is free, and I will not pretend otherwise.

My own field sometimes dismisses the medications by saying that when patients stop taking them, weight often returns. That is true. But surgery is forever too—and so is a surgical complication.

Stopping a drug is a decision a patient may be able to revisit next month. A stricture, leak, nutritional injury or chronic complication may not be reversible.

Medication carries its own risks, costs and uncertainties. Metabolic surgery has unmatched durability for many patients and a long evidence base, but it also asks the patient to accept an irreversible intervention. When timing is crucial or disease severity demands a decisive intervention, surgery may be the first option. In most other cases, it should follow a fair trial of medication and comprehensive nonsurgical care—fairly offered, adequately supported and not declared a failure simply because coverage ran out.

Histotripsy is genuinely noninvasive, but it is new, not appropriate for every tumor, and its long-term oncologic role is still being defined. TAVR avoids a sternotomy, but age, anatomy, valve durability and future procedures still matter.

Every advance carries a ledger of risk, benefit, durability and reversibility.

The honest clinician puts both columns on the table. Then the patient, the evidence and the biology choose together. The Human Principal—the person whose life and health are at stake—carries both the authority and the responsibility for the decision.

That honesty is not a brake on innovation. It is what earns the trust innovation requires.

Why this is a systems story

When a decisive, less-invasive treatment exists but patients cannot reach it, the remaining barrier is often no longer scientific.

It is architectural.

We built a health system that pays, in perpetuity, for the consequences of disease: insulin, dialysis, repeat admissions, complications and rescue. At the same time, it places interventions that may prevent those consequences behind coverage cliffs, prior-authorization mazes, fragmented benefits and geographic treatment deserts.

That arrangement made a grim kind of sense when decisive options did not exist.

Increasingly, they do.

Keeping the old payment and delivery architecture in place is now a choice—and an expensive one.

This is the heart of what I mean by Health 4.0: a system organized around prevention, early detection and durable treatment at the level of mechanism, delivered through the precise intervention that fits the person. Sometimes that intervention will be behavioral. Sometimes pharmacologic. Sometimes surgical. Sometimes it will be a beam of sound delivered from outside the body.

The point is not that one tool wins.

The point is that the system must stop forcing every patient through the same door.

Cutting the Knot

This essay begins a series.

In Cutting the Knot, I will take the metabolic knot apart strand by strand: type 2 diabetes, hypertension, dyslipidemia, fatty liver disease, PCOS and obesity-associated cancer. I will place the surgical and pharmacologic evidence side by side, including what works, what fails, what lasts, what reverses and what we still do not know.

This is not an argument for surgery over medication, or medication over surgery.

It is an argument against making patients wait for rescue because the system is more comfortable paying for consequences than changing mechanisms.

Alexander understood something medicine forgot for a century: some knots are not meant to be patiently untied.

They are meant to be cut.

Across medicine, we have finally begun to build the swords.

The question now is whether we will build a system with the courage—and the honesty—to use them.

Where are you seeing this shift in your own field? What disease are we still managing strand by strand when we should be changing the mechanism? I would like to hear the examples I am missing.

Sources and further reading

In 2019, in the Bulletin of the American College of Surgeons article, “Surgery and the Gordian Knot of Metabolic Syndrome”my coauthors and I described metabolic syndrome through an image from antiquity: the Gordian knot.

• Surgery and the Gordian Knot of Metabolic Syndrome

• Mercy patient experience with histotripsy

• FDA De Novo classification of the Edison System

• Focused Ultrasound Foundation: more than 1,000 liver-tumor patients treated

• HistoSonics kidney-tumor FDA submission

• HOPE4KIDNEY pivotal study

• Pancreatic-tumor histotripsy feasibility study

• Prostate BPH histotripsy study

• Exploratory abdominal-tumor histotripsy study

• Technical review of histotripsy and acoustic access

• Preclinical transcranial histotripsy tumor study

• Preclinical histotripsy osteosarcoma study

• Twelve-year outcomes after gastric bypass

• Gallup: 2026 GLP-1 use and obesity trend

• Five-year outcomes of TAVR versus surgery

• Six-year outcomes of TAVR versus surgery

• National Cancer Institute: durable immunotherapy responses in a subset of advanced cancers

• CDC: hepatitis C treatment and cure

For more information about Health 4.0 - a novel health system for the 21st Century go to https://robinblackstone.com

In 333 BC, Alexander the Great marched into Gordium and found a wagon lashed to a post by a knot so tangled that no one had ever loosened it. An oracle had promised the rule of Asia to whoever could. Alexander studied it for a moment, drew his sword, and cut it through in a single stroke.

I have been thinking about that knot for most of my career, because it is the truest metaphor I know for the way medicine has approached chronic disease — and for the way that approach is now, finally, changing.

One disease wearing many masks

Seven years ago, my coauthors and I opened an article in the Bulletin of the American College of Surgeons with that story. We were writing about the metabolic syndrome: obesity, type 2 diabetes, high blood pressure, abnormal cholesterol, fatty liver disease, polycystic ovary syndrome, and several cancers. For a century, medicine treated each of these as its own problem, with its own specialist, its own drugs, its own clinic — picking at one strand of the knot at a time, indefinitely.

Our argument was heresy at the time and is becoming common sense now: these are not separate diseases. They are the downstream expressions of one disease, wound around a single control point in the brain and gut that defends body weight the way a thermostat defends temperature. And if that is true, then the whole knot can be cut with one decisive stroke rather than untied strand by strand.

We had proof. Metabolic and bariatric surgery does exactly that — it resets the body’s defended set point and sends type 2 diabetes into durable remission in about half of patients even twelve years later. It was, and remains, one of the most effective durable therapies in all of medicine.

The country did not want to hear it. Surgery was “drastic.” It was what you did when you had failed to control yourself. The stigma that coats obesity — the lie that it is a matter of willpower — coated the treatment too. So the most effective durable therapy we had reached roughly one in a hundred of the people who qualified, and we told ourselves that was simply how things were.

Then the drugs arrived, and everything changed — though not in the way the headlines say.

The number that had never moved

In 2026, for the first time in the modern era, the American adult obesity rate went down. Gallup measured it at 36.4 percent, off from a record 39.9 percent in 2022, as the share of adults taking a GLP-1 medication for weight loss rose from 3 percent to 11 percent in two years.

Sit with that thought for a second. Every anti-obesity effort of the past forty years — the food pyramids, the gym memberships, the willpower sermons — failed to bend the national curve even slightly. A class of molecules did it in four years.

The individual results are staggering by the standard of everything that came before. Semaglutide takes off around 15 percent of body weight and, in a landmark trial, cut heart attacks and strokes by a fifth in people with obesity and heart disease. Tirzepatide does better — about 20 percent — and in the first head-to-head trial it beat semaglutide cleanly. There is now a pill, orforglipron, that reaches roughly 12 percent weight loss with no needle at all. And a triple-hormone agent, retatrutide, is posting numbers in trials that approach what surgery achieves.

Here is the part almost everyone gets wrong. The drugs did not solve the biology. Surgery had already proven the biology, years earlier. What the drugs solved was the politics.

A weekly injection did what fifty years of surgical outcomes could not: it made it sayable — in the exam room, the boardroom, the policy hearing — that obesity is a chronic, biological disease and not a character flaw. That is the real revolution. GLP-1s moved the Overton window, the invisible boundary of what a society is willing to say out loud. The moment it moved, everything my colleagues and I had argued in 2019 stopped sounding radical and started sounding obvious.

I will take the win. But I am not celebrating the way you might expect, because the most important part of this story is not about obesity at all.

The pattern is bigger than obesity

A few weeks ago, a patient had a liver tumor destroyed while lying still on a table. No incision. No needle. No radiation. A focused beam of sound liquefied the tumor from the outside — a technology called histotripsy. She went home the same day. Since the FDA cleared it in 2023, more than a thousand people have had liver tumors treated this way, and this spring the same system was submitted for kidney tumors. The technology may be effective in a broad range of solid tumors, treated earlier. That is not an isolated marvel. It is the same move, in a different organ.

Look at the shape of what is happening across medicine right now. In severe aortic stenosis, TAVR — a replacement valve delivered through a catheter in the leg — now matches open-heart surgery at five and seven years, without ever cracking the chest; it has quietly become the default for a huge share of patients. In advanced cancer, immunotherapy is turning what used to be a death sentence within months into a chronic, survivable disease for many. In hepatitis C, a pill cures more than 95 percent of patients in eight to twelve weeks, where we once waited helplessly for cirrhosis and a transplant.

Different organs. Different specialties. Different technologies. But the same underlying move every single time: stop managing the consequences of a disease for a lifetime, and make one precise, less-invasive, decisive stroke at its mechanism. Stop untangling the knot. Cut it.

Obesity is simply the loudest, most public instance — the one that put the whole shift on magazine covers. But it is one case of a general transformation, and that is why it matters far beyond the people who take these drugs or have these operations. We are watching medicine change its fundamental posture toward chronic disease, from lifelong management to decisive cure.

The honest ledger

None of this is free, and I refuse to pretend otherwise, because dishonesty is exactly how you squander the trust these tools will require.

My own field has a reflex: wave the drugs away with stop taking it and the weight comes back. It is true — about two-thirds of the lost weight returns within a year of stopping. But said by itself, that is a double standard, and it is beneath us.

Surgery is something a patient lives with for the rest of their life, too. The difference is the direction of the door. Stopping a drug is a decision you can revisit next month, when your circumstances or the science change. A surgical complication does not offer that. A stricture, a bleeding ulcer at the surgical connection, chronic malabsorption, the reactive hypoglycemia that can follow a bypass, the new alcohol use disorder that a reshaped gut can unmask — those do not reverse when you change your mind.

We tell patients a drug is a life sentence because you have to keep taking it. We almost never tell them that a complication is a life sentence because you can never stop having had it.

I say this as a bariatric surgeon who believes deeply in the operation. As one who has personally operated on thousands of people. Saying both halves of the truth does not weaken the case for surgery; it is the only thing that makes the comparison honest. There is one knot, and one decisive stroke that cuts it, and that stroke now comes in more than one form. The operation and the molecule are not rivals fighting over patients. They are two blades of the same sword, and the honest clinician reaches for whichever one a particular human being can actually live with — and often it is both.

The same honesty applies to every item in that table. Histotripsy, TAVR, immunotherapy, direct-acting antivirals — each carries its own ledger of risk, cost, and durability. The point is not that the new stroke is always better. The point is that a decisive, less-invasive option now exists where it did not before, and patients deserve to weigh it with both columns visible.

The barrier was never the biology

If disease after disease now has a decisive answer, then the real scandal is no longer scientific. It is architectural.

We built a health system that pays, in perpetuity, for the consequences of disease — the insulin, the dialysis, the stents, the repeat admissions, the long downstream bill of conditions we let advance. What it resists paying for is the treatment that prevents all of it. GLP-1 coverage is a patchwork of cliffs and prior-authorization mazes. Surgery sits behind network exclusions and long stretches of the country with no program at all. The newest precise therapies are approved faster than they are covered.

That architecture made a grim kind of sense when the decisive options didn’t exist. They exist now. Keeping the old payment structure in place is a choice — and, as the obesity data are beginning to show at population scale, an expensive one.

This is the core of what I mean by Health 4.0: a system organized around durable, mechanism-level treatment, delivered by whichever precise stroke fits the patient, and paid for as the chronic-disease care it actually is — continuously, equitably, and without forcing patients to prove their worthiness at every renewal. Increasingly pulling upstream from rescue.

The series ahead

This essay is the opening argument. Over the coming weeks I am going to take the knot apart strand by strand. Each piece will pair the surgical evidence with the new pharmacologic and technological evidence, honestly, complications and all:

- Diabetes — the disease that first proved an operation could outperform a lifetime of pills.

- Hypertension and dyslipidemia — the quiet comorbidities that resolve when you cut the knot.

- Fatty liver disease (MASH) — where drugs, surgery, and even focused-energy tools are converging.

- PCOS — metabolic disease wearing a reproductive mask.

- Obesity-associated cancer — the strand we least like to talk about.

Same cadence, same discipline: the data, the mechanism, the honest ledger, and the policy fix.

Alexander understood something the field forgot for a hundred years. Some knots are not meant to be untied. They are meant to be cut. Across medicine, we have finally built the swords.

The only question left is whether we will build a system with the nerve to use them.

Subscribe to follow the series. Another version is on linkedIn newsletter: https://www.linkedin.com/newsletters/architecting-future-health-7191851130182672385

Original article is available on my website: robinblackstone.com

If you’re seeing this shift in your own corner of medicine, tell me the example I’m missing — I’m collecting them.

There is a version of American health care running on your PlayStation or internet right now. Several versions, actually. I’ve been playing some of them for years, and it took me an embarrassingly long time to notice that the argument I’ve spent a career making in clinics and boardrooms was sitting in my hands all along, rendered in polygons and frame rates and the specific 2 a.m. despair of a boss you cannot beat by trying harder.

I want to walk you through three games. If you’ve ever held a controller, you already understand what I’m about to say. You earned it. Nobody explained the lesson to you and nobody had to.

The world we have

Start with the one everyone knows. The loot game, Diablo. Kill things, they drop gear, the gear lets you kill bigger things faster, and when the incoming damage overwhelms you, the answer is never why is this much damage coming in. The answer is a health potion on cooldown and a better weapon. There is no prevention role. Nobody in the party is watching the trajectory. There is only the next pack of monsters, the next drop, the next season that wipes the board clean and asks you to do it again.

You can out-DPS anything if your numbers are high enough. Continuity is a fiction. The meter counts one thing: damage done.

I don’t need to draw the map. That is the health system we have. We reward the dramatic downstream intervention — the rescue, the procedure, the admission — and we pay for volume, and when chronic disease overwhelms the population we reach for more force, more intervention, more powerful gear. We built a health system that plays like a loot grinder and then we act surprised that we cannot out-DPS diabetes at scale.

The tell is that there is no role, anywhere in the design, for the person who keeps the damage from landing.

The world we could build

Now the game I mained for years. World of Warcraft. I played a Priest. More specifically, I healed — which means I learned early that survival is not about power. It is about timing, attention, and knowing the fight before the damage lands.

WoW looked like fantasy, but it behaved like American life. It was a working model of a society: people sorted into roles, resources rationed under pressure, success measured on public meters, failure blamed on the person standing closest to the damage — all while the deeper rules of the world went mostly unquestioned.

World of Warcraft even staged, by accident, one of the strangest public-health parables of the internet age: the Corrupted Blood incident. Some players tried to heal. Some warned others away. Some hid. Some spread the contagion on purpose. Designers did not set out to build a pandemic laboratory, but the world became one anyway.American health care is not a puzzle game. It is an MMO — a massive multiplayer system where every player’s choices are constrained by the rules of the world.

And inside the healer role is a quiet moral choice. You can play reactive: let the boss hit, let the raid bleed, then cast the big, beautiful, expensive heal that lands just in time. The meter lights up. The save is visible. The healer gets the credit. Or you can play prevention: shield before the pull, read the damage before it comes, spend resources early so the crisis never blooms. Prevent the wound instead of performing the rescue. In all those years I never picked the prevent option. After all, as a surgeon I was rescue incarnate.

And prevention has a public relations problem. When you do it perfectly, nothing happens. The bar does not crash. The siren does not sound. The family does not panic. The hospital bed is never filled. The death certificate is never written. No one applauds the healer for the catastrophe that never arrived.

Rescue has a scoreboard. Prevention has silence.

Here is what that game taught millions of people, years before anyone handed them the vocabulary: you cannot out-heal bad mechanics. If the tank stands in the dragon’s fire, no amount of downstream healing saves the raid. The damage is generated by the environment and the behavior. You have to change the fight, not just treat its casualties.

That is the entire argument. Prevention is a real role. It is more efficient. And our instruments cannot see it, so we do not pay for it.

The difference between WoW and the loot grinder is not better potions. It is that prevention exists as a role at all, and that the combat log makes it legible. The MMO solved a measurement problem healthcare hasn’t: it built the instrument that lets a shield show up on the record, so the person preventing the crisis is finally visible. Give me the log that proves the trajectory bent, and the question why don’t we pay for shields answers itself.

The world that is true

And then there is the third game. The strange one. The one that is haunting.

Elden Ring.

I started playing it a year or so ago, and I tried to play it the way I play everything — like a healer, reacting, topping off, managing resources against a legible threat. It refused me. It punished the reflex. And months later I am still turning it over, which is how you know a thing taught you something it never said out loud.

Three features of that game are the truest things I know about health.

The constraint is real in every move. Not the fake constraint of a cooldown that refills. The permanent kind. Every flask you drink is gone. Every choice forecloses another. The resource runs out, and the game never once pretends otherwise. You cannot rescue your way through a fight you are playing wrong — you have to change how you play, or you keep dying. American healthcare runs on the opposite fiction: infinite flasks, rescue always in stock, the cost hidden until the resource is empty and someone dies off-screen. This game puts the cost in your hand every second. It is honest in a way our system has never been.

The roles are not labeled. This is the haunting part. The one I thought about after dying yet again. In the raid, the healer is a class — assigned, legible, queued for. Here, help arrives as the summon sign of a stranger you will never speak to, a companion whose allegiance you cannot read, a spirit that fights beside you and then is gone. And harm arrives the same way — the kindly figure who betrays, the thing in the fog that turns out to be either the ally you needed or the enemy that harms you. You cannot tell in advance. Support and threat are not pre-marked. You have to be paying attention to know which is which.

It creates uncertainty deep in your subconscious. You are off balance.

That is not bad design. That is the actual condition of being a patient in a health system where the help you need arrives from unexpected quarters and vanishes uncredited because no log records it.

The feedback is merciless and it does not lie. When you fail, the screen tells you plainly. It does not hide the trajectory or launder the failure into a quality metric. You died. Now learn the tell — the wind that comes up before the blow, the pattern you could have read if you had been watching the fight instead of the health bar. Survival is not throughput. It is earliness born of attention. The players who win are not the ones who react faster. They are the ones who saw it coming and prepared to have a different outcome. They may not be right every time, but they live longer.

That is prevention arrived at through a harder truth than the raid ever taught. Because it doesn’t offer, up front, the choice to prevent. You are in the game, as you are. The results of the culture you inherited, the biology of your character, the environment. Not “intervene sooner” as a tactic. Something closer to a discipline: you attend to the fight, you learn its patterns, you spend your finite resources on not taking the hit, because you cannot afford to take the hit. If you do, you die.

Where the machine goes

Now — the part the gaming world will feel before anyone else, because you have already lived the distinction and found it good.

In every one of these worlds there is an intelligence that sees more than you can. The boss mod that calls the incoming mechanic three seconds early. The overlay that reads the raid-wide state no single human can keep in working memory. The accumulated knowledge of ten thousand prior deaths — every player who fell to this boss before you — distilled into the sense that lets you finally read the tell.

That intelligence is not the healer. It never touches the fight. You still cast. You still dodge. You still make the call and own it. What it does is expand what you can perceive and how early you can act.

The intelligence is sight.

Not hands.

This is the thing I most want the people building the future to understand, because it dissolves the fear that has made the whole AI conversation stupid. Nobody who has cleared a hard raid believes the boss mod replaced them. The callout did not take the win. It meant nobody died to a mechanic they could not see coming. The human is still the one who acts — more capable, seeing further, because the machine held the pattern that no single mind can hold.

That is the role for AI in medicine. Not the same role as the physician/human health ally. AI is the sight.

Doctor AI is the accumulated knowledge of every prior death, surfacing the tell so a human clinician can act before the damage lands — reading the population’s and the individuals trajectory, calling the wind-up, making the human health ally a better healer without ever being the healer. Accounting for the context of the individual players uniqueness.

And here is why the suspicion about AI is so backwards. The machine draws more scrutiny than human intuition not because it is more dangerous but because it is more inspectable. You can read its callout in the log. You can audit the tell it flagged. You can inspect the health trajectory it plotted and judge the outcomes, see where the deviations and “life happens” actually impact the curve of health for any individual. Contrast that to the human hunch that has run medicine for a century. It leaves no such trace. The fragmentation of records, the misrecording of diagnosis and conversation, the hoarding of data that doesn’t belong to them.

We are training our suspicion on the one participant we can actually see.

The GPUs that render these worlds, the same compute now sitting in hospital basements and research clusters, the accelerated infrastructure the gamer world knows by name — are not building a replacement clinician. Why would they want to? They are building the boss mod for chronic disease. The overlay that sees the population-wide state. The sight that lets a finite, exhausted, deeply human healer act three seconds earlier, when three seconds earlier still means finite flasks, unlabeled allies, and a fight you cannot rescue your way out of.

Disease is a formidable boss.

The turn

So here is the whole argument, and many reading this — the gamers and the developers — already own it.

We built American healthcare like a loot grinder. Drink on cooldown, swing harder, no role for prevention, a meter that counts only the damage we do. We are surprised we cannot out-DPS a chronic disease epidemic.

We could build it like a raid, where prevention is a real role and the log makes it legible, so the shield finally shows up on the record, and we pay for the wound that never opened. Many incumbent’s in the current health game, advocate for this model. They get to keep grinding as they extract their tolls.

But the truest game is the third one, and it is the one we are actually playing whether we admit it or not. Finite resources in every move. Allies and adversaries that arrive unlabeled, health and harm you have to be present enough to tell apart. A feedback loop that will not lie to you about whether you are winning. And a fight you cannot brute force, where the only path is to see sooner, spend wisely, and change how you play the long game.

People keep dying to the same boss.

Not because we lack force. Because we refuse to learn the tell.

The flask runs out.

Learn the tell.

Change the game.

This essay is part of the H4 Alliance’s work to author a new architecture for American health, one designed to bend trajectories rather than reward rescue. If it resonated, the conversation continues at the H4 Alliance — website coming soon. Author — robinblackstone.com

The Republic Has a Pulse. It’s Time for a National Health Exam.

America turns 250 this week. On every other measure we celebrate the milestone — so why won’t we check the vital signs of the people inside it?

By Robin P. Blackstone, MD

On July 4, the United States turns 250. We will light the sky, read the old words aloud, and tell ourselves the story of survival. And we should. But I want to make a different case this week, one I’ve spent a career arriving at as a physician: a birthday is not a diagnosis. At 250, America is not simply due for a celebration. It is due for a national health exam.

Let me tell you what I mean — and why I think this is the decade we either build a health system worthy of this country, or admit we chose not to.

Old is not the achievement

The number itself is not the miracle. Other civilizations are older. States have survived for centuries by centralizing power, silencing dissent, controlling information, and subordinating the individual to the state. A country can be old and unfree. A government can last without listening. A system can persist long after it has stopped serving the human beings inside it.

The American achievement is different. A large, diverse, constitutional republic has survived repeated crises while continuing to argue over freedom, citizenship, equality, power, and the meaning of “We the People.” Its survival was never guaranteed. Its progress has never been linear. Its promises have never been fully delivered. And yet — across revolution, slavery, civil war, depression, migration, industrial upheaval, scientific discovery, technological transformation, and cultural fracture — the republic has continued to revise itself.

That is not perfection. It is vitality.

Like a human body, a nation’s health is measured not only by what it has survived, but by the direction of its trajectory: whether it can recognize disease, repair injury, adapt under stress, renew the systems that sustain life, and make a freer, healthier future possible.

In health, the diagnosis is clear: we built one of the most powerful medical rescue systems in human history. Now we must build a health system.

What we built, and what we didn’t

Across two and a half centuries, medicine moved from holding a person’s hand while they suffered to changing the course of suffering itself. We learned to see inside the body, prevent diseases that once terrified families, repair trauma, transplant organs, relieve pain, change the course of heart disease and stroke, treat infections, replace joints, sustain premature infants, and offer hope in the face of cancers and conditions that once felt beyond reach.

America did not accomplish this alone. No nation did. Discovery has always crossed borders. Science has depended on inheritance, migration, collaboration, courage, sacrifice, and the willingness to be corrected by what reality reveals.

And America is not diverse by accident. It is diversity itself: First Peoples whose nations long predate the Republic; descendants of enslaved Africans brought here by force; immigrants and refugees who came by hope, hunger, exile, ambition, or necessity; and generations born into the unfinished argument over what “We the People” must mean. However ancient the first human migrations to this continent, the moral and political truth is clear: Indigenous nations were here first. They were not immigrants to the United States. The United States arrived into a world they had already made.

This human complexity is one of the country’s deepest sources of intelligence. Diversity did not slow discovery; it gave it velocity. Different bodies, lineages, languages, diseases, environments, injuries, exposures, diets, beliefs, cultures, and ways of seeing helped medicine move faster than any closed system could. Collaboration did not make science simpler. It made science more honest.

That achievement is real. It deserves recognition. It deserves gratitude. But gratitude should not become self-congratulation.

A nation can honor what it has built and still tell the truth about what it has failed to protect. We have generated healing at extraordinary scale, even as too many people have suffered inside systems organized around power, profit, neglect, exclusion, fragmentation, and delay.

We have produced miracles, but not fairness. We have built institutions of rescue, but not an architecture of health. We have extended life, but not always protected the conditions that make life livable.

At 250, the question is: can America become worthy of its best work?

Rescue is not the same as health

A rescue system responds when crisis arrives. A health system sees people before crisis defines them. A rescue system treats the episode. A health system understands the trajectory.

But a human trajectory does not unfold in abstraction. It unfolds inside culture.

Culture is not decoration. It is not etiquette. It is not a translation service added at the end of care. Culture is how people make meaning of illness, suffering, birth, aging, disability, finances, healing, food, pain, family, authority, death, and hope. It shapes whether a person trusts a clinician, whether a mother feels safe in childbirth, whether a veteran speaks of trauma, whether a child’s symptoms are believed, whether mental illness carries shame, whether prevention feels empowering or intrusive, whether technology feels like help or surveillance.

A health system that ignores culture does not become neutral. It defaults to the culture of the institution. It mistakes difference for noncompliance, silence for agreement, mistrust for irrationality, and efficiency for care. That is how much of American health care still operates: institutional culture mistaken for universal care.

In a republic this diverse, culture is not a side concern. It is national health architecture.

A rescue system asks what procedure, code, bed, claim, authorization, or discharge plan is needed now. A health system asks what kind of life the person is trying to live, what risks are accumulating, what support is missing, what trust has been broken, and what would make health more possible over time. A humane health system asks one more question: what culture gives this person’s life meaning, and how can care honor that meaning without turning it into a stereotype or a cage?

Rescue is sacred work. A society that cannot rescue the person in crisis has failed at the most basic level of care. But a society that can rescue and still refuses to prevent avoidable arrival at the cliff has failed at a deeper one.

We know this in our own lives. A body does not become healthy because the emergency room is excellent. Health depends on earlier signals, trusted relationships, memory, prevention, repair, adaptation, belonging, and the daily conditions that make resilience possible.

So does a republic.

These are vital signs, not side issues

The health of a nation is not separate from the health of its people. It lives in the mother entering childbirth, the child breathing polluted air, the veteran trying to come home whole, the elder aging alone, the worker delaying care because the bill may break the family, the clinician answering one more message after midnight, the rural hospital on the edge of closure, the community carrying grief across generations, and the patient whose story is scattered across systems that do not remember them.

These signals are vital signs. The vital signs are not good.

Health 4.0: why AI makes the exam impossible to avoid

Artificial intelligence now makes the national health exam impossible to postpone. We have tools that can see patterns across time, summarize complexity, support clinicians, translate information, identify risk earlier, and help make continuity real. Used wisely, AI can become an instrument of memory, coordination, prevention, and accountability. Used poorly, it becomes another layer of surveillance, denial, bias, opacity, and institutional control.

This is the choice at the center of what I call Health 4.0 — the shift from a century of rescue toward a system designed for human flourishing, with intelligent tools finally serving the person rather than the institution.

Culture matters more in the age of AI, not less. Culture cannot be reduced to a demographic field, inferred by a machine, or assigned as a label. It must be carried by the person, interpreted through relationship, protected by consent, and used only to make care more humane. AI may help translate language, reduce burden, and surface missed patterns. But it must never turn culture into a stereotype or use prediction to narrow a person’s future.

The question is not whether AI will shape health. It will. The question is whether we will govern the institutions, incentives, data flows, payment structures, and technologies that determine whether AI serves human beings — or systems alone.

A larger vision of freedom

The goal of the American health project in the next century should be clear: to ensure each person holds greater authority over their own future. That is not a smaller vision of health care. It is a larger vision of freedom. Freedom is strengthened when people have the capacity to choose health: to live, work, love, recover, care, create, and build.

At 250, America needs a structure capable of keeping its promise. A health system worthy of this republic would make trust a design requirement, not a slogan. It would make failure visible. It would make institutions answerable. It would honor clinicians as human allies, not endless shock absorbers for broken systems. And it would treat people not as billing events, risk scores, data sources, or passive recipients of instructions, but as human principals: the rightful holders of their own lives, stories, choices, cultures, and futures.

The first vital sign of a republic is whether it can tell the truth about itself. In health, the truth is clear: America has built miracles of rescue, but not yet a system worthy of human flourishing.

That is the work now.

The republic has a pulse. We must listen.

I’m Robin Blackstone, a physician who has spent a career inside America’s rescue system — and who believes we can build something better. This essay is part of the argument I make in my forthcoming book, Doctor AI Reimagining Healthcare Rebuilding Trust Delivering Health 4.0, published April 7, 2026. I make the case, this country still owes its people a health system. Transforming what feels politically impossible today into what becomes obviously necessary tomorrow.

If that’s a conversation you want to be part of, subscribe. Share this with one person who’s tired of a system that only shows up at the cliff’s edge. And tell me in the comments: what would a real health system have caught in time for you or someone you love?

— robinblackstone.com

In December 2024, a CEO was shot in midtown Manhattan, and the response told us something we already knew: Americans no longer trust the for-profit infrastructure of their healthcare.

In the year and a half since, the FTC has filed a formal complaint against the three largest Pharmacy Benefit Managers — Caremark, Express Scripts, and OptumRx, which together administer roughly 80% of US prescriptions — over their drug-pricing practices. UnitedHealth Group has faced antitrust scrutiny and federal civil and criminal investigations into its Medicare Advantage billing, alongside a class-action wave that hasn’t slowed. “Private equity in healthcare” has become a slur in the Wall Street Journal, not just in left-leaning policy magazines.

This is the political environment my book, Doctor AI Reimagining Healthcare Rebuilding Trust Delivering Health 4.0, arrived in. And it produced a reasonable question from a reviewer at Kirkus, who suggested that my proposal for a future system built on artificial intelligence and physician partnership in the private corporate world might be “ideological laundering” for the same free-market machinery that produced UnitedHealth and the PBMs in the first place.

I don’t think the reviewer is unreasonable to ask. I think the question deserves a precise answer. So here it is.

The question behind the question

The skeptical reader’s worry runs like this: every era of American healthcare reform — private or governmental — has promised that this time the combination of markets and policy will finally deliver care, dignity, and access. HMOs were sold that way. Managed care was sold that way. The ACA was sold that way. Vertically integrated conglomerates were sold that way. Each delivered less than it promised, often while extracting more than it should. And now I arrive with a new book, a new framework called the Triangle of Trust, and a new entity called the H4 Alliance.

How exactly is this different?

That is a fair challenge. The answer turns on a piece of legal architecture most readers — even most physicians, even most policymakers — do not know in detail: the difference between an ordinary corporation, a Public Benefit Corporation, and a Trust. The H4 Alliance is structured as a combination of the second and third. The combination is not decorative. It is the entire point.

Why wouldn’t the government be the best fiduciary?

A reader who agrees that the for-profit corporate form has failed American healthcare might ask the next question: if shareholders are the wrong fiduciary, why not relocate the obligation to a government that owes a fiduciary duty to the public? Why build a PBC paired with a Trust when a federal program is already a public-fiduciary arrangement?

Fair question. The Triangle of Trust is structurally compatible with most forms of government healthcare. The H4 Alliance is not an argument against government healthcare. It is an argument that the architecture we need cannot be built by waiting for government to build it, and cannot best be held by government once it is built. Several reasons:

Timing. The AI revolution in medicine is happening now, on a timeline no federal legislative process can match. Federal cycles — appropriations, rulemaking, regulatory review, agency budgeting — run on year-scale calendars. The AI capability that will define the next decade of clinical practice changes on month-scale calendars. By the time a federal AI-in-medicine framework is drafted, debated, passed, regulated, and operationalized, the technology will be three generations ahead of what the framework was designed to govern. Waiting for the right federal program is not a strategy. It is a delay.

Track record. Medicare, Medicaid, the VA, the Indian Health Service, and the federally qualified health centers are themselves under serious strain — fee-schedule pressure, prior-authorization games, fraud and abuse, chronic underfunding, uneven quality. These are not arguments against government healthcare in principle. They are observations that adding more healthcare to government delivery does not automatically solve the structural problems that already plague it. A new federal program would inherit those constraints, not escape them.

Regulatory fragmentation and capture. Federal healthcare frameworks must navigate 50 distinct state regulatory environments — different licensing rules, scope-of-practice doctrines, insurance regulations, and corporate-practice-of-medicine standards. What applies in one state often cannot apply in another, which limits the portability and scope of any federal program. And when federal frameworks do arrive, they are typically shaped by the regulated incumbents — insurers, hospital systems, PBMs — through the slow accretion of comment cycles, lobbying, and regulatory capture, until the framework constrains the market to conform to what incumbents need to survive.

The H4 Alliance, operating at the level of corporate form and platform architecture rather than federal legislation, is subject to neither constraint. The PBC is incorporated under a single state’s law — typically Delaware — and operates in all 50 states through standard foreign-qualification processes that don’t require the host state to have its own PBC statute. The Trust, holding 501(c)(3) federal tax-exempt status, transcends state lines as a federal designation. State healthcare regulations still apply to the Alliance’s operations the way they apply to any multi-state healthcare entity — but to its activities, not to the public-benefit fiduciary structure itself, whose mission is locked at incorporation, not negotiated in comment periods.

International track record. Government-stewarded systems abroad — the UK’s NHS, Canada’s Medicare, France’s Sécurité Sociale — outperform the United States on most population-level metrics, including life expectancy, infant mortality, and per-capita cost. But that is not the same as saying they are the better steward. The NHS is operating under unprecedented strain: A&E (emergency) waits routinely above twelve hours, resident doctors (the grade England renamed from “junior doctors” in 2024) in recurring strikes, GP access collapsing across rural England, and an elective backlog still around seven million. Canadian Medicare faces a chronic primary-care shortage — roughly six million Canadians without a family doctor — and Quebec is drifting toward a two-tier model that strains the system’s equity principles. France faces accelerating physician burnout, déserts médicaux, and rising public dissatisfaction. None of these systems is a model of effortless public-fiduciary success. Each carries its own structural failures — different from the US, but failures all the same. The H4 Alliance is designed to address both sets, not to trade one for the other.

Political volatility. A government program is subject to political change every election cycle: funding shifts, coverage rules shift, priorities shift. A Health Principal relying on the federal system relies on contours that can be redrawn by the next administration — an insufficient floor of governance. The PBC + Trust form is not subject to that volatility. The Trust’s fiduciary obligation to its beneficiaries does not change with elections. The PBC’s stated public benefit is locked in its incorporating documents.

Trust itself. A significant portion of the American public does not trust the federal government to deliver healthcare. That distrust is not symmetric across the political spectrum, but it is real and widespread. The Triangle of Trust is, by definition, about rebuilding trust. Rebuilding it by relocating delivery from one untrusted institution — for-profit corporate healthcare — to another untrusted institution — federal bureaucracy — is not a successful trust strategy. The H4 Alliance is, by design, a third option: a credentialed, mission-bound, fiduciary-constrained alternative to both the corporate and the governmental defaults.

Precedes, not precludes. The H4 Alliance does not prevent government action. If the United States eventually moves toward single-payer or another form of expanded government healthcare, the Triangle of Trust can operate within it. The Health Principal, the Human Health Ally, and Doctor AI do not depend on a particular payment model. They depend on the relational and operational architecture of how care is designed and delivered. That architecture has to be built. It will not be built by federal legislation. It will be built by a focused operating entity demonstrating that the architecture works, at sufficient scale to make the argument irrefutable. The H4 Alliance is that entity.

Government already plays a role in American healthcare, and reasonable people disagree about how much more it should. The question is who builds the architecture of trust, intelligence, and partnership that any future system — government, private, or hybrid — will need in order to deliver. That cannot be deferred to a legislative process. It has to be built now, by an entity structured to build it.

What an ordinary corporation is

Under standard American corporate law, the directors of a for-profit corporation owe their fiduciary duty to shareholders. This is the doctrine underlying almost every familiar pathology of American healthcare. When a private-equity firm acquires a hospital chain and strips its assets, the directors are not betraying their duty — they are fulfilling it, because their duty is to shareholders, and shareholders want returns. When a PBM negotiates rebates that pad its margin while pushing more expensive drugs onto patients, the executives are not deviating from their mandate — they are executing it. When an insurer’s algorithms deny claims in patterns that correlate with profitability targets, the system is not broken. It is working as designed.

Any honest argument for healthcare reform has to begin by acknowledging that the corporate form most American healthcare entities operate under has legal obligations that point in only one direction: the shareholder.

This is the dynamic that produced the past decade’s wave of private-equity catastrophes. KKR’s Envision Healthcare ended in 2023 bankruptcy after becoming synonymous with surprise billing and emergency-physician staffing extraction. Cerberus Capital’s Steward Health Care ended in 2024 bankruptcy and the closure of community hospitals across multiple states. Joel Freedman bought Hahnemann University Hospital in 2018 and closed it less than two years later, displacing roughly 2,500 employees, orphaning more than 570 medical residents and fellows — the largest such displacement in US history — and ending a safety-net teaching program at the heart of Philadelphia. None of these were failures of strategy. They were successes of fiduciary obligation — directors fulfilling their duty to shareholders, even when fulfilling it destroyed institutions.

A clarifying note on labels. KKR is a publicly traded private-equity firm; Envision under KKR was a private corporation. Steward under Cerberus was a privately held for-profit corporation. Hahnemann under Freedman was a privately held corporation. The labels vary — “publicly traded,” “private,” “private-equity-owned” — but the legal form is the same: a for-profit corporation, directors owing fiduciary duty to shareholders. Private-equity ownership does not change that obligation. It intensifies it. PE funds have explicit return targets and five-to-seven-year exit horizons, so the pressure to extract is sharper, not weaker, than under public-company ownership. The corporate form is the constant; the extraction pressure varies with who owns the shares and how soon they want their money back.

What a Public Benefit Corporation is

A Public Benefit Corporation — sometimes called a Benefit Corporation, depending on the state — is a distinct legal form, recognized in more than 35 US states, in which directors are required to consider, alongside shareholder returns, a stated public benefit. The corporation must specify that benefit at incorporation. Directors are legally obligated to balance it against profit. Shareholders can sue if directors fail to weigh it. And the corporation must report publicly on its performance against the stated benefit, typically annually.

Allbirds is a PBC. So is Kickstarter. So is Patagonia — and Patagonia is worth pausing on, because in 2022 it went a step further than the form usually goes. Its benefit corporation is now held by a purpose trust: the Patagonia Purpose Trust holds the voting shares and guards the mission, while the economic interest sits with a mission-bound nonprofit, the Holdfast Collective. That is not a distant analogy to what I’m describing. It is nearly the same architecture — a benefit corporation that does the work, bound by a trust that guards the mission. One of the most admired companies in America reached, independently, for the same structure I am proposing for American healthcare.

A PBC is not a nonprofit. It can take investment, distribute returns, and operate commercially. What it cannot do is pretend profit is the only thing that matters. The corporate form itself constrains the optimization.

A third form that already exists

The structure I’m describing is not novel, and it is not utopian. It is three proven forms, assembled around health for the first time.

The first is the pairing at its core: a mission-locked trust holding a company that does the work. This is how some of the most durable, trusted enterprises on earth are owned. The Novo Nordisk Foundation controls Novo Nordisk through a dual-share structure — roughly a quarter of the equity but about three-quarters of the voting power — precisely so the company cannot be sold off or steered away from its purpose. Bosch, IKEA, and Tata are owned the same way, by foundations that own themselves and answer to a mission rather than to shareholders. Patagonia, as we have seen, reached for the same architecture in 2022. The foundation or trust holds the mission; the company does the work; neither can be stripped for parts.

The second is health stewardship that sits outside both market and state — and it is already working. Germany has run its health system this way for more than 140 years, since Bismarck’s 1883 law, through self-governing funds that are neither government agencies nor private companies. Israel covers every resident through non-profit funds that are forbidden by law from turning anyone away. And this is not only a foreign idea: in the United States, HealthPartners — the largest consumer-governed nonprofit health system in the country — has been run since 1957 by a board its own members and patients elect, delivering care and coverage to well over a million people across Minnesota and Wisconsin. HealthPartners proves that consumer-governed health delivery works in America, at scale. These are not thought experiments. They are functioning systems, governing the health of millions, on exactly the principle the H4 Alliance is built on: neither extraction nor the state.

The third is a perpetual endowment that no administration can raid — the durability and independence that lets the mission outlast any single government, market cycle, or leader.

America has built independent, mission-bound institutions before. The H4 Alliance is not asking the country to imagine something it has never seen. It is taking three forms that already work — a trust that holds the mission, a corporation that does the work, and an endowment no administration can raid — and assembling them, for the first time, around American health.

What a Trust is

A Trust is older and stricter. Trusts originated in English common law and have been refined for centuries to do one specific thing: bind an entity to a stated purpose so its assets and operations can only serve that purpose. The trustees owe fiduciary duty not to investors or shareholders but to the beneficiaries — the people or causes the trust was established to serve.

Trust law is, for our purposes, the most restrictive form of fiduciary obligation in American jurisprudence. A trustee who diverts assets from the trust’s stated purpose can be removed, sued, or in extreme cases criminally charged.

Trust law is also unusually uniform across states — descending from a shared common-law tradition rather than state-by-state statutory innovation. For the H4 Alliance Trust, that uniformity is a strategic advantage to the public: every Health Principal in every state is held by the same fiduciary obligation, under the same body of law, regardless of where they live or how their state regulates healthcare. A Health Principal in Alabama, Texas, Wyoming, or California is held to the same obligation by the same Trust. The protection the Alliance offers does not vary by ZIP code.

What about medical societies?

A reader might reasonably observe that the United States already has a different institutional form: the medical society. The American Medical Association is a 501(c)(6) trade association. The American College of Surgeons is a 501(c)(3) educational and scientific nonprofit. The American College of Physicians, the American Society for Metabolic and Bariatric Surgery, the American Hospital Association, the American Nurses Association, the American Academy of Pediatrics — all member-governed nonprofits, most under one of the two tax-exempt forms recognized for professional and scientific work. They have no shareholders. They are governed by boards elected by members. Their fiduciary obligations run to their missions, not to financial returns.

These organizations have done — and continue to do — important work: setting clinical standards, accrediting training, publishing journals, running continuing education, convening the profession. The Triangle of Trust is in conversation with that tradition.

But medical societies are not operating entities. They do not deliver care, run platforms, or employ clinicians at scale. They are convening, credentialing, and advocacy bodies, correctly designed for those functions. Asking the AMA to operate the architecture of American healthcare delivery would be asking it to do something it was not built for. The H4 Alliance is built for the operational work medical societies cannot do, while remaining accountable to the same public interest the best of them have always represented.

The H4 Alliance Trust is itself a 501(c)(3) — the same form the American College of Surgeons and most major scientific and educational medical organizations use. The difference is what sits beneath it. The H4 Alliance pairs the Trust with a Public Benefit Corporation that can do the operational work — build and run platforms, employ technologists and clinicians, generate revenue, scale — that pure nonprofits cannot. It is, in a real sense, what you would get if you took the mission focus of a medical society, combined it with the operational capability of a corporation, and bound the result with a Trust to prevent drift toward UnitedHealth-style extraction.

What the H4 Alliance is

The H4 Alliance is a Public Benefit Corporation paired with a Trust.

The PBC is the operating entity. It can do commercial work — develop the Health 4.0 platform, license intellectual property, partner with hospitals and health systems, employ physicians and technologists, generate revenue. Its public benefit, written into its incorporating documents, is the architecture of the Triangle of Trust: a system in which the Health Principal, the Human Health Ally, and intelligent tools are aligned in cooperation rather than extraction.

The Trust sits above the PBC. Its beneficiaries are the people the Triangle of Trust is built to serve — Health Principals across the United States, present and future. The Trust holds strategic stewardship of the mission and constrains the PBC from drifting away from its stated public benefit over time. Trustees have a fiduciary duty to the Triangle, not to investors and not to the Alliance’s executives.

This is a deliberately constrained structure. It does not allow the H4 Alliance to behave like UnitedHealth even if every executive wished it would. The legal architecture forbids it.

A crucial aspect is the integrity of the people who serve on each of these nine-person boards and in leadership (CEO, Chair). Every officer and director must disclose their relationships and ensure there is no influence from the healthcare lobby, which is extensive. In 2024, the health sector spent roughly $744 million on federal lobbying — the only sector to top $700 million — with its PACs contributing tens of millions more directly to congressional campaigns. A structure that does not screen out that influence at the board level is a structure that will eventually be captured by it.

Two further commitments make the accountability concrete. The people govern: a guaranteed share of every board belongs to the people the system serves — candidates nominated by them, screened against published conflict rules the Trust cannot weaken, then drawn by lot from everyone who qualifies. It is a gate, but no gatekeeper to capture. And the people own it: every person holds a non-tradeable share — like the community ownership of the Green Bay Packers — that pays nothing and cannot be sold, so the system can never be carried off. This is the safeguard ordinary member-owned cooperatives lack. Because the share cannot be sold and the Trust’s mission cannot be voted away, no member majority can ever demutualize the system or cash it out — the failure mode that has hollowed out mutual insurers and health co-ops before. The people govern it and the people own it, but neither members nor capital can ever sell it. Capital can fund it. Capital can never own it.

The entities: H4 Alliance, Inc. (PBC) and the H4 Alliance Trust (501(c)(3) under §170(b)(1)(A)(vi), EIN 39-4429422, issued September 17, 2025).

What this means in practice

A few concrete things follow from this structure that distinguish the H4 Alliance from the entities the Kirkus reviewer was worried about.

First, the H4 Alliance cannot be acquired by a private-equity firm or a conventional corporation and stripped for parts. The PBC form has anti-acquisition provisions, and the Trust’s fiduciary obligations to its beneficiaries would block any acquisition that removed the public-benefit constraint. The structure is, in a real legal sense, acquisition-resistant.

Second, when the PBC makes decisions — pricing, product design, partnerships, hiring — the directors are legally required to weigh the impact on Health Principals against the profit consequence. This is enforceable. Shareholders who object can sue. Trustees who object can intervene. The optimization function is not single-variable.

Third, the Alliance must report publicly on its performance against the stated public benefit. This is not a marketing claim; it is a regulatory requirement of the corporate form. Annual reports must address whether the Triangle of Trust is being served, and how.

Fourth — most relevant to the Kirkus critique — the H4 Alliance does not “rely on free-market capitalism to solve the problems capitalism created.” It relies on a constrained form of commercial enterprise, paired with a fiduciary stewardship layer, to do work that pure nonprofits cannot scale and pure for-profits cannot align. It is, deliberately, neither pure market nor pure mission. It is the legal form created for exactly the kind of work this moment requires.

The honest limits of the argument

I want to acknowledge something the Kirkus reviewer was right about, even if I think the conclusion missed.

Structure does not guarantee outcome. A PBC’s directors can still make bad choices. A Trust can still be poorly stewarded. The architecture I’ve described constrains the direction of decisions, not their wisdom. The H4 Alliance can fail. The Triangle of Trust can be poorly executed. Doctor AI can be deployed in ways that re-create some of the problems the framework was built to solve.

What the structure does is make those failures legally recoverable. Beneficiaries can sue. Trustees can intervene. The state can examine. Failure modes invisible in standard corporate form are visible and actionable in PBC + Trust form. That is the protection the structure offers — not a promise that nothing will go wrong, but that when things go wrong, the apparatus to correct them exists.

This is, to my mind, and after a long career in surgery, hospital administration, and corporate medicine, the most honest stance an American healthcare reformer can take. Any reformer who claims their entity will never drift, never extract, never disappoint, is lying. The question is whether the legal form lets you catch it when it does. The H4 Alliance is designed so the answer is yes.

Back to the reader

If you have made it this far, you now understand a piece of legal architecture that perhaps 95% of healthcare commentators — most journalists, most reviewers, most readers — do not. You can now distinguish the H4 Alliance from UnitedHealth, from a private-equity rollup, from a conventional health-tech startup, and from a nonprofit advocacy group. The differences are not rhetorical. They are statutory.

Trust in America’s public institutions, healthcare organizations, and government is at an all-time low — for understandable reasons. The system is largely irrelevant to the millions without coverage, a source of severe financial hardship for the many it leaves in debt, and an exhausting machine for a physician, nurse-practitioner, and physician-assistant workforce asked to fill an ever-expanding role against overwhelming data.

The only response the incumbents offer is to graft some new widget onto a foundation that has expired. We need a new health system for a new era. It is called Health 4.0.

The Triangle of Trust is the conceptual architecture of what I am trying to build. The PBC and the Trust are the legal architecture that holds it in place. Together they answer the question every thoughtful reader of Doctor AI deserves to ask: how is this not just another version of the system that produced the mess we are in?

The answer starts in the corporate form.

If you agree, share this essay with someone who is skeptical. If you disagree, push back. The work needs both true believers and questioners. We get to a better healthcare system only by arguing about the structure of the entity that will deliver it. The H4 Alliance is built so that argument can happen in public, on the record, under enforceable fiduciary obligations.

That is the case for it. That is what it actually is.

If this resonated

Subscribe for more on Trajectory Engineering, Health 4.0, and Doctor AI — where I make the case, evidence first, for rebuilding trust and redesigning the American healthcare system from the corporate form up.

My book, Doctor AI Reimagining Healthcare Rebuilding Trust Delivering Health 4.0, is available on Amazon. Learn about the H4 Alliance at H4alliancetrust.org. More at robinblackstone.com. Papers on SSRN.

A version of this essay first appeared in my LinkedIn newsletter.

There is a line in medicine we do not talk about enough.

On one side of that line is Rescue. The operating room. The ICU. The transplant list. The catheterization lab. The heroic intervention after disease has advanced, anatomy has changed, organs have failed, or risk has finally declared itself.

Modern medicine is very good at Rescue. We have built entire institutions around it. We celebrate it, fund it, train for it, measure it, and tell stories about it. And we should. Rescue saves lives.

But the future of health will increasingly be defined by what happens before Rescue is necessary — by preventing the disease that would demand it, and by preventing harm within the procedures we cannot avoid.

That is the pulse of the next era of medicine. Not less expertise — earlier expertise. Not less technology — better-timed technology. Not the abandonment of Rescue — the redesign of health so fewer people arrive there.

It is the deeper pattern behind some of the most important changes now emerging in medicine. They may look unrelated at first: GLP-1 medications reshaping obesity and metabolic disease; robotic systems entering the operating room; safety tools that help surgeons recognize danger before an irreversible act. Underneath, they are the same shift.

Medicine is moving upstream. Not away from expertise. Away from preventable harm.

From late rescue to earlier intervention

We are already seeing this in obesity and metabolic disease.

Obesity and chronic disease are the defining medical issues of American healthcare. Chronic disease drives roughly seven of every ten deaths in the United States and the large majority of its healthcare spending — and much of that burden traces to risks we already know how to prevent.

As GLP-1 medications have expanded, metabolic and bariatric surgery use has fallen sharply. A 2026 JAMA Surgery analysis of nearly 31.7 million surgery-eligible adults found that metabolic and bariatric surgery use declined 46.4% — from a peak of 0.17% of eligible patients in the fourth quarter of 2022 to 0.09% by the third quarter of 2025. Over the same period, the prevalence of semaglutide and tirzepatide prescriptions among eligible patients rose to 24.17%.

That does not mean surgery is obsolete. It does not mean medications are right for every patient. And it certainly does not mean obesity is solved.

But it shows something important: when treatment moves earlier in the course of chronic disease, the architecture of care changes. Referral patterns change. Operating rooms change. Workforce needs change. Economics change. The point of intervention moves.

The deeper story is not that one treatment replaced another. It is that the whole cascade can be interrupted earlier. Obesity sits upstream of a vast share of the procedural burden in American medicine: type 2 diabetes and its complications, hypertension and heart disease, osteoarthritis and joint replacement, sleep apnea, fatty liver disease, several cancers. A treatment that meaningfully reduces obesity does not avoid one operation — it bends the trajectory away from most of them, sparing the diabetes that never advances, the joints and arteries and liver never harmed, and the long sequence of later operations that never becomes necessary. A shift in how we treat weight is not a niche story about bariatric surgery. It is a story about the shape of the entire system.

This is what I mean by moving upstream from Rescue.

The democratization of medical intelligence

Part of this shift is patient-driven, and that is not a bad thing. For the first time, patients have access to much of the same data and intelligence that once lived only with clinicians — and that access confers agency. Many will choose a medication they can start and stop over an operation they cannot undo. The trade is real: bariatric surgery is a one-time intervention with durable effects for many, while GLP-1 therapy is ongoing, and weight often returns when it stops. A reversible pill and an irreversible operation are not the same bet — and patients, increasingly informed, are the ones placing it.

This cuts both ways. The same access that confers agency can deliver noise instead of signal, misinformation instead of evidence. The answer is not to withhold intelligence from patients — it is to make trustworthy intelligence the easiest to reach. That is its own design problem, and its own responsibility.

Moving upstream does not erase surgery — it redistributes it

Here is the part the headlines miss. Moving upstream is never a clean subtraction. When prevention rises in one place, demand does not simply vanish; it moves. And the gallbladder is where that becomes vivid.

The same medications now reducing bariatric surgery appear to increase the risk of gallbladder disease. A 2022 meta-analysis of 76 randomized trials found GLP-1 use associated with a higher risk of gallbladder and biliary disease, with a relative risk of 1.37; acute cholecystitis was roughly one and a half times more likely. Two mechanisms drive it: GLP-1 medications slow gallbladder emptying, and rapid weight loss itself promotes gallstone formation. Both are classic triggers for exactly the disease that leads to the operation at the center of this essay.

The absolute risk is modest — on the order of twenty-seven additional cases per ten thousand people each year — and the long-term picture is genuinely uncertain, because obesity is itself a major cause of gallstones. Durable weight loss may lower lifetime risk even as the active treatment phase raises it. But the direction is striking. The very drug class moving obesity upstream may be pushing more people toward the one operation this essay is about. Prevention in one disease can generate procedures in another.

And this is not unique to the gallbladder. It is an early, visible instance of a pattern now unfolding across medicine.

In cardiology, statins and aggressive risk-factor control move the point of intervention upstream of the catheterization lab; as medical therapy improves, the case for stenting stable disease narrows, even as other procedures evolve to take its place.

In oncology, screening, vaccination, and increasingly powerful systemic therapies shrink, delay, or even replace operations that once defined cure — sometimes preserving the very organs surgery would have removed.

Each of these is its own essay, and I will take them up in turn. Together they point to one conclusion: the volume and shape of surgery are not fixed. They are downstream of how well we intervene earlier. Move upstream, and you do not abolish the operating room. You change what walks into it — and you redistribute where skill and judgment are most needed, including how we will have to teach and train the next generation of surgeons.

For decades, too much of medicine has waited until disease becomes procedural. We intervene after the joint is destroyed, after the artery is blocked, after the ventricle has failed, after metabolic disease has reshaped the body, after risk has accumulated into crisis.

The next era will ask a harder question: how much suffering can we prevent before Rescue becomes necessary?

That same question is now coming into surgery itself.

A gallbladder is not always a small operation

A recent discussion about autonomous gallbladder surgery prompted me to look more closely at the field. I have personally evaluated and operated on many hundreds of patients with gallbladder disease.

At first glance, gallbladder surgery may seem an odd place to focus. Laparoscopic cholecystectomy is common, usually safe, and often same-day. But every surgeon knows the phrase “routine case” can be dangerous. The operation is simple until it is not.

The complication that matters most is bile duct injury. Major bile duct injury after cholecystectomy is uncommon, but not rare enough — large studies report major injury rates around 0.2% to 0.6%. With roughly 750,000 to 1,000,000 cholecystectomies performed annually in the United States, even a 0.3% major injury rate translates to roughly 2,300 to 3,000 major bile duct injuries every year.

That number should be viewed as a problem to be solved.

A bile duct injury is not a footnote in an operative report. It can become a life-altering disease: bile leak, biloma, sepsis, cholangitis, repeated ERCPs, percutaneous drains, reoperation, hepaticojejunostomy, long-term strictures, recurrent hospitalizations, disability, litigation, reduced quality of life, increased mortality. It affects the life and well-being of both the patient whose health is at stake and the physician and team caring for them. It invites legal review that is often hostile.

A gallbladder operation can be small. A bile duct injury is not.

Cholecystectomy is one of the most common operations in the United States — and unlike obesity or heart disease, gallstone disease has no scalable pharmacologic alternative. There is no pill that dissolves the problem at population scale; the definitive treatment is to remove the gallbladder. That is precisely why, here, moving upstream cannot mean prevention by medication. It has to mean moving upstream inside the operation itself.

The real question is not whether the robot can cut

The public conversation about autonomous surgery often jumps to the wrong image — a robot replacing a surgeon. That is not the right frame.

The question is not whether a robotic system can hold an instrument, place a clip, or cut tissue. The real question is whether it can know when it is not safe to do those things — because that judgment, not dexterity, is what could one day let it improve on the human.

In gallbladder surgery, that question lives in Calot’s triangle — more precisely, the hepatocystic triangle — where the surgeon must identify the anatomy before dividing any ductal or arterial structure.

This is where the operation becomes a judgment problem. Not a dexterity problem. Not a gadget problem. A judgment problem.

The critical view of safety is the discipline surgeons use to reduce the risk of misidentification. SAGES describes three required elements: the hepatocystic triangle must be cleared of fat and fibrous tissue; the lower third of the gallbladder must be separated from the liver to expose the cystic plate; and two, and only two, structures should be seen entering the gallbladder.

Only then should the surgeon proceed.

That pause matters. It is the quiet moment when the operation asks the surgeon: Do you truly know what this is? Have you earned the right to clip? Have you earned the right to cut?

If the answer is no, the right response is not confidence. It is restraint.

No critical view, no clip

This is why the most important sentence in the future of autonomous gallbladder surgery may be very simple:

No critical view, no clip. No clear anatomy, no cut. No confidence, stop.

That is not a slogan. It is a safety architecture.

The most consequential failure in gallbladder surgery is usually not the mechanics of applying a clip. It is misidentifying anatomy — mistaking the common bile duct or common hepatic duct for the cystic duct, and proceeding when the field has not been made safe.

The future of surgical robotics will not be decided by whether a machine can move smoothly. It will be decided by whether the system can refuse to proceed when the anatomy is uncertain. That is the difference between automation and safety.

What the Johns Hopkins work actually shows

The Johns Hopkins SRT-H work is important. The team reported a robotic system that completed a long sequence of cholecystectomy-related tasks across eight ex vivo gallbladders with 100% success, operating without direct human intervention.

That is a milestone. Let’s examine the space between what they did and what may be possible, because the nuance matters.

The related ImitateCholec dataset, published in Scientific Data, was built around the clipping and cutting phase of robotic cholecystectomy. It includes more than 18,000 demonstrations from 34 ex vivo porcine cholecystectomy models, totaling about 20 hours of data across 17 surgical tasks.

Here is the key surgical detail: before that clipping and cutting phase, Calot’s triangle was dissected manually to achieve the critical view of safety. Tissues with anatomic abnormalities were excluded; only tissues with two separable tubes were selected.

That does not weaken the work. It tells us exactly where the frontier is.

The system demonstrated that once the anatomy had been prepared and constrained, a robotic platform could perform a bounded, high-stakes phase of the operation. But the harder problem remains. Can the robotic system prepare the field for that step? Can the system recognize when the field is not safe? Can it identify uncertainty? Can it stop?

That is the line between impressive mechanics and real surgical reliability.

Technology alone does not make surgery safer

“Robotic” does not automatically mean safer.

A 2025 JAMA Network Open study of 737,908 Medicare beneficiaries found that bile duct injury rates were about three times higher after robotic-assisted cholecystectomy than after laparoscopic cholecystectomy — roughly 0.7% versus 0.2% — and that this gap held across low-, medium-, and high-risk groups.

That finding should sober the conversation. A robotic platform is not a safety system by itself. Better instruments do not automatically produce better judgment. A more advanced interface does not automatically produce better anatomy recognition.

The central issue is not whether there is a robot in the room. It is whether the system reduces avoidable harm. That requires discipline. It requires verification. It requires a design that can pause, warn, document, and refuse. At this time, it requires a human surgeon.

And anything that depends on a human surgeon scales only as far as we can train them. For common diseases, treatment has to scale to the number of people who have them and the urgency of their need. That is one more reason medicine keeps moving upstream: prevention scales in a way that Rescue cannot.

The upstream move inside the operating room

In chronic disease, moving upstream means preventing the need for Rescue. In surgery, it means preventing the wrong cut from ever being made. That is the connection.

GLP-1 medications and autonomous surgical robotics look like different stories — one a drug for metabolic disease, one a machine in the operating room. They are the same move at two points on the same timeline: the drug prevents the disease that sends people to Rescue; the safety system prevents the harm once Rescue is unavoidable. Prevent the journey, or make the journey safer — both push the point of intervention earlier.

The most important innovations will not simply make Rescue more dramatic. They will move the point of intervention earlier. Earlier in the disease process. Earlier in the risk curve. Earlier in the operation. Earlier than the irreversible mistake.

In metabolic disease, that may mean treating obesity before advanced complications require procedural intervention. In surgery, it may mean building systems that recognize danger before a duct is clipped, before tissue is divided, before the complication exists.

That is the future worth building. Not because it is flashy. Because it is humane.

The surgeon does not disappear

This is where the conversation often goes wrong. The future is not “robot versus surgeon.” That is too crude.

The surgeon’s highest value is not hand motion. It is judgment: patient selection, operative strategy, anatomic interpretation, recognition of danger, management of bleeding, decision-making under uncertainty. Knowing when to image. Knowing when to convert. Knowing when to perform a subtotal cholecystectomy. Knowing when to call for help. Knowing when to stop.

Those are not minor parts of surgery. They are surgery.

The best future does not remove the surgeon from responsibility. It surrounds the surgeon with systems that make the safest action easier, the dangerous action harder, and the uncertain action impossible to ignore.

That is how high-reliability fields evolve. They do not depend indefinitely on individual heroism. They design around human variability. They standardize what can be standardized. They measure what can be measured. They create pauses before irreversible action. And they preserve judgment where judgment matters most.

Medical education has to change

If judgment is the surgeon’s highest value, then the institutions that form judgment matter most of all — and they are among the slowest in medicine to change. Medical education was built for the Rescue era. It still selects for memorization, rewards procedural volume, and venerates individual heroism — the very things this shift is quietly making less central. The movement toward simulated care scenarios and teaching without endangering people is the right step, and is becoming more sophisticated.

But there can be no mistake that the next era of medicine asks for something different. If machines will increasingly handle recall and dexterity, then training has to cultivate what they cannot: anatomic interpretation, recognition of danger, the discipline to stop. That cannot be left to apprenticeship and case volume alone — especially as the case mix shifts and the number of any single operation a trainee sees shrinks and redistributes. It is undetermined but possible that the live cases they do take part in will be more unusual and more difficult, with patient expectations higher. Tomorrow’s clinicians must also learn to work alongside systems that pause, warn, and refuse — to supervise intelligence rather than compete with it — and to practice prevention and longitudinal care, not only rescue. There is no haven in the Global South for surgeons of yesteryear; the Global South is going to leap over this era straight into the next, as it has with payment models.

None of this is how the established system is built. Accreditation, residency design, reimbursement tied to procedures, and a culture that still tells its best stories about the save — all of it resists. But a system engineered to produce rescuers will not, on its own, produce the clinicians the next era needs. Reforming medical education is not a side project of this transformation. It is a precondition for it. That is a subject I will take up on its own.

The future of health will be judged by what it prevents

The promise of autonomous surgery is not that a robot can cut. Cutting is not the moral center of this story. The promise is a system that may one day help prevent the cut that should never have been made.

That is the bigger lesson. The next era of health will not be judged only by the brilliance of its rescues. It will be judged by how often Rescue was no longer necessary — by disease prevented before it begins, and by harm prevented inside the procedures we cannot avoid.

How many operations were avoided because disease was treated earlier. How many complications never occurred because danger was recognized sooner. How many patients were spared the long tail of preventable harm. How many clinicians were supported by systems designed for reliability rather than heroics.

That is the architecture of future health. Not replacement — reliability. Not spectacle — safety. Not Rescue alone — prevention before Rescue.

And in gallbladder surgery, that future begins in a small triangle where the oldest surgical wisdom still applies:

No critical view, no clip. No clear anatomy, no cut. No confidence, stop.

If this resonated

Subscribe for more on Trajectory Engineering, Health 4.0, and Doctor AI — where I make the case, evidence first, for rebuilding trust and redesigning the American healthcare system from prevention up. The same shift is unfolding in cardiology and oncology, and I’ll take up each of those next.

My book, Doctor AI: Reimagining Healthcare, Rebuilding Trust, Delivering Health 4.0, is available on Amazon. More at robinblackstone.com. Papers on SSRN.

References

1. Centers for Disease Control and Prevention. Fast Facts: Health and Economic Costs of Chronic Conditions. https://www.cdc.gov/chronic-disease/data-research/facts-stats/index.html

2. Rohde SC, Chao GF, Abdel-Rasoul M, Elsaid MI, Sweigert PJ. Trends in Metabolic and Bariatric Surgery Use During the GLP-1 Receptor Agonist Era. JAMA Surgery. 2026;161(5):546–549. doi:10.1001/jamasurg.2026.0049. https://jamanetwork.com/journals/jamasurgery/fullarticle/2845817

3. He L, Wang J, Ping F, et al. Association of Glucagon-Like Peptide-1 Receptor Agonist Use With Risk of Gallbladder and Biliary Diseases: A Systematic Review and Meta-analysis of Randomized Clinical Trials. JAMA Internal Medicine. 2022;182(5):513–519. doi:10.1001/jamainternmed.2022.0338. https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2790392

4. Bile Duct Injury. StatPearls (NCBI/NIH). https://www.ncbi.nlm.nih.gov/books/NBK546703/

5. SAGES. Safe Cholecystectomy Multi-Society Practice Guideline. https://www.sages.org/publications/guidelines/safe-cholecystectomy-multi-society-practice-guideline/

6. Kim JW, et al. SRT-H: A hierarchical framework for autonomous surgery. Science Robotics, 2025. https://www.science.org/doi/10.1126/scirobotics.adt5254

7. ImitateCholec: A Multimodal Dataset for Long-Horizon Imitation Learning in Robotic Cholecystectomy. Scientific Data, 2025. https://www.nature.com/articles/s41597-025-06526-z

8. Mullens CL, et al. Patient Complexity and Bile Duct Injury After Robotic-Assisted vs Laparoscopic Cholecystectomy. JAMA Network Open, 2025. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11937934/

Healthcare has many names for the person at the center. Patient. Consumer. Member. User. Covered life. Chart. Risk score. Data source. Each word reveals something about the system that uses it.

“Patient” belongs to the clinical encounter. “Consumer” belongs to the marketplace. “Member” belongs to the insurer. “User” belongs to the platform. “Covered life” belongs to actuarial calculation. “Chart” belongs to the medical record. “Risk score” belongs to the algorithm. “Data source” belongs to the machine.

But none of these words is strong enough.

The person whose life, body, data, dignity, family, finances, and future are at stake is not merely a participant in healthcare. That person is the Human Principal— the one healthcare exists to serve.

This is more than a semantic adjustment. It is a moral correction. Because every system eventually reveals who it believes the principal really is.

If the principal is the hospital, the system optimizes for institutional throughput. If it’s the insurer, for risk and cost containment. If it’s the technology platform, for engagement, data capture, and scale. If it’s the investor, for return. If it’s the government, for population metrics, compliance, or budgetary control.

All of those actors matter. None can be ignored. But none of them is the principal. The principal is the human being whose life is actually at stake.

That may sound obvious. But much of modern healthcare behaves as though it has forgotten it. And the age of AI will make the question unavoidable. Before we ask what artificial intelligence can do in healthcare, we have to ask something more fundamental: on whose behalf will it act?

AI changes the autonomy question

The arrival of AI in healthcare doesn’t merely add a new tool. It changes the architecture of agency.

An AI system can summarize a record, suggest a diagnosis, flag a risk, draft a message, recommend a treatment pathway, predict deterioration, prioritize outreach, shape scheduling, influence insurance decisions, and determine which human beings are seen as urgent. Even when it isn’t “making the decision,” it structures the field in which decisions are made.

It can make some options visible and others invisible. Some patients legible and others statistically obscure. It can make institutional goals appear clinically neutral. It can turn a person’s story into a prediction.

This is why the Human Principal must come before Doctor AI.

Begin with the technology, and we ask: what can the model do? Begin with the institution, and we ask: what can it optimize? Begin with the market, and we ask: what can it monetize? Begin with the Human Principal, and we ask: what form of intelligence would actually serve this person’s health, dignity, agency, and future?

That is the better question.

AI should not turn the human being into a more efficient object of management. It should help restore the conditions under which a person can be known, heard, protected, and supported. It should make care more understandable, not more opaque; systems more accountable, not harder to challenge; relationships stronger, not replaced by automated indifference. It should help clinicians see the whole person rather than reduce the person to a risk score.

AI in healthcare will either deepen autonomy or quietly diminish it. The difference will be governance. The difference will be design. The difference will be whether we know who the principal is.

The principal is not always alone

To call the person the Human Principal does not mean they must make every decision alone. That would be a cruel misunderstanding of autonomy.

The principal may need a surrogate, a family member, a trusted advocate. They may need a clinician to recommend, not merely list options. They may need protection from coercion, misinformation, cognitive overload, or predatory systems. Autonomy is sometimes exercised directly, sometimes supported, sometimes represented — and sometimes reconstructed from prior values, advance directives, and careful ethical judgment.

In Cruzan v. Director, Missouri Department of Health, the Supreme Court recognized that a competent person has a liberty interest in refusing unwanted medical treatment — while grappling with what happens when a person can no longer choose, and a surrogate must act in their place. That tension sits at the center of medicine.

The human person is capable of autonomy. But autonomy exists within vulnerability. A serious health system has to be built for both. It must respect the person who can speak and protect the person who cannot. It must listen to the voice that has historically been discounted, and support the person whose choices are constrained by poverty, racism, disability, trauma, geography, language, age, or institutional power. And it must never confuse silence with consent.

The moral failure of reduction

The central danger in healthcare is reduction.

The person becomes the tumor. The pregnancy becomes the risk category. The elder becomes the bed. The woman becomes the reproductive organ. The person with mental illness becomes the diagnosis. The person with a disability becomes the cost. The person of color becomes the adjustment factor. The patient’s story becomes unstructured data.

Reduction makes systems easier to manage. It also makes people easier to harm.

The Human Principal is a refusal of reduction. It insists that this person is more than the disease, more than the claim, more than the code, more than the model output, more than the cost center. The person has a body but is not only a body. Has data but is not only data. Carries risks but is not only a risk profile. May be vulnerable, but is not without dignity. May require help, but is not morally subordinate. May be ill — but is still the principal.

What the Human Principal requires

Take the Human Principal seriously, and a set of obligations follows.

It must be intelligible. A system that cannot be understood cannot be trusted. No one should need professional training to understand what is happening to their body, their data, their bills, or their care plan.

It must be accountable. When a decision affects a person’s diagnosis, treatment, coverage, or risk profile, there must be a way to know who made it, on what basis, and how it can be challenged.

It must be relational. Information alone is not care. A portal is not a relationship. A chatbot is not trust. A discharge packet is not continuity. People need human beings who can listen, interpret, accompany, and advocate.

It must protect dignity in data. Health data is not ordinary information — it carries the person’s body, history, family, vulnerabilities, and future. To use it without a dignity-preserving framework is to treat the person as raw material.

It must recognize constrained autonomy. There is no meaningful choice when the only options are unaffordable, inaccessible, culturally unsafe, or administratively impossible. Autonomy requires real conditions, not just formal permission.

And AI must be fiduciary in spirit — designed around loyalty to the person’s legitimate interests. Not loyalty to engagement. Not to billing. Not to institutional convenience. Not to hidden commercial incentives.

The Human Principal is the test. Does this technology help the person understand? Does it help the clinician care? Does it reduce fragmentation, protect privacy, make decisions more transparent, improve safety, honor the person’s values, strengthen trust? If not, intelligence may be the wrong word for what we are building.

A new word for an old truth

The Human Principal is a new phrase for an old truth. The person is not the property of the healer, the instrument of the researcher, the subject of the state, the product of the platform, or the asset of the data economy. The person is the one on whose behalf the system is supposed to act.

That is the thread connecting bodily autonomy, informed consent, human-subjects protection, patient-centered care, shared decision-making, health equity, data ethics, and AI governance. Each names part of the same moral architecture: the body must not be touched without permission; the mind must not be bypassed; the vulnerable must not be exploited; the patient must not be silenced; the algorithm must not become an unchallengeable authority.

The next era of medicine will be judged by whether it remembers this. Care may become more computational, more predictive, more personalized, more continuous. That future could be extraordinary — earlier detection, better coordination, less burden on clinicians, expertise extended to people who never had it. Or it could become colder, faster, more opaque, more extractive. The difference will not be the technology alone. It will be the moral order around the technology. And moral order begins with a single question: who is the principal?

The first principle of trustworthy healthcare

The first principle is this: the human being is not an object of the healthcare system. The human being is the principal.

The hospital, the insurer, the clinician, the AI model, the data platform, the payment system, the regulator, the investor, the research enterprise, the public health authority — all have legitimate roles. None should displace the human being at the center.

The future of healthcare is not really about making medicine more digital, or more efficient, or even about making AI more accurate. It is about whether we can build intelligent systems around trustworthy human relationships. And that begins by remembering the person — not as a patient alone, not as a consumer, a user, a chart, or a data source, but as the Human Principal. The one whose life is actually at stake. The one healthcare exists to serve.

So I’ll leave you with the question this whole essay turns on:

What would change if healthcare treated the person not as a patient, consumer, member, user, risk score, or data source — but as the principal? And what would a trustworthy health system have to do differently if that were truly its starting point?

Interested in learning more: robinblackstone.com.

Somewhere along the way we decided the word human was common.

Not common as in shared — common as in ordinary, lowercase, unremarkable. A category. We capitalize the names of companies, of products, of the smallest startup with a logo and a runway. Google. Pfizer. Epic. We grant the capital letter to anything we’ve agreed to take seriously. And then we write human in lowercase, the way you’d write thing, or item, or unit.

I’ve started capitalizing it. Human. And I want to tell you why, because it is not a typo and it is not an affectation.

A capital letter is the smallest act of respect a language has to offer. It costs nothing. It changes nothing you can measure. And yet we reserve it with great care — for the proper, the named, the things we’ve decided are not interchangeable. To capitalize a word is to say: this one is not a category. This one is a someone.

We already know this. English capitalized Black — the AP made it official in 2020 — because a people is not a color swatch, and the capital letter carried a recognition the lowercase one withheld. The Deaf community capitalizes the Dfor the same reason: to mark a culture, not just a condition. These were not grammatical decisions. They were moral ones, made in the only place where respect and language touch. The capital letter was the recognition.

So here is my small, stubborn proposal. The Human is the proper noun of medicine. Not the patient — a role. Not the case — a chart. Not the provider — a billing code with a stethoscope. The Human. The someone who is not interchangeable, who cannot be swapped for the next someone in the next room, whose health is a trajectory across a life and not a service rendered at a moment.

When you lowercase the human, you make the rest of it easy. A lowercase human can be a throughput metric. A lowercase human can be a fifteen-minute slot. A lowercase human can be discharged on paper while still frightened in the body. The word permits the system. Language always goes first; the spreadsheet follows the grammar.

This is the same move I keep coming back to. We were renamed providers and the job quietly shrank to fit the smaller word. The Human is the renaming in the other direction — restoring the capital letter that medicine, of all things, should never have surrendered. It is why the role I’m building is the Human Health Ally, and why Human comes first in it. Not human as a function. Human as a someone, with a someone walking beside them.

I know the objection. Grammar says common nouns stay lowercase, and I’m breaking a rule. I am. On purpose. Every capital letter that now feels obvious was once a rule someone chose to break for a reason that mattered more than the rule. I think this one matters more than the rule.

A system that cannot bring itself to capitalize the Human will build everything else accordingly — the metrics, the codes, the fifteen-minute slots, the quiet assumption that you are a function and your job is to step aside. A system that does capitalize it has to build differently. It has to defend the body the way we defend the air, the water, the food. It has to treat continuity as care and not as overhead. The capital letter is small. What it commits you to is not.

So I’ll keep writing it this way. Human. If that looks like a mistake to you, sit with why a capital letter feels like too much to give — and to whom we give it without a second thought.

That second thought is the whole argument. Come think it through with me.

They named us providers. As if the whole of medicine is dispensing — a vending machine with a diploma. You insert a complaint, a service comes out, next patient.

I want to talk with you about that word, because words are never just words. There was a day it appeared on the badge — no announcement, no meeting. Physician became provider. Surgeon became provider. It seemed small. A billing convenience. A box on a form.

It was not small.

You don’t strip a profession of its judgment, its memory, its right to know a patient over time by taking those things directly — people would notice. You do it by changing what you call it. Physician carries an oath and a relationship. Provider carries a transaction. And a transaction is interchangeable, which means the person performing it is interchangeable, which means their presence over the years is overhead to be trimmed. The word did the work the spreadsheet wanted done.

That’s the trick, and it’s worth saying plainly: rename the person, and you can quietly redefine the job.

So here’s mine.

Yes — I’m renaming them too. That’s exactly the point. The same lever that was used to make us smaller can be used to make us whole. They used the word as a cage. I’m using it as a door.

We are not providers. We are Human Health Allies.

Read that name slowly, because every word in it is a correction. Human — not a function, not a billing code with a stethoscope; a person, knowing another person. Health — not rescue at the edge of the cliff, but the long upstream work of keeping someone off it. Ally — not a vendor on the other side of a transaction, but someone on your side, over years, when dangerous and benign look exactly the same and the only thing that tells them apart is having been there, and seen it all along.

An ally is one corner of a triangle of trust: the Human Principal, the human who walks beside them, and the intelligence that now amplifies what that human can see. Take away the ally and the triangle collapses into what we have now — a patient and a machine and no one who remembers them. The whole of Health 4.0 depends on that third name being ally, not provider. You cannot build trust on a transaction.

I wrote last week about an airplane aisle — silver hair, a hitch in my hip, a cabin full of people assuming I couldn’t lift my own bag before I’d said a word. The assumption arrived before I did. “Provider” is that assumption, printed and laminated. It decides who you are, and what you’re allowed to keep, before you’ve spoken.

I’m not interested in being a provider. Not the woman with the silver hair, and not the profession being told its judgment is overhead.

If you remember why you became a physician — if you’ve felt the word provider fit you like a uniform two sizes too small — there’s a door. Walk through it here.

We are not done.

And we were never just providers.

(The piece this grew out of, “Not Done Yet,” is here.)

A bridge between the Human Principal and the Human Health Ally — the two human corners of the Triangle of Trust.

This week we met the Human Principal — you, the person whose health it is, restored to the head of the table.

Human Principal: the person whose health and life are at stake, holding final authority over both. Words chosen from agency law: a principal is the party on whose behalf all agents act and from whom their authority derives. Doctor AI advises, remembers, and navigates; the Health Ally counsels and accompanies; but consent, purpose, and the definition of what this life is for belong to the Principal alone — and cannot be delegated, only exercised or surrendered.

Next week we meet the Human Health Ally — the human being who stands beside you there. Before we make that introduction, I want to give both of them a list. It is short. It is researched. And it is non-negotiable.

In 2024, two researchers at MIT — Isabella Loaiza and Roberto Rigobon — asked a question most of healthcare has been too nervous to ask plainly: when the machines arrive in force, what work remains irreducibly human? Not sentimentally human. Not nostalgically human. Demonstrably — in the labor data of an entire economy — human.

They found five capabilities, and they gave them a name that happens to describe the moment we are living in: EPOCH.

Empathy. Not the recognition of emotion — machines can do that now, and do it well. The sharing of it. To be genuinely affected by another person’s suffering is a biological event between two nervous systems. There is no API for it.

Presence. A hand on a shoulder. A body in a room. The thing nurses have always known and billing codes have never captured: some part of care only transfers in person, between people who showed up.

Opinion. Judgment, and the accountability that judgment carries. A machine can rank options; it cannot own a decision. Responsibility is not a computation — it is a commitment made by someone who can be called to answer for it.

Creativity. The leap that is not recombination — the question nobody thought to ask, the treatment plan that breaks the pattern because this patient is not the pattern.

Hope. The capacity to hold a vision of a future that the data does not yet support, and to lead another person toward it. Every clinician who has ever sat with a frightened patient knows that hope is not a mood. It is a clinical intervention. It may be the oldest one we have.

Here is what the MIT data showed, and it should reorder how we think about the next decade: work rich in these five capabilities grew faster*over the last decade, and the new tasks emerging in the economy score higher on them than the old tasks they replace. The market — not the philosophers, the market — is already paying a premium for what only humans do.

And there is a sixth thing, unlisted because no economist can put it in a column: culture. Culture is uniquely interpretable by humans. A machine can map its surface — the foods, the festivals, the words for pain. But to read what health means inside a culture — which meal is love, which stoicism is survival, which silence is consent and which is fear — requires a reader who lives inside one. Empathy, presence, opinion, creativity, and hope are the five capabilities; culture is the language they are spoken in. That is why we say it as plainly as it can be said: culture is health, uniquely interpretable by humans.

Now look at the Triangle of Trust through this lens.

Doctor AI takes the codified work: the memory across years, the vigilance that never sleeps, the navigation of a system too complex for any human to hold. The data too complex to sort, and to apply or not to the human in the triangle. That is its corner, and it is vast.

But the other two corners are human, and EPOCH is their shared job description. The Human Health Ally owes the Principal all five: empathy that is real, presence that is physical, opinion that is owned, creativity that sees this patient, hope that leads somewhere. And the Human Principal — this is the part we say too rarely — holds the same five commitments toward their own life: empathy for the body they have sometimes treated as an adversary, presence in their own care instead of absence from it, opinion as the final word on what their life is for, creativity in how a health is built, and hope as the discipline of believing their trajectory can bend.

The machine cannot be the moral agent of care. That sentence is on the wall of everything we are building. EPOCH is the evidence behind it — five capabilities, two human corners, one triangle held together not by technology but by the things technology cannot do.

Next week: the Human Health Ally — who they are, what they owe, and why the future of medicine belongs to the humans who master what cannot be automated.

*Loaiza, I., & Rigobon, R. (2024). The EPOCH of AI: Human-Machine Complementarities at Work. MIT Sloan School of Management.*

*Culture is Health.*

— Robin Blackstone, MD

In American emergency departments this morning, a physician put on a white coat and went to work knowing — statistically — that someone in that building was going to threaten, harass, or hit them before the day was over. Ninety-one percent of emergency physicians personally experienced or witnessed workplace violence in the prior year. Reported harassment of providers more than doubled between 2018 and 2022. We have crossed a line, and very few people are willing to say it out loud.

I went looking for the comparison data, because I wanted to know how bad this gets when it gets worse. The answer is China. There, at least 362 doctors have been injured, disabled, or killed in the past decade, and in the most serious documented assault cases, 12.8% ended in the physician’s death. The Lancet has called it a crisis for the practice of medicine itself.

This is personal for me.

In 2012 I went to China, on behalf of Johnson & Johnson, to talk with the leadership in Endocrinology and Metabolic/Bariatric Surgery about establishing the data and the quality-of-care framework necessary to implement a quality system. While in Nanjing, a patient burst into our meeting and started yelling. He had come to Nanjing to have a gastric bypass to cure his diabetes and had just been told he had to go home without the surgery. The medical staff and surgeons were upset. His son came and took him out of the meeting room. The whole team spoke about his case. They had just started operating, and at a BMI of 58 with complex medical problems, he wasn’t a good place to start their experience.

All seemed calm until we left the room and he was waiting outside. He came right up to me and — through his son, who spoke English — we talked about his father’s situation. Gradually his father calmed down and began to understand. He thanked me, bowed to me, and then he and his son left.

It was a graphic demonstration of the expectations people bring to medicine, and how dangerous it can become when those expectations meet a system that cannot — or will not — explain itself. That memory was very much on my mind when the issue of violence was raised by a colleague on LinkedIn.

Two countries, one story.

Two countries, two very different health systems, one identical story underneath: when patients lose trust in the people and the institutions that are supposed to care for them, violence is what comes next. China shows the endpoint. The United States is on a slower version of the same curve.

The argument, in one sentence: physician assault is a trust problem dressed up as a security problem, and Health 4.0 — AI in service of the physician–patient relationship, not in place of it — is the most credible plan we have to reverse the curve.

What the Numbers Are Telling Us

China shows the endpoint. When a health system becomes overstretched, opaque, and transactional, doctor–patient trust collapses — and violence becomes the symptom. The Lancet has called it a crisis for the practice of medicine itself.

The United States is on a slower version of the same curve. Murders remain rare, but the daily baseline of assault, intimidation, and harassment is now structural. Reported harassment of providers more than doubled between 2018 and 2022.

The mechanism is the same in both countries: patients who feel unseen, unheard, or wronged by a system they no longer believe is on their side. Dissatisfaction with care is the single most-cited motive in U.S. physician homicides.

Health 4.0 as the Trust Intervention

Treating physician assault as a security problem — metal detectors, panic buttons, federal felony statutes — is necessary and insufficient. It hardens the perimeter of a system whose center is failing. The opportunity, and the argument of Doctor AI, is to change what healthcare is for the patient: from a transaction they survive, to a relationship the AI-augmented physician can finally restore.

Give time back to the doctor. Ambient AI documentation, decision support, and pre-visit synthesis pull the physician out of the screen and back into the room. Eye contact is the cheapest trust-repair tool we have.

Make the system legible to the patient. Explain the diagnosis, the cost, the alternatives, and the wait — in the patient’s language, on demand. Opaque systems breed grievance. Grievance is the gateway to violence.

Surface risk before it escalates. Predictive models can flag the conditions that precede serious incidents — long waits, repeated unresolved complaints, unmanaged pain, untreated psychiatric distress — and route human attention there first.

Rebuild measurable trust. Trust is not a feeling; it is a system property. Health 4.0 lets us measure it, report on it, and hold institutions accountable to it.

The Bottom Line

Violence against doctors is the loudest possible signal that a healthcare system has lost the consent of the people it serves. China’s data is a warning. America’s data is a forecast. Doctor AI argues that Health 4.0 — AI in service of the physician–patient relationship, not in place of it — is the most credible plan we have to reverse the curve before it bends further.

Read it. Share it with anyone who runs a hospital, writes healthcare policy, or thinks the violence in our EDs is somebody else’s problem.

— Robin Blackstone, MD

Sources

• Cai R. et al., Serious Workplace Violence Against Healthcare Providers in China, 2004–2018 (NIH/PMC) — https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7841458/

• Restoring Doctor–Patient Trust to Curb Violence Against Doctors (JMDH/PMC) — https://pmc.ncbi.nlm.nih.gov/articles/PMC9507972/

• Violence against doctors in China — The Lancet — https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(14)61438-0/fulltext

• Council on Foreign Relations — Rising Violence Against Doctors in China — https://www.cfr.org/articles/rising-violence-against-doctors-china

• American College of Surgeons — Violence Escalates Against Surgeons & Healthcare Workers (2024) — facs.org link

• American Hospital Association — The Burden of Violence to U.S. Hospitals — https://www.aha.org/costsofviolence

• Physician Leaders — Murders of Doctors by Patients: 123 cases since 1860 — physicianleaders.org link

• ACEP — Emergency Physicians’ January 2024 Workplace Violence Poll — emergencyphysicians.org link

• Stateline (Oct 2025) — Doctors and nurses are punched, attacked, even shot — stateline.org link

• HIPAA Journal — SAVE Healthcare Workers Act 2025 — https://www.hipaajournal.com/save-healthcare-workers-act-2025/

Doctor AI: Reimagining Healthcare, Rebuilding Trust, Delivering Health 4.0 by Robin Blackstone, MD. (April, 2026) Amazon Books.com

American Health - Who Gets Paid by Robin Blackstone (January, 2026) Amazon Books.com

BEDSIDE A podcast by Robin Blackstone, MD - available everywhere

MIT’s Work of the Future initiative has published a careful, evidence-based study of how generative AI is changing professional work across more than twenty companies in healthcare, retail, finance, and manufacturing. The findings are honest. The recommendations are sober. And for healthcare specifically, the paper is necessary but not sufficient — for a reason that is itself the argument I want to make.

The diagnosis is precise. Generative AI gets deployed against three recurring problems: bottlenecks, where routine tasks crowd out higher-value work; the cafeteria problem, where workers must synthesize input from multiple experts; and the learning curve problem, where novices need help performing like veterans. Across all three, workers are being shifted into supervisory control roles — the human in the loop — while AI executes more of the underlying process. The authors then offer six lessons: minimize drudgery, promote learning, preserve teamwork, design interfaces well, maintain domain expertise, and establish accountability.

These are the right concerns. I would defend every one of them inside the frame the paper sets up. Mental offloading is a real risk. Domain expertise is irreplaceable. Accountability creates the incentive structure that makes vigilance pay. Their formulation about expertise needing to be sufficient to validate and scrutinize what AI provides is one of the cleaner sentences I have read on this question.

And the frame is the workplace. The unit of analysis is the organization. The implicit question is: how do we design jobs around AI inside the institutions we already have.

For healthcare, that is the wrong grain.

Healthcare’s problem is not that its workers need better tools. Healthcare’s problem is that the institutions paying those workers are organized around the wrong thing. American healthcare pays for procedures, encounters, and codes. It does not pay for health. Every accountability mechanism MIT recommends — every guardrail, every audit, every interface tweak — will, deployed inside the existing system, sharpen the workforce’s vigilance about the things the system already rewards. Coding accuracy will improve. Documentation will improve. Throughput will improve. None of those things are health.

You can adopt all six MIT lessons and still produce worse health outcomes, because the container is misaligned with the outcome. This is not a critique of the paper. It is a statement about where its evidence applies and where it stops applying. Job redesign inside a misaligned institution is not the same intervention as redesigning what the institution is for.

Inverting the supervisory relationship

The supervisory control framing — clinician supervises AI — also preserves an assumption worth naming: that the clinician is the principal actor and the patient is the recipient. That is not how health is actually produced. Health is produced across years and decades, mostly outside any clinical encounter, by a person living a life inside a context. The clinician sees flashes of that arc. The patient is the only continuous actor in it.

A system designed for actual health, rather than clinical throughput, has to invert the supervisory control relationship. The patient — the Health Principal — is the person whose trajectory is being engineered. Clinical expertise, AI tools, longitudinal data, family context, environmental signals: these are instruments serving that trajectory. The Health Principal, supported by a Human Health Ally and by Doctor AI, is the one in the supervisory seat. The clinician is summoned when the trajectory demands clinical judgment, and is invaluable in that role, but is not the architecture’s pivot. This is a meaningful break from the workplace frame. Their human in the loop is a worker overseeing automation. Mine is a person overseeing their own health arc, with the entire apparatus — clinical, computational, social — accountable to that arc rather than to a billing code.

The offload already happened

The paper also worries about mental offloading: workers using AI may lose skill. The concern is real for any field where mastery matters, including medicine.

But for patients in the current system, the offload already happened — long before AI showed up. The American healthcare apparatus has spent decades transferring health understanding away from the people whose health is at stake and into an opaque, fragmented, professionally-gated set of institutions. Most patients cannot read their own labs, do not know their own trajectory, and have no instrument for sustained engagement with their own physiology between encounters. That is not a failure of patient effort. It is the predictable output of an architecture that does not treat the patient as the principal. Generative AI deployed inside that architecture risks deepening the offload — making the system more efficient at producing patients who are even more dependent on it. Generative AI inside a Health 4.0 architecture does the opposite: it returns understanding, signal, and continuity to the Principal. The technology is the same. The container determines whether it concentrates power or redistributes it.

Governance has to be capture-resistant

Every prior reform that aimed at this kind of redistribution was assimilated by the system it sought to change. Telehealth was going to redistribute care; mostly it became another billing modifier. Patient portals were going to give people their data; mostly they became another inbox the patient ignored. The institutions paying for these tools absorbed them and kept their incentives. Health 4.0 only matters if its institutional shape is structurally incapable of being absorbed in the same way. That means a non-profit foundation that does not benefit from utilization, a public-benefit corporation whose duties to the Principal are written into its charter, federal-reserve-style governance designed for capture resistance over decades, and active stewardship against the inevitable misreadings and re-framings that are how reforms get neutered. The architectural question and the institutional question are the same question — you cannot have a Principal-centered architecture inside a sickness-monetizing institution, and you cannot build the institution without committing to the design that prevents capture.

The Ally and the question of capacity

A meaningful share of the population, at any given time, cannot fully be the principal of their own healthcare in any given moment: acute illness, cognitive decline, mental health crises, children, end of life, or simply working three jobs and exhausted. Returning understanding to the Principal cannot become an unfunded mandate on the patient. The unit of supervision is Principal-plus-Ally-plus-Doctor-AI, not Principal-with-AI.

The architecture handles this through three commitments. First, capacity is decision-specific and time-bound, not a person-level switch. Someone can be incapacitated for one decision and fully capable for another. Second, a Health Charter — authored by the Principal during clear moments — captures values and standing instructions that govern interpretation during cloudy ones. Most people, most of the time, can articulate values even when they cannot articulate a specific decision in a specific moment. Third, the Ally’s role is interpretive rather than substitutive — the faithful translator of the Principal’s drift, fear, exhaustion, and hope into the most accurate reading of what the Principal would want if they were clear, with fiduciary duty to that Principal and no other allegiance. Outside acute emergency or sustained incapacity, this works for the vast majority of decisions a Principal will face across a lifetime. For the residual cases, the Charter governs by default and the Ally has standing; in the rare case neither exists, clinical judgment carries forward with retrospective Charter governance once the Principal returns to capacity.

This is not a softer version of the current system. It is a different architecture. The current system treats capacity as binary and defaults the absent-Charter case to maximum intervention regardless of what the Principal would have wanted, and that default is not accidental.

Accountability geometry

MIT is right that accountability is what makes vigilance pay. Inverting the supervisory relationship does not eliminate the accountability question; it relocates it, and the relocation has to be designed. The current system locates accountability where decisions visibly happen — the clinical encounter — and places malpractice liability on the clinician as the enforcement layer. That made sense in an architecture where decisions clustered at encounters and information lived with the clinician. In Health 4.0, decisions happen continuously, distributed across the Principal, the Ally, Doctor AI, and the episodic clinician. Accountability has to distribute too — without becoming the no-accountability that distributed schemes usually devolve into.

The geometry has six pieces. Capacity-based allocation: each actor is accountable for what they have the capacity to know and do, and only that. Fiduciary standard for Doctor AI and the Ally: a duty of allegiance to the Principal above all other interests, enforceable in court, designed to cut off the paths by which AI advisors get captured by payer or commercial incentives. Algorithmic transparency infrastructure: every Doctor AI recommendation is logged with inputs, model version, confidence, alternatives considered, and uncertainty disclosed; the reasoning is inspectable by the Principal at any time and by an investigator after any adverse event. A no-fault arm for stochastic outcomes: biology is irreducibly variable, the best decision can produce the worst outcome, and a compensation scheme on the model of vaccine injury compensation lets the Principal recover without forcing the question of fault into cases where nobody actually erred. A public learning body on the model of the NTSB: serious adverse events investigated systematically, findings published, the system actually learns, which is something the current system structurally cannot do. A public backstop for catastrophic financial harm: the Principal carries existential risk because they always have, but cannot carry catastrophic financial risk without re-creating the cost-shifting that disempowers Principals in the first place.

This is not a small architecture. It is, however, the architecture the rest of the case implies. A Principal-centered system with no fiduciary floor becomes a marketing channel. A transparent algorithm with no public investigation body produces no learning. A no-fault arm without a financial backstop becomes another way to settle cheap. The pieces only work together.

The third grain: culture

There is a third grain underneath both the workplace and the system, and honesty requires naming it. Healthcare institutions, however configured, produce a relatively small share of population health compared to what public health calls the social and structural determinants — housing, food, work, relationships, sleep, environment, the racial geography of zoning, the conditions in which life is actually lived.

What we are calling social and structural determinants is, more accurately, culture.

The standard public health framing treats these as discrete inputs to be adjusted with policy patches. That framing has been mostly absorbed without changing very much, because the determinants are not floating variables. They are emergent properties of a culture. Housing policy is a cultural artifact. Food systems are a cultural artifact. The way work consumes a life is a cultural artifact. Even biology, at the population level, is shaped by culturally determined exposures and behaviors. Naming this as culture forces the conversation to be about who we are and what we value, which is harder, slower, and actually generative. Naming it as “social determinants” lets the conversation stay technocratic, which is why it has changed so little.

This reframes Health 4.0 itself. The Principal-Ally-Doctor AI architecture is not primarily a delivery reform. It is a cultural artifact that embodies different assumptions about agency, authority, knowledge, and trust — and once embodied at scale, it becomes part of the cultural production of health. The architecture does not just deliver health; it encodes a different culture, and the encoded culture changes what determinants look like across generational time. That is the level at which Health 4.0 has to be evaluated. Not “does it deliver care more efficiently” but “does it embody and propagate a culture in which health is more likely to occur.”

Two boundary conditions are worth naming directly, because the equation can be misused in both directions. Culture operates on generational timescales; some determinants harm on weekly ones. Lead pipes, contaminated water, particulate air, food deserts as they exist today — these are downstream of culture but operate as physical facts on any given child this year. The cultural argument is not a license to defer the material one. It is an argument about where durable change comes from, and it is true alongside, not instead of, the work that has to happen at every other level. And the cultural framing must not become cover for structural actors who would prefer the conversation stay vague. Concentrated capital, regulatory capture, and industry consolidation produce culture as much as culture produces them, and a Health 4.0 argument that names culture without also naming those forces is performing the abdication it claims to oppose.

Three conversations

MIT is doing important work at the workplace grain. Inside any given organization, deploying AI well versus deploying it badly produces measurably different outcomes for workers and the business. Their six lessons are real lessons, and any leader integrating these tools should take them seriously.

Healthcare requires a second conversation, at the system grain. At that grain, the question is not how should clinicians use AI inside hospitals but what should the institutions of American health be organized to produce, and what architecture — clinical, financial, technological, regulatory, governance, accountability — would make them produce it. Workplace-grain answers applied without system-grain thinking will produce AI-augmented versions of the wrong system. That is not progress. That is acceleration in the wrong direction.

And healthcare requires a third conversation, at the cultural grain. Clinics do not cure poverty. A Principal-centered architecture is not a substitute for the cultural conditions of health; it is the architecture that no longer obscures them, and at scale becomes part of producing different ones. The promise of Health 4.0 is not that it delivers health. The promise is that it stops actively damaging health and extracting wealth, returns understanding to the people whose health is at stake, and makes the cultural conditions of health visible and addressable rather than invisible and inevitable.

The temptation, given how much capital and political energy are currently flowing into healthcare AI, is to treat the workplace-grain question as the whole question. It is not. The institutions buying these tools are themselves the problem the tools are being asked to compensate for. No amount of interface design will fix an incentive structure that pays for sickness instead of health. And no amount of incentive redesign will, by itself, change the culture that produced the incentive structure in the first place.

MIT has given us a serious account of how AI is changing work. What healthcare needs next is a serious account of how the work itself, the institutions organizing it, the governance protecting them, the accountability geometry holding them honest, and the cultural conditions in which a healthy life is actually possible — must change, together. Those are different conversations. We need all of them. Right now we are mostly only having the first one.

Riding the subway in New York City is an experience not to be missed.

It is crowded, irritating, intimate, democratic, and miraculous all at once. A city of millions runs because strangers agree, mostly without speaking, to move together through narrow spaces. The subway is not just transportation. It is a daily act of civic choreography.

Not long ago, I got on the 6 train in the late afternoon, just as people were getting off work. The car was packed. Standing room only. Bodies pressed together. Bags shifted. People adjusted their shoulders and their eyes in that practiced New York way — close enough to feel one another’s breath, distant enough to pretend not to notice.

A man was standing near the door.

At every stop, he had to move aside. People pushed past him to get off. Others pushed past him to get on. He stayed by the door anyway, turning his back to the opening with mounting irritation, as if the entire city had arranged itself against him.

Then another man stepped onto the train carrying a large toolbox.

I have seen many men like him on the subway in the morning and late afternoon, carrying the physical infrastructure of the city in buckets, bags, boxes, and tired shoulders. This toolbox was big. The space was small. He tried to make himself smaller.

The man by the door began muttering.

At first it was under his breath. Then louder. The words were not meant to be kind. The man with the toolbox ignored him, or perhaps did not fully understand him. But tone does not require translation. Posture does not require translation. Aggression announces itself before language catches up.

The rest of us stood there.

We all knew what was happening. We all felt the same tightening in the air. We all made the same silent calculation.

Will this pass?

Will it escalate?

Is it my place?

Will saying something make it worse?

This is the bystander’s bargain.

Stay quiet. Stay safe. Hope the next stop solves it. Hope someone else steps in. Hope the problem is not yet bad enough to require you.

Then the man’s voice rose. He was speaking directly over the head of the man with the toolbox.

And suddenly I heard myself say, “Excuse me — have you ever seen that movie on the poster?”

The angry man stopped.

He looked at the poster. Then he looked back at me.

For one second, the whole train seemed to hold its breath.

“No,” he said. “I haven’t seen it.”

I asked, “Do you know how many more stops it is to Hunter College?”

“Three,” he said.

He was right.

Two stops later, the man with the toolbox got off the train. Nothing happened. Or rather, something happened. The trajectory changed.

When I got off, I looked up at the man by the door.

He said, “Thank you.”

That moment has stayed with me, not because it was heroic. It was not heroic. It was not dramatic. It was not even especially brave.

It was simply an interruption.

And sometimes an interruption is the difference between watching and acting.

I have thought about that subway car often because I think American healthcare is now inside a version of it.

The pressure is rising. The space is shrinking. People are frightened, frustrated, exhausted, and increasingly close to one another’s breaking points. Costs rise. Trust falls. Clinicians burn out. Families delay care. Employers absorb expenses they cannot control. Hospitals struggle. Patients ration decisions no family should have to ration. Medical debt follows people long after the clinical event is over. Artificial intelligence is arriving faster than the governance needed to make it trustworthy.

Everyone can feel the escalation.

And still, most people have been assigned the role of bystander.

Wait.

Comply.

Appeal.

Navigate.

Call again.

Log into the portal.

Accept the denial.

Trust that someone, somewhere, is in charge.

But no one is truly in charge of the whole health trajectory.

That is the failure.

We have built a system in which the person whose life is most affected is often the least structurally empowered. We call that person a patient when sick, a consumer when billed, a member when enrolled, and a user when touching technology. Each word captures a fragment. None captures authority.

The patient waits.

The consumer shops.

The member belongs to a plan.

The user clicks.

But who governs the arc of the life?

Who has standing over the long trajectory — before disease, during disease, through treatment, recovery, disability, aging, and the financial consequences that can trail behind illness for years?

Too often, the answer is no one.

Or worse: everyone, in fragments.

That fragmentation trains people into passivity. It teaches them to believe that healthcare is too complicated to question, too technical to understand, too bureaucratic to challenge, and too expensive to escape. It turns confusion into compliance.

And when people are afraid to engage, they do something else.

They perform certainty.

They inherit slogans. They cling to partial truths. They turn personal wounds into universal explanations. They decide that the problem is doctors, or insurers, or hospitals, or government, or pharma, or patients, or personal responsibility, or technology, or politics — and sometimes they are partly right. But partial truth becomes dangerous when it hardens into total explanation.

American healthcare is full of people with strong opinions and very little shared map.

That is not a moral failure of the public.

It is a design failure.

A democracy cannot redesign healthcare if its people have never been taught how healthcare actually works. A person cannot exercise autonomy inside a system whose rules remain hidden. A family cannot make wise decisions when the architecture only becomes visible after the bill arrives.

This is why the language matters.

The person at the center of health is not merely a patient, consumer, member, or user.

That person is the Health Principal.

Principal, not principle.

The Health Principal is the person whose life, body, data, culture, finances, family, risk, and future are at stake. The Health Principal is not expected to become their own doctor, economist, lawyer, engineer, or pharmacologist. Expertise still matters. Clinical judgment still matters. Institutions still matter. AI will matter profoundly.

But authority has to be properly ordered.

The person whose life is at stake cannot remain a bystander to the system acting upon that life.

This is not a new lesson. History teaches it again and again, though never in exactly the same form.

The U.S. Holocaust Memorial Museum defines a bystander as someone present but not taking part in what is occurring, while also warning that the category can flatten very different degrees of power, knowledge, fear, responsibility, and complicity. We should not use history carelessly. We should not compare unlike harms as though they are the same. But we can still learn from the moral category: passivity is not always neutral when harm is escalating.

During the AIDS crisis, activists refused the role that institutions had assigned them. They did not wait quietly for research, regulation, and access to move at the ordinary pace while people were dying. They forced a different conversation about urgency, trial design, drug access, and the standing of patients in the production of medical knowledge. FDA’s own historical account notes that ACT UP protests publicized patients’ concerns, pushed FDA toward accelerated approval regulations, and helped expand access to promising therapies.

The patient safety movement taught a related lesson from another direction. When To Err Is Human reframed medical error as a systems problem, it challenged the comforting fiction that harm was only the result of bad individuals. The National Academies describes the report’s argument plainly: the problem was not “bad people in health care,” but “good people” working in systems that needed to be made safer.

Often, harm is the output of architecture.

Good people can work inside bad systems. Bad systems can normalize preventable harm. And once we see the architecture, we are no longer innocent bystanders to its consequences.

That is where we are now.

We are standing in a crowded car, feeling the pressure rise, pretending not to notice how close we are to the next escalation.

We argue about healthcare constantly, but we do not understand it together. We argue about cost without understanding incentives. We argue about access without understanding coverage. We argue about prevention while paying mostly for rescue. We argue about AI without first deciding who should have authority, who should be accountable, and whose life the system is meant to serve.

The stakes are no longer abstract. Chronic diseases are the leading causes of death and disability in the United States and remain major drivers of national healthcare costs. KFF reports that four in ten U.S. adults have some form of debt due to medical or dental bills. The Commonwealth Fund’s 2024 comparison of ten high-income countries ranked the United States last overall in health system performance, despite spending more than its peers.

So this July, around a personal milestone, my colleagues and I will begin something we have been building toward for a long time: a free public course called We Are Not Sheep: Authority and Autonomy in Healthcare.

The timing matters to me personally, but the work belongs to something much larger.

It belongs to every person who has sat in an exam room with questions they were afraid to ask. Every person who has opened a medical bill they could not decode. Every family member who has watched someone decline while the system moved too slowly. Every clinician who knows the system is broken but has never been given a structural vocabulary to explain why. Every employer, teacher, mayor, caregiver, student, journalist, policymaker, board member, and citizen who senses that health is no longer simply a private matter.

Health is the infrastructure of human freedom.

The course will not turn anyone into a doctor, economist, lawyer, technologist, ethicist, or policy expert. It will not pretend healthcare is simple. It will not tell people what to think.

But it will offer a map.

Because autonomy cannot survive in confusion. Authority cannot exist without standing. And people cannot become Health Principals if they are never taught to see the system acting on their lives.

That is the deeper meaning of “We Are Not Sheep.”

It is not an insult. It is not a demand that people become louder, angrier, or more certain. In fact, it asks almost the opposite.

It asks us to become less certain where we are uninformed. Less passive where we have standing. Less afraid to ask basic questions. Less willing to accept confusion as the normal price of care. Less willing to mistake compliance for trust.

The opposite of a bystander is not a hero.

The opposite of a bystander is a principal.

Someone with standing. Someone with authority. Someone who understands that their body, data, money, family, culture, risk, and future are not passive objects in someone else’s system.

The lesson of the subway was not that everyone should become heroic. It was smaller and more demanding than that.

When a trajectory is moving toward harm, someone has to interrupt it.

American healthcare is moving toward harm. The cliff is visible now. The pressure is rising. The crowd is uncomfortable. The old instructions — wait, comply, appeal, endure — are no longer enough.

We do not need a nation of heroes.

We need a nation of Health Principals.

We are not sheep.

We are not bystanders.

And we do not have to keep riding silently towards the cliff.

Hope is not a strategy.

Wendell Potter’s piece this week, building on the American Economic Liberties Project’s coalition work and the bipartisan Warren–Hawley Break Up Big Medicine Act, is doing necessary work. The diagnosis is right.

UnitedHealth, CVS, and Cigna have spent a decade buying every link in the supply chain — the insurer, the PBM, the wholesaler, the physician group, the pharmacy, the surgery center. The federal government now funnels over a trillion dollars a year to insurance contractors that have failed at their basic mission. Independent practices and pharmacies are closing in waves. Patients face delays, denials, and rising out-of-pocket costs. A criminal investigation of UnitedHealth for Medicare Advantage fraud is underway. None of this is in dispute.

So I want to be clear at the outset: the people pushing this bill are doing necessary work, and I support the fight. The conflicts of interest at the heart of insurer-acquired provider networks are exactly what they describe. The break-up is overdue.

But the bill, as drafted, has two problems that need to be named — one immediate, one structural. And once we name them, a larger question opens up that the bill, by design, cannot reach.

The first problem: not all integration is the same.

The Break Up Big Medicine Act makes it unlawful to commonly own a “provider” and an “insurance company.” It defines “health plan” but not “insurance company” — which means the legal scope is ambiguous, and likely sweeps in physician-led integrated systems with subsidiary health plans.

That’s Cleveland Clinic. That’s Mayo. That’s Geisinger, Kaiser, Intermountain, UPMC, Sanford. The exemplars. The systems that pioneered salaried-physician care teams precisely because the alternative was the fee-for-service mill that produced the consolidation crisis in the first place.

These are not the targets. They are, in many ways, the model.

A Cleveland Clinic cardiologist on salary, working in an integrated team with a unified record and capital for innovation, is the most autonomous physician in American medicine when the metric that counts is *clinical autonomy at the point of decision*. An “independent” practice physician strangled by prior authorization, declining reimbursement, and 800-line EHR templates is not free. They are atomized.

The bill conflates two fundamentally different things. Optum buys physicians to harvest premium dollars and pad diagnosis lists for Medicare Advantage. Cleveland Clinic salaries physicians to align them with patient outcomes. Both are “vertically integrated.” Only one is the disease.

The disease is *integration in service of financial extraction*. The cure includes *integration in service of clinical outcomes*. The bill, as written, can’t tell them apart.

The second problem: 80% is not a defection. It’s a verdict.

The bill’s findings note that almost 80 percent of physicians now work for a corporate parent, and frames this as evidence of capture. It is — partly. But it is also a verdict on what the practice of medicine has become.

Independent solo practice did not collapse only because it was crushed by monopolists. It collapsed because the regulatory, technological, and financial overhead of running a small practice in the United States now exceeds what any reasonable person should bear: EHR mandates that consume hours per patient, prior authorization that requires dedicated staff, billing that requires consultants, malpractice premiums, declining reimbursement, capital costs.

Most of the physicians now working for corporate parents did not lose a fight. They made a choice — and many of them prefer integrated systems with infrastructure, support staff, predictable schedules, and capital for the tools they need. The Marcus Welby fantasy is not what they want restored.

“Restoring power to providers” is a beautiful slogan that makes sense only if you misunderstand who providers are now and what they actually want. The fight worth having is not for solo practice. It is for clinical autonomy inside whatever structure physicians work within. That is a different fight, and it requires different tools.

The deeper question: configuration is not architecture.

Now the larger point. Suppose the bill passes in a tightened form that hits Optum, CVS, and Cigna without sweeping up Cleveland Clinic. Suppose the structural separation works. Suppose insurance is once again separate from provision, PBMs once again separate from pharmacies, wholesalers once again separate from specialty practices.

What have we got?

We have the same financial chassis: insurance pays providers per procedure, in a system designed to pay more when Americans get sicker. We have the same hospital revenue model: filled beds, performed surgeries, completed encounters. We have the same employer-based coverage architecture, inherited from a 1942 wartime wage freeze. We have the same paradigm that treats “health” as the absence of currently-billable disease.

The conflict of interest at the entity level is removed. The misaligned incentive at the system level remains.

Antitrust addresses *configuration*. It does not address *architecture*. The Break Up Big Medicine Act is a defensive structural reform — necessary to clear the field of monopolistic distortion. We also need an offensive structural reform — one that builds something different in the cleared space. These are different fights with different tools, and confusing one for the other is how we end up doing the easier one and calling it the war.

The Two-Option Trap, in a new disguise.

I’ve written before about the Two-Option Trap that collapses every structural critique of American healthcare into:

A) Defend the existing private system from break-up, or

B) Break it up — usually meaning replace it with a single-payer government version.

The Break Up Big Medicine fight reproduces the trap in a new form:

A) Defend Big Medicine consolidation, or

B) Break up vertical integration.

Both options accept the underlying premise that the architecture of American healthcare — sick-care fee-for-service rescue medicine paid by insurance contractors — is the thing we are fighting over. The argument is about who owns it, who profits from it, and how it is configured.

There is a third option. It is not single-payer (option B drawn larger). It is not the existing private market with sharper antitrust (a stronger version of A). It is a different architecture entirely — one in which the financial logic of the system is aligned with health rather than against it.

That is the work I have been writing about as Health 4.0.

Trajectory Engineering, briefly.

In a Health 4.0 architecture, the unit of payment is not the encounter but the trajectory. Care is organized around the path a person is on — toward cardiometabolic disease or away from it, toward functional decline or away from it, toward the trajectory that produces the next $200,000 hospitalization or away from it. The financial logic of the system rewards the bend, not the rescue.

This is not utopian. The data infrastructure exists. The clinical models exist. The integrated delivery systems that already work — Cleveland Clinic, Geisinger, Intermountain — are operating *fragments* of this architecture inside the constraints of fee-for-service. The Veterans Health Administration, with all its problems, has more longitudinal trajectory data than any private system in the country. The technology to support clinicians in trajectory-aligned decisions is here.

What is missing is not the tools. What is missing is the architecture. And no antitrust bill will produce one.

Where this leaves us.

The Break Up Big Medicine Act is necessary work. I will support a tightened version that hits Optum-style extraction without breaking up the Cleveland Clinics of the world. AELP and the people pushing this fight are allies. Wendell Potter has been telling the truth about insurance industry capture for longer than most of us have been paying attention. The structural-separation logic is sound where it applies.

But after the break-up, the architecture problem still waits. Defensive reform clears the field. Offensive reform builds something different in the cleared space. Both are needed.

The next question is not “who owns what?” It is “what is healthcare *for*?” The first question is what antitrust can answer. The second question is what Health 4.0 begins to answer.

The fight against extraction is an ally’s fight. The fight for architecture is the one I am writing about. They go together.

*Robin Blackstone, MD, FACS, FASMBS is a bariatric surgeon, past president of the American Society for Metabolic and Bariatric Surgery, and founder of the H4 Alliance. Her book on healthcare architecture, [Doctor AI: Reimagining Healthcare, Rebuilding Trust, Delivering Health 4.0](https://www.amazon.com/dp/B0F5PCS9QZ), launched April 7, 2026.)

In the past month, the largest AI company in the world published a healthcare policy blueprint. The American Medical Association continues refining its position. Epic, Oracle, and a half-dozen payer organizations have circulated white papers of their own. CMS is iterating on its Health Technology Ecosystem initiative. HHS is signaling on information blocking. Reporters are pulling apart each document to determine who benefits.

They are all asking a version of the same question: how should AI fit into American healthcare?

It is the wrong question.

The harder question — the one no incumbent stakeholder can afford to ask out loud — is whether the architecture AI is being asked to fit into is the architecture we should keep.

The shape of the current debate

The OpenAI document that prompted this past week’s reporting is, on its face, reasonable. Make health data more portable. Clarify which AI uses are low-risk enough to skip FDA review. Invest in research infrastructure. The proposals are not unhinged. As David Blumenthal, a former national coordinator for health IT, told STAT, the company is trying to sound responsible while keeping its markets open. That is what corporate policy papers do, and it is not a scandal.

The deeper observation is what almost no one is saying out loud. Every major AI policy framework now in circulation — from technology vendors, from medical societies, from federal agencies — accepts the existing American healthcare architecture as the substrate. The arguments are about how AI should be deployed inside fee-for-service reimbursement, employer-sponsored coverage, fragmented data ownership, and an FDA review process designed for medical devices when most of the people writing the new frameworks were children.

The architecture is treated as a given.

The architecture is not a given.

It is a series of accumulated decisions compounded over decades, most of them made before the technologies that now define modern medicine existed, and most of them optimized for the wrong outcomes. To accept the architecture as the substrate is to accept the outcomes the architecture produces. To accept, in other words, the failure of a healthcare system on the brink.

That is not what you will hear from the healthcare incumbents — though if you listen, their own employees say it. It is not what you will hear from the policy papers. But ask anyone in the street: the man, the woman, the child. Not one will tell you American healthcare is doing fine. They may be grateful for rescue when it comes. But somewhere inside, they know there was a moment when they could have avoided all of it.

Where the policy papers go quietly wrong

Consider the proposal — common to several frameworks, OpenAI’s included — that the FDA should clarify which AI uses are “low-risk” and therefore exempt from review.

On its surface, this is a regulatory efficiency argument. Don’t slow down innocuous tools. Reasonable.

Underneath, it inverts something foundational. It says: deploy first, establish trust later. Sort out the scaffolding once the products are in clinical use.

Trust does not work that way. It cannot be retrofitted onto a deployed system after the fact.

The patient asked to share their wearable data with a clinical model does not have the option to revisit that decision once the model is integrated into hospital workflows. The clinician asked to act on AI output does not have time, in a fifteen-minute encounter, to interrogate whether the underlying training data represented the patient sitting in front of them.

A serious AI healthcare framework would put trust scaffolding before deployment, not after. It would treat patient data sovereignty, clinical accountability, and algorithmic transparency as preconditions rather than refinements. It would name explicitly who is the principal and who is the agent in any AI-mediated clinical decision. The principal is the person whose life and health are at risk. Not the institution that captured the data. Not the company that built the model. Not the human health ally or physician working with the principal.

None of the current frameworks does this. They treat trust as an output to be optimized rather than the input that determines whether the system functions at all.

Why the architecture cannot absorb AI safely

American healthcare has two operating modes. One is rescue: high-cost, late-stage intervention after a body has already failed. The other is the simulacrum of prevention — fifteen-minute visits, billable codes, a system organized around episodes rather than trajectories. They look like different modes. They are not. Both are billing models dressed as medicine. The combination produces the worst aggregate health outcomes of any wealthy country, at roughly twice the per-capita cost.

AI deployed inside this architecture amplifies whatever the architecture already does. If the architecture is built to bill for sickness, AI gets very good at billing for sickness. If the architecture treats clinical data as the property of the institution that captured it, AI gets very good at extracting value from data the patient did not know was being used. If the architecture has no mechanism for longitudinal accountability across a human life, AI introduces none either.

The technologies are not neutral. They take the shape of the container.

The third option

There is a different framework available, and it begins from a different question. Not how AI should fit, but what architecture would make AI’s deployment a net good for the people it claims to serve. I have spent the last several years building it out, clinically, structurally, in policy terms. I call it Health 4.0. The name is the least interesting part. The architecture is the point, and the architecture is what the rest of this work will be about.

The structural features are not science fiction. Patient sovereignty over health data — not as a portability checkbox but as the legal and technical foundation. Longitudinal accountability — a system organized around the trajectory of a life rather than the episodes of a billing cycle. Trust as scaffolding, designed in, not legislated after deployment.

And in every clinical encounter, a triangle: the patient as principal, the clinician as human partner, the AI as ally to both. The patient is the one whose life is at stake and whose data the system runs on. The clinician is the one whose judgment and presence cannot be automated away without something essential being lost. The AI earns its place by serving the other two, not by replacing either or extracting from both. None of the three corners can be compromised without breaking the structure.

The components exist. The intellectual work is done. What is missing is the political will to ask the architecture question instead of the fit question — and the recognition that the longer we delay asking it, the more deeply AI will be cemented into the container we already have.

The window

The policy frameworks being written this year will determine what AI in American healthcare looks like for the next twenty. The incumbents understand this. They are writing fast. They are writing with their lawyers, their lobbyists, and their market positions in mind, and that is exactly what one would expect them to do.

Look at what is actually on the table. The OpenAI blueprint accepts the architecture and asks for friction reduction. The AMA’s recent positioning accepts the architecture and asks for clinician protections inside it. CMS’s Health Technology Ecosystem initiative accepts the architecture and asks for better plumbing. HHS’s information-blocking signals accept the architecture and ask for cleaner pipes. None of these documents is wrong on its own terms. All of them are answering the fit question.

The question is whether anyone outside the incumbent perimeter is writing with the same urgency, on the same timeline, with the same clarity about what is actually at stake.

Fit, or replace.

The fact that almost no one in current policy circles is naming the second option is itself the most important thing to notice. Have they chosen the easy path or the right one?

Robin Blackstone, MD is a physician and the author of Doctor AI: Reimagining Healthcare, Rebuilding Trust, Delivering Health 4.0, (April, 2026). This essay is the one of a series on the architecture of American healthcare and what AI in medicine should actually become. Subscribe?