Six months after its App Store debut, Sora had half the users it peaked at, was burning a million dollars a day, and Disney found out the app was shutting down less than an hour before the public did. A billion-dollar deal, dissolved. No money ever changed hands.
This is not a story about OpenAI.
This is the story of every well-capitalized company that has ever looked at healthcare.
The Assay
In mining, an assay is the chemical test that tells you whether the ore is worth extracting. You drill, you analyze, and you get a number: the deposit is real, the concentration is meaningful, the value is there.
Healthcare has an assay result that has been cited so many times it has become part of the air. Three trillion dollars. Four trillion, depending on who’s rounding. Either way, the ore is there. Anyone who has spent a day inside the American healthcare system — provider, patient, payer — understands immediately that it is bloated, broken, and ripe for something better.
The error is not in the assay. The error is in what happens next.
An addressable market and an accessible market are different instruments measuring different things. Confusing them is how you end up raising a Series B on a problem statement and a spreadsheet, before a single clinician has touched the product. Capital allocation decisions get made at the board level. Go-to-market plans get built. Teams get hired. The assay said the ore was there. It said nothing about the cost of extraction.
The Geology Nobody Mapped
Between the assay result and a functioning mine, there are four layers that swallow companies whole.
Regulatory terrain. People describe FDA clearance and CMS reimbursement pathways as walls. They’re not walls. They’re three-dimensional obstacle courses that change shape as you move through them. The requirements for clearing a device are different from the requirements for getting it reimbursed, which are different from the requirements for getting a health system to actually pay for it on formulary. You can clear all three and still not get used. Companies hit this layer expecting a gate and find a maze that regenerates.
Workflow geology. Clinical workflows are not inefficient processes waiting to be optimized. They are load-bearing structures that evolved under pressure — decades of workarounds and institutional memory encoded into behavior that nobody can fully explain and nobody wants to break. I wrote about this directly in The Physicians Are the Lanes: the physician as bottleneck is a choice the system keeps making without realizing it’s a choice. Companies that parachute in with a streamlined solution tend to discover that what they optimized was holding something else up. You solve one workflow problem and three escalation events appear downstream.
Incentive fault lines. In consumer tech, the person who pays, the person who benefits, and the person who absorbs the friction are often the same party. In healthcare, they are almost never the same party. The hospital buys the software. The physician absorbs the training burden. The patient may or may not benefit. The insurer controls the reimbursement that makes any of it economically viable. Misread the incentive map and you will build something the buyer loves, the user resents, and the payer ignores.
The IT substrate. Legacy health IT infrastructure is not technical debt. It is the terrain itself. EHRs were not designed for interoperability. They were designed for billing. Every integration project involves negotiating with systems built in a different era for a different purpose, and any company that assumes a clean API into a major health system is in for an education. This is part of why the physician agent architecture question matters so much — without a genuine model of the operational environment, you’re building on assumptions about the substrate that the substrate will quietly disprove.
The Mine Closures
Amazon Care launched in 2019 as virtual primary care for employees, expanded to employers across the country, and shut down three years later. Haven — the healthcare venture Amazon, JPMorgan, and Berkshire Hathaway built together — disbanded after three years without a product in market. Google has announced and quietly retreated from more healthcare initiatives than most companies have launched.
These were not incompetent organizations. They were some of the most sophisticated operators on earth, with access to capital, talent, and distribution that most healthcare incumbents will never see. The assay was valid every time. They had read the ore correctly.
They just hadn’t looked at the map.
Sora gets one sentence here as the newest entry in a long line: the capability was real, the market was real, the extraction problem was fatal. And as I wrote about Eight Sleep’s recent $50M raise — detection capability without a care pathway is just an alert nobody acts on. The same logic applies one level up. A valid assay without extraction infrastructure is just a data point.
What a Mine Actually Looks Like
The companies that extract at scale in healthcare don’t look like tech companies at launch. They look boring. They run the commissary.
Epic sells workflow software to hospital administrators. Veeva sells compliance infrastructure to pharma sales teams. McKesson moves boxes. None of these were pitch-deck stories about disrupting a trillion-dollar market. They were operational bets made by people who had internalized the terrain before they started building. They shared three things: deep workflow integration, incentive alignment with the operator — not just the buyer — and tolerance for the geological pace of change in healthcare.
They did not show up with a solution and wait for the system to adapt. They adapted first, then found the leverage.
The implication for builders is uncomfortable but important. The advantage is not the ore. Everyone can see the ore. The advantage is extraction capability — and that is built inside the terrain, not on a product roadmap.
The deposit is real. The $3 trillion is real. The question was never whether the ore exists.
The question is whether you’re building a drill press or a mine.
For clinicians: the geology is not the enemy. It is the moat.
For builders: if your go-to-market assumes the terrain is flat, you have not looked at the map.
What have you seen get swallowed by the geology? Drop it in the comments.
One way to break the status quo is stop the supply. Especially in chronic disease. Treating cancer is hard, it is even harder and unpredictable with addition of comorbidity. It is more likely easier to reduce chronic disease than to reengineer healthcare. Do both.