Almost nobody funding the rebuild of health came out of medicine. This is why that keeps happening, what the retreat market looks like once you map it properly, and the thing about our own build that most companies here prefer not to mention.
A body scanning company founded in 2018 by Daniel Ek together with an engineer, Hjalmar Nilsonne, took a further $700 million this July and now sits at a valuation approaching $7 billion. Mark Zuckerberg is among the personal investors. Last November, Function Health closed $298 million at $2.5 billion; its founders are a technologist, Jonathan Swerdlin, and a physician, Mark Hyman. Midjourney, better known for generating pictures, is putting a body scanner inside a spa due to open in San Francisco. Then, in April, came the most ambitious version of all: $500 million over five years from Biohub, the organisation Zuckerberg set up with the paediatrician Priscilla Chan, for something called the Virtual Biology Initiative. The aim is AI that can simulate a human cell precisely enough to anticipate how illness starts and how to head it off. What they say they are after is helping to prevent or cure every disease there is. Every one of them.
Look at who these people are. Software, hardware, consumer products. Almost none of them medicine. I do not think that is a passing enthusiasm. I think something is being corrected.
This is an educational and strategic perspective, not personal medical advice. The views are the author's own and not statements by Atlas Cove Lda.
The half that medicine never took on
At keeping a person alive, medicine is remarkable. At keeping a person well, considerably less so.
In 2024 JAMA Network Open carried a cross-sectional study out of the Mayo Clinic that put a figure on this across 183 WHO member states. Take a life anywhere on earth and, on average, 9.6 years of it are lived in poor health rather than good. Across 2000 to 2019 that distance stretched by 13 percent. Nowhere is it wider than in the United States, at 12.4 years. For women it runs 2.4 years wider still, and what drives that is a heavier burden of chronic, noncommunicable disease.
Roughly a decade of an ordinary life, spent unwell. Length of life improved. Quality of it did not keep pace.
No clinician is to blame for that. The apparatus was designed around a different task, which is finding illness and dealing with it once present, and at that task it performs. Everything upstream of the first breakage, all the strength and sleep and recovery and capacity, falls outside the design. It has no department. It has no billing code. Responsibility for it belongs to nobody. A problem that size with no incumbent standing over it is exactly what someone trained in software learns to look for, which is why they keep arriving.
A market with two ends and a hole between them
But why retreats?
Because of the shape that emerged when we worked out how this market is actually laid out. By the Global Wellness Institute's count, tourism in this category came to $894 billion during 2024. Vast. And piled almost entirely at the two poles.
The cheaper pole sells a week of yoga or pilates, one facilitator, a single modality. Frequently charming. Seldom personal. Measured, essentially never. You return rested, and a fortnight later nothing about you has moved.
The costly pole sells the medicalised estate. Diagnostic clinics in Germany. Ayurvedic hospitals across India. Imaging suites attached to a week that costs more than a car. Their equipment and their reading of it are first rate. Along with the equipment they have inherited the same blind spot medicine has: outstanding at naming the problem, threadbare on whether the person does anything about it once they are home again.
Price is not really what separates the two poles from the middle. Purpose is. Very few people have built a stay whose product is the doing: numbers taken from you on arrival, a plan written specifically from those numbers, the work carried out with someone watching, the identical measurements taken again before departure, and then a year of somebody keeping you on it.
Which is a peculiar thing to be missing, because the doing is where the whole difficulty lives. Anyone who can pay at either pole generally knows the theory already. The podcasts have been listened to. There may be bloodwork. There is occasionally a scan of the entire body. The missing piece is never the knowledge. Take a plan that succeeds ninety-nine times out of a hundred, leave it undone, and its success rate is zero.
Why it is the two of us
My biography is not the argument here, so I will keep this short. Still, you are entitled to know what makes two people out of tech this invested.
My brother Tom and I have each carried chronic illness through nearly all of our adult years. In my case: Hashimoto's, PCOS, endometriosis. In his: Gilbert's syndrome, cold urticaria, migraine, raised autoimmune reactivity, inflammatory joint pain. Nothing on either list will kill us. Everything on both lists sets the terms of a week. And a fair number of them sit in the poorly researched corner of medicine, which, predictably, is the corner where the conditions affecting mostly women ended up.
No plan was ever handed to us. What we have was put together slowly, from the literature, from getting it wrong, and from a great many consultations that concluded in a shrug and a prescription pad. The knowledge of how to make a body carrying a chronic condition strong and capable and genuinely energetic is knowledge we had to source ourselves. None of that is an accusation aimed at anyone who treated us. It is simply what the apparatus was and was not built to do.
The week we run in Sesimbra is that self-assembled thing, converted into something purchasable. Six days beside the Arrábida coast. Day one is measurement, taken on validated instruments we carry in ourselves: strength, fitness, sleep, recovery, blood pressure, stress. By the following morning those readings have become a protocol belonging to you. The work happens with a coach present each day. Come day six, everything is measured a second time and placed alongside day one, which means any movement is something you read for yourself instead of something we characterise for you. After that, the twelve months that follow: a guided month of return, included in the price, plus a membership that keeps both your numbers and your protocol alive in the intervals.
Any week that collapses on contact with a normal calendar was only ever theatre. We don't diagnose, we build.
The sentence this industry avoids
An AI generates our protocols.
I would rather write that sentence than let it be discovered, and most of this sector will not write it. What is ours is the method: the reasoning that decides what a particular set of readings calls for, assembled from the literature and from something like twenty years spent managing our own conditions. That reasoning has been encoded, and what the encoding does is convert your day-one assessment into your week. A doctor then goes over the result remotely, sets it against your medical history and your readings, and approves it before you ever see it. No guest has anything run on them that a clinician has not put their name to.
For a while we did not talk about this, and I now think the silence was a mistake. The reason for it is obvious enough: tell somebody an AI wrote their health plan and they flinch, and the flinch is reasonable. But there is no coherent position in which we argue publicly that AI belongs in health while being cagey about our own use of it. That stance falls apart at the first serious question.
So, the argument, put as squarely as I can. The best evidence available to me comes from Sweden, from the MASAI trial: randomised, controlled, more than 105,000 women, the first study of its kind, with the final results carried by The Lancet in January. Radiologists reading mammograms with AI support found 29 percent more cancers. False positives did not rise. The burden of screen reading dropped by 44 percent. Fewer aggressive cancers turned up in the intervals between screening rounds.
What that trial refrained from doing matters more than any of those figures. The radiologists stayed. They were given a better instrument, and the two together outperformed either on its own. We copied that arrangement and shrank it to fit ten rooms: something that is not fatigued at eleven at night, and does not miss a contraindication four pages into a history, supervised by a person who carries responsibility for what comes out. It is worth registering that Biohub, which is the boldest AI undertaking in this entire field, has a practising paediatrician co-leading it. Wherever this is done seriously, a clinician is inside it.
What arriving from outside is actually good for
It is not a belief that you could do everybody else's job. It is clarity about which job is yours.
The system is ours, and so is being present behind it. Every cohort is hosted by Tom and me in person, together with specialists in whichever disciplines that week rests on, for the plain reason that a method with nobody in the room is a document. Anything beyond what the system can do, and beyond what the two of us can do, goes to somebody who trained for it.
The other thing we imported is a reflex that medicine's structure tends to flatten: ship it, measure it, revise it, repeat next week. Scanners are where Ek applies it. Cells are where Biohub applies it. Bloodwork is where Function applies it. Ours goes on the least glamorous stretch of the whole route, the one running from knowing to doing, and my claim is that this stretch is where health loses the most money. It is not the scan nobody ordered. It is the follow-through nobody arranged. The cost of that stretch is something I have written about before, in terms of energy and mental load.
Everything I am not yet able to claim
Argue for measurement and then get vague about your own, and you have earned every bit of scepticism coming. So here is that part, at the length it deserves rather than the length that flatters us.
Nothing here is proven. The first full cohort arrives in September. Design-wise, the same markers get captured at both ends of the week on the same instruments, and each guest leaves holding their own pair of columns. Outcome data, though, is the one thing I have nothing of, since none has been collected at any scale that would mean anything. Whatever figures appear on our method page are estimates taken from research into comparable programmes, and the page states that they are. I would sooner set that down myself than have somebody turn it up and conclude we were counting on them not checking.
It is also worth being exact about what those two columns can and cannot support. Six days will move certain things and comes nowhere close to moving others. A comparison between day one and day six describes a change in one person across six days, uncontrolled and unblinded, which makes it a description rather than a finding. It cannot tell you how much came from the protocol and how much came from a week of decent sleep, decent food and no commute. That limitation is the reason the following year exists at all, and the reason the claim I will stand behind today concerns how the thing is designed rather than what it achieves.
There is one signal I do have, and it is what persuaded me the hole in the market is real rather than something I had convinced myself of. The waitlist passed a thousand people before any full-price cohort had happened. As proof that the method works, that is worthless. As evidence, it says a great many people have already bought at the bottom of this market and at the top of it and are still hunting for a thing that neither pole stocks.
Which splits the honest position cleanly in two. On the category, I will argue today and I will argue hard. On outcomes, I will have numbers by next summer, drawn from cohorts running every month from November. I intend to be held to that.
Starting an argument with an industry from inside ten rooms is a strange posture, so let me be precise about the scale. There is a single site, on an estate belonging to a partner, and it is deliberately the smallest thing we could have built. What we intend is a network across Europe of sites we run ourselves. Beginning this small is a choice: a method that has not yet met real cohorts has no business being replicated into five buildings, and the constraint of ten rooms and one cohort each month is precisely what keeps the measuring honest. Evidence first. Expansion afterwards.
The Zuckerberg schedule for curing all disease extends to the close of the century, which is a well-mannered way of ensuring no living person ever audits it. What we are promising covers six days and then a year, and that is far harder to say out loud, because anyone can go and check it.
The direction is the thing I hold with confidence. People the existing version of health failed are rebuilding it, with instruments that version was too slow to take up, and with clinicians positioned as the check rather than the gate.
So let me put the question to you. Does your own gap sit on the knowing side or the doing side? And answering honestly, how long has it been since more knowledge changed a single thing?
Questions we get asked
Is Atlas Cove a clinic?
No. We don't diagnose, we build. The diagnostic end of this market is genuinely good at establishing what is wrong, and our work begins after that point, at the question of whether anything is then done.
Is an AI really generating the protocols?
It is, and we would rather state that than be slippery about it. The reasoning encoded in the system belongs to us. A doctor reviews what comes out remotely, against your medical history and your day-one readings, and signs it off before it reaches you.
Is there proof that the week works?
Not yet. September brings the first full cohort, and the numbers currently shown on our method page are estimates drawn from research into comparable programmes, labelled as estimates. Data should exist by next summer.
What does a waitlist of a thousand people demonstrate?
About the method, nothing whatsoever. What it indicates is that a lot of people have already paid at both poles of this market and have not stopped looking.
Why does the year afterwards matter more than the week itself?
Because the week is the easy half. A protocol that cannot withstand an ordinary calendar was never going to alter anything, which is why a guided month of return and a membership holding your numbers between visits are part of the thing you buy rather than something sold to you later.
Sources
Garmany A, Terzic A. Global Healthspan-Lifespan Gaps Among 183 World Health Organization Member States. JAMA Network Open. 2024;7(12):e2450241. DOI: 10.1001/jamanetworkopen.2024.50241
Global Wellness Institute. Global Wellness Economy Monitor 2025. Miami, November 2025. Wellness tourism figures for 2024.
Gommers J, Hernström V, Josefsson V, et al. Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study. The Lancet. 2026;407(10527):505-514. DOI: 10.1016/S0140-6736(25)02464-X
This is an educational and strategic perspective, not personal medical advice. The views are the author's own and not statements by Atlas Cove Lda.
Lisa Wuerden · Co-Founder
Co-founder, brand and product