The way AI is entering health isn’t the way you think.

Here is a headline that sounds like science fiction and isn’t: Stripe, Anthropic and the OpenAI Foundation are among the funders putting half a billion dollars behind a new nonprofit whose stated goal is to end the common cold — and eventually to get rid of respiratory viruses altogether. The organisation is called Intercept. Bill Gates is in. So, reportedly, are a handful of traders from a quantitative hedge fund. It is a genuinely serious effort, led by the Stripe executive who previously ran a $1.8 billion carbon-removal programme.

It would be easy to file this under ‘tech billionaires buy themselves a moonshot’ and move on. We don’t think that’s the right read. The more useful thing to notice is what this story tells you about how AI is actually entering health — because it is not the way the wellness internet keeps telling you it will.

what Intercept is actually doing

The pitch is straightforward and, on the evidence, fair. We spend roughly 5% of our lives fighting colds and flu, and almost nobody works on preventing them — because the sniffles are caused by more than two hundred different viruses, and no drug company can make the maths work on a vaccine for any single one. There is no commercial incentive, so the resources never show up. Intercept exists to fund the work the market won’t.

And the toolkit it’s betting on is the interesting part. Not one vaccine, but broad countermeasures designed to work against many viruses at once: RNA drugs, engineered antibodies, and computational protein design — including the idea of virus-grabbing proteins you’d spray into your nose to catch an infection before it starts. Alongside the biology, there’s a deeply unglamorous bet on the air itself: large-scale ultraviolet cleaning systems for schools and offices, the way a city treats water before it reaches your tap.

Computational protein design is the phrase to sit with. The reason this is fundable now and wasn’t a decade ago is that AI made designing proteins tractable. That is AI entering health — but notice where it’s entering. In a lab. In an air-handling unit. In a nasal spray that doesn’t exist yet. Nowhere near your phone.

the two altitudes of AI in health

It helps to think of AI entering health at two completely different altitudes, because they get constantly confused — usually on purpose, by people selling you something.

1. The infrastructure layer — capital, drug discovery, protein design, public-health systems, the air in buildings. This is enormous, slow, expensive, and almost entirely out of your hands. Intercept lives here. You cannot DIY a broad-spectrum antiviral, and you shouldn’t try. When this layer works, it works for everyone at once, quietly, in the background.
2. The personal layer — the data you already generate, the decisions you make weekly, the questions you ask before and after you see a clinician. This layer is small, fast, cheap, and almost entirely in your hands. It’s where your sleep, your labs, your training and your symptoms actually live. No nonprofit is coming to run it for you.

The mistake nearly everyone makes is assuming the first layer will eventually hand them the second — that once the big labs are ‘in health’, a magic copilot will descend and manage your body for you. It won’t. The infrastructure layer is busy ending the common cold. It is not coming to read your bloodwork.

“AI entering health from the top doesn’t reduce the value of understanding your own. It raises it. The more capable the system gets, the more the scarce skill becomes knowing what to ask it and when to ignore it.”

why this is genuinely good news — and not the whole story

We are not cynical about Intercept. A world with fewer respiratory infections is a straightforwardly better world, and using AI-era biology to chase the diseases the market ignored is close to the best version of what this technology could do. If virus-grabbing nasal proteins and UV-clean classrooms arrive, take them. Gratefully.

But here’s the honest caveat, and it’s the whole reason we exist. Even in the best case, the infrastructure layer gives you a cleaner environment and better drugs. It does not give you judgement about your own health. It will not tell you whether your fatigue is your training load, your iron, or your sleep. It will not weigh a supplement’s evidence for you, or turn three years of scattered data into a question worth asking your doctor. That work stays yours — and it’s the work that determines most of your day-to-day health long before any nasal spray ships.

the pattern underneath all of this

Step back and the Intercept story rhymes with everything else happening in AI right now. The capability keeps moving up the stack — into the labs, the infrastructure, the tools you already use — and the thing that stays scarce, the thing that keeps its value, is human judgement. Knowing which problem to point the machine at. Knowing when its confident answer is wrong. Knowing what only a qualified human should decide.

That’s as true at the personal scale as the planetary one. The individual version of ‘back a nonprofit to end the common cold’ is much humbler: read your own data well enough to make better decisions and ask better questions. It doesn’t cost five hundred million dollars. It costs a bit of literacy and a refusal to outsource the one layer that’s actually yours.

1. Take the infrastructure wins when they come — vaccines, cleaner air, better drugs. You don’t have to build them to benefit from them.
2. Own the personal layer deliberately. Keep your health data somewhere you can actually read it, not scattered across apps that each show you a sliver.
3. Use AI as a thinking partner for that data, grounded in your own records — to understand and to prepare questions, never to diagnose. The human in the white coat is still the point.
4. Treat judgement as the skill worth compounding. The tools will keep changing. Knowing what to ask, and what to ignore, is the part that lasts.

Stripe and the AI labs are buying a quieter world, one virus at a time, and they may well succeed. It’s a good thing to root for. Just don’t mistake it for the thing that manages your health. That layer was always going to be yours — and the smarter the machines around it get, the more that ownership is worth.