Your lived life is your differentiator. Your uniqueness is the one thing that can’t be downloaded.
There is a cost to everyone using the same few AI models, and for once it has nothing to do with privacy. It is this: we are starting to sound the same.
When a handful of cloud models are trained on everyone’s data, and everyone then leans on those same models to think, write and decide, the output drifts toward an average. The model is trained on the crowd, so it answers as the crowd. Ask a thousand people to write the difficult email, name the strategy, or summarise the book – route every one of them through the same intelligence – and the thousand answers converge on the median. Fluent, competent, interchangeable. The more useful the tool becomes, the harder it pulls toward the mean – a rising tsunami of mediocrity: competence without character, at scale.
Call it homogenisation, and notice that it is the quiet flip side of a convenience nobody wants to give up. It erodes the only thing that was ever yours to trade on: the specific set of things you have read, said, heard and lived through, in a combination nobody else holds.
So the question for a person in the AI age is not only is my data private? It is sharper than that. How do I keep my edge when everyone is renting the same averaged mind?
The paradox, reframed for a person
Satya Nadella recently set out what he called the Reverse Information Paradox. His frame is the enterprise buyer, but read it as a person and it lands harder, not softer. “In the AI age,” he writes, “the buyer risks giving away knowledge, just in order to use what they bought.” You end up paying twice: “once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful.”
How does the knowledge leak out? As exhaust. The prompts you write, the tools your agents reach for, “and especially,” Nadella notes, “the corrections people make when the model is wrong.” Every time you tell the model no, not like that, like this, you are teaching it – trace by trace, correction by correction – the exact judgement that made you worth listening to. For a company that is trade secrets. For a person it is more intimate than that: it is taste, memory, the hard-won feel for your own life.
His conclusion is the part worth pinning to the wall. “In consuming intelligence, you are creating intelligence. And what you create should belong to you.” He calls it your particular intelligence, in Hayek’s sense: the knowledge of time, place, and circumstance that no one else can hold. That is not a description of an enterprise. That is the definition of a person. Your particular intelligence is your lived life.
Nadella draws one more line that maps cleanly onto what we build. Distinguish, he says, the veteran from the generalist: the accumulated, hard-won capability that is yours, versus the general-purpose model you happen to be renting this year. The test is simple – if the generalist is taken away tomorrow, does the veteran stay with you? Rent your intelligence from a cloud and the answer is no; when the model changes, your accumulated context resets to zero. Keep the veteran on your own machine and the generalist becomes what it should always have been: swappable. Ostler is the veteran that knows you. The frontier model underneath is a part you can replace.
A right principle from an uncomfortable source
Around the same time, Palantir published a nine-point “AI sovereignty” manifesto. I’ll be candid: I find the company, and its leadership, hard to admire, and there is a heavy irony in a business built on mass data-analysis lecturing the rest of us about keeping our data to ourselves. But a principle can be correct even when it arrives from an uncomfortable messenger, and two of the nine survive the translation from institution to individual almost word for word.
The first: “Sovereignty is the precondition for choice. Relinquishing sovereignty transfers the future choices of your institution to others, who are likely to exploit it for their gain and your loss.” Swap institution for life and it is simply true – the choices you give away are the ones someone else gets to make for their benefit, not yours.
The second speaks directly to homogenisation: “Data retention is your treasure. Transfer it at your own peril. Your ability to win is dictated by your ability to recognize and use your unique edges, and you keep winning by compounding the underlying data to generate new insights.” Read for a person, your unique edges are your specialist knowledge and your lived experience, and compounding the underlying data is exactly what a second brain does: it gets sharper the more of your reading, your conversations and your messages it holds. Hand that data to someone else and, in the manifesto’s own terms, you hand over both your existing winning plays and the means to make new ones. Their blunter formulation is that controlling your weights is controlling your fate, because weights are the distilled form of hard-won, accumulated knowledge. A person’s hard-won knowledge should not be distilled into somebody else’s weights.
Why the people saying this can’t sell you the answer
Here is the twist worth sitting with. These are enterprise arguments, made by enterprise and Big-Tech players, and they are correct. But the incumbents making them structurally cannot offer you the individual version, because the shape of their business pulls the other way.
This isn’t an accusation of bad faith. The serious cloud providers offer opt-outs, enterprise data-boundaries and contractual controls, and many take them seriously. The problem is more structural than any single policy toggle can reach: to use a rented cloud mind, your data has to leave your machine and be processed on theirs. However good the controls, you are trusting a boundary you don’t own, operated by a business whose revenue depends on your continued use of it. On-device doesn’t improve those settings – it changes the category. There is no boundary to trust because nothing crosses it, and no meter running because there is no server on the other side. Ostler can make that promise for one unglamorous reason: it runs on your Mac, and there is no cloud to feed.
You can watch the ground shift in real time. As I write this, an open-weight model from a Chinese lab – Kimi K3 – has landed trading blows with the best closed models at a fraction of the cost, and the closed labs did not exactly welcome it. OpenAI’s own head of strategic futures branded open-weight models “inherently decelerationist” and mused about using regulatory uncertainty to scare enterprises away from them; the White House’s AI czar shot back that the leading labs “want the government to eliminate their open-source competition.” Strip out the politics and the signal is plain: the model itself is becoming a commodity, and the people who sell you access to one know it. When the generalist is nearly free, the only thing left worth owning is the veteran – the particular intelligence of your own life. The incumbents are fighting to keep the generalist scarce precisely because they can feel the value migrating to whoever holds the context.
Sovereignty is going mainstream. That is a tailwind and a countdown.
“Own your data” is going mainstream fast, and that cuts both ways. It legitimises the idea – helpful. It also starts a clock. Apple and Microsoft will happily adopt the language of sovereignty while remaining, in every respect that matters, the hyperscaler. Expect “your data never leaves” on a great many billboards belonging to companies whose revenue depends on it quietly leaving.
So the durable point was never the word. It is the depth of what you actually own: cross-source, on-device context that no policy page can approximate. Not one app’s slice of you, but the whole of it – everything you have read, said and heard, compiled into a single graph that lives on your machine and answers only to you. That is the moat a marketing department can’t co-opt, because it isn’t a promise. It is an architecture.
Which brings it back to where we started. The averaged mind is coming for everyone’s output, and it will make a great deal of work faster and more forgettable at the same time. The one thing it cannot average away is the particular intelligence of a particular life – if you keep it. A second brain with access to everything you’ve ever read, said and heard doesn’t dissolve you into the crowd’s mean. It sharpens the intelligence you use with the specifics only you have.
Your lived life is your differentiator. Rent the same mind as everyone else and you give it away by degrees. Keep it on your own machine, and it compounds. That is the whole bet, and it runs on your Mac.
If that is the future you want, you can read how the architecture works here – and if you’ve already spent years feeding the cloud, you can start to take it back.
Arguments, corrections and disagreements – [email protected].