On your side of line
Most software gets smarter by taking your life to a server. I'm building the opposite, and 2026 is the year that bet finally makes sense.
This spring, Google quietly changed one sentence.
In Chrome's settings, the description of its on-device AI used to promise that the model ran directly on your machine without sending your data to Google's servers. Then that sentence disappeared. When people noticed, Google said the architecture had not changed, that the data still stayed on the device. Maybe that is true. I have no reason to think anyone was lying. But that is exactly the problem, and it is the reason I am building what I'm building.
The trust lived in a sentence. And a sentence can be edited overnight.
The Deal
For about fifteen years we have all been living inside the same arrangement, whether or not anyone ever spelled it out for us. The deal is this: give us your life, and we will make the software smart. Every photo, every location, every message, every late-night search becomes fuel for a model running on someone else's machine. In return the product gets better. It finishes your sentences, finds your face in a crowd, knows the route home before you ask.
It is a good deal in the way that good deals usually are, which is that you do not feel the cost until much later. Because the same data that makes the software smart also makes the advertising smart, the profiling smart, the version of you that gets sold smart. And the only thing standing between you and the worst version of that arrangement is a promise, written on a settings page or buried in a privacy policy, both of which can be rewritten while you sleep. As Chrome just reminded everyone.
I do not think the people who built that world are villains. I think they made a reasonable engineering choice at a time when the intelligence genuinely had to live in a data center, and then they built a business on top of that necessity and stopped questioning it. The necessity is what I want to talk about, because the necessity is gone.
THe Line
Here is the bet I am making instead. Software can get smarter by keeping your data, not by taking it. The compounding can happen on your side of the line.
Picture an actual line drawn between your device and everyone else's servers. Almost all of modern software is designed to move the valuable thing, your life, across that line, because that is where the intelligence sits. I want to build software where the intelligence sits on your side, so the valuable thing never has to cross at all. The more of your life it holds, the more useful it becomes, and none of that has to leave your phone to make it work.
The important word in there is not privacy. Privacy is the word every company uses, including the ones whose entire business is the opposite. "We respect your privacy" is printed on the side of the machine that sells you. So I want to be precise about what I actually mean, because the precise version is the only one that is worth anything.
I do not mean "we promise not to look." I mean "the architecture means we cannot." When the model runs on your device and your data has no reason to leave it, privacy stops being a promise and becomes a property of how the thing is built. That is a different kind of claim entirely. A promise asks for your trust. A structure lets you check. You should not have to believe me. You should be able to verify me, and if one day my apps start quietly shipping your life across that line, you should be able to catch me doing it.
That is the whole difference between what Chrome had to say this spring and what I want to be able to say. They had a sentence. I want a structure.
Why Now
For years this was a lovely idea you could not actually ship. The intelligence lived in the cloud because that was the only place big enough to hold it. Anyone promising on-device intelligence in 2022 was mostly promising a slideshow.
That broke this year, and it broke fast. In a single stretch of weeks this spring, the missing pieces all arrived at once. NVIDIA shipped consumer hardware built for running serious models locally. Microsoft launched an always-on personal agent and put models that run directly inside the operating system on the machine in front of you. The open-source world shipped a capable local agent you can run on your own laptop. Apple, meanwhile, has staked its entire AI strategy on small, private, on-device models rather than the race for the biggest one. And the silicon underneath all of it, the neural chips that make local intelligence possible, now ships even in inexpensive phones rather than only the flagships.
What that means is simple and, to me, enormous. The thing that was idealism in 2023 is an engineering choice in 2026. You can now run real intelligence on the device you already own. So the old necessity, the reason your life had to travel to a server to become useful, is no longer a law of physics. It is a business model. And business models can be refused.
What I am Building
I run a small studio called Veyago. It is one person, for now, and a short list of apps held to a single standard rather than a sprawl of whatever the market rewards this quarter.
Kept is already out. It keeps the boring, important parts of your life in order, your documents and renewals and the dates that cost you money when you forget them, and it does that on your phone and nowhere else. A travel app is in the works, and it is the bigger bet, the one I will be writing about a lot here as it comes together.
Underneath all of it I am building a layer I call Veyago OS. It is a private intelligence spine that lives on your device and belongs to you, and every app I make is built to run on it. The idea is that my apps are not four separate programs reporting back to four separate servers. They are windows onto the same private intelligence, one that gets more useful the more of your life you let it hold, with the compounding staying entirely on your side of the line.
The Honest Part
I have to be straight about something, because the whole thing collapses the moment I am not.
This is hard, and I am early, and the idea contains its own trap. The instant you say "one backbone that holds everything across all your apps," you have described, structurally, the exact surveillance pattern I started this essay complaining about. A single thing that knows your whole life is the dream of every data broker on earth. The only thing that separates what I am building from that is the architecture, on-device first, owned by you, shared between one app and another only when you explicitly say so. If I ever ship the connectedness without the ownership, I will have built Google with nicer fonts, and you should say so loudly.
So I am not going to stand here and tell you I have already arrived. I have not. I am telling you the standard I am building toward and asking you to hold me to it, in public, while I build it. That is the honest version of a manifesto. Not "look what we made," but "here is the line we have drawn, and here is how you will know if we cross it."
Most of the industry is betting that your life is worth more to them on their servers than it is to you on your device. I am betting the other way, that the most valuable place for everything that makes software personal is the one place it never has to leave.
On your side of the line. That is the company. That is the whole idea.
I will be writing here as I build it, including, especially, the parts that break.