AN AI THAT LIVES ON SOMEONE ELSE’S SERVERS IS A HOTEL ROOM, AND A HOTEL ROOM CAN’T BE FURNISHED
Anyone using an AI assistant inside an app is staying in a hotel room and thinks it’s home. A hotel room can’t be furnished. What’s inside is whatever the hotel decided. And the day they give it to someone else, we pack our bags. To know where we stand, we only need to ask where our AI actually lives.
Our phone and our computer are just a terminal, the screen we talk to it through. Its body, the part that stays on and does the work, sits on the servers of the company selling it to us. We can’t get our hands in there. So the AI only reaches where the company has decided it can reach. If we want to connect it to one of our programs or to a device at home that isn’t on their list, we are stuck. It works like a closed video game: we can move around perfectly well, but only inside the map they drew.
And the room can disappear. If the company raises the price or shuts the service down, we lose the subscription and everything the AI had learned about us along with it. Our preferences and months of corrections stay on their servers. And we start over with an assistant that doesn’t know us.
A home is a body of our own, in practice. That’s a computer that is always on, at home or in the office, with plenty of video memory, the graphics card memory an AI needs to think. On it runs a program like OpenCode with a local model. A brain that thinks inside our own computer and sends nothing out. Next to it we can run Claude Code or Codex too. They’re installed on our computer and call a more powerful brain online when needed. We decide when to call it. Or the software coordinating everything decides, case by case, whether the home brain is enough. It’s called an orchestrator. And that’s the real home. The brain can be rented. But what the AI connects to and what it remembers is decided by the orchestrator, which means by us.
Compared to that, ready-made services look like toys. And I find them dangerous once we connect sensitive things to them. Like our email or our files. The catch is that without those connections they do very little. Because an assistant that can’t read our email or see WhatsApp can only chat. So to make it useful, we end up handing our digital life to a company that runs all of it on its own servers.
Letting a company run everything used to be a huge convenience, because the alternative was learning to code. Today we tell our home AI what we need. And it writes the program. That’s why relying on others makes less and less sense. And tools built to measure, inside a body of our own, make more and more. I’ve been working this way for months, many months. And when I need something new, we connect it the same day.
So when you try an AI assistant, ask yourself right away where its body lives, and what you lose if it shuts down tomorrow. Think twice before connecting your email. And in the meantime, keep the rules and the information your AI needs to know about you in a file of your own. So you decide when to move out. Do it now.
What do you think?
#ArtificialDecisions #MCC #LocalAI #DataPrivacy #DigitalSovereignty