Zuckerberg Promised an AI That Works for You, Not for Meta

On Monday Mark Zuckerberg published a 6,500-word essay on Meta’s website built around the AI assistant he already lives with. It monitors his sleep and his training and coaches him on both. His daughter loves to bake, so it plans a recipe for the two of them each weekend, orders the ingredients, and offers tips while they cook. Soon, he writes, everyone will have one, working around the clock on "your relationships, health, career, finances, home management, hobbies, and more."

Longtime readers have met this assistant before. We opened this series with Courtne Smith, who grew up with Drake, knew his tastes and his patterns better than anyone, and handled what he needed before he asked, all the way to a $400 million career. That is what Zuckerberg is describing, a Courtne for everyone, and it is exactly the future we expect.

What’s missing from the essay

Buried in the technical middle is the line that matters most. Meta views alignment, Zuckerberg writes, "as ensuring that agents share a person’s goals and values, not our company’s." Alignment, as he uses the word, comes down to who your AI assistant actually works for, and his answer is you. None of his rivals have said it that plainly, and it is the correct standard for an assistant that will manage your health, your money, and your daughter’s weekend baking.

Courtne’s loyalty was never in question, because Drake was the one paying her (and they were friends since childhood). Meta’s assistant is built and run by Meta, and everything it learns about you accumulates on Meta’s servers. Zuckerberg promises that nobody else will be able to read your information, and pledges a private mode where even Meta cannot see it. That covers who can look at what the assistant knows. Nowhere in 6,500 words does he say you can inspect it, correct it, keep your own copy, or move it to a better assistant next year. As long as Meta holds everything your assistant knows about you, its loyalty isn’t all that clear.

The cornerstone of personal AI

Everything an assistant learns about you over years of use, from your habits to the reasoning behind your decisions, is your context, and it is the piece the whole system runs on. Everything else is becoming a commodity. Meta alone is spending $145 billion on AI this year, and every dollar buys the same things its rivals are buying: chips, data centers, and models that leapfrog each other every few months. What can never be bought or rebuilt is the accumulated knowledge of your own life. That is the cornerstone of a personal AI, and it is the only part of the system an ordinary person can own.

That ownership is the only way you control a personal AI. You decide what goes into your context library and what never does. You decide how much each assistant is allowed to see, the baking recipes but not the medical history. And if you want to move to a better assistant next year, the whole library moves with you, so the new one starts its first morning already knowing everything the old one knew. Zuckerberg asks you to trust that your assistant shares your goals, and a library you own is how you make sure of it.

Keep the library on your side

Meta will ship the assistant Zuckerberg described, and it will be very good, until a version from OpenAI or Google or a lab nobody has heard of yet is better. The assistants will keep trading places for years, but your context is the part that makes them work for you. The essay describes the right destination, an assistant that takes your side, but leaves out how you get there.

The only way to accomplish it is to own your context library, held by you and answering to nobody else. Building that is the problem we work on at Bot Food, and the right to everything ever collected about you is what the Open Memory Alliance exists to establish. Zuckerberg has already listed everything this assistant will need to know about you. The only question he left open is whether any of it will belong to you.