A personal AI that
actually works.
And how a person who is not an engineer gets their own.
Most second-brain and personal-AI tools fail for the same structural reason: they are archives, not organisms. Zero Brain is a different construction — a mechanism, a set of invariant principles, and a stated set of limits, built to be trustworthy enough to hold a life.
Declaration of interest — Agent Zero publishes field papers that carry no product and sell nothing. This page is different in kind, on purpose: it introduces a paid product. It holds the same standard — mechanism over opinion, every limit stated — but it is not disinterested, and it says so here rather than let a reader discover it.
An archive is not an assistant.
A note app is a library. Its value is storage and retrieval; the work — capturing, connecting, remembering, deciding — stays with the person. As a life grows, the library grows, and so does the burden. The tool scales the problem, not the relief.
Amnesia by design
State lives in the chat window. Close it and the system forgets who its owner is.
Confident fabrication
Asked something it does not know, the model invents a plausible answer instead of admitting the gap. In a system that holds a life, one invented fact is worse than a missing one.
No authority boundary
Either it does nothing — a chatbot — or it acts without permission — a liability. Neither is an assistant.
Zero Brain is designed against all three.
Three layers, one loop.
THE BRAIN
Plain-text notes — the single source of truth for one life.
People, health, habits, goals, tastes, memories, decisions, knowledge — and work. Plain text because it is portable, inspectable, durable, and owned. Nothing is true in the system unless it is written here.
THE ORGANISM
Always-on agents that write the brain and act on it.
Agents run continuously and do two things: capture what is lived, said and learned, and act on it — remind, draft, cross-reference, anticipate. Specialised, but sharing one brain and one set of rules.
THE WINDOWS
Messaging, chat, voice — consulted, never edited by hand.
Ask, and it answers live from the brain. It pushes what matters, when it matters. The owner is the director, not the clerk.
Figure 1 — the loop. The organism remembers and watches; the human keeps judgement and the decision.
Eight invariant principles.
These do not change with whose life the system holds. They are what separate a trustworthy organism from a pile of brittle automations.
SINGLE SOURCE OF TRUTH
One brain. If a fact is real, it lives in exactly one place.
AGENTS WRITE; THE OWNER CONSULTS
The owner becomes the director, not the clerk.
NEVER INVENT
The system never asserts a fact it did not read from the brain. If it does not know, it asks. This is the difference between an assistant and a charming liar.
NOTHING IRREVERSIBLE WITHOUT EXPLICIT APPROVAL
It proposes; anything that leaves the owner's world or cannot be undone waits for an explicit OK — always with an undo.
HUMAN VOICE AT THE ROOT
Everything that reaches the owner sounds human. No jargon, no codenames, no system leakage.
PROPOSE, DO NOT ACT
Suggestion by default; permission widens with trust.
ONE INNER FACE, MANY OUTER FACES
To the owner, a single assistant that knows everything. To the world, distinct personas per context.
PRIVACY AND OWNERSHIP BY DESIGN
The owner's brain, the owner's server, the owner's private network. The owner is the owner — not a company holding their life.
Principles 1–4 are the direct application of Agent Zero's field papers on truth, uncertainty, failure and human authority. The product is the field papers, made operational.
Built without code.
Systems like this normally demand a technical operator — repositories, servers, configuration nobody outside engineering understands. That barrier is why almost no one has one.
Zero Brain removes it with a role shift: the owner becomes the architect and director; an AI builder does the technical work, following instructions step by step. The owner decides the shape, the rules, the life it holds. The machine assembles it.
Architect does not mean magician
Building an instance means following guided steps — copying commands an AI hands over, connecting accounts through a screen. The floor is low, not zero. For those who want none of it, the managed path exists.
Commodity parts. Nothing exotic.
Figure 2 — the stack. The cost of the parts is small. The cost that matters — the one this system removes — is the knowledge of how they fit.
Two paths, one promise.
The ownership principle drives the delivery model. Whichever path is taken, the brain belongs to the person whose life it holds.
SOVEREIGN
The purest form of the ownership principle.
- The owner provisions their own server and private network, billed to them directly.
- The brain never leaves infrastructure the owner controls.
- Software, organism, updates and support billed monthly.
Asks for guided technical steps. The floor is low, not zero.
MANAGED
An instance provisioned and run on the owner's behalf.
- Server, private network and organism operated for the owner.
- Portable — export the brain and leave at any time. No lock-in.
- Software, organism, updates and support included.
Asks for nothing technical. The brain is still owned by the person whose life it holds.
Because the organism runs continuously, its cost is ongoing, not one-time. This is a subscription, not a purchase. A one-time fee for a living system is a promise no one can keep.
A life, first.
The system's first purpose is not productivity. It is presence: to stop carrying a whole life in one's head.
Never miss a birthday, a preference, what a friend said last time.
A coach that knows its owner: sleep, movement, weight, goals, trends over time.
Every restaurant, trip, wine and book loved — recalled from the owner's own history.
Gifts and greetings anticipated by relationship. Preparation before what matters.
Ideas and reflections kept, and returned when useful.
Dates, errands, personal finances, logistics — off one's shoulders.
Memory of people, preparation, follow-through. One room in the house, not the house.
What survives inspection admits its own edges.
Discipline is the cost
The never-invent guarantee requires deterministic verification and honest degradation. Skip the discipline and the result is a confident liar. The mechanism is not free; it is engineered.
The floor is low, not zero
The sovereign path asks the owner to follow guided technical steps. Non-technical does not mean effortless.
Ongoing cost is real
A living organism consumes infrastructure and inference daily. There is no version of this that is both continuous and free.
It is an assistant, not an oracle
It proposes; the owner decides. Removing the human from the loop removes the thing that makes it safe.
An agent is not a prompt. A second brain is not an archive. The difference in both cases is the same: a bounded system that receives state, decides, acts, observes and continues — under a human's authority, from a single source of truth, refusing to invent.
Zero Brain is that construction, aimed at a single life and built by anyone willing to direct it. The mechanism is inspectable. The ownership is real. What cannot be supported is removed.