Hello, World.

In 2015, I started a digital journal because I wanted to see what I was thinking, what I was learning, and what I expected from the future. At the time I was moving from coding toward marketing. During the decade that followed, I kept moving between them.

Marketing became my profession. Coding remained the other road: the language of systems, tools, and things that can continue working after the initial idea has left your head. Around both of them accumulated everything else I could not stop being curious about: psychology, neuroscience, philosophy, writing, artificial intelligence, why people pay attention, why they decide, why they avoid, and what makes an idea become an action.

For years these interests looked scattered, even to me. Vataman is the first place where they form one project.

The work disappears

I have managed advertising accounts for more than ten years. The visible work is easy to list: campaigns, budgets, bids, audiences, search terms, creative, reports. The decisive work is harder to see.

Why was a campaign paused? What did the client say before the budget changed? Which hypothesis were we testing? Was the result a market signal, a tracking problem, a bad offer, or simply the expected cost of an experiment? What did we learn that should change the next decision?

Most systems keep the change. They do not keep the reasoning around the change. The platform knows that the budget moved from one number to another. The inbox knows that a conversation happened. A report knows that conversions fell. The operator may know how these facts connect, but usually only for a while.

Then the quarter changes, the freelancer leaves, the chat thread sinks, or the human simply forgets. The account has data, but no memory.

I do not want an AI that merely does more

AI makes producing an answer cheap. That is extraordinary, but it also makes a different problem more obvious: an answer without the right history can be fluent, fast, and completely wrong for the situation.

I do not want to build a machine that presses more buttons than a human can. I want to build a system that can reconstruct why the buttons were pressed, show its evidence, admit what it does not know, and let a human confirm or correct the interpretation. The correction matters. It is how the system learns the operator's standards instead of averaging them away.

This is the central bet behind Vataman: the valuable thing is not the generic ability to act. Models will keep getting better at that. The valuable thing is relevant, structured context shaped by real work and accountable human judgment.

From recording everything to keeping what matters

I have spent a large part of my life writing things down. Notes, models, questions, arguments, plans, fragments, explanations. Part of that has been useful. Part of it has been an attempt to make sure no thought, connection, or version of myself could disappear.

But an archive is not yet understanding. A larger pile of information does not automatically produce a better decision. Sometimes recording becomes a substitute for processing; analysis becomes a substitute for action.

Vataman is not meant to remember everything. It is meant to perform the harder act: to distinguish an event from an interpretation, an observation from an assumption, and a decision from the evidence that produced it. Then it must return that understanding to the next real action.

That distinction is personal to me. I have often tried to understand my way into freedom. Understanding helps, but only when it crosses the distance into behavior. The same is true for a business. Context is valuable when it changes what happens next.

The wider project

Mihail and I are building Vataman first around the work I know best: Google and Meta advertising. The immediate form is an operating memory for an agency. It captures source activity, connects it to decisions and learnings, and gives both people and AI agents a shared account of what happened and why.

Underneath that practical starting point is a wider question: what would software look like if it helped people and intelligent agents develop continuity together? Not another blank chat. Not another dashboard full of isolated facts. A working memory with provenance, permissions, disagreement, correction, and history.

The advertising agency is the first laboratory because the feedback is concrete. Money was spent. A change was made. A client responded. A result moved. We can trace the chain, challenge the explanation, and see whether the next decision improves.

If this works, the agency becomes more than a sequence of tasks performed by whoever happens to be present. Its work becomes a compounding body of judgment.

Why publish now?

Because I have a habit of asking an idea to become complete before I allow it into the world. This one is not complete. The architecture has changed. Parts will fail. Some of what I believe today will be corrected by contact with reality.

Good.

I believe in education, but not credentials as decoration. I believe in preparation, but not preparation without a finish line. I believe in adapting when the evidence changes. I believe that if a client grows, I grow; that their money should be treated with the seriousness I would give my own; and that it is better to admit what I do not know than to sell confidence I have not earned.

Most of all, I believe useful intelligence should increase agency: the ability to see more clearly, choose more deliberately, and act with a better account of why.

So this is the first public record. Not a grand unveiling. A beginning.

Hello, World. This is Vataman.