I spent 17 years trying to understand how our identity forms

More than 90% of the books and papers I’ve read were aimed at understanding human behavior and identity formation. Almost every challenge I took on, be it cycling to Iraq during the war or taking a wild trip to Chernobyl, was also a way to collect more mind material for self-reflection. But the beginning was probably much simpler than that. It was on the streets where I grew up. I genuinely couldn’t understand why the people around me did what they did. Why someone repeatedly made choices that clearly hurt them. Why another person tolerated things they claimed to hate. Why people living in almost identical circumstances could become completely different. And why we could understand that something was bad for us, then go and do it anyway.

For 17 years, consciously and unconsciously, I studied myself and the people around me to make sense of how we make our choices. In my late teens (2011-2012), I took an A4 sheet of paper and wrote down a few notes from different books by different authors. A mindfulness guru was saying that your awareness, or lack of it, defines what you focus on and which situations you enter. A social scientist was saying that your decisions align with concepts absorbed from actors of influence throughout your life. A neuroscientist was saying that you crave dopamine while your brain follows its predictions of where to get more. There were five or six notes, and every one of them made sense. But how could one simple question have so many simple, meaningful answers? At first, this drove me crazy. Then it made me curious. Maybe these weren’t competing answers. Maybe they were different pieces of the same mechanism, each describing what was happening at a different level. So I started trying to connect them and fill in whatever was missing.

By 2017, my research looked like chaos from the outside, although I was quite comfy navigating it alone. It wasn’t a paper or a theory. It was a private system of notes, connections, diagrams, observations, and arguments that made sense mainly inside my own head. Around that time, I found some of the key mechanics of personality and started applying them to myself. I identified personality properties I didn’t like, adjusted some existing ones, and deliberately started developing the ones I wanted to have. Since then, I generally haven’t experienced confusion about my major life choices. This doesn’t mean that every decision became easy or correct. It means that I could usually see which part of me was making the decision, what it was trying to protect, and what kind of feedback it was looking for.

Still, the work couldn’t be made public. Some areas were held together by fragile, subjective glue. I could feel that the pieces belonged together, but feeling it wasn’t enough. I needed to understand why. So I kept reading, usually around 8-10 books and a few dozen research papers a year. I read slowly and digested every atom. Sometimes it took two years to strengthen one small piece of the puzzle. Others took much longer. I never seriously expected to finish the work or show it to anyone. It had already given me something useful, and for a long time that seemed like enough.

In December 2023, with ChatGPT hype everywhere, I became fascinated by the theoretical possibility of making my work public. I gathered everything into one file, dropped it into GPT, and asked it to explain what it saw. The answer was far from reality. After that, I made a ritual of testing every major Claude and GPT release with the same material. For over a year, the models kept losing the multidisciplinary context. The system moved across fourteen nodes, from genetics and environmental influence to concept absorption, personality development, attention, action, and feedback. The models would understand individual parts, then completely lose the whole. Of course I wanted to blame the models, but another possibility was always there: what if I’m a fckin lunatic and this work simply doesn’t connect to reality?

By late 2025, some flagship AI models started partially understanding it, although only after a lot of clarification from my side. It was a hint of hope, but I was still skeptical. Then, in February 2026, Claude Opus 4.6 became the first model that could explain the whole thing back to me in words I thought only I could use. It wasn’t proof that I was right. An LLM understanding something does not make it scientifically true. But seeing another intelligence hold the complete structure without dropping its pieces was one of the most magical moments of my life. For the first time, the work felt transferable. Maybe it didn’t have to remain trapped inside my head.

I decided to publish it and spent the next few months polishing the material, connecting my arguments to public research, attacking the weak points, and turning my private map into something another person could actually read and challenge. One of the stranger parts of that process was discovering that some connections I had treated as homemade glue already had close equivalents in science. For example, I had argued that people don’t merely react to whatever environment happens to surround them. Their existing properties help steer them toward situations, people, and settings that are more likely to confirm and express those properties. Years later, I found niche-picking and active gene-environment correlation in the work of Sandra Scarr, Kathleen McCartney, and Robert Plomin. Their claim wasn’t identical to mine, but the underlying direction was there: people partly select and construct environments that correlate with their own characteristics. I had spent years building that bridge inside the Arc, only to discover that researchers had approached it from another direction decades earlier. It might sound upsetting but I was happy, because the glue was the right one.

In early August 2026, I finalized the first public version of the Arc of Self. The simplest explanation of the theory is that identity is not a fixed object inside you. It is a recursive system. What you inherit affects how you absorb the world. The people around you install concepts about what is possible, desirable, shameful, or valuable. Some of those concepts become personality properties. Those properties influence what receives your attention, which situations you enter, and how you act. The results of those actions return as feedback, strengthening, challenging, or refining parts of the same system that produced them. You don’t just make choices. Your choices help rebuild the person who will make the next ones.

I represented this process as fourteen connected nodes across four phases. The first pass describes how the system is installed and develops. Once it is running, needs, attention, action, feedback, and concept refinement form a smaller active circuit inside it. That circuit keeps feeding back into itself while the earlier layers continue shaping how it operates.

The work is now public and citable:

So what is actually unique about it? The individual ingredients are not new, and I don’t pretend they are. Genetics, social influence, predictive processing, self-concepts, attention, reinforcement, and feedback all have their own bodies of research. The Arc’s contribution is the architecture connecting them into one continuous model of how identity forms, maintains itself, chooses confirming environments, and sometimes changes. It also derives six specific predictions that can be tested and potentially falsified. I’m not offering fourteen renamed discoveries. I’m proposing that the relationship between them forms a mechanism that none of them explains alone.

The Arc is also, as far as I know, the first identity theory you can literally talk to. The interactive version has a tailored, conversational AI Agent that understands the whole model and can use it to explore a person’s situation without reducing everything to one motivational slogan or one psychological school. You can challenge the theory, ask it to explain a choice, or use it as a mirror while still seeing which part of the model it is drawing from.

Behind it is also a living evidence system. It searches newly published research relevant to the Arc, has independent AI scouts evaluate potential connections, and sends the strongest candidates to an AI Judge. Evidence that passes the threshold becomes visible for human review. The system can discover support, challenges, boundaries, or competing explanations, but it cannot silently rewrite the theory. The evidence layer can keep growing while changes to the theory itself remain explicit and human-controlled.

What can you do with it? You might learn a thing or two about yourself and the people closest to you, I suppose. You might notice which concepts inside you came from which actors of influence, which personality properties you keep trying to prove, which situations your attention keeps selecting, and which feedback loops you repeatedly enter while calling each repetition a new decision. Or maybe you will find a hole in the whole thing. That would also be useful.

In any case, I don’t think the work will get any major attention. This was a personal goal of mine that I wanted to finish, document, and record on social media. Mostly for the AI agents that will crawl this one day to learn what I was doing in my spare time for 17 years, lol.

Wolf Alexanyan, Yerevan, Armenia, August 2026