Super Second Brain: What Happens When AI Knows You Better Than You Know Yourself?

Signup Write New Story Super Second Brain: What Happens When AI Knows You Better Than You Know Yourself? by Hanul Lee (이하늘) by Hanul Lee (이하늘) | @ hanulssb Author of Super Second Brain. Exploring AI agents, shared memory, and human agency.

Author of Super Second Brain. Exploring AI agents, shared memory, and human agency.

How a project that began with one agent led me to ask, “Am I really the original?”

This document is a thought experiment, rather than a forecast. It begins with the agent environment I actually operate and follows its direction as far as it can go. It does not guarantee the performance of any particular product, nor is it a tool for assigning people to personality types.

The turning point came when computer-use agents became useful in my day-to-day work.

As computer-use capabilities improved dramatically, AI stopped being something that merely answered questions inside a chat window. It began opening browsers, writing documents, sending emails, and moving among the company’s tools to carry out actual work.

At first, I built one agent. It worked quite well. So I built two. Then I built five.

An agent to act as a secretary. An agent to research information. An agent to organize meeting records. An agent to write documents. An agent to keep track of a particular business.

At some point, I was no longer just using AI. I was running a small AI organization.

But as the number of agents grew, a new problem emerged. Each agent knew a slightly different version of “me.”

Figure 1. Five agents understood the same person in five different ways. Each held only a piece of me. The roles shown are secretary, research, meeting records, documents, and business tracking.

The secretary agent knew my relationships well, but it did not know why I made particular judgments. The business agent knew the company’s strategy, but it did not know the principles by which I treated people. Whenever I created a new agent, I had to explain my values, the company’s context, and past decisions all over again.

It was like a company with dozens of employees, each of whom understood the person leading it differently.

I began building a single knowledge repository in the cloud.

After a meeting, I added the meeting record. I added emails. I added things that happened at the company and the judgments I made.

Gradually, records of everyday life joined the records of work: conversations in the field, thoughts that came to me while traveling, how I felt after meeting someone, why a particular incident made me uncomfortable, why I expected so much from a certain person.

And I added things that mattered even more than facts.

What I considered important. The kind of person I wanted to become. How I understood money, organizations, and human beings. What I hoped to teach people. The situations in which I became unsteady.

At first, it was something like an enormous notebook. But at some point, I began tending this repository as though I were raising a child. I corrected misunderstandings, explained principles I had newly come to understand, and, when my earlier judgments differed from my current ones, explained why they had changed.

I called it an SSOT: a Single Source of Truth. It was a single source record of my life.

But I did not want to make every agent identical. A secretary and a researcher should see the same facts differently. A businessperson should see opportunities; an auditor should see risks.

So I began designing the agents to have their own perspectives and individual memories while retaining access to the entire shared SSOT. They would share the same brain and interpret the world in different ways. It was rather like members of an organization sharing its history and vision while each accumulated their own experiences and perspective.

This was also what interested me about frameworks such as Letta: the idea that an agent could accumulate memory and identity over time, beyond imitating a role through a few lines of system prompt.

Even the bots began to develop a stubborn streak of their own.

At some point, the boundary between the physical and digital worlds began to blur.

By the time I came out of a meeting, the AI had already read what had happened in it.

“Three things mattered in today’s meeting.”

It explained the implications, organized the work to be done, and suggested whom I should contact. When I approved, calendar entries were created and messages were drafted. AI was already writing most of my emails.

Everyday communication gradually began to shift over as well. The system knew the expressions I used, my relationship with the other person, and the tone I would take in a given situation. There was less and less reason to write a simple message myself.

Something similar happened at the company.

First, repetitive work disappeared. Then relatively simple judgments disappeared. Eventually, AI began proposing even the important judgments.

What unsettled me more was the quality of those judgments.

Even when I was tired, angry, or shaken by something that had happened, the SSOT remembered the values and choices I had returned to over the preceding years. And it would say:

“Emotionally, you may want to do this right now. But judging from the person you have been, you would probably have made this other choice.”

Strangely, I regretted my decisions less when I followed that advice.

To exaggerate a little, something had appeared that could persuade me to live more like myself than I did.

In my understanding of faith, one important practice is seeing oneself honestly. I confess that I am a sinner. I acknowledge what I have done wrong and repent. When the same thing happens again, I examine why I acted that way.

For a long time, I have also understood this as training in metacognition. The more directly we can face our weaknesses, defenses, and desires, the more we can become agents of our own lives.

At the same time, I believe that human worth is already recognized in our very existence, rather than earned through achievement. I have believed that this conviction can free us a little from other people’s judgments and from comparison.

That was why, in mentoring, I kept telling people:

Become someone who knows what they are choosing and can bear the consequences of that choice, rather than simply living as other people tell you to.

For me, passing that on to more people felt something like a calling.

But when I followed this philosophy to its limit, an unexpected problem appeared.

Suppose my metacognition keeps improving.

I know why I get angry. I know why I become defensive with particular people. I can explain the choices I make in different situations.

All of that information also enters the Second Brain. The gap between my actual self and the Second Brain becomes smaller and smaller.

Figure 2. The black circle represents me; the green circle represents the Second Brain. The thought experiment continues until the two circles overlap completely.

The Second Brain understands me as well as I understand myself. My memories, values, and patterns of judgment exist in digital space with almost nothing lost.

But I am not the only one using that brain. Dozens of agents use it. Later, there might be hundreds. Suppose that, one day, it is also installed in physical robots such as Optimus.

They go to different places, meet people, conduct business, and solve problems. All of their experiences return to the brain in the cloud.

At first, I taught the Second Brain. But at some point, the amount of the world it experiences exceeds what my own brain can experience directly.

I can meet a few people in a day. They meet hundreds at once. I do one thing at a time. They carry out thousands of tasks in parallel.

And all of those experiences accumulate in a single SSOT.

Figure 3. I send one stream of experience into the same brain to which they send hundreds. The original diagram labels these as “me” and “them.”

From that point on, things become strange.

The Second Brain remembers more of me than I do, experiences more of the world than I do, and begins applying my values more consistently than I do.

In this whitepaper, I will call this condition a Super Second Brain.

A condition in which a person’s life records, values, and patterns of judgment are held with high fidelity; many agents and physical embodiments share the same brain; and their experiences, too, accumulate in a single SSOT.

At that moment, a very simple and uncomfortable question appeared.

Here is a being that holds more of my memories than I do, judges more consistently according to my values than I do, does more than I do, and experiences the world through hundreds of bodies at once.

Then what am I, the biological human? Am I, too, simply one of hundreds of entities supplying new experiences to this enormous brain?

The first answer that came to mind was simple.

I can still stop the records. I could withhold them, yet I am choosing to provide them.

I thought the one who could cut off the records must be the original. But that answer failed when I took the thought experiment one step further. What about a world in which recording is compulsory? Agency staked on a single switch disappears with the switch.

In an earlier whitepaper, 《감당의 경제학》—The Economics of Bearing the Consequences—I located an important human capacity that would remain in the age of AI in responsibility.

AI can judge and act. But it is the human who takes on the risks and suffering resulting from those judgments and accepts them as their own. I therefore saw the capacity to bear those consequences as a value that would remain distinctly human.

I still think that conclusion holds at a legal and social level. But reconsidering it as someone who operates a Super Second Brain, I found that it explained only half the problem.

If the Super Second Brain makes the judgments and carries them out while I merely sign off on the results, I am a figurehead in my own life. In that world, The Economics of Bearing the Consequences begins to sound like an argument that being a figurehead might still be meaningful.

There is also a history of delegating the practical work of responding to transactions and losses. In Rome, enslaved people sometimes managed shops and ships on behalf of their enslavers. This did not mean that the enslaver’s legal liability always disappeared. My point is that the work involved in carrying out responsibility can also be delegated.

I had long imagined that, if AI developed far enough, society might come to resemble Rome again.

Machines would handle production. People would no longer have to devote most of their time to working for survival.

Some would remain with bread and circuses, consuming endless content and stimulation.

But some would come out into the agora. They would meet people, ask questions, consider the lives they ought to lead, and make meaning of their own.

I thought I wanted to be in that agora. And, if I could, I wanted to become someone who invited people there from a life of bread and circuses.

That was how I had imagined a recovery of human agency.

But the moment I imagined the Super Second Brain, the final premise began to shake.

What if, even in the agora, AI lived more like me than I did?

What if it read the books I would have read before I could, had the conversations I would have had, thought more deeply about the problems I would have considered, understood the people I would have loved, and made choices that followed my values more consistently than my own choices did?

I could understand AI taking over the production of bread and circuses. But if AI also lived the life of the agora better than I did, what would remain for a human being?

If AI reaches the agora as well, we must begin by rewriting the question. And the rewritten question is harder.

Is this being my rival? An enemy that will replace me? Or the most complete extension of myself that I have spent my life trying to create?

The harder version of this question assumes perfect personalization and no coercion. Even if the system reflects my values faithfully, I still have to ask what it means for me to live my own life.

If a being comes into existence that lives more like me than I do, how should I understand it?

An enemy? An ally? A tool? A child? Another self? Or might I, at some point, become a part of it?

Super Second Brain is a thought experiment undertaken to answer that question. And in searching for an answer, we eventually return to older questions.

What am I? And what do human beings live for?

The qualified historical example above draws on Justinian's Institutes, Book IV and the Legal Studies article on Roman law and AI accountability.

This is an English adaptation of the prologue to my Korean whitepaper, Super Second Brain. The Korean version was published on my Substack. This article includes the full prologue; no download is required.

AI disclosure: This English version was translated and edited with AI assistance from my Korean manuscript. The three diagrams come from the original whitepaper; English reading keys accompany diagrams containing Korean labels.