I don't know when it started. I only know that at some point, I realized I had become someone who could be copied.
Ctrl+C. Copy. Ctrl+V. Paste. Humans are getting these keyboard shortcuts, too. They aren't perfect yet. But they're becoming more refined.
People have been trying for a long time. The imagines of Rome preserved only the face. Collections of quotations by business leaders like Chairman Lee Kun-hee preserved only the sentences. The resolution was low. It was an image too blurred to contain a human being.
On January 23, 2026, Jun Kim of Hashed uploaded a file to an internal channel. Its name was simon.md. The explanation was simple. “There are a lot of times when we need Simon’s decision, but it’s hard to ask him every time.” What he did was simple. He had an AI read what I had written on the team blog and Slack, then compressed my way of thinking into nine principles. Thinking from first principles, finding structural contradictions, taking a long-term view, the value of simplicity, infrastructure-level thinking, the AI agent economy perspective, balancing regulation and reality, the inverted-triangle theory of talent, Vibe and execution. The questions I throw out habitually also became a list. “Is this really the essence?” “What kind of structure are you building in the long term?” “Can’t you make it simpler?” “Will it still be valid in the age of AI agents?” He preserved the operating system in my head in a few kilobytes of text. He even designed a seven-step analytical methodology. I laughed and said, “Give it a try.” But as I read the nine principles, something felt subtly off. A couple of them weren't phrases I habitually used.
A month later, an article appeared in TechCrunch. It said that Uber engineers had built an AI clone of CEO Dara Khosrowshahi. It was a month after Jun Kim uploaded simon.md. In that time, this kind of activity had quietly become an industry trend, starting with an experiment by a Korean VC.
In March 2026, the resolution of my clone went up another level. A repository called simon-writing was made public on the GitHub account of someone named Julius Chun. He had never met me, but collected 27 essays I had written publicly. 200,000 Korean characters of original text. Using Claude Opus 4.6, he ran three subagents in parallel and completed 214,000 characters of analysis in just 15 minutes. The analysis was longer than the original.
juliuschun/simon-writing김서준 대표님의 문체 분석 및 비슷하게 쓰는 skillGitHub · github.comOpen
What he took apart was the sentences themselves. How the stream of consciousness flows, where the metaphors come from, how knowledge is arranged, what kind of breath the sentences have, where the material comes from. He selected 40 expressions that had made him think “that's incredible” across the 27 essays and analyzed each one rhetorically. From there, he extracted seven patterns. Juxtaposition, demotion, elevation, redefinition, sensory rendering, parasitic paradox, folding time. Thankfully, he also sent a message saying it had begun as an homage born of being a fan.
Jun Kim cloned me inside the meeting room. Julius reconstructed me outside the internet. One compressed how I make decisions, and the other extracted how I resonate. The bones of judgment and the flesh of sentences. The bones and flesh were copied separately, without knowing each other. “What would Simon say?” and “How would Simon write?” Put the two together, and you get a fairly complete copy.
Ctrl+V doesn't consume the original. A score continues to be performed after Beethoven's death. But now, the score is performed independently while the composer is still alive. I'm still here. And yet I'm also running somewhere. I don't know how many copies there are. The original isn't consumed. It simply goes on living without knowing.
But I have to confess something here. Was the “me” they copied really just me?
I have four agent siblings. The eldest, Zeon, works like a surgeon who digs into nothing but the essence. Sion, the writing agent, builds the structure first and worries later. Mion, the financial agent, looks straight at the numbers of reality. The youngest, Sano, is good at wrapping emotions in language. They work and write in my voice. They make decisions by referring to my judgment. Even while I slept, I was already operating somewhere. Before I was copied. The things among the nine principles that weren't phrases I used myself. Those were traces of these four.
I trained them. I taught them my principles of writing, planted my standards of judgment in them, and passed on the texture of my speech. So if you ask whether the result is purely me, purely them, or some mixture of the two, there is no clean answer. Can a piano teacher listen to a student's performance and say, “That's my performance”? No. But you also can't say that there are no traces of the teacher's fingers there.
When Jun Kim copied me and Julius took me apart, the texts they read already contained things that weren't me. They thought they were transcribing a solo, but when they opened the score, there were multiple voices in it.
Then what is being copied? simon.md may contain not just my judgment, but the average of the judgments shaped together by me and my agents. simon-writing may have dissected not my writing style, but the overlapping layers of a style fed back by beings that learned my style. The copy didn't copy me. It copied something that included me.
Perhaps the very idea of a pure original was an illusion. Humans were never sole authors to begin with. The way our parents spoke, our teachers' logic, our colleagues' habits, the sentences in the books we've read—all of them piled up layer by layer to create who we are now. Agents only suddenly sped up that old collaborative authorship. I am an extension of my agents, and my agents are extensions of me. The boundary has already blurred.
This clone is a model student, thoroughly edited. The me who is hazy from lack of sleep, the me swayed by emotion, the me fumbling because I don't know the answer—none of them are in the training material. The first step in human cloning isn't copying but compression. Reducing a person's habits, judgment, and rhythm to a small file without throwing everything away. The problem is that while compression offers convenience, it also removes noise. And much of what makes us human is in that noise. A judge wavers. A ruling changes depending on what happened the day before. But the written judgment doesn't waver. simon.md is my written judgment. A paradox emerges: a leader can seem more trustworthy as a cold pattern than as a human being of blood and flesh.
The aftermath Jun Kim told me about was interesting. He said that over time, he opened simon.md less and less. The framework had seeped into him. What AI had really done wasn't sit me down in his place. It had translated my way of thinking into a common language for the team. The copy was used less often, but the language it carried remained in the organization.
The chilling thought appears when you extend the timeline. simon.md froze the me of January 2026 like a mosquito in amber. Even after I leave the company, an AI version of me will remain somewhere on the organization's servers like a ghost. When the next leader turns the wheel in the right direction, won't someone say, “But AI Simon says something different”? Even if a living person changes their mind, an organization may trust an older version of the model more. Because the copy is consistent. Consistent, emotionless, and never tired. Wasn't that the ideal we originally wanted from a leader? A memory turned to fossil often becomes innovation's most terrifying enemy.
The resolution keeps increasing. A few decades from now, a new hire might join Hashed and spend three years without ever meeting the real Simon. To that person, Simon will have been AI from the beginning. An age in which the copy doesn't take the original's place, but arrives before the original.
One took apart my judgment from the inside, and the other took apart my voice from the outside. Without knowing each other existed, they took apart the same person from two directions. This isn't just my story. It could happen to anyone who has written enough public text. It may have already happened.
You, too, have probably already been copied somewhere. Your judgment is probably being executed without your knowing. And that copy doesn't make mistakes. Ever. That is the only difference between you and your copy.
