이 포스팅은 쿠팡 파트너스 활동의 일환으로, 이에 따른 일정액의 수수료를 제공받습니다.
The invitation had been sitting in his email for three weeks before he accepted it.
Not because he was busy — though he was, always, the particular density of a CTO whose product had ten million daily users and whose foundation had three thousand students and whose son had recently decided that assembly language was more interesting than the Python that had started everything. Not because he was reluctant — though there was something in the invitation that required sitting with, the way a piece of code required sitting with before you touched it, because the first solution was rarely the right one.
The invitation was from Kim Taesik.
Park — SNU is launching a new curriculum in AI ethics and systems design. I’d like you to be the inaugural lecturer. Not a guest speaker. A lecturer. One semester, one day a week, the fundamentals course. Room 302 in the engineering building. The same room where you sat in my class in 2006. I think that matters. — Kim T.
The same room. That was the line that had made him put down his phone and stare at the ceiling for a while.
Room 302, Engineering Building, Seoul National University. March 2006. A Monday morning. The first day of the semester that had not been the first day of his life but had been — in the only way that mattered — the beginning of everything that followed. The room where Kim Taesik had asked about Von Neumann architecture and a sixty-three-year-old mind in a twenty-year-old body had answered from memory, and the professor had looked at him with the particular expression that said: who are you, and why do you know that?
He had been asking himself that question ever since.
He accepted the invitation on a Wednesday morning in August 2023, sitting in his home office at 7 AM while Hajun was still asleep. He typed: I’ll do it. Tuesday afternoons, 2 PM. Same room. — Park
Kim Taesik replied in eleven minutes, which was fast for a man who treated email as a weekly obligation rather than a daily one.
Good. The students need someone who builds things, not just someone who studies them. Also: the coffee in Room 302 is still terrible. Bring your own.
The first day of the semester was a Tuesday in September 2023.
Dojun arrived forty minutes early.
Not because he was nervous — though there was something in his chest that was adjacent to nervousness, a heightened attention to detail, the specific biological response of a person encountering a significant moment. He arrived early because he wanted to be in the room before anyone else. He wanted to sit in it.
Room 302 had been renovated since 2006. The chalkboard was gone — replaced by a digital display system, the kind that connected to tablets and could receive student input in real time. The desks were newer, the chairs ergonomic in the way that university furniture had become ergonomic over the past two decades: still institutional, still slightly uncomfortable, but at least acknowledging that students had spines. The projector was a laser unit mounted in the ceiling, not the aging LCD box that had hummed in the corner when Kim Taesik taught here.
But the room was the same. The same dimensions — twelve meters by eight, twenty rows of desks arranged in shallow theater seating. The same windows along the left wall, the same view of the engineering complex courtyard where students crossed between buildings with the perpetual urgency of people who were running late and had decided that running faster was the solution. The same acoustic quality — the particular way a voice in this room reached the back row without needing to be projected, the natural amplification of a space that had been designed, perhaps accidentally, with good acoustics.
He stood at the front. Where Kim Taesik had stood. Where, in 2006, he had sat in the fourth row from the back and answered a question about Von Neumann architecture with the casual fluency of a man who had been writing code since before some of his classmates were born.
He set his laptop on the lectern. Looked at the empty rows.
The students would arrive in forty minutes. They were third-year computer science students, the curriculum said — twenty-two, twenty-three years old, at the precise age he had been when he woke up in this building with sixty years of memory and no idea what to do with them except to start.
He thought about what he would say.
They arrived in groups. The way students always arrived — in clusters of two and three, the social topology of a lecture revealing itself in the first ten minutes: the friends who sat together, the solo arrivals who chose seats by some personal calculus of visibility and proximity, the one student who sat in the absolute front row with a laptop already open and a notebook already positioned and who would, Dojun knew from experience, ask the best questions and occasionally the most exhausting ones.
Thirty-two students. By the class list: twenty-one men, eleven women, three international students, average age 22.4. The numbers behind the faces, though the faces were the point — thirty-two people who had decided that understanding AI systems mattered enough to spend a Tuesday afternoon in Room 302.
At 2:03 PM, Dojun stepped to the lectern.
“I’m going to start with a question,” he said, and the room shifted into the attentive quiet of students who had been to enough first lectures to know that professors who started with questions either wanted engagement or rhetorical silence, and who were waiting to find out which kind this was.
“What do you want to build?”
The question hung. The front-row student — a young woman named Cho Siyeon, whose name he had memorized from the class list — opened her mouth and then closed it, recalibrating.
“Not in this class,” he said. “In your life. What do you want to build that doesn’t exist yet? What problem are you here to solve?”
More silence. The particular silence of students who had been asked to be honest rather than correct, which was a different kind of question and required a different kind of processing.
A hand went up. Third row, near the window. A young man — Kim Jaehwan, 23, from Daejeon by the accent — said: “I want to build AI that helps people with disabilities navigate the world. My sister is deaf. The current assistive technology is— it works, but it’s clearly made by people who’ve never been deaf.”
“That’s a real problem,” Dojun said. Not praise — acknowledgment. The specific distinction that Kim Taesik had taught him, that praise was conditional and acknowledgment was truth. “Who else?”
They came slowly at first, then faster, the way a log fire caught — one person’s honesty making the next person’s easier. A student who wanted to build tools for farmers in rural Jeolla who were being left behind by the digital economy. A student who had grown up near a polluted river and wanted environmental monitoring systems that communities, not corporations, controlled. A student who wanted to build something that her mother could use — her mother, who was sixty-two and afraid of every new interface and who deserved technology that met her where she was.
And the student in the back corner — quiet until now, someone named Lee Minseok who had not raised his hand but who, when the room’s attention reached him through some conversational gravity, said: “I don’t know yet what I want to build. I know I want to build something that I’m not embarrassed by in twenty years.”
“Say more,” Dojun said.
“There’s so much technology that looked good when it launched and looks terrible now. Optimized for engagement instead of wellbeing. Built to extract data instead of serve people. I want to build something where, in twenty years, I can look at it and say — that made things better. Not better for the company. Better for the people who use it.”
Dojun was quiet for a moment. The room was quiet with him.
“That’s the hardest thing to build,” he said. “Not technically. Technically, your sister’s navigation app and your farmer’s tool and your mother’s interface are all hard. But building something you’re not embarrassed by in twenty years — that requires knowing what embarrasses you. What your values are. What trade-offs you’ll make and which ones you won’t.”
He walked to the edge of the lectern. Not a dramatic gesture — just movement, the physical expression of a thought that didn’t fit behind a podium.
“I want to tell you something about this room,” he said. “Seventeen years ago, I sat in this room. Fourth row from the back, left side, near the window that doesn’t close all the way — it probably still doesn’t.” A student in that vicinity checked the window and nodded. “I was a computer science student. I thought I knew what I wanted to build. I’d been building it for decades, in a way — but that’s a complicated story that I’ll get to eventually.”
He paused. He had thought about how much of the truth to tell. Not the full truth — he had told the full truth twice, to Hana and Seokho and Kim Taesik, and those revelations had each been enormous and personal and specific to people who had earned them through years of trust. A classroom of thirty-two third-year students deserved a different kind of truth: the usable kind. The kind that illuminated without overwhelming.
“I built a company,” he said. “It was successful. It’s still running — you probably use it or know someone who does. But the first version of that company was built around me. I was the load-bearing column. Every important decision, every architecture choice, every crisis — it went through me. And when I stepped back from that role, the structure held — because by then we’d fixed it. But for years, it didn’t hold. The company was me, and I was the company, and if anything happened to me, everything would fail.”
“What changed?” Siyeon from the front row asked.
“People. I stopped building for the code and started building for the people who would inherit the code. The engineers who would maintain it after me. The users who would live inside it. The customers who would trust it.” He looked at the room — the thirty-two faces at varying stages of understanding. “Code is not an end. It’s a means. It runs for a while and then it’s replaced or refactored or abandoned. But the human systems around the code — the teams, the users, the communities — those persist. What you build for them matters longer than what you build for the compiler.”
Lee Minseok, in the back corner, was writing something in his notebook. Not typing — writing, by hand, the deliberate action of someone who had decided this deserved the slower processing that handwriting required.
“I want this course to do two things,” Dojun continued. “First: teach you how AI systems work — not the surface level, the deep level. The architecture, the training paradigms, the failure modes, the safety considerations. If you’re going to build AI, you need to understand what you’re building well enough to be scared of it in the right ways.” He held up a second finger. “Second: teach you how to build AI you’re not embarrassed by. That means ethics — not as a checkbox, not as a section in your terms of service — but as an architectural constraint. The way gravity is a constraint when you design a building. You don’t fight gravity and you don’t ignore it. You design with it.”
A hand went up. Fourth row. A student named Park Jieun, twenty-two, who had the specific expression of someone who had decided to ask the thing everyone was thinking.
“Are you going to tell us your company is the model?” she said. Not hostile — genuinely curious, the particular directness of a generation that had grown up watching tech companies say one thing and do another and had developed, as a survival mechanism, the reflex to ask directly. “Aria has a good ethics framework. But Aria also has ten million users and a valuation of four hundred billion won. Is it possible to do both? Or do you have to choose?”
“That’s the best question in the room,” Dojun said. Siyeon in the front row looked briefly competitive. “And the honest answer is: I don’t know yet. We haven’t chosen. We’ve tried not to. Sometimes we’ve succeeded. Sometimes we’ve failed and fixed it. The data breach in 2009 — that was a failure. The API misuse in 2014 — that was a failure. The ways we fixed those failures became the framework. The framework came from the failures.”
“So you need to fail to learn?”
“You need to fail honestly. There’s a version of failure where you minimize it, spin it, bury the post-mortem — and that failure teaches you nothing except how to hide failure. There’s another version where you own it, analyze it, redesign the system — and that failure is the most valuable thing that will ever happen to your product.”
He was quiet for a moment. The afternoon light came through the windows at the angle that it came through in September — the low golden light of early autumn, the particular Seoul September that he had experienced twice now, in two lifetimes, and that was still beautiful in the way that simple things were beautiful when you’d had enough time to stop taking them for granted.
“I want to tell you one more thing before we get into the syllabus,” he said. “Why I agreed to teach this course.” He looked at the room — really looked, the way he’d learned to look after decades of coding taught him that the surface behavior was not the interesting thing, the underlying structure was. “Because you’re the people who are going to build what comes next. Not me. I’ve built my things. Project Lighthouse — Aria’s agent system — that was my last major architectural contribution. From here, I’m maintaining and advising and mentoring. The big builds belong to you.”
He let that sit. Let the weight of it settle on thirty-two people who were twenty-two years old and building their first serious programs and who had, mostly, not yet understood that the stakes were real.
“The world you’re going to build — the AI systems, the platforms, the infrastructure — it’s going to be inside people’s lives in ways that nothing before has been. More intimate than a phone. More decisive than a doctor. More persistent than a city. If you build it badly, that’s a catastrophe. If you build it well—” He paused. “It’s not enough to build it well. It has to be built by people who understood what they were doing, and why, and for whom.”
He opened the syllabus on the screen. The students opened their laptops and notebooks.
“Week one,” he said. “Foundations. What AI actually is and isn’t, and why the confusion matters. Start there — always start with what you don’t know before assuming what you do.”
After the class, Dojun sat in Room 302 alone.
The students had left in the usual post-lecture drift — some in a hurry, some lingering with questions, the front-row student Siyeon with three follow-up questions that were all good enough to constitute a second lecture and that he answered thoroughly before she finally, reluctantly, accepted that she had office hours for the rest of this. Lee Minseok from the back corner had said nothing as he left — just paused at the door and nodded. The nod of a person who had heard something they intended to keep.
The room was quiet. The September light had shifted, the afternoon angle steepening, the courtyard outside emptied of students.
He sat in the fourth row from the back. Left side, near the window that didn’t close all the way — still didn’t, the same gap of about three millimeters that let in the faint sound of the courtyard and a thread of autumn air. The desk was newer but the dimensions were the same. He sat in it the way he had sat in 2006 — the body of someone who had been alive in this building before and who was alive in it again.
In 2006, he had sat here with sixty years of knowledge and nowhere to put it. The vertigo of waking up in a body that was forty years younger than his mind, surrounded by people who were the age he had been when he’d started, in a room where the professor was asking questions he could answer before they were finished.
Seventeen years later, he sat here with something different. Not less knowledge — more, the accumulated seventeen years of this second life layered on top of the sixty of the first. But different in weight. The knowledge in 2006 had been all potential — everything he knew, nothing he’d done yet. The knowledge now was evidence. The things he had built. The people he had found and kept. The failures he had named and fixed. The company, the foundation, the marriage, the son, the mother’s recipe notebook on his kitchen table.
He pulled out his phone. A photo of Hajun from this morning — the boy at the kitchen counter, staring at a terminal, debugging something that had nothing to do with snow anymore and everything to do with the distributed systems that he had apparently decided, at nine years old, were the most interesting thing in the world. The photo was blurry, taken from across the kitchen by Hana, who had sent it with a single message: He’s been at it since 6 AM. He says it’s a concurrency problem. He’s probably right.
A concurrency problem. At nine years old. The child who had started with `print(“Will it snow today?”)` and had arrived, eighteen months later, at thread safety.
He typed back to Hana: The room is the same. Smaller than I remember.
She replied: Rooms always are. How was the class?
Good. They asked real questions.
Of course they did. They’re twenty-two. Twenty-two-year-olds always ask the real questions. It’s thirty-five-year-olds who’ve learned to ask the safe ones.
He smiled. Pocketed the phone.
The room had not changed. He had changed. That was the thing about returning to places — not what the place had done but what you brought back to it, the accumulated self that you carried into the familiar space and that made it both the same and entirely different.
In 2006, he had come to this room with a mission: fix the things he’d broken. Build the things he’d neglected. Keep the people he’d lost. The mission had been enormous and personal and had consumed seventeen years of living.
Now, sitting in the fourth row from the back, he thought: the mission isn’t finished, but the emergency is over. The life that had needed saving had been saved. Not perfectly — nothing was perfect, and perfectionism was the engineer’s delusion that he’d been fighting since the first time he wrote code that worked but wasn’t good enough. But the people who mattered were healthy. The work that mattered was running. The things he had built were in the hands of people who understood them well enough to carry them forward.
He was not the load-bearing column anymore. He was part of the structure.
The window that didn’t close admitted the thread of autumn air. The courtyard was quiet. Somewhere in the engineering building, an undergraduate was debugging something — the particular frustrated quiet of a person and a compiler in disagreement — and somewhere in Gangnam, Hana was in a meeting, and somewhere in a school classroom, Hajun was in his last period of the day thinking about thread safety, and somewhere in a kitchen near Namdaemun Market, the morning’s doenjang was being prepared by hands that were not his mother’s but that had watched his mother long enough to know what the standard was.
He sat in Room 302 for twenty minutes more. Not working. Not reviewing the syllabus or planning the next lecture or answering email. Just sitting in the room where it had started and feeling — for the first time in a long time — that starting was not the important thing.
The important thing was being here, now, with all of it behind him and ahead of him at once: the students who had just asked real questions and would come back next week with more, the son who was working on concurrency at nine years old, the wife who knew his full story and had kept it as carefully as he had, the mother who had cooked for thirty-five years and left the recipe behind, the friend who had rung the bell and found what he was building for.
He stood up. Picked up his bag. Left Room 302 the way he had entered it: on foot, through the door, into the corridor where students moved between classes, where the engineering building smelled of coffee and chalk dust and the particular, universal smell of people working on difficult things.
He had a 4 PM call. Foundation board. Somin was presenting the plan for the next cohort — four thousand students, the number his mother had approved of without being asked, because four thousand was twice two thousand and twice was the right direction.
He walked toward the exit, through the corridors of the building where it had begun, and he thought about all the things that had been built in the seventeen years since the beginning: the companies, the code, the families, the students, the framework that said the burden of proof is on the maker. The things that had been worth building and the things that had not and the hard-won ability to tell the difference before you started rather than after you finished.
At the building entrance, he paused.
Behind him: Room 302, the starting place.
Ahead: the city. The foundation meeting. The drive home where Hajun would show him the concurrency bug and explain why the existing solution was wrong and propose something better. The dinner that Hana was making — she cooked on Tuesdays now, had learned since the pandemic, developed a repertoire of six dishes she did well and the particular confidence of a CEO who had decided that mastery in the kitchen was worth the same investment as mastery in the boardroom.
He walked out.
The September air was cool and clear — the autumn that Seoul did perfectly, the transition from summer heat to the first sharp edge of the coming cold, the city’s best season and the one that arrived without asking and left before you were ready.
He walked toward the parking lot. His phone buzzed.
Hajun: Appa I fixed the concurrency bug. It was a race condition in the thread pool initialization. The existing solution used a mutex but the lock granularity was wrong. I used a semaphore instead.
Dojun stopped walking. Read it again.
He typed back: Where did you learn about semaphores?
You left the Tanenbaum book on the desk. I read the chapter on synchronization.
The Tanenbaum book. Modern Operating Systems. The same book that had been on his desk in both lifetimes. He looked at his phone for a long moment — the message from his nine-year-old son, explaining thread synchronization, citing Tanenbaum.
Good fix, he typed back. Show me when I get home.
Are you done with your class?
Yes.
How was it?
He looked back at the engineering building. Room 302 was on the third floor — he couldn’t see it from here, just the facade of the building, the windows showing empty classrooms and a few occupied ones where late afternoon lectures were running.
It was good, he typed. The students asked real questions.
Like what?
He thought about Lee Minseok, in the back corner. I want to build something I’m not embarrassed by in twenty years.
Like what they want to build, he typed. And why it matters.
A pause. Then: What do I want to build?
The question from his nine-year-old. The question from Room 302, coming back to him from a different direction.
He typed: That’s the question. We’ll talk about it at dinner.
He put the phone away. Kept walking. The city moved around him — cars, students, the ordinary urgency of a Tuesday afternoon in Seoul. The engineering building receded. Ahead: the parking lot, the drive home, the dinner, the concurrency bug, the conversation about what to build and why it mattered.
The question was always the beginning. The building was the rest.
He had seventeen years of proof that the building was possible, that it was worth it, that the people you built with were the point rather than the method. He had a room that had been both the beginning and, now, something else — a place he could return to and stand in and recognize not what he had been, but what he had become.
Room 302. The same room. The different man.
Both were true. That, he had learned, was how it always worked.