이 포스팅은 쿠팡 파트너스 활동의 일환으로, 이에 따른 일정액의 수수료를 제공받습니다.
The board meeting was on a Tuesday in June 2015, and Dojun had been rehearsing his resignation speech for three weeks.
Not resignation, exactly. Transition. The word mattered—resignation implied failure, implied being pushed, implied the specific corporate drama that business journalists loved to write about and that never captured the actual human decision underneath. Transition implied choice. And this was a choice. The most deliberate choice he had made since the morning he woke up in a university lecture hall nine years ago with sixty years of memories and a twenty-year-old body.
He was sitting in the Aria conference room—the big one, the one they’d named “Debugging” because Hana had insisted that every meeting room needed a name that reminded people what they were actually doing—and the Seoul morning light was cutting through the floor-to-ceiling windows in clean diagonal lines across the table. Outside, the Gangnam skyline stretched in both directions, all glass and ambition and the particular Korean energy of a city that never quite stopped building.
The conference room seated twenty. Today, it would hold seven: the five board members, Hana, and him. Seven people who would decide the future shape of a company that now employed one hundred and forty-two people across four offices in three countries.
His phone buzzed. Hana.
You ready?
Been ready for three weeks.
That’s not what I asked. I asked if you’re ready, not if you’ve prepared.
He stared at the message. She was right, of course. Preparation and readiness were different animals. He had prepared meticulously—the presentation, the transition plan, the org chart, the timeline. But readiness was a feeling, not a deliverable, and feelings were the one category of human experience that sixty years of living and nine years of reliving had not taught him to manage efficiently.
I’m ready, he typed back.
Liar. But that’s okay. I’m nervous too. See you in ten.
He put the phone down and looked at the empty chairs. In forty minutes, those chairs would contain the people who had invested money and trust and reputation into the company he’d built, and he was going to tell them that the company’s founder was stepping down as CEO.
The word “stepping down” was wrong too. He wasn’t stepping down. He was stepping sideways. Into the role he should have been in from the beginning.
The first six months after the hearing had changed everything, and nothing.
Everything: the AI ethics framework had become an industry standard. Three other Korean tech companies adopted versions of Aria’s four-pillar model within six months. The Ministry of Science invited Dojun to chair an advisory committee on AI regulation. TechCrunch ran a profile titled “The CEO Who Testified Against His Own Product—and Won,” which was flattering and inaccurate in equal measure, because he hadn’t testified against the product. He had testified about responsibility. Those were different things.
Nothing: Dojun still spent his evenings writing code.
This was the problem. Not the code itself—the code was brilliant, innovative, the particular kind of architecture that happened when a mind with sixty years of perspective had access to 2015’s tools and 2025’s insights. The problem was what the code represented: where his heart actually lived.
Every morning, he sat in the CEO’s chair. He reviewed financials with Jiyoung. He took calls with investors. He approved marketing budgets and negotiated enterprise contracts and made hiring decisions and did the thousand small, essential, fundamentally non-technical things that a CEO did. And every morning, some part of him—the part that had spent forty years writing code before the cancer came, the part that still dreamed in syntax and woke up with solutions to architecture problems he hadn’t consciously been working on—that part sat quietly in the CEO’s chair and waited for the day to end so it could go home and do what it actually wanted to do.
Hana had noticed first. Of course she had.
“You’re writing code at 2 AM again,” she said one night in March, standing in the doorway of his home office with Hajun asleep on her shoulder—the toddler’s face pressed against her collarbone, the specific dead-weight sleep of a two-year-old who had spent the day discovering that gravity applied to all objects, including himself. “The commit timestamps don’t lie.”
“I had an idea for the prediction engine—”
“You always have an idea for the prediction engine. That’s not the issue. The issue is that you’re doing CEO work during the day and CTO work at night, and you’re sleeping four hours, and you have circles under your eyes that could hold rainwater.”
“I sleep five hours.”
“Dojun.”
“Five and a half.”
“The number is not the point. The point is that you’re doing two jobs because you can’t admit that you only want one of them.”
He leaned back in his chair. The monitor’s glow made the room feel like a submarine—blue light, the hum of the machine, the world outside reduced to darkness and the distant sound of Seoul traffic. Hajun shifted in Hana’s arms, murmured something that might have been “appa” or might have been a dream about trucks, which was his current obsession.
“I built this company,” Dojun said. Not defensively. Factually. “I’m the CEO because I’m the founder.”
“You’re the CEO because no one questioned it. You started the company, so you sit in the big chair. That’s startup logic. But we’re not a startup anymore, Dojun. We have a hundred and forty people. We have offices in Seoul, Tokyo, and Singapore. We have enterprise clients who expect their CEO to be a CEO, not a programmer who occasionally attends board meetings.”
“I attend every board meeting.”
“You attend them the way Hajun attends dinner. Physically present. Mentally somewhere else. Last meeting, David asked about our go-to-market strategy for Southeast Asia, and you answered with a three-minute explanation of how the new caching layer would reduce latency for users in Indonesia.”
“Latency is a go-to-market strategy.”
“For engineers. For business people, go-to-market means pricing, partnerships, distribution channels, and localization. You answered the wrong question because you were thinking about the right problem—for a CTO. Not for a CEO.”
He was quiet. Outside, a siren passed—the Seoul night soundtrack, distant and constant and part of the city’s breathing.
“I know,” he said.
“You know what?”
“I know I’m a better CTO than CEO. I’ve known since the hearing.”
Hana shifted Hajun again. The toddler’s weight was distributed unevenly—twenty-six pounds of sleeping human concentrated in the specific, physics-defying way that toddlers managed, where all the mass seemed to migrate to whichever arm was already tired.
“What happened at the hearing?”
“I stood in front of the National Assembly and talked about AI ethics and regulatory frameworks and corporate responsibility. And I did it well—I know I did it well, the committee responded, the public responded. But the entire time, in the back of my mind, I was thinking about a code refactor. A specific refactor—the consent mechanism in the prediction engine. I was testifying about the most important corporate crisis we’d ever faced, and part of my brain was writing code.”
“That’s not a character flaw, Dojun. That’s your identity.”
“A CEO’s identity should be the company. All of it. Not just the code.”
“Then maybe you shouldn’t be CEO.”
The sentence landed in the room like a stone in still water. Not angry. Not accusatory. Just true, in the way that true things were often simple and quiet and obvious only after someone said them out loud.
“Maybe you shouldn’t be CEO,” Hana repeated, softer this time. “And maybe that’s okay.”
The conversation continued over the next three months. Not as a single discussion but as a series of moments—late-night talks, morning walks with Hajun in the stroller, a weekend trip to Namhae where the sea was flat and gray and conducive to honest conversation.
They talked about what the company needed. Aria had grown beyond the startup phase. The product was mature—version 4.2, serving 1.8 million users across twelve countries. The engineering was world-class. What the company needed now was not better code but better strategy, better storytelling, better relationship-building with the enterprise clients who were signing seven-figure contracts and expecting a CEO who spoke their language.
They talked about Hana. About how, during Dojun’s paternity leave, she had run the company not just competently but brilliantly—closing the partnership with TaskFlow, negotiating the Series C terms that David Yoo later called “the best founder-friendly deal I’ve seen in Korean tech.” About how enterprise clients consistently rated their satisfaction higher when Hana led the meetings. About how the team—the hundred and forty people who showed up every day—had a different energy when Hana was in the room. Warmer. More cohesive. More willing to argue, which was paradoxically a sign of trust.
“You make people want to build,” Dojun told her during the Namhae trip, standing on the rocky beach while Hajun investigated a tide pool with the focused intensity of a scientist discovering a new ecosystem. “I make people want to solve problems. Both are important. But the first one is what a CEO does.”
“And the second one is what a CTO does.”
“Yes.”
“So you’re saying—”
“I’m saying you should be CEO. And I should be CTO. And we should have done this two years ago.”
Hana picked up a flat stone and skipped it across the water—three bounces, the last one barely a kiss on the surface. She had always been better at skipping stones. One of the hundred small things he knew about her that had nothing to do with business and everything to do with why he trusted her with the most important thing he’d ever built.
“The board won’t like it,” she said.
“David will love it. Eunji will be cautious but supportive. The other three will need convincing.”
“And Seokho?”
“Seokho will say something sarcastic and then admit it makes sense.”
“And your mother?”
Dojun smiled. “My mother will ask if you’re getting a raise.”
“Am I?”
“You’re getting my job. That’s either the biggest raise or the biggest punishment, depending on the day.”
Hajun looked up from the tide pool. “Appa,” he said, holding up a small crab with the triumphant expression of a conqueror presenting captured territory. “Bug.”
“That’s a crab, Hajun.”
“Bug,” Hajun insisted, with the absolute certainty of a two-year-old who had classified all small creatures into a single category and saw no reason to complicate the taxonomy.
“He gets the stubbornness from you,” Hana said.
“He gets the confidence from you,” Dojun said.
“Same thing.”
The board meeting started at 10 AM.
David Yoo was the first to arrive—early, as always, with the particular punctuality of an investor who valued time because he’d spent thirty years learning that it was the only non-renewable resource. He wore a suit that cost more than Dojun’s first year of university tuition and carried a leather portfolio that he never opened because he kept everything in his head.
“Park,” David said, shaking Dojun’s hand. “You look serious.”
“I have a proposal.”
“Good proposals don’t require that expression. That expression means you’re about to do something that makes business sense but feels emotionally complicated.”
“How do you know that?”
“Because I’ve been an investor for thirty years, and every founder who changes the company’s leadership structure makes that exact face. It’s the ‘I know this is right but I need you to agree’ face.”
Choi Eunji arrived next—still sharp, still direct, still wearing the particular no-nonsense energy of a venture partner who had bet on Aria when it was five people in a room and had never once regretted it. She nodded at Dojun, sat down, and opened her laptop with the efficiency of someone who intended to be productive regardless of what the meeting contained.
The other board members—Park Jinsoo from Meridian Capital, Lee Soojin from the Korean Development Bank, and Tanaka Kenji from Sony Ventures—arrived in quick succession, each carrying the polite, slightly-wary expression of board members who had been called to an unscheduled meeting.
Hana entered last. She was wearing the blazer—the specific, navy blue, tailored blazer that she wore to meetings where the outcome mattered. Dojun had learned to read her wardrobe the way meteorologists read cloud formations: the blazer meant preparation, significance, the intention to be taken seriously not through argument but through presence.
She sat beside him. Not across from him—beside him. A deliberate choice. Whatever he was about to say, they were saying it together.
“Thank you for coming on short notice,” Dojun began. “I’ll be direct. I’m proposing a change in Aria’s leadership structure. Effective September 1st, I would like to transition from CEO to Chief Technology Officer. And I’m recommending that Lee Hana be appointed as CEO.”
Silence. The specific, board-meeting silence that meant everyone was calculating simultaneously—financial implications, market perception, investor confidence, talent retention.
David Yoo was the first to speak. “Why?”
“Because the company needs a CEO who leads with vision, communication, and relationship-building. Hana does all three better than I do. And it needs a CTO who can focus entirely on the technology that differentiates us. That’s me.”
“You’ve been both.”
“Poorly. I’ve been a decent CEO and a frustrated CTO. The company deserves an excellent CEO and an excellent CTO. This gives it both.”
David looked at Hana. “And you want this?”
Hana’s voice was steady. The blazer voice—calm, clear, the voice of someone who had thought about this for three months and arrived at a conclusion that felt not like ambition but like alignment.
“I’ve been doing parts of this job for two years,” she said. “During Dojun’s paternity leave. During the hearing preparation, when he was focused on the technical framework and I was managing investor relations and media. During every enterprise meeting where the client wanted to talk strategy and Dojun wanted to talk architecture.” She paused. “I’m not proposing to change the company. I’m proposing to formalize what’s already happening.”
Eunji leaned forward. “The market will react. A founder stepping down as CEO—even laterally—signals instability.”
“Only if we present it as stepping down,” Dojun said. “We’re presenting it as specialization. The company has grown beyond the stage where one person can be both the technical visionary and the business leader. This is a maturation signal, not a distress signal.”
“The stock—”
“We’re private.”
“The investors, then. The Series D conversations.”
“Will go better with Hana at the table. She closed the TaskFlow partnership. She negotiated the Series C terms. She managed the hearing crisis communications. Every data point says she’s the better CEO.”
Park Jinsoo from Meridian cleared his throat. “I have a question for Lee Hana. What’s your vision for Aria in three years?”
Hana didn’t hesitate. “Three years from now—2018—Aria will be the platform that every knowledge worker in Asia uses, not because their company tells them to, but because they can’t imagine working without it. We’ll have three million users, forty billion won in annual revenue, and we’ll be preparing for either an IPO or a strategic partnership that values us at three hundred billion won or more.” She folded her hands on the table. “Dojun built the technology. I’ll build the business. Those are not competing goals—they’re the same goal seen from two sides.”
Tanaka Kenji from Sony Ventures, who had been silent, spoke in his quiet, precise Korean. “At Sony, we had a similar transition in 2003. The founder’s grandson stepped back from operations. It was presented as a failure. The stock dropped twelve percent in a week.” He paused. “But the company that emerged was stronger. The question is not whether this is the right decision. The question is whether you can execute the transition without the narrative becoming about departure rather than evolution.”
“That’s a communications problem,” Hana said. “And communications is what I do.”
The vote was 4-1. Eunji voted yes with reservations. David voted yes without any. Park Jinsoo voted yes after asking three more questions. Tanaka voted yes with a slight bow toward Hana. Lee Soojin from KDB voted no—not because she disagreed, she explained, but because she believed the timing should be delayed until after the Series D closed.
“September 1st,” Dojun confirmed. “Three months to prepare.”
“Three months,” Hana agreed. She looked at him across the conference table—the particular look that meant we just changed everything again—and he felt, for the first time in months, the specific relief of a person who had been sitting in the wrong chair and had finally found the right one.
The announcement went out on July 15th. Hana had spent three weeks crafting the narrative—not because she was spinning the truth but because she understood that truth, like code, needed architecture. The same facts, arranged differently, told different stories. “Founder Steps Down” was one story. “Company Matures: Specialized Leadership for the Next Phase” was another. Both were true. Only one was useful.
The response was, by startup standards, remarkably positive. The Korean tech press—which had been following Aria since the Showcase days—ran profiles of Hana that focused on her design background, her role in the hearing response, and the increasingly-cited “Hana Layer” that industry analysts credited with Aria’s unusually high user satisfaction scores. Bloomberg ran a brief noting that founder-to-CTO transitions were “increasingly common in maturing Asian tech companies” and cited Aria as “a textbook example of strategic role specialization.”
Seokho called on the day of the announcement.
“So you finally did it,” he said, his voice carrying the particular tone of someone who had been waiting for this moment and was mildly annoyed that it had taken so long.
“Did what?”
“Admitted that you’re an engineer, not a CEO. I’ve been saying this since 2009.”
“You’ve been saying a lot of things since 2009.”
“Most of them correct. This one especially. You belong in front of a compiler, Dojun. You always have. The CEO thing was a costume you wore because nobody else fit it at the time.”
“And now?”
“Now Hana fits it. She’s always fit it. She’s better with people, better with strategy, better with the parts of the job that require being human rather than being brilliant.” A pause. “I mean that as a compliment. To both of you.”
“I know.”
“Nova’s IPO roadshow starts next year. Come to the pre-filing dinner. Bring Hana. I want my investor base to see what a real tech company looks like when it has the right people in the right seats.”
“That’s the nicest thing you’ve ever said to me.”
“Don’t get used to it. I have a quarterly earnings call in twenty minutes and I need to get back to being terrifying.”
September 1st arrived with the quiet inevitability of dates that have been circled on calendars for months.
The transition was not ceremonial—no speeches, no champagne, no corporate rituals. It was, instead, a Tuesday. Hana moved into the corner office. Dojun moved into the engineering wing—a desk among his team, surrounded by monitors and whiteboards and the particular organized chaos of a development floor. He had specifically requested no private office.
“The CTO should be in the middle of the code,” he told the facilities team. “Not watching it from behind glass.”
His new desk was between two junior engineers—Yoon Jihye, who had joined six months ago from POSTECH and wrote backend code with the elegant minimalism of a haiku, and Kang Minho, who had been with the company for three years and whose debugging instincts were, in Dojun’s private assessment, better than his own.
“Isn’t this weird?” Jihye asked on his first day, with the candor of a twenty-four-year-old who had not yet learned to filter honesty through corporate hierarchy. “The founder sitting with the grunts?”
“You’re not grunts. You’re the people who build the thing that everyone else sells. And no, it’s not weird. It’s where I should have been the whole time.”
“But you were CEO.”
“I was. And now I’m not. And now I’m going to spend my time on the thing I’m actually good at.”
“Which is?”
“Making the code do things that nobody thinks code can do.”
Minho glanced over from his monitor. “Is that a promise or a threat?”
“Both.”
The first thing Dojun did as CTO—officially, formally, with the title on his business card and the engineering team’s Slack channel renamed from “Engineering” to “Engineering (now with adult supervision)”—was kill three projects.
Not dramatically. Not vindictively. But with the specific, surgical clarity of someone who could finally see the engineering roadmap without the CEO’s obligation to say yes to everything that had revenue potential.
Project Horizon—an enterprise analytics dashboard that three clients had requested and that engineering had been building reluctantly for six months—was the first to go. Not because it was bad. Because it was mediocre, and mediocre consumed the same resources as excellent without producing the same results. The same twelve engineers could, Dojun calculated, build something transformative in the time they were spending on something adequate.
“The dashboard works,” Minjae protested, because Minjae was the kind of engineer who measured success by functionality and was offended by the cancellation of anything that functioned.
“It works the way a sedan works. Reliably, competently, forgettably. Nobody will switch to Aria because of a dashboard. They’ll switch because of something they’ve never seen before.”
“And what have they never seen before?”
Dojun pulled up a whiteboard. He drew a single circle in the center and wrote one word inside it: AGENT.
“Imagine,” he said, “an AI that doesn’t wait for you to ask it questions. An AI that observes how you work—your patterns, your habits, your recurring problems—and proactively solves them before you notice they exist. Not a tool. Not an assistant. An agent. Something that acts on your behalf with your permission and your values.”
The engineering team—fourteen people gathered around the whiteboard in the way engineers gathered around whiteboards, which was with skepticism and interest in equal measure—was quiet.
“That’s five years away,” Minjae said.
“Three,” Dojun said. And he knew it was three because he had seen it—not this exact technology, not in this exact form, but the underlying architecture, the approach, the specific combination of machine learning and user modeling and behavioral prediction that would, in his first life, become the foundation of every major AI platform by 2025.
But this was 2015. The tools were primitive by 2025 standards. The compute was expensive. The data infrastructure was adequate for analytics but nowhere near sufficient for real-time behavioral modeling. Building an AI agent in 2015 was like building a skyscraper with Bronze Age tools—theoretically possible, practically insane, and exactly the kind of challenge that made Dojun’s brain light up in a way that quarterly earnings reports never had.
“I’m not asking you to build it tomorrow,” he said. “I’m asking you to start thinking about it. The prediction engine we built for version 3—the one that anticipates user needs based on calendar and email patterns—that’s the seed. An agent is the tree.”
Jihye raised her hand. The gesture was oddly formal for a team meeting, but Jihye was oddly formal in general—the POSTECH precision, the habit of treating every conversation as a technical review.
“You said ‘acts on your behalf.’ That implies autonomy. Autonomy implies trust. We just spent six months rebuilding trust after the stalking incident. An autonomous AI agent is the stalking problem multiplied by a thousand.”
“Yes,” Dojun said. “Which is why the ethics framework isn’t a constraint on this project—it’s the foundation. Every capability the agent has will be built on consent, transparency, and user control. Not as afterthoughts. As architecture.”
“Ethics as architecture,” Minjae murmured. “That’s either the most important sentence in AI development or the most naive.”
“It’s both. The important things usually are.”
He called his mother that evening.
“Eomma. I changed my job.”
“You got fired?”
“No. I moved from CEO to CTO.”
A pause. The particular pause of a woman who had spent her life in Namdaemun Market selling banchan and who processed corporate terminology the way Dojun processed recipe instructions—with effort, suspicion, and the general sense that the whole thing was unnecessarily complicated.
“Is CTO higher or lower than CEO?”
“It’s… different. CTO means I focus on the technology. The computers. The code.”
“So you went back to the computers.”
“I never left the computers, Eomma.”
“But before you were also the boss.”
“Hana is the boss now.”
Another pause. Longer this time. He could hear the market in the background—the distant clatter of Mrs. Kang’s stall, the vendor three doors down shouting about persimmons, the particular late-afternoon soundscape of Namdaemun that had been the backing track of his childhood.
“Hana is a good boss,” his mother said. Not a question. A judgment rendered with the absolute confidence of a woman who had evaluated Hana’s character through the only metric she trusted: how the girl treated food.
“She is.”
“And you’re happy? With the computers?”
“I’m happy, Eomma.”
“Happier than when you were the boss?”
He thought about it. About the three months of preparation, the board meeting, the announcement, the move to the engineering floor. About sitting at his new desk between Jihye and Minho, surrounded by code and conversation and the particular energy of people who built things for a living. About the whiteboard with AGENT written in the center, and the fourteen engineers who had looked at it with skepticism and excitement, and the feeling—the specific, physical, almost-tangible feeling—of being in the right place.
“Yes,” he said. “Happier.”
“Then it was the right decision. When you do the right thing, your shoulders go down. I can hear it in your voice. Your shoulders are down.”
“How can you hear shoulders in a voice?”
“Because I’m your mother. I’ve been listening to your shoulders for twenty-nine years.”
“I’m twenty-nine, Eomma. Not—” He caught himself. In this life, he was twenty-nine. The correction was automatic, the reflexive arithmetic of a man who carried two ages and occasionally forgot which one was real.
“Come Saturday. I’ll make galbitang. And bring Hajun. Mrs. Kang has been asking about him.”
“Mrs. Kang sees Hajun every week.”
“And every week she asks about him again. That’s what grandmothers do. We ask about children we already know about because the asking is the point.”
He smiled. The particular smile that only his mother could produce—not the polished, public smile of a founder-turned-CTO, but the unguarded, slightly crooked, deeply-personal smile of a son who had lived twice and still found his mother surprising.
“Saturday,” he said. “Galbitang.”
“And Hajun.”
“And Hajun.”
On his third week as CTO, Dojun stayed late on a Friday.
Not because of a crisis. Not because of a deadline. Because he wanted to.
The engineering floor was empty—the team had gone home at reasonable hours, because one of Dojun’s first CTO mandates was that nobody worked past 8 PM unless something was literally on fire, and even then, they should call the fire department first and debug second. The lights were dimmed to the evening setting, the monitors casting blue rectangles on the empty desks, the server room humming its constant low note behind the wall.
Dojun sat at his desk and opened the codebase. Not to fix anything. Not to build anything. Just to read.
He scrolled through the Aria architecture—the layers of code that he and his team had written over five years, from the first prototype in a Yeouido apartment to the current platform serving 1.8 million users. He read the comments, the commit messages, the function names. He traced the logic from user input to prediction output, following the data through the pipeline the way a river followed its bed—naturally, inevitably, the shape of the channel determining the flow.
The code was good. Not perfect—no code was perfect, and the pursuit of perfection was the specific engineering delusion that produced beautiful systems nobody could maintain. But good. Solid. The kind of architecture that would hold weight as the company grew, that would bend without breaking when new features were added, that would age the way good buildings aged—with character, not decay.
He opened a new file. Typed a comment at the top:
// Project Lighthouse
// An AI that acts, not just answers
// Started: September 2015
// Goal: Make the user's life better without being askedThen he began to write. Not the agent itself—that was years away, layers of research and testing and ethical framework-building away. But the foundation. The data model. The permission architecture. The specific, careful, consent-first framework that would allow an AI to observe behavior and take action without becoming the thing they’d testified against at the National Assembly.
The code came easily. Not because it was simple—it wasn’t. It was some of the most complex architecture he’d ever designed, in either lifetime. But it came easily because he was doing the thing he was meant to do, in the place he was meant to be, with the particular clarity that arrived when a person stopped pretending to be something they weren’t and started being what they were.
The cursor blinked. The server room hummed. Seoul breathed outside the windows, ten million people and their ten million stories, and somewhere in a market in Namdaemun, his mother was closing the banchan shop for the night, counting the day’s earnings with practiced fingers, not knowing and not needing to know that her son was sitting in an empty office writing the first lines of something that would change how humans and machines worked together.
He wrote until midnight. Then he saved the file, closed the laptop, and went home to his family.
The right seat. Finally.