이 포스팅은 쿠팡 파트너스 활동의 일환으로, 이에 따른 일정액의 수수료를 제공받습니다.
Chapter 54: The Return
Dojun returned from paternity leave on a Monday in February 2013 and discovered that Aria had not, as he had secretly feared, fallen apart without him.
It had, in fact, grown.
“We added twelve enterprise clients while you were changing diapers,” Minjae reported, projecting the dashboard with the particular pride of a man who had held the fort and enjoyed every minute of it. The dashboard — Prometheus Labs’ internal metrics display, a custom-built tool that Dojun had designed three years ago and that Minjae had improved in his absence with the specific, unsupervised, I-finally-got-to-touch-the-boss’s-code energy of an engineer who had been waiting for permission to optimize — showed numbers that were, by any startup metric, excellent. “Consumer users crossed 800,000. Monthly recurring revenue: 620 million won. And—” He paused for dramatic effect, the engineer deploying theatrical timing with the same awkwardness that all engineers deployed theatrical timing. “TaskFlow’s CEO requested a meeting.”
“TaskFlow wants to meet?”
“Not compete. Meet. Their head of business development reached out to Jiyoung last month. They want to discuss ‘strategic alignment in the Asian market.’”
“Strategic alignment is corporate code for ‘we want to stop fighting you because fighting you is expensive.’”
“Or it’s corporate code for ‘we want to buy you.’”
“Same thing. Different payment schedule.”
The office had changed in his absence. Not structurally — the Yeouido space was the same second-floor, inadequately-ventilated, twenty-five-person workspace it had been when he left. But the energy was different. The team had operated without the founder for eight weeks and had discovered, through the specific, liberating, slightly-terrifying experience of functioning without the person who usually made the decisions, that they could function without the person who usually made the decisions.
This was, Dojun knew from his first life, the most important transition a startup founder could make. The moment when the company stopped needing the founder in the room. The moment when the organism developed its own immune system, its own decision-making capacity, its own ability to solve problems without deferring upward.
In the first timeline, this transition had never happened. Dojun had built Prometheus Labs around himself — the central node, the decision hub, the person through whom all code reviews and architecture decisions and hiring calls and investor meetings and product pivots and fire drills flowed. The company had been successful. The company had been him. And when he was gone — when the cancer took him and the company needed to function without the person it had been built around — the company had fractured. Not immediately. Not dramatically. But slowly, structurally, the way a building fractured when you removed the load-bearing column.
This time: the column was optional. The team had held the building for eight weeks. The building had not fallen. The building had grown.
“You did good,” Dojun said to Minjae. To the team. To the twenty-five people who had shown up every morning and made decisions and fixed bugs and closed deals and done the thousand small, unglamorous, essential things that a company required and that a founder’s absence had forced them to do themselves.
“We did our jobs,” Minjae said. “That’s different from doing good. Doing good implies exceeding expectations. Doing our jobs implies meeting them.”
“Doing your jobs while I was gone is exceeding expectations. My expectation was that you’d call me every day with emergencies.”
“We had emergencies. We solved them.”
“That’s the point. You solved them without me. That’s growth.”
“Or it’s evidence that you were the bottleneck.”
“Those are the same thing.”
Hana was at her desk. The co-founder’s desk, beside his, the side-by-side configuration that had been the company’s spatial foundation since the original Yeouido office. She was working — the headphones on, the design tools open, the specific, Hana-level, do-not-interrupt focus that the team had learned to respect the way you respected a weather system.
She did not look up when Dojun sat down. This was not rudeness — this was the Hana-protocol. When the headphones were on, the world was outside. The headphones-on state was the state in which Hana produced her best work, the deep-focus, single-task, empathy-to-pixels concentration that generated the user experiences that were, as the industry had begun to recognize, Prometheus’s actual competitive advantage.
Not the code. The experience. The code was excellent — Dojun’s architecture was decades ahead of the competition, literally. But the experience — the way the software felt, the way users described using Prometheus’s products not as “efficient” but as “pleasant,” the specific, measurable, Hana-generated quality that made a dashboard “feel fast” even when the load time was 2.7 seconds — that was what kept clients. That was what differentiated Prometheus from the hundred other AI analytics startups that had emerged in the 2011-2013 boom and that were building technically-comparable products without the Hana layer.
The Hana layer. Minjae had named it. The rest of the team had adopted it. It was now a standard term in Prometheus’s internal vocabulary: “Does it have the Hana layer?” meant “Does it feel right?” and “No Hana layer” meant “It works but nobody will love it” and the distinction — the distinction between working and being loved — was the distinction between a product and a brand.
Hana took off the headphones. Looked at him. The co-founder look — the specific, five-year, we-built-this-together look that carried, in its steady gaze, the accumulated history of a team project and a startup and a company and a relationship that was not romantic but was, in its own way, the most intimate relationship that either of them had.
“How’s Hajun?” she asked.
“He discovered his toes.”
“Important milestone.”
“He discovered his toes and then screamed about them for twenty minutes. The discovery was not entirely positive.”
“That’s the engineering process. Discovery, confusion, screaming, eventual acceptance.”
“You’ve just described every product launch we’ve had.”
“I’ve described every product launch in history.” She leaned back. The chair — the same budget office chair that everyone had, because Hana refused ergonomic upgrades until the whole team got them and the whole team getting them was “not in the budget, Dojun, the budget is for servers” — creaked with the familiar protest of overloaded furniture. “The TaskFlow meeting. I have opinions.”
“You always have opinions.”
“My opinions are data. The data says: TaskFlow is approaching from weakness. Their consumer product peaked six months ago — user growth stalled at 3 million. Their enterprise pivot is behind schedule. They’re not looking for strategic alignment. They’re looking for acquisition.”
“They want to buy us.”
“They want to buy our technology. Specifically, they want to buy our recommendation engine and our Hana layer. The engine gives them an AI advantage. The layer gives them user retention. Both are things they can’t build internally because their team is — I’m going to use a technical term here — bad.”
“Bad is not a technical term.”
“Bad is a precise description of a team that shipped a search algorithm with O(n-cubed) complexity in production. I reviewed their codebase. It’s publicly available. The variable names alone are a human rights violation.”
“You review competitors’ code?”
“I review everything. Design requires understanding the ecosystem. The ecosystem includes the code. The code is terrible. They need ours.”
“And your opinion on selling?”
“My opinion is: no. Not because the offer won’t be good — it will be excellent. Not because TaskFlow is bad — they have distribution and brand recognition and a hundred million dollars in annual revenue. My opinion is no because selling means integrating, and integrating means compromising the Hana layer, and the Hana layer is not a feature that can be integrated into another company’s product. The Hana layer is a philosophy. And philosophies don’t survive acquisitions.”
“In the other timeline, Prometheus sold. To a different company. Different circumstances. But the pattern was the same — the technology transferred, the philosophy didn’t. The products got faster and cheaper and less human. The Hana layer dissolved.”
“Then we don’t sell.”
“We don’t sell.”
“Good. Now tell me about the toes.”
The TaskFlow meeting happened on a Thursday. Not at Prometheus — at a neutral location, a conference room at the Grand Hyatt that cost more per hour than Minjae’s monthly salary and that communicated, through its marble floors and its view of the Han River and its specific, hotel-conference, money-is-not-a-concern aesthetic, that TaskFlow took the meeting seriously.
TaskFlow’s CEO was American. James Whitfield. Stanford MBA. The specific, Silicon Valley, I’ve-disrupted-three-industries-and-I’m-forty-two confidence that came from a career of building things fast and selling them faster and that was, in Dojun’s experience (both timelines), the marker of a man who understood growth but not value.
“Prometheus is impressive,” Whitfield said. The opening. The CEO-to-CEO opening that was, in every business meeting Dojun had attended in two lifetimes, the same: a compliment designed to establish a power dynamic in which the complimenter was the evaluator and the complimentee was the evaluated. “Your AI engine is — honestly, it’s the best I’ve seen outside of Google. And your user experience is — Hana-san, your work is extraordinary.”
“It’s Hana-ssi, not Hana-san,” Hana said. “And thank you.”
“My apologies. Cultural differences.”
“Cultural differences are the reason your Korean market share is 4% and ours is 31%. But yes, thank you.”
The meeting continued. Whitfield presented TaskFlow’s vision: global AI productivity platform, combined analytics and task management, leveraging TaskFlow’s existing user base (120 million globally) with Prometheus’s technology. The proposal was not acquisition — not explicitly. The proposal was “partnership,” which was the word that acquisitions wore when they wanted to seem friendly.
Dojun listened. In his first life, he would have been tempted. The numbers were large. The distribution was enormous. The partnership-that-was-really-an-acquisition would have given Prometheus access to 120 million users and the capital to scale the AI engine to a size that independent operation would require years to achieve.
But Dojun had lived the sequel. He knew what happened when a Korean startup partnered with a global platform: the technology migrated, the team dispersed, the Hana layer — the thing that made the product human rather than efficient — was optimized away by engineers who measured everything and felt nothing.
“We appreciate the interest,” Dojun said. “We’re not looking for a partnership at this time.”
“Every company is looking for the right partnership.”
“The right partnership doesn’t require one partner to become the other partner’s subsidiary. We’ll build our user base independently.”
“Building independently in the current market is — ambitious.”
“Ambitious is a compliment when you say it about your own company and a warning when you say it about someone else’s. We’ll take the compliment version.”
Whitfield was quiet. The specific, American-CEO, I’m-recalculating quiet of a man who had expected the small Korean startup to be grateful for the attention and who was discovering that the small Korean startup’s founder had the negotiating disposition of a man who had already built and sold a company in another lifetime and who was not, under any circumstances, going to make the same mistake twice.
“The offer stands,” Whitfield said. “When you’re ready.”
“We’ll call,” Dojun said. Knowing that they would not call. Knowing that the independence was the thing. The thing that the first life had sacrificed and that the second life would protect.
Saturday. Namdaemun Market. The pilgrimage.
Hajun was there. Four months old. Strapped to Dojun’s chest in a baby carrier — the specific, Korean-dad, I’m-carrying-my-son-through-a-market configuration that Dojun’s mother had insisted on (“the baby needs to smell real food, not that formula”) and that Dojun had adopted because the feeling of his son’s small body against his chest was, in the ranking of physical sensations that Dojun had experienced across two lifetimes, the best.
Better than the first successful compile. Better than the standing ovation at the Prometheus IPO. Better than the moment of waking in the 2006 lecture hall and realizing that the universe had given him another chance. Better than all of it. The weight of a four-month-old boy against his chest. The warmth. The small, sleeping breath.
His mother was at the stall. The banchan arranged. The japchae in the front. The eternal, non-negotiable, Namdaemun Market arrangement that had survived two timelines and thirty years and that would, Dojun suspected, survive the heat death of the universe.
“The baby is sleeping,” Younghee said. Looking at Hajun with the specific, grandmother, I-will-hold-that-baby-and-you-cannot-stop-me expression.
“He just fell asleep.”
“Then don’t wake him. Sit. Eat. Quietly.”
Dojun sat. He ate. Quietly. The japchae (Mr. Hwang’s sesame oil, the premium press, the family recipe). The kkakdugi (his mother’s, the specific fermentation that Dojun had never tasted at any restaurant in any city in any timeline because no restaurant could replicate the specific conditions of his mother’s refrigerator and his mother’s bacteria and his mother’s thirty-year relationship with the fermentation process).
He ate and his mother watched and the baby slept and the market hummed and the world — the 2013 world, the world of smartphones and social media and the specific, accelerating, everything-is-changing reality of an era that Dojun had already lived through once — continued around them with the indifferent, relentless, non-stop-for-anyone momentum of time.
“The company meeting,” Younghee said. “The American.”
“How do you know about the American?”
“Mrs. Kang’s grandson. Joonho. He tells his grandmother everything. His grandmother tells me everything. The information chain is more efficient than your company’s internal communications.”
“Joonho wasn’t in the meeting.”
“Joonho heard from Eunjin who heard from Minjae who heard from you. The company has twenty-five employees. Privacy is theoretical.”
“We said no. To the American.”
“Good.”
“Good?”
“If the American wanted to buy your company, the company is worth buying. If the company is worth buying, it’s worth keeping. You keep it.”
“That’s — surprisingly strategic.”
“I’ve been running a banchan stall for thirty years. I know when someone is trying to buy my recipe versus appreciating my food. The American wanted your recipe. He didn’t appreciate your food.”
“Mom, that’s the most accurate business analysis I’ve heard this year.”
“Business is food. Food is business. The only difference is the serving size.”
Hajun stirred. The small, four-month-old movement — the specific, mid-sleep, I’m-about-to-wake-up-and-be-hungry shift that babies performed and that parents learned to read the way meteorologists read cloud patterns. Dojun adjusted the carrier. The baby settled. The grandmother watched.
“He looks like you,” Younghee said. “When you were a baby. The same jaw. The same — the same quietness. You were a quiet baby. You watched everything.”
“I remember.”
“You can’t remember being a baby.”
“I remember being told I was a quiet baby. By you. Many times.”
“I told you because it was true. You watched. You computed. Even as a baby, you were computing.”
“Babies don’t compute.”
“My baby computed. My baby looked at the world like it was a problem to solve. And now my baby’s baby is looking at the world the same way. Look at his face.”
Dojun looked. Hajun’s face — four months old, the features still forming, the jaw that was his jaw and the eyes that were closed but that would be, when open, the specific, dark, watching eyes that Dojun had inherited from his mother and that Hajun would inherit from him. The baby’s face was not computing. The baby’s face was sleeping. But the sleeping face had, in its small, relaxed, trusting expression, the specific quality that Younghee recognized: attention. Even in sleep, the face was attentive. The face was receiving. The face was, in the way that all babies’ faces were, a frequency that was tuned to the world and that was recording everything.
“He’ll be like you,” Younghee said. “But better. Because he’ll have something you didn’t have.”
“What’s that?”
“A father who is present. You didn’t have that. I gave you everything I could — the food, the stall, the sesame oil, the japchae-in-the-front. But I couldn’t give you a father. And the absence — I know it shaped you. The absence is in your code. The absence is in the way you build things alone, at 3 AM, because you learned early that the only person who would be there at 3 AM was you.”
“Mom—”
“I’m not criticizing. I’m observing. The way you observed. The way Hajun will observe. The difference is: Hajun will observe a father who is present. Who comes to the market on Saturdays. Who carries the baby on his chest. Who says no to the American because the keeping is more important than the selling.”
Dojun looked at his mother. At the woman who had raised him alone, who had run a banchan stall and a household and a single-parent life with the specific, non-dramatic, daily heroism of a Korean woman whose options were limited and whose determination was not. She had never complained. She had never said “your father should be here.” She had made japchae and sold banchan and sent him to college and supported his incomprehensible career with the sustained, thirty-year, non-conditional love that was her specific, Park-Younghee, sesame-oil-scented version of infinite capacity.
“Thank you, Mom.”
“For what?”
“For the japchae. For the stall. For the thirty years. For — for making me someone who can carry a baby through a market on a Saturday and know that this is the most important thing I’ll do this week.”
“More important than the American?”
“Infinitely more important than the American.”
“Good.” She adjusted the japchae containers. “Now eat. The baby is sleeping. The japchae is optimal. The window for both is limited.”
He ate. The baby slept. The market hummed. The web held. The japchae went in the front.
And Park Dojun, twenty-seven years old, sixty-eight years of memory, seven years of second chances, sat in a market stall with his mother and his son and felt, with the specific, two-lifetime, earned-not-given certainty of a man who had built everything and lost everything and was building again, that the most important code he had ever written was not in any codebase in the world.
It was in the small, warm body of a four-month-old boy named Park Hajun, who was, at that very moment, sleeping against his father’s chest, breathing the specific, Namdaemun, japchae-and-sesame-oil air of a market that had been here before the Internet and would be here after whatever came next, dreaming dreams that no amount of future-knowledge could predict and that no amount of past-experience could improve upon.
The dreams of a baby whose father was present.
The best code is not the code you write. The best code is the life you compile.
And the life was compiling. Line by line. Meal by meal. Saturday by Saturday.
Exactly as designed.