The Accidental COO

The first game ended quickly. I had been playing janggi since boyhood, taught by an older brother who was better than I would ever be, and my new college roommate was good-natured about losing. Then I won the second game, and his face flushed in a particular way, not anger exactly, more the humiliation of watching yourself be beaten. We had to share an apartment for the rest of the year.

I let him win the next two. Not conspicuously; just enough to give him something to stand on. For years I remembered the move as a small generosity. I am less sure of that now. To lose on purpose is to exercise the same power that winning would have, only with a softer hand, and the room I was protecting was as much mine as his. The instinct was already there: when the choice was between filling a room and making space in it, I made space. It would decide how I held the most consequential job I ever had, and it would decide how I lost it.

I never wanted to be a chief operating officer. For most of my career I wrote software, and I took a leadership role only when the change I wanted could not be made from a keyboard. Once the change was made I tried to step back. The limelight has never appealed to me.

When the offer came I was at SmugMug, a company that took care of its people in the ordinary daily ways that are rare enough to notice, and that paid well besides. Leaving it was the hardest resignation of my life, and I did it for one reason. Someone was offering me the chance to build a company like that in Korea, where the working norm I knew was to spend engineers hard and retire them early, most people leaving the core job of their careers in their early 50s,¹ and where a company that put its people first would be, I thought, its own argument.

The someone was close to me. He ran a parent company with several subsidiaries beneath it, and this would be one more; we called it illuminarean. He had the ideas, and he wanted a team of developers good enough to turn them into products. My conviction was the one I was carrying from SmugMug, that the way to attract and keep such people was to put them first and let the rest follow.

The arrangement was stranger on paper than it was in the room. In the Korean filings I was the person in charge of illuminarean, and the parent gave me a director's title, though the other subsidiaries were his to run and I never saw inside them. I had no interest in either title. Day to day I ran illuminarean as its second: he held the vision, and I held the conditions that let it survive. My name was on the paperwork. His hand was on the wheel. I thought of that as a division of labor. It was also a division of accountability: if illuminarean failed, the man answerable for it on paper would be the one who had not been steering. I did not notice, then, that the arrangement had put me on that side of the line.

Nathan Bennett and Stephen Miles, writing in the Harvard Business Review after interviewing dozens of executives, found that the chief operating officer's job takes 7 distinct forms depending on what the chief executive needs, and that the one thread running through all of them is trust between the two people at the top.² I read that and recognized myself. I did not read it as a warning.

The pandemic years were when putting people first paid off in a way I could count. The market for developers was the hottest I had seen in 2 decades of building teams; startups in Seoul were offering signing bonuses of 100 million won to land a single senior engineer, and losing the candidate anyway.³ We grew to 30 in that stretch. A chief executive I knew from a university alumni group, whose small company had spent months trying to hire a third developer to sit beside the two it had, asked me over coffee how we were doing it. I did not have a tidy answer. What I had, it turned out, was a team that had been quietly recruiting for me, vouching for the place by name to people they had worked with before. After the pandemic the market turned fiercer still, every company trying at once to build the software teams it had put off building, and we kept hiring into it, until the payroll stood at close to 70.

That is what trust looks like from the inside, and it is the best thing I have ever helped build. People who felt seen did not just work harder. They stayed, and they brought others like themselves. I trusted my team leads completely, and they repaid it with the kind of work you cannot demand. When one of them told me a project needed another engineer, I believed it. When another said the new hire was worth the salary, I believed that too. I ran a tight ship on nearly everything. On this, I ran it on their word.

The end announced itself before it came. We were at our largest when the first signs of strain reached us from the parent, and we let about 10 people go, one of them a senior video and animation developer whose work I had admired from the day he arrived. I had come into that conversation to be the one doing the hard thing. He thanked me for the chance to have worked at illuminarean, and said he was glad of it, even now, even as we parted. I left the room having been consoled by the man I had just let go, and the wrongness of that order has not worn off.

illuminarean closed in April 2025. The proximate cause was above us: the parent company had grown faster than its cash could carry, and when the cash ran short, so did we. From where I sat it looked like weather, something arriving out of a sky I could not see into, and I could not have prevented it. I have stopped pretending otherwise to myself.

But I have also stopped pretending the failure was entirely elsewhere. Close to 70 was my number. I approved it one hire at a time, and some of those hires were approved because someone I trusted asked, not because I had checked. A second in command is many things, but he is supposed to be the one person in the building who verifies, precisely because everyone else is free to believe. The visionary believes; that is his job. The engineers believe; that is what makes them worth having. The operator's job is to be the person who is not free to, and I had made an exception for exactly the people I loved working with most. A leader with a harder edge and a sharper instinct for self-preservation would have carried a smaller payroll into that spring. Some of the people I let go might still have had jobs. I do not know that. I suspect it is partly true, and I do not want to write the version of this essay that leaves the suspicion out.

I am not saying trust was the mistake. The trust was the company. It was why close to 70 people came and stayed and did work I am still proud of, and it is why several of them are still in my life more than a year after the company is not. Nor is the lesson to trust less. The lesson is smaller and harder: trust and verification are not opposites, and an operator who treats them as opposites will end up choosing trust, because it feels like the better kind of person to be. It is not a failure of character to check the numbers behind a friend's request. It is the thing a friend in that chair owes him. Today no team lead would get an additional hire approved by me on the strength of the ask alone. The request would come with the current state of the workload and the daily progress against it, in writing and with the data to support it, so that the need is on record before the hire is, and so that the person we hire can be matched to the need rather than to the wish.

The chess game reads differently to me now. I used to think I let my roommate win so he could keep his footing. I think I let him win so I could keep being the kind of person who lets people win. There is nothing wrong with that person. But he needed a partner, the one who checks the numbers before he makes room, and for a while at illuminarean I sent only the first of them to work.

In the months after the closing, the letters came. One former employee wrote of having joined us at a point in her life that felt like a long tunnel, and of how the team's warmth had made it possible to take a step again. A team lead of more than 3 years wrote that no company would replace illuminarean in memory, and then, in the same message, apologized for not having been more help to me. Not every letter was kind; some were critical, and some of the criticism was fair. But I keep the one with the apology in it, because it has the order of things exactly backward, and because that is what a company looks like when the trust was real.

What remains is not small. People grew, and work went out into the world that is still doing its job in places I cannot see. Relationships outlived the organization. A leader's responsibility was never to make the company last forever. It was to make sure that what got built was worth building, and that the people who built it left better equipped than they came, and by that measure we did not fail. By the other measure, the one on the balance sheet, we did, and I owned a share of it that I once assigned to the weather.

The second in command is not asked to win. In the moments that matter he is not even asked to be in charge. He is asked to make the room worth the time the people in it are giving, and to be the one person in that room who never stops counting.