2025 Retrospective

flex team

Until July, I led and executed a project in the Web Client Platform Team (WCP) to end the regular release cycle, which had been the biggest bottleneck in our web product delivery process. Judging only by the outcome, it was a good thing. But it required making extremely sound decisions on matters that would determine the future of the codebase. The process was long and grueling, and many moments of execution were risky. I lived in constant tension, gradually getting mentally exhausted. It was an extension of the state I wrote about in last year’s retrospective. “If I didn’t pull myself together and make good decisions in a good state, I couldn’t do anything properly.”

After it wrapped up, I gave a talk at a developer conference in August. Monorepos and code governance are topics that make for good talks, so the audience response on site was strong. It was the largest crowd I’ve ever spoken in front of in my life. I was delighted that we embedded a QR code for our team’s hiring post into the YouTube video. Around then, my mental energy was depleted, and every day felt slightly like I might throw up. I think it was because there wasn’t really a clear moment to catch my breath even after finishing something big.

In the fall, I took a single LLM-calling API from the server side and, in something like a playground, kept building demos with a fellow PM and engineers—whatever they suggested like “Could we do this or that with AI inside the product?”—to get people’s heads turning. In the flex product, you could say, “Make an Acid Rain game in the name of a member of our company,” and the AI would make it—stuff like that. I wrote a lot of prompts and started thinking seriously, around that time, about how to integrate AI into a product well. It was the most unusual kind of work I’d done within the team. It reminded me that making software is, roughly speaking, a way to turn imagination into reality—and that I like that. I’m most wary, in life, of cynicism; good imagination lives the farthest away from cynicism.

In the last quarter, I joined a squad building an AI Agent product. After 3 years and 9 months in the developer platform organization, I transferred into a product-building squad. Starting in October, I switched into a mode of working long hours to ship. While working 50+ hours a week, I’m exploring—and trying various things—around how to live happily without being filled with anger, and how to savor the beauty of the world. I’ll probably live in this mode through the entire first half of next year. When I first joined the platform team, the entire platform org was three people; now it’s over twenty. I’ve also contributed a lot—head-first—to the team’s goal-setting, systems, ways of working, and DNA. I hope many colleagues feel fulfillment and grow well on top of that foundation.

Development

Like many people, I can’t talk about development this year without talking about AI. The second half of the year was when I fully introduced AI assistants, led by Claude Code, into my workflow. The trigger was something I heard when meeting a developer friend: “An intern at our company doesn’t know development at all, but they’re vibe-coding all the way to git commits—and the product is getting built. It’s insane. If developers can’t ride this wave, they’ll become Heungseon Daewongun.” I’d tried it only in a half-hearted way and gotten disappointed a lot, but that was when I decided I needed to go all in.

I tried to shift my workflow to delegate as much as possible to AI. I consciously tried to take my hands off the keyboard, even at the cost of short-term productivity loss. I wrote instruction prompts in the ticket body in Linear and let it vibe. If I didn’t like it, I’d delete everything, revise the ticket body, and make it start over from the beginning. I repeated this until I got an output I was satisfied with. After living like that for about a month, a plausible CLAUDE.md emerged from the process, and Claude Code latched tightly onto my flow. Now I use about 200 million tokens per week (per ccusage). It’s great and fortunate that I work at a company where I can use Claude Code practically without limits.

In the first half, I was very interested in the latest approaches to Micro Frontends and in strengthening polyrepo governance, and I tried to keep up. In the second half, AI became my primary exploration path. I was interested in using AI for development, but I was also interested in attaching AI to an existing product—something I felt was comparatively less explored in the ecosystem. I also studied how to build Generative UI well and how to design chat UIs. Toward the end of the year, I returned to a product squad, and it’s been a long time since I last built product. I opened Figma for the first time in a long while, and it’s been a while since I last talked with designers. I’m trying to combine the latest approaches to global state management, styling, and data management with the approaches I prefer, and with what AI products require. I feel a bit disoriented after being assigned to the squad, but I want to do well even in this newly given environment and opportunity.

I’m slowly working on a personal project. Right now it’s an AI application that takes notes on articles and books I’ve read, uses them as material to write prose, and sends the result via email. I’m building it to automate blog writing and to publish AI-written posts on the blog as well. To delegate the work to AI, I had to make my writing process concrete and explainable. I ended up thinking a lot about how I write. If I just threw in references and said, “Write an article about this,” the quality was never high. I had to explain everything: from choosing the topic, to selecting references that fit the topic, to choosing a writing method aligned with both the topic and references, to choosing and explaining an editing method aligned with that writing method. Only then could I prompt in a way similar to how I write. Now I’ve split those processes out and am improving prompt quality. Writing is far harder than coding because it’s difficult to set a verifiable reward, and it’s very labor-intensive. This year, I want to produce something I’m satisfied with and show it to people.

People

Even in 2025, if someone asked for a coffee chat, I usually showed up. I basically fixed the location at the Mahogany Cafe on the first floor of Homeplus at Jamsil Station and kept meeting people. In the first half, I met people I see every year, and it was nice to see people who had moved from our team to other companies.

After getting hyped from the conference, I rode the momentum and did about 13 private coffee chats. Then I met too many people and reached a state where I could no longer talk to anyone. I’m a very introverted person (INFJ), but I also like talking and going out, so I receive all the mental shock and fatigue in full—an odd personality. From 4Q, I couldn’t even reply properly; I’m very sorry. It’s because I pushed myself too hard. I have no promised timeline, but I will reply to everyone…

A lot of people struggled at work, and hiring isn’t going well. In situations like that—when someone else is having a hard time—I often wondered how I could possibly dare to give advice. To people struggling at work, I said: challenge yourself enough, but protect yourself. For job hunting, I had nothing to say. In most cases, even by my own judgment, there wasn’t anything more they “should” be doing—they were already preparing extremely well. And yet, if they still want to build a career as a developer, there’s nothing else to do but keep going and keep knocking. It’s really hard.

I want to meet more people from roles other than developers. In the summer, I attended a homecoming event for seniors from my university’s independent press, where I used to be an editor-in-chief. I received dozens of business cards from people who had already become reporters. For them, meeting developers was so rare that they found me fascinating and didn’t know what to ask. In that sense, the way conversations worked changed completely, and I felt a big sense of freshness from having conversations I’d never had before. I also felt respect—so many people are walking a path I gave up.

An absurd number of people got married. When I counted, I went to 13 weddings this year. One of my closest friends—someone I’ve known for over 10 years—also got married, and I gave my highest wedding gift amount ever. I still enjoy seeing one of the happiest scenes in people’s lives. But after paying wedding gifts so often, I started wondering if I’ve become part of this wedding ecosystem—since some unknown point in time—without realizing it. Will this strange sense of belonging end up making me get married? I don’t know. I’ll think about it next year.

Health

I ran 950 km this year. After switching into a heavy-work mode near the end of the year, I kept strength training consistently, but I didn’t have the psychological slack to run. Running has a higher “cost.” But I didn’t set a goal for how many kilometers I “should” run anyway. Next year I’ll run whenever I feel like it—and in April, I’m running a half marathon.

I want to go out at night and also study personally, but once I work out after work, the day is just over, and I can’t stand that. So I tried to make morning workouts a habit. After fighting myself desperately for about half a year, it became a bit easier to shake off that cosmic force of not wanting to exercise immediately after waking up. This might be the biggest gain of the year. I either run 8 km in the morning or do strength training at a gym near the office.

My herniated disc improved a lot, thankfully. I caught a few colds. Maybe because I’m getting older, my body feels a bit drier. Next year, I want to make meditation a routine too. I still hate hospitals so, so much, but I want to get better at going.

Finance

I spent some money on traveling, but I succeeded in cutting spending quite a bit compared to last year. Part of it was trying not to buy many things. Also, I’ve been living in my current home for well over three years now, so there isn’t much I particularly want or need anymore.

I improved the constitution and weightings of my portfolio a lot, so compared to last year, I built an environment where I can invest with relatively more peace of mind. At the same time, I raised my equity allocation a lot and greatly increased the amount per buy in a single name. I truly felt how important cash is, and I think I’ll keep thinking deeply about how not to dismantle my investment allocation when large expenses arise. I say the same thing every year at year-end. “I was satisfied with my investment return this year. Next year I have no confidence, and I’m worried about the market situation, so I still don’t know what strategy to take. I want to invest with a bit more peace of mind, but it really isn’t easy.”

I want to study real estate a bit. People around me bought homes occasionally this year. I’m living pretty roughly right now, but housing is something you can’t avoid forever. There’s probably value in knowing certain things in advance.

Life

It was a year spent almost entirely together with my partner. I tried really hard to have fun on weekends. We went around a lot together. We spent the most time in Sindang, Jamsil, and Seokchon, and also enjoyed Yongsan, Haebangchon, Yeonnam, Hannam, Yaksu, Geumho, Seongsu, Seoul Forest, Konkuk Univ., Anguk, Seochon, and Maebong. We went to Cigarettes After Sex and Coldplay concerts together, looked at a lot of furniture in showrooms, drank about six bottles of wine together, and ate a lot of peaches, tangerines, strawberries, mochi bread, Shine Muscat, and Lalasweet pint ice cream. Hanging out with my partner was explosively fun, to the point where I once came back from a developer coffee chat and rambled about how insanely funny and awesome my partner is—how can someone be like that…

I took three domestic trips. During a two-week refresh vacation, I walked 160 km along the East Coast over 10 days. The difficulty was lower than I expected, so next time I think I should aim for about 20 km a day. I’ll probably keep doing a bit each year toward the goal of completing the Haeparang Trail. I already planned everything for this year too. Since I don’t travel that much, I didn’t really have a “taste” for it, but I realized I like trying to deeply understand the humanistic environment and subtle points of the places I travel to. I like learning that Yangnam-myeon in Gyeongju is a National Geopark, understanding what kind of volcanic activity created columnar joints, and why there are so many volcanic landforms along Korea’s East Coast. Compared to going to somewhere like Europe once in a while, I think it’s much more fun to visit places or countries with some shared “homogeneity” multiple times.

I did so, so many things that by year-end I was completely in a groggy state. And then I got put into a busy project at work, so from the start of 4Q until now I’ve been working 50+ hours a week. While reflecting that maybe I became a bit cynical from being tired, I also learned that after getting tired this time, I’ve built a certain mental resilience to these situations. I’ll keep trying to maintain high intensity and density at work while still making efforts to have fun.

I’m moving next year. I’ve lived alone for 10 years, and I decided to stop doing that. It’s an enormous decision in my life.

Writing

I published 14 blog posts.

I Should Stop Growing / Someone I Want to Work With in flex Team WCP / Types of Cohesion Between Software Modules / I’ve Never Seen Contribution Rotate Well in a Dev Team / The Dependency Protocol of Node Package Managers / Good Engineering Opportunities Themselves Are Unequal / Rewatching The Newsroom 10 Years Later / Module Federation Changes by Version (1.0, 1.5, 2.0) / May 18 for Me, My Family, and My Neighbors / How Will AI Change Software UI? / Agent2Agent Protocol / Interviews I Listened to More Than Twice This Year / How to Build an App That Generates UI with AI / Haeparang Trail Courses 1–10 Trekking Review

I wrote quite a lot. I consciously tried to write regularly. I ended up writing several topics in parallel, very slowly. I want to publish more “nonsense” posts too, not just dev posts. But in some ways, that’s even harder than writing dev posts. Next year, I want to renew my blog with vibe coding.

Reading

I read 19 books.

Freedom with a Pint of Beer Between Love and Control / Concept Class / 2024 Year-End Tax Settlement Guide for Office Workers / How to Interview / Back Mechanic: Diagnosis Edition / Back Mechanic: Treatment Edition / The Best Prompt Engineering Lecture / Slow Work / Beyond Death, Beyond the Darkness of the Times / Building a Chatbot Using the ChatGPT API / Zero to One / The Evolution of Investing / Running LangChain / On the Whiteness of the White Whale / Silicon Valley Team Leads / Dirty Work / Real Estate Investing That Changes Your Fate — Basics / The Myth of Sisyphus / Anyway, Interview

I read less than last year. I keep adding to the list of books I want to read, and next year I think I should read more engineering books too. Eating lunch alone at work while reading became a bit of a routine. I didn’t follow through as much as I expected on reading during leisure time. I think I watched a lot of short-form content. I didn’t go to the library much. I always felt like I wasn’t confident I’d finish everything if I borrowed books.

How I Want to Live Next Year

Next year, I don’t expect much from myself. Reading more and writing more nonsense posts are things I plan every year and fail every year, so it’s probably better not to even discuss them. I hope I can preserve my mental stamina well and keep diligently digging down below the surface into essentials. I’d like to use small pockets of time better. Something like that? Overall, I think I’m living well and working hard, so if I just keep doing what I’m doing, it should be fine.

As for things that aren’t fully up to me, I have quite a few wishes. I hope my partner does well. I hope good things happen at the company. I’m moving next year, and I hope I find a good place safely and do it well. I hope I can navigate the fairly big life change coming next year.

I’m doing fine; I just hope luck and circumstances are good. I’ll live the way I’ve been living.

Awards

• Albums of the Year: Ahn da Young - Where is My Friend? , Jennie - Ruby, Lorde - Virgin, NECTA - Seoul Bizarre
• Crush of the Year: Tate McRae
• YouTube Channels of the Year: No Jeongseok, Banjiha Yujeongsu, SPNS TV, Minumsa TV, Starting Today, I’m an Accountant
• Places of the Year: The top of Ogeum Park, Ulsan National Industrial Complex, Sindang Dasan Children’s Park playground
• Shoes of the Year: NNormal Tomir 2
• Books of the Year: On the Whiteness of the White Whale, Dirty Work, The Power Law
• Drink of the Year: Pépin Blanc
• Stocks of the Year: SK Hynix, Shopify, Reddit
• Screen Media of the Year: One Battle After Another, Stop Making Sense, Attack on Titan
• Purchase of the Year: Patagonia Torrentshell 30L Jacket
• Item of the Year: 3M Nexcare Tape
• Snack of the Year: Gukhee Apple Sand