March – August 2026

Clevon

Starting in March, I took on a series of outsourcing projects as a freelancer at Clevon, my friend’s company. This half-year brought an unusual number of healthcare-related commissions: an app for a dermatology clinic that scans a person’s face to build a 3D model for consultations, then a healthcare app for a yet-to-be-announced hospital franchise. The VR project still in progress arguably belongs to healthcare as well.

In August, I evolved from freelancer to full-time employee. We’re friends who have worked together for a long time, so beyond the outsourcing projects we had always talked a lot about Clevon’s growth and future — and by converting to employment, I now also get to take part in the company-strengthening work: company-wide strategy, process improvement, and so on.

Back to School

Spending long hours hunched over with a pen in hand was something I hadn’t done since cramming for the college entrance exam, and it was rough. It got to the point where, during midterms, my neck felt stiff — and then I properly strained it in my sleep, so I spent the entire exam period going to physical therapy. Still, studying something while being chased by exams and assignments wasn’t all bad. The motivation is certainly there, and conversely, when you’ve done your best during that period, even a bad result is at least over — you sigh once and move on. There were even a few courses I could feel genuinely proud of.

Orange Curation Club

As a personal project, I built a restaurant app called Orange Curation Club. It started from the idea, “Why do restaurant recommendations from friends never fail?” and works as an invite-only community. I started it in February — back when, fresh out of the military and reeling from LASEK surgery, I could barely keep my eyes open — as a kind of coding rehabilitation.

I treated it as half toy project, half product with various ambitions. Checking while writing this post, as of August 31, 2026 it has 46 users, 125 restaurants, and 141 reviews. It operates as a small community where about 70% are my own acquaintances and 30% are friends-of-friends they invited.

I learned a lot from this project — painfully. Testing devours an enormous amount of time, good UI/UX is quite expensive to build, and referrals don’t magically work just because a referral feature exists. I had assumed that once the product reached a certain level of quality, friends would invite friends and referrals would drive a fair amount of user growth. I was spectacularly wrong. The ratio of loud users to quiet users was about what I expected, but far fewer people invited their own friends than I had imagined.

Business Trip to Shenzhen

At the request of a Korean client, I visited a partner company in Shenzhen, China. It was a short trip, so I didn’t get to see much, but even brief glimpses of the streets and shops surprised me. In the area I mostly moved around in, the streets had quite a few expensive foreign cars, and true to the city’s nickname as China’s Silicon Valley, skyscrapers housing tech company headquarters dotted the skyline.

At the first meeting, there was an employee on the Chinese partner’s side who looked young — around my age — and sure enough, they were the SWE. Interestingly, perhaps because of the image the SWE role carries, both Korean and Chinese companies seem to like it when you’re young. I doubt it would be much of a plus in mechanical or other engineering fields… After the schedule wrapped up, I had a short chat with their SWE. The most memorable moment was when I asked which LLM they use and they said OpenAI’s Codex. I had assumed they would use a Chinese open-source LLM like Qwen or Kimi — a big bias on my part.

I drank pu’er tea, drank Moutai, and drank plenty of other alcohol-bearing liquids besides.

Sole Proprietorship

I was reminded once again of the greatness of Korean administrative efficiency. The morning after I made up my mind to register as a sole proprietor, everything — business registration, e-commerce business report, opening a bank account — was done in half a day. The e-commerce report was the best part: five minutes after I submitted it, the Seongbuk-gu district office called to kindly walk me through a missing document in my application. I was genuinely moved.

Truth be told, I agonized over this decision for a long time. My friends have probably heard me obsess over it to the point of neurosis. Staying a “prospective founder” rather than an established one is advantageous for government support programs later on, and even once you decide to register, there’s the choice between sole proprietorship and incorporation — I received a lot of advice on that and went through even more anguish, but in the end I registered as a sole proprietor. Once it was all over in half a day, though, I felt a little hollow.

The main reason boils down to a growing conviction that I need to actually get someone to pay me, even a single cent. The failure of Orange Curation Club taught me that building a product people open their wallets for poses a different set of challenges from building a decent product. So for Daisy, the vocabulary-memorization app I’m building now, I’ve concluded it should be sold as a paid product from day one.