Donghyun Kang.

AI Product Engineer

★ Case 05 / 7Product Engineer Donghyun Kang

Han River Social Dating App (Personal Product)

When code takes 5 minutes to write, deciding what to build takes days — a solo product engineer turned one Han River scene into a shipped product in 3 weeks.

개인 프로젝트Full-Stack2026.04

Landing page App Store Google Play

01.Discover

When code takes 5 minutes, the hard part is the decision.

I was eating chicken with my family at Ttukseom Han River when I noticed the next blanket over. Two guys in suits were asking to join. My wife and I cracked up. That scene was the start. The first product I built without a real user pain — I just thought the Han River picture was funny.

No persona analysis, no market research, no team meeting. I was the one who normally chants "MVP, MVP," and yet I could not stop myself. Pair matching or group matching — that was the decision that took the longest.

Cursor and Claude Code were always on, so I wrote code fast. In exchange I let the decisions go slow on purpose. In a workflow right after AI dropped the cost of code to zero, what a product engineer is left with is the decision.

Field notes

Note 01

Starting without user pain

One scene at the Han River was all there was. Zero personas, zero market research.

Note 02

I could not stop myself

The one who normally chants MVP was me — and that is the weight of deciding alone.

Note 03

Decisions took longer than the code

Code in 5 minutes; decisions across several days.

02.Build with AI

Slow on the decision, instant on the build — AI was not a tool, it was the cycle accelerator.

Decision 01

Drop pair matching

The decision I sat on the longest. Pair matching was too elaborate, not an MVP. The hardest part was that I — who normally chants MVP — could not tell myself to stop. That one decision turned into the core asset of the launch.

Decision 02

AI-first workflow (Cursor + Claude Code)

Code fast, decide slow. The risk of 5-minute code was that you could go 5 minutes in the wrong direction. I used AI as the cycle accelerator, not as a tool.

Decision 03

The weight of deciding alone

BeforeLooks fast from outside — no meetings, no reviews

AfterActually fast because the moment a decision lands, the hands move

tradeoffEven showing the work to my wife or a colleague does not move me unless the direction matches what I already want — both the limit and the weight of solo decisions.

Decision 04

Stack choice for shipping three platforms together

Flutter for iOS and Android, Next.js for web. Supabase cut backend infra cost and time. The thing that let one person carry three platforms at once was the stack choice itself.

03.Prove with metrics