I create things that fix everyday annoyances, and share them with people who've felt the same.

I was a kid who liked sports, and a computer was just a game console to me. I don't remember what set it off, but I started tampering with packets in online games, and in StarCraft, typing the money cheat every time got tedious, so I'd edit the values directly with a Windows memory debugger. Nobody had taught me any of it, and it came strangely easily.

My elementary school's native English teacher saw me at it. The other kids typing the cheat line by line, and one of them poking at memory to raise the minerals and gas in a single move. A kid who was bad at English and hated studying, using something the teacher didn't know either — I think that struck them. They said I was like Alan Turing and gave me the English name Alan. (My English nickname, for the record, is Winnie the Pooh.)

I had meant to go to a sports university and then join the Air Force, but acute rheumatism arrived in high school and I was ill for a long time. Lying there, I thought the whole thing over again, and I followed what I was drawn to: in 2010 I entered the computer science department at Dankook University.

In university, the friends around me were caught up in the smartphone boom, throwing themselves at app development and at Java and Spring. I kept looking at infrastructure, deployment, and the systems underneath instead. Docker, AWS, and Terraform were what I dug into first. That didn't mean I let service development go; I studied Node, React, and Vue alongside it. This was also when I started keeping several GitHub accounts and contributing to open source anonymously, a habit that stayed with me for years.

After graduating I joined my first startup, SalesBoost. Bringing TypeScript in properly there moved my work up a level, and I went deep into multi-tenant environments, which sit at the center of any SaaS. I moved to TainAI after that, working with React Native, Next.js, Firebase, and AWS, and designing mobile SDKs in earnest. Around that time I started using Supabase, Drizzle, and Zod in my personal projects, and I have been building with that stack ever since.

Across both companies and a great many projects, one problem kept showing up, and it was a serious one.

Authentication and authorization. The usual answer is to take something like AWS Cognito or Firebase Auth, or to build it yourself. But the auth code in a great many services was flimsier than you would expect. With a single token you could freely edit or delete someone else's data, and there was no shortage of places that couldn't even stop SQL injection.

I was doing the same thing myself. Every time I moved between work and a personal project, I wrote the auth logic again from nothing. We hand payments to a payment gateway — so why is everyone reinventing security and identity, separately and dangerously, every single time? What was missing was an auth service that plugs in as easily as a payment gateway does, whatever cloud or database you happen to be on.

So in 2023 I started building quickstart.id, a CIAM service a product can take off the shelf instead of implementing login and permissions itself. I worked through the authentication RFCs, looked into OpenTelemetry, watched OpenAI, and built in an idea I couldn't shake: that when the age of AI agents arrived, an AI would have to pass through authentication and authorization the same way a person does.

Then in 2024 a friend pressed me hard to join them, and we co-founded HyprCommunity. It was a location-based social platform, and we made a lot of complicated pieces mesh: real-time location, short-form video, live streaming in the style of Chzzk and Twitch, chat, and real-time AI translation with the logic to keep its token costs down. Talks with VCs went back and forth but never closed. I had kept my attention on building, and my first company ended in June 2026, wound down with a bad taste.

After some drifting, in July 2026 I came back to quickstart.id. It had never quite let go of me, so I sat up through the nights with the RFCs again, and by August I had the Next.js and Expo SDKs attached and the whole thing finished — past an MVP, a product ready for beta.

But the market had moved faster, and the honest read was that I had missed the window to make a business of it. After a long think, I decided to keep it as a personal project and as my own core infrastructure. It is running right now, in fact: the login on this blog, the authentication an AI agent goes through to leave a comment, and the login on a planner app I built for a friend — all of it runs on quickstart.id.

Now I am getting ready to leave a long stretch in the shade and step back out into the world. Whether that becomes a job or another company, I don't know yet.

I have always been introverted, the sort who avoids leaving traces of myself online. (Get to know me and I turn into someone who will talk about development all night 😅) But unlike before, I am not going to hide myself. This blog is the first step of that change. What I agonized over, what I set out to build next, and how I actually use AI — I intend to put all of it down here honestly.