I turn ideas into working technology.

From an early age, I took apart toys and machines and tried to put them back together. A friend once told me that in kindergarten I drew a laptop in a sketchbook and pretended to type on it. I do not remember the scene clearly, but I was interested in machines long before I knew what coding was.

In elementary school, I turned that curiosity into experiments with inherited Galaxy phones: installing custom ROMs and rooting them. I also assembled a PC from used parts because I wanted to play PUBG. Searching, trying, and fixing things myself became familiar long before I started building software.

I played a lot of games, so naturally I wanted to make one. I imagined an open-world game set in Korea, bought a C++ book, and installed Unreal Engine. It was far more difficult than I could handle at the time.

I set game development aside and changed direction to the web. I watched Nomad Coders and other programming channels, followed along with HTML and CSS, and discovered how satisfying it was to see a page change immediately through clone-coding projects.

I built the first Sungblab with basic HTML and Bootstrap. The layout looks awkward and the content is thin in hindsight, but at the time I was proud simply because I had made my own website.

I published it with GitHub and a free domain. Unlike hardware, software could be made once and immediately shared with anyone, and I loved that feeling.

When I first used ChatGPT, what mattered most was not the answer itself. It was being able to move from not knowing to the next question. Starting in 2024, I began using Claude to write code, fix errors, and expand what I could build on my own.

I started the Codelab programming club. I turned a friend's handwritten attendance process into face-recognition and QR/RFID attendance systems, and built a multi-LLM chatbot after seeing friends struggle to understand the differences between AI models.

At university, I built Scrim Manager, UnivMind, and OpenCairn. UnivMind worked but lacked a strong reason to return, while OpenCairn had a vision that ran ahead of its implementation. I shut the latter two down in June 2026 and learned that building an MVP with AI is different from building a product people will keep using.

I now run this portfolio and blog while experimenting with Devflow Native, a way to carry an AI coding agent's work records and verification results into the next session.

When I start something new, I try to ask who repeatedly faces the problem, what they use today, and why they would come back. I also see explaining and verifying the result myself as part of building the product.