I Wasn't Crazy After All
In early May 2025, I was in my final semester of college. Graduation was getting closer, but my path forward was still nowhere within reach.
In early May 2025, I was in my final semester of college. Graduation was getting closer, but my path forward was still nowhere within reach. The year before, I had said I would study to become a certified accountant. So when my parents asked me, "What are you doing?" and "Aren't you studying for the accounting exam?", it was, in a way, completely natural. From their point of view, it was worry. It was something I had said I would do in the first place, and I still had nothing to show for it.
But that day, those words hurt in a way they hadn't before. They didn't just sound like nagging. They sounded like a blade aimed straight at the fact that I still hadn't proven anything. I got into a huge fight with my parents. To be exact, we fought hard, and then I grabbed my bag and walked out of the house. If I had stayed any longer, I felt like I would suffocate. I wanted to explain what I was seeing, why I was trying to walk this road, why I was trying to build an AI app right now instead of studying for the accounting exam or preparing for a job—but I couldn't explain it. Because I had no results yet.
Words were too weak. No matter how many times I said "I think this is going to work," without results I could hold in my hand, it all came back to worry in the end. It's not that I didn't know that worry came from love. But the me of that day could no longer stay inside that worry. I grabbed my bag, walked out, and took the dawn bus. On the bus to Seoul, I was practically gritting my teeth. I thought I had to make money, fast. I thought I had to become independent, fast. I thought I had to make my parents stop worrying about me. Instead of explaining myself, I thought I had to produce results.
There's a record of a conversation I left behind back then.
Reading it again now, it makes me a little dizzy. That was less a goal than it was practically a declaration of survival. 700,000 won a month in rent, a 20-million-won deposit, 2 million won a month in living expenses. I had even written down the numbers. I think I felt like I could only move if I pinned reality down with numbers.
The me of that time was too impatient, too sharp, too desperate. But thinking about it again now, I had no choice. It was my final semester of college. I was still a student, but society was already at the door. There are plenty of people who don't get a job right after graduating, and plenty who take time before they start earning money. Looking back now, that wasn't such a strange thing. But the me of that time couldn't allow my life that kind of time.
There was an idea I latched onto back then. An app where you capture a KakaoTalk conversation and upload it, and the AI reads the emotional flow, analyzes the other person's tone, and recommends what you might say next. Its first name was Maeumgyeol.
It might sound silly now. A student in his final semester of college—not studying for the accounting exam, not preparing for a job—said he was going to build an app that uses AI to analyze KakaoTalk conversations and give relationship advice. I understand, to some degree, why the people around me looked at me like I was crazy. But the me of that time didn't feel it was far-fetched. If anything, it was the most realistic thing I could hold onto.
I knew how often people lose their words in front of relationships. There are moments when you have the heart to speak but don't know how to say it. Moments when you're hurt but don't want to fall apart, when you want to hold on but don't want to look like you're clinging. Moments when you read a single line from the other person over and over, wondering alone whether this person likes you or not, whether it's okay to reply, and if you do, what you should say—until you wear yourself out.
Maeumgyeol was an app for those moments. Not an app that loves on your behalf, but an app that helps you with the next single sentence, so a person can reach another person a little better.
At first, everything was crude. I put up a screen with Streamlit, hooked up the OpenAI API, and added OCR to read the KakaoTalk captures. I tried to save the emotion records in Firebase. By today's standards, I was far too clumsy. I got stuck because of a single API call method, the OCR couldn't read the KakaoTalk text properly, and even the Firebase setup was a struggle. I even had to agonize over whether to turn the VPN on or off on public Wi-Fi. Problems I feel I could clear in ten minutes today were, back then, battles that ate up an entire day.
Still, I didn't stop. When I got stuck, I asked, fixed it, and ran it again. When I got stuck again, I asked again. And that's how Maeumgyeol's first screen came to be.
There's a line I kept holding onto throughout that process.
It still holds true today. This app wasn't simply an app that generates replies. It had to be an app that reads a person's heart, grasps the situation, and helps with the next single sentence. So the recommended message couldn't be an explanatory note. Advice like "Try being honest with the other person" is something anyone can give. What mattered was words you could drop straight into the actual KakaoTalk window. A sentence that isn't too heavy but carries real feeling, that doesn't corner the other person but nudges the flow of the relationship a little forward, a single sentence the user could copy and send without it feeling awkward. That's what I wanted to make.
There was one more thing I considered important.
This, too, still holds true today. Romantic conversations are hard to judge from the text alone. The same "haha" means different things depending on the situation, and the same "I'm busy" reads completely differently depending on the context of the relationship. So the app couldn't end at simply analyzing the conversation. It had to ask the right questions to get the context it needed. What kind of relationship are the two people in, what's the mood been like lately, what does the user want, does the other person usually talk this way. I thought you had to gather that kind of information for the recommended sentence to truly reach a person's heart.
The me of that time didn't know technology well. But I think I was seeing the problem pretty accurately. People lose their words in relationships. AI can read the emotion and context of that moment and help with the next single sentence. That single sentence isn't just information—it can become an interface for handling the heart. This idea has barely changed even now.
The May version of me started Maeumgyeol on a dawn bus. Over the summer, I built an MVP, attaching OCR, the GPT API, Firebase, user emotion analysis, and reply recommendations. And in September, I built an on-device AI runtime for the first time. At first it was an app that leaned on a server to analyze the heart, but gradually I came to think like this: if it's an AI that handles the heart, shouldn't it run somewhere closer? Shouldn't it run in the user's own hand, in a faster, quieter, more private way?
KakaoTalk conversations, emotional flow, the other person's psychology, follow-up questions, the next reply, personalization, and a more private AI. Looking back now, the direction was almost the same from the very beginning.
So these days, when I read those records again, a strange feeling comes over me. The direction back then wasn't completely wrong. No—it was almost exactly the same. A year later, Maeumgyeol led to love.maeum.ai. It went up on Google Play under the name Campfire of the Heart. And from March 2026, actual revenue started coming in.
My first goal, "economic independence by November 2025," failed. I didn't fully become independent within the deadline I'd set back then. I didn't make the 20-million-won deposit, or a stable monthly income, or results at the speed I'd imagined. But Maeumgyeol didn't disappear. The form changed, the name changed, the technology changed, but the problem I cared about stayed the same.
For a while, I lived almost without thinking about that time. I just built the next feature, prepared the next release, watched user reactions, checked revenue, and fixed things again. I don't think a person can keep living while constantly dwelling on painful memories. More precisely, it's not that you forget—it's that you cover it up in order to live. But when I look at the records again, time suddenly folds in on itself. I really was desperate back then. I really was lonely back then. I really was trying to build this back then. Belatedly, thoughts like that come rushing in.
The thing that tormented me most in those days was the sense that no one understood.
It really was like that. In my mind, I was right. The direction—AI reading a person's conversation, analyzing the flow of emotion, and helping with the next words—clearly seemed like it was coming. People waver so often in front of love, and I felt there was something AI could do in that moment. And yet no one understood.
The people around me only saw the surface. Final semester of college, not studying for the accounting exam, doesn't look like he's preparing for a job, still hasn't made money, says he's building an AI relationship-advice app. Looking at that combination alone, it would have looked crazy. I know. But the essence I was looking at was different. People lose their words in relationships. AI can help in that moment. It can read emotion and context and help with the next single sentence, so a person can reach another person better. That's what I was looking at.
But people couldn't see it.
Back then, I really was treated like I was crazy. It wasn't just that I didn't get encouragement. The very direction I was looking at seemed, to the people around me, like a choice that made no sense. That was the loneliest part. A state where I didn't seem to be completely wrong, yet no one believed me, and the more I tried to explain, the crazier I looked. That's why I wanted to build it even more. I wanted to show it through results, not explanations.
Looking back now, I think that was the first scene of a founder. Of course, at the time I didn't think I was starting a company. I didn't have some grand vision, and I didn't have a slick business plan. I just built it to survive. But in hindsight, the whole journey of a founder was in there. Conflict with family, anxiety about the future, money pressure, an idea no one understands, a shabby MVP, technical fumbling, user logs, the first revenue, and even the retrospective you look back on later. I didn't know it then, but that was my first scene as a founder.
Of course, there's still anger I feel even now. Why could no one see it? Why did they look only at the surface? Why did they say that money, titles, exams, employment, and stable routes were the only reality? The Korean-style sense of reality I ran into in those days was too narrow. When you try to look at something new, they treat you like you're crazy, and until you've proven it, they don't believe a single thing you say. The moment you step outside the set path, they immediately see you as an unstable person. The me of that time hated that atmosphere.
But I don't want to hold onto that feeling for long now. Anger can be a spark, but it can't be a home.
Instead, this is what I want to leave behind.
It wasn't completely wrong. It was almost exactly that direction. And so, I wasn't crazy after all.
The me of that time was immature, impatient, sharp, and trying too hard to become an adult too fast. But I wasn't talking nonsense. I just didn't have a product yet. I just didn't have revenue yet. I just hadn't yet proven it in a form anyone could recognize. But there was a direction.
That direction became Maeumgyeol, Maeumgyeol became the Campfire of the Heart, and the Campfire of the Heart is, even now, lighting a small fire beside someone's heart. At first it was something I made in order to survive. But at some point, it had become an app that helps someone's heart.
On June 28, 2026, I read the conversation record from a year ago again. I got choked up. I felt irritated, I felt sorrowful, and I felt a little proud, too. The me of that time wasn't perfect. The goals slipped, the plans were clumsy, and there was a lot I didn't know. But one thing is clear. That day, I didn't stop. I got into a huge fight with my parents, grabbed my bag and walked out of the house, took the dawn bus, and on the way to Seoul, I started building Maeumgyeol.
I thought I had run away back then. But looking at it now, I don't think I ran away—I think I started building. To prove that I wasn't crazy. To show that the direction I was looking at wasn't completely wrong. And above all, to take hold of my own life with my own hands.
That was Maeumgyeol's first grain. It wasn't a finished product, nor some great success. But that day, for the first time, I made up my mind that I would build my own life myself.