Snowflake World Tour × Bloom: The Tools You Build Next Are Our Textbooks

Snowflake invited six of Korea's leading tech communities to World Tour, its biggest annual event. We were there alongside heavyweight communities like the AWS and Microsoft groups, which made it even more of an honor, and I'm genuinely grateful for the experience. There was also a bit of a competition between communities over who could bring in the most people. We came in first and won about 500,000 won (roughly $370) in food and drink coupons. At first I worried about how we'd ever spend that much at a booth, so instead Snowflake agreed to put it toward supporting our next event. Thanks again to Snowflake on every front.

The event at COEX (a huge convention center in Seoul) draws tens of thousands of people, so we thought hard about how to make the most of the slot. We ended up preparing two sessions. One introduced the Bloom community. For the other, we decided it would mean more to put a community member on stage than to have one of us speak again. (In a way, the only reason we were recognized as a representative community at all is our members.) So we opened applications inside the community and picked a speaker fairly.

About 20 people raised their hands, which I really appreciated. One of them stood out: a 17-year-old who attends Next Challenge School, a high school for aspiring founders.

I first met them at an event we ran with AWS. Those events usually draw people in their 30s and 40s, maybe 50s at most, so I was a little surprised to see someone that young show up alone. Then we talked for a few minutes, and at the end they came up as team lead and presented for their group. Watching that, I thought: I really want to give this kid a shot.

So this time I wanted to make a choice with a bit of a twist. Put someone from the next generation on the biggest conference stage we'd ever been part of, and have them show how natively teenagers use AI. I figured it would be a real jolt for a lot of people, and a nice shot of energy in a room full of senior folks.

Honestly, people around me weren't sure about it, and while I had faith, this was our first time preparing a talk together, so I had some worries too. Then I watched them on stage, and it may have been the most confident, well-organized talk from any speaker I've brought in so far. That's when I knew my instinct had been right.

Here's my write-up of what that 17-year-old high schooler said.

Dropping out was a calculation, not an emotion

"I'm a teenage founder who dropped out of school. I might be the only person in Korea carrying all three of the keywords you're not supposed to have."

The audience laughed at the first line. They'd been an ordinary middle schooler who studied for tests and went to hagwon (Korean cram school). By now they'd be prepping for midterms with friends. Then one day they saw a news story about AI writing code and making music. The thought that hit them was this: if AI is moving this fast, what am I actually going to come away with after six years of elementary school, three of middle school and three of high school?

The slide for this part was the most striking thing I saw all day. "I gave up school, and got a deadline instead." Under it, two numbers side by side. Three years: how long it takes to validate something the school way (learn, graduate, then try it). Three months: how long it takes them to validate something now (build, ship, watch the reaction). The question was whether checking the same thing takes three years or three months. Looking at that slide I kind of froze. None of the usual story we attach to the word "dropout" was there. No rebellion, no failure. It was just a conversation about cycle time. Today they attend Next Challenge School, an alternative school where students study for the high school equivalency exam while actually learning how to start companies.

Two things a 17-year-old is building

They're working on two products, one as CTO and one as CMO. The first is Lessgo, a mission-based screen time app that received investment from JA Korea. The setup is simple. You set a goal, prove it by screenshotting your screen time, and if you fail, it turns into a donation. You run the challenge with friends. I liked how they explained the design. They knew screen time addiction is stronger than drug addiction, so willpower won't cut it. The slide read: "Because we know willpower doesn't work, we built something that doesn't use willpower." What it taps into instead is a teenager's need for recognition. You don't hold out alone. You hold out because you hate losing in front of your friends.

The second is Yeochideul, which makes an acne diagnostic patch and AI-personalized skincare. It won the top prize at a youth startup competition. In a survey they ran themselves, 89 percent said they weren't sure the products they use actually suit their skin, and they'd lived it too. The dermatologist says oily, the skin test at Olive Young (Korea's biggest beauty store chain) says dry. AI skin analysis apps give different results depending on lighting and camera. A precise exam at a dermatologist is accurate but costs over 500,000 won (about $370), which no teenager can afford. So they built a patch you wear while you sleep that directly measures pH and squalene. The goal is to analyze your skin and recommend the right skincare for 5,000 to 6,000 won (around $4).

Teens don't learn AI. They use it.

This is where the room went the quietest.

They said teenagers don't learn AI. That sounds odd when there are endless meetups out there on how to code well with AI or how to vibe code. Their explanation: we're the generation that lives alongside AI, and we code and make songs with it. Even when school teaches how to use ChatGPT, kids don't learn it that way, because they'd rather grow with it and build things together with it.

They broke down how they actually use it. It used to be that when a math problem got annoying, you copied the answer key. Now you ask ChatGPT to explain why that's the answer and keep pushing until you get it. Questions before the answer key. Then there's stuff that's awkward to ask a friend, like "I had a fight with my mom and I don't want to make up, is this my fault?" Late at night they'll write honestly about school stress and friend drama. Opening up to AI more than to friends. They added that this trend worries them a bit too. And they use it to find the latest information in areas they care about, polish sentences in personal statements, and turn ideas in their head into real business ideas, raising the quality of their own thinking. They wrapped it up with one line: "For people my age, AI isn't a difficult technology. It's closer to a useful tool that helps with everyday life."

"And this is where we got stuck"

The second half is what made this session special. Usually when a teenage founder gets on stage, it ends with wins and potential. Juwon gave an entire slide to failure, titled "And this is where we got stuck." Reading the patch's color fell apart when the lighting changed even slightly; the same color read differently. They took the ingredients Claude suggested all the way to a university lab, ran the experiment, and it failed badly. Where domain knowledge was needed, the AI only gave answers that sounded plausible. Things that worked on their phone broke on a friend's phone, in cases they'd never thought to ask about.

Then came the high point of the day.

"I didn't know why it wasn't working, so I couldn't fix it. I couldn't even come up with the question to ask the AI."

They're vibe coding Lessgo, and they get alerts that something's wrong with the database. But they're not someone with ten-plus years of development behind them, so they can't find where or how the problem started. They said they felt the paradox of using AI to fix problems AI created. This reminded me of a session from the week before: the more you delegate to AI, the more your judgment muscles shrink, and the scariest part is not noticing you're getting duller. A 17-year-old was already aware of it, and brought it up first, on stage.

The last slide was a single sentence.

"The tools you build next are our textbooks."

They added that AI is a community we live alongside, our friend, and in a way it could become our teacher. The room was full of people who build data infrastructure and AI products. A 17-year-old had just redefined what their work means.

Getting the school lunch menu on KakaoTalk

The remaining 15 minutes were a panel talk.

I started with the story that impressed me most when we first met. The school lunch menu was only posted on the school website, and checking it every time was a pain, so they hooked up an agent to send it to them on KakaoTalk (Korea's main messaging app) every morning. The tech isn't impressive. The way of thinking is. There's something annoying in my life, how do I fix it, oh, I'll just hand it to an agent. When thoughts flow in that order, that's what AI native looks like to me. Most of us solve things the old way first and think of AI later. But Juwon drew a line here. Teens don't use AI as specifically as adults imagine. Its biggest role is answering questions and satisfying curiosity. I took that to mean it's less about using it with a precise goal and more that it's just woven into life.

For their friends, day to day, that means asking AI instead of a search engine, and keeping AI open on the side while watching online lectures, like a personal tutor. They used to text a tutor on KakaoTalk about problems they couldn't solve. Now you take a photo of the problem, put it into AI, and it explains it more accurately for free, so people are moving over. I asked whether that means less talking with teachers and friends. They said yes. Honestly, that answer left me with mixed feelings.

Thinking in my head, execution by AI

Schools can't block AI, so where should the line be? I really wanted to ask this. Next Challenge School encourages AI use. Asking it about things you don't understand while doing an assignment, or having it make presentation slides, is allowed. But there's a line. In a book discussion, you don't use AI to find your supporting evidence. In a debate, when you're supposed to say what you think, you don't pass off the AI's thinking as your own. In other words, they hold back on using it for anything where they're supposed to do the thinking. The school drew the line this way: the thinking comes out of your own head, and expressing it and carrying it out can be left to AI.

Companies are exactly the same. New hires used to be trained by senior colleagues. Now they learn from AI, and then they stop thinking, hand everything to AI, and turn in the report it produced as is. That keeps happening at companies right now. Schools and companies are each looking for answers to the same problem. A side note: when they mentioned that if school hands out a lecture link, they put it into AI, pull out only the key concepts, and memorize just those, they added, "I'm only saying this because there are no teachers here," and everyone laughed.

Sending the whole conversation instead of a Notion doc

I asked what adults use that their generation doesn't. The biggest one, they said, is Notion. At companies you organize documents, put them in Notion and share them. Their team made a shared team account on an AI tool and sends the conversation they had with the AI itself. They set up projects under each person's name and say, "I had this conversation, come into my files and check it once for me." Instead of a polished report, they hand over the raw context in a form an agent understands better. If someone asks me what an AI-native organization looks like, I think I'll describe this scene from now on.

Their sense of "AI feel" was different too. At companies, if a report reads like AI wrote it, people dislike it because it seems like no effort went in and it might contain hallucinations. Teens have no problem with it, they said, because everyone uses AI anyway and they think AI is smarter than they are. AI-made music and images still feel a bit off to them. Though they added that they've been seeing AI-made songs and images since middle school, so if anything those feel more familiar now.

So what percentage is your industry at?

When I asked if they had any questions for the audience, they'd come prepared. Friends talk all the time about developers disappearing or researchers disappearing, but they'd almost never met people who actually work at companies. So they asked people to share their job and how far AI has gotten into their industry right now. Watching a 17-year-old throw that question at a room full of senior people, I thought, yes, I'm glad we did this session.

They also told a story about a friend. This friend had been coding since they were young, worked on backend, and was preparing to major in computer science. Lately the friend keeps saying, "I'm in trouble. At this rate I won't get into college or get a job. AI is doing all of it and there's nothing left for me to do." The education we describe as shaky was a much more concrete fear for the people already inside it. Finally I asked about their plan for the next year. They said you have to dream big to catch even a piece of it, and that they'd sell the service to companies and the public sector and hit 1 billion won (about $740,000) in revenue. Then right away they added: if not 1 billion, I at least want to earn enough that 1 million won (about $740) lands in my bank account. People who can state a wildly ambitious goal and a very realistic one in the same breath are rarer than you'd think.

If I keep just one thing

That day Bloom also had a booth in the Community Village, and at noon on the main stage I talked about how a single blog post I'd started, about the AI blues (Claude Blue), turned into a community. Then at 2:20 p.m. I sat back and watched a 17-year-old pick up that story and carry it forward. Our booth sign-ups were also the highest of all the communities. A community that's a little over four months old got here because of the members who show up every week.

If I take only one thing from that day, it's this sentence: "I couldn't even come up with the question to ask the AI." We usually talk about how to use AI well, how to write prompts and how to feed in context. But where you really get stuck is the moment the question won't form, and that comes down to how well you know the domain. What a 17-year-old learned by botching an experiment in a university lab is exactly what we're learning at work. Thank you to Snowflake for the opportunity, to Juwon for taking the stage, and to the members who made this possible. Whenever a good stage comes along, I plan to keep handing it back to our members.

Originally written in Korean on Brunch, August 29, 2026.