EP 98
In the Age When AI Executes, What Remains for Humans Is 'Intent' (Hashed CEO Seojoon Kim)
Opening: The Age Where Only Intent Remains and Agentic Organizations 0:00
Seojoon Kim When I think about our lives, about 5 or 10 years from now, I think almost only intent will remain.
If you look at people working at large corporations, according to the intent that comes down from above, most of the work is execution. Intent is extremely limited. So if you look at more efficient startups.
they simplify the hierarchy to just about three or four levels, and they make efforts to reduce the latency of decision-making, judgment, and all of that. That is startup culture; in startups, the space for intent grows.
Chester Roh That is such a good expression. If you go to the extreme of an agentic-native company.
Seojoon Kim each person has the intent to fully cover their own entire domain, and rather than team members, it becomes a company where people operate with 100 agents each. In the past, there was something called the 10,000-hour rule. But now, just putting in those 10,000 hours is not what matters. It feels like we are in an environment where someone who has thought for 100 hours with intent can outperform people who thought for 10,000 hours without intent. Today, we have invited a very special guest.
Introducing Hashed CEO Seojoon Kim and Today’s Main Agenda 0:55
Chester Roh I worked very hard to bring on someone who is extremely well-known. We have Hashed CEO Seojoon Kim with us.
and Hashed has recently been doing a lot of major work in AI.
Also, CEO Kim writes essays on Facebook, and every single one of those essays is very deep. I felt that CEO Kim has some kind of clear visual of the future.
I often use the expression “Emerald Castle” for this. Whether someone is a founder or an investor, people who have first gone into the future they dream of and believe is right, and who have thought about it constantly, tend to have that image.
So asking CEO Kim today about the Emerald Castle he dreams of is our main agenda.
Seojoon Kim This is a show I have always enjoyed watching, so it is an honor to appear on it like this. Thank you for coming.
Chester Roh I think we should start the conversation now. Given your background, Web3 really swept through in a huge way once, didn’t it? I feel a bit bad saying “swept through,” but there is also an atmosphere that it may be starting again now. Because of AI and the agent economy, and in the areas where those connect, you are the person with the most scenarios. When should I ask you was the moment when you felt that AI and blockchain would fit together perfectly and that click happened?
Andrej Karpathy’s Vibe Coding and How It Differs from Copilot 2:28
Seojoon Kim Andrej Karpathy first brought up the idea of vibe coding in February of last year. I started taking an interest from then as well. But when I originally tried what is called vibe coding, the term that existed before vibe coding was Copilot. The biggest difference I felt between Copilot and vibe coding was that with Copilot, I basically had to do everything, and it would help a bit as an assistant. So even when coding, I would set up the basic structure of the code, and it would do some debugging, or create a few unfinished parts for me. That was about the feeling. And for example, when Copilot was added to Microsoft tools, it felt like I was basically making the slides and it would help a little, maybe polish the design a bit. That was about it.
Still, the term vibe coding itself meant coding in natural language, and demos came out, and a lot of impressive things appeared on screen. To be honest, early last year, I think that was a bit of exaggerated advertising. I tried experimenting with it too, But it did not work. Right, it did not work as well as I expected.
My major was computer science, and there was a time when I worked as a developer, but it has been a long time since I stepped away. In practical terms, since I last sat at a terminal and coded, more than 10 years had really passed. But when you watch vibe coding videos and look at the concept, it is so interesting. So I tried it every now and then, but it did not work well. I periodically thought that spending time studying this again would not produce enough ROI. So I would try it once every two or three months, and then I just dropped it all. The timing was not right yet. But then there was one major turning point. There is a company called Across.
Across and GPTO as Signals of the Shift to AI Search 4:08
Seojoon Kim It operates a service called GPTO, and it is run by CEO Lee Jaehong, who is not a developer. CEO Lee graduated from KAIST, but has no computer science or development background at all. The work CEO Lee used to do was only in areas like planning and marketing.
Chester Roh CEO Lee worked at Cheil Worldwide, right? Yes, that is right. But at some point, CEO Lee disappeared from Korea.
Seojoon Kim And videos kept being posted of CEO Lee living almost like a hippie. Doing yoga in Bali, and in the meantime, many photos of CEO Lee’s activities at Network School were also posted.
Chester Roh What is Network School? There is someone named Balaji. I think many people would agree.
Balaji Srinivasan’s Network State and Network School 4:42
Seojoon Kim but among the thinkers in the San Francisco Bay Area, many people consider Balaji to be one of the absolute top minds. So Balaji was also a general partner at Andreessen Horowitz, and was also the CTO of Coinbase. Balaji has been holding a conference called Network State for the past three years. Balaji also published a book with the same title, and the concept is this.
nations are rebuilt on a network basis. The old concept of nationality was something acquired by birth, and much of a nation’s power or energy came from hard power. Things like military strength, or the hard infrastructure provided by the state. Things like that defined a nation to a large extent, and changing nationality was also extremely difficult.
But as many people probably feel, that is changing. Moving nationality has become much easier, and I often say that nationality is now turning into membership. In a way, it is changing into something like the relationship between a platform company and its customers. That is right. People are actually moving a lot in search of various regulatory environments.
or market environments or tax environments that they prefer more. Companies are doing the same and changing their nationality too. Korean companies also do a lot of flipping these days. The concept at almost the far end of that trend is the concept of the Network State. So the idea is to make the nation itself network-based from the start, among people who share ideologies, beliefs, and values, first creating a community online, and then having those people later choose some physical space. The Malaysian prime minister was inspired by Balaji’s concept, and since it was unused land anyway, said, “Just try whatever you want to do,” and gave him the land.
Chester Roh Legally, as a state, it still belongs to Malaysia. Yes, so Network School is not a state concept.
Seojoon Kim but the government gave them a community space where they could experiment with it. So if you look at Network School now, including the hotel, there are large spaces that can be used as offices, and even the dead real estate sites next to it, so a large village is being created. So now about 200 people are working in spaces there, like coworking spaces.
Reverse-Engineering Rankings in the AI Search Era and AEO GEO 7:07
Seojoon Kim This was before keywords like AEO and GEO became hot in Korea. But CEO Jaehong built a site called GPTO and said he was looking for customers, and posted something on Facebook. When I saw that and took a look around, it made so much sense. Because this was already about six or seven, seven or eight months ago, but it was a trend that was happening. People barely use Google Search anymore. Traffic is now all moving to AI search, and when people ask AI, it recommends at most five things, so if you do not make it in there, there is no room to be chosen at all. But one of the biggest industries on the internet is the advertising market surrounding the SEO market. So this really makes sense. But how did he build this? I was so curious.
So I was so curious that I immediately asked to meet, and we met for the first time in about 10 years. But this friend had quit everything he was doing for two years and studied development while traveling around places like Bali, Malaysia, and Singapore. For each of the four major LLMs, the way they receive new information and turn it into a database, and how they prioritize and show it when people send queries, they do not disclose those methods. But I think he figured those things out by running countless tests himself.
Chester Roh In fact, AI also brings that information from somewhere, grounds the context, and shows it, so search is actually running behind the scenes, and it does depend heavily on that search ranking algorithm. But looking at what CEO Jaehong did, he had really hacked deeply into how this system works. How do you do it? It is also kind of a trade secret.
Seungjoon Choi Avoiding the secrets. If we talk while avoiding the secrets.
Chester Roh when you ask ChatGPT, it gives you an answer. But underneath that process of giving the answer, if you ask, “Why did you find this? What did you search with?” it does show all the tool calls it ran.
Seungjoon Choi So it is reverse prompting, reverse engineering. Yes, exactly. So for each LLM, when it does web search.
Seojoon Kim there are preferences for what kind of web search it uses. By reverse engineering that.
Seungjoon Choi You are reverse-solving the query itself. Yes, so like Hansel and Gretel.
Seojoon Kim he keeps laying cookies along the paths where the LLM is likely to follow.
Seungjoon Choi In the end, is it a kind of prompting? Or is there a more technical aspect?
In the end, is it a kind of prompting? Or is there a more technical aspect?
Seojoon Kim but in the process of optimizing it well overall and spreading it well, he is using a lot of AI.
Fairness in Information Distribution and Decentralizing Ranking Power 9:37
Seojoon Kim But CEO Jaehong clearly has a sense of the problem too. Anyway, then is the information we know being distributed fairly? The truth is simply what wealthy American media magnates set up.
what they set up politically and create as truth. Is that truth really fair? It is not something that is, after all, created by people fairly. Yes, from a crypto perspective.
Chester Roh that ranking power is also excessively concentrated in some ways. I also agree with the view that it needs to be decentralized a bit more fairly.
Seojoon Kim So, one of the models Across is preparing is content created by people. I mean, people use Naver Blog a lot because if they create good content, they can make money from it.
A Content Ecosystem Built on Stablecoin-Based Rewards 10:16
Seojoon Kim And the reason more people in Korea are using Twitter these days is also because now Twitter can make money. Once you exceed a certain level of impact, they start paying you. Yes, I found that interesting. Yes, so because of things like that, people eventually start spending money because of economic rewards, but instead of keeping that part circulating only within an ecosystem of a small number of advertisers, on channels where many people can help AI learn, and even on already-famous social media platforms, people can receive money in stablecoins while creating content that can raise the relevant rankings. He also has this vision of decentralizing this market.
Chester Roh So blockchain is not just pursuing technology. There is something like a manifesto. I think the fact that ideology and cutting-edge technology are combined is a very interesting and powerful point. Because a kind of doctrine has been written around society’s problems, the state’s problems, existing systems, and things that somewhat go against the establishment, it is attractive when you look at it. But it does not just end as writing. They have implemented one layer of it.
Seojoon Kim An LLM makes it possible for someone who was a non-developer to do something that looks very technical. In fact, what he essentially did was reverse engineer the way LLM web search and search ranking work. From one perspective, this is something that seems like a brilliant engineer would do, but the fact that a complete non-developer, someone who knew nothing about development until two years ago, was able to do it was itself so interesting. I had never seen anything like that before either.
The second thing is, he built a company, but he is doing everything alone. That was also really fascinating. I do a lot of investing, starting from angel investing and also running a venture capital firm, but when people start a company, there is almost a formula. Usually there is a founder, and then, even for the engineering team, almost like a formula, there is one backend developer, one web developer, one iOS developer, one Android developer, or one UX designer, one marketer. But he is doing all of that alone. We also have a U.S. office, and we had been watching related deals and trends in San Francisco, but when I witnessed someone I knew doing that right in front of me, I thought, now what was coming has arrived. Seeing a complete non-developer solve such technical problems, do everything alone, and even start making money, I thought, this is serious, VCs are finished now. I saw how he was building it.
Non-Developer Solo Building and the One-Click Moment in Vibe Coding 12:36
Seojoon Kim But then it turned out everything was being done through vibe coding. So I thought, now even things like this can be built with vibe coding. A non-developer can take a topic that looks this technical and solve it to this extent now. Just as I was starting again, Gemini 3 came out. Gemini 3 was quite an advanced engine. Those who have tried it would feel this, but there was a quantum jump. From the vibe coding perspective. That was last November. Yes, it was early last November.
The Quantum Jump After Gemini 3 and Opus 4.5 13:23
Seojoon Kim But a week after that, Opus came out.
Chester Roh 4.5?
Seojoon Kim Yes, Opus 4.5 came out. What was really lucky for me was that, after meeting CEO Jaehong, just as I was thinking I really needed to start vibe coding, I tried Gemini 3 on the exact day it came out, and right after that, I used Opus 4.5 from the first day it was released. In a way, the before and after that Opus has created is really significant now. And I was able to experience that without being too late.
Chester Roh Then specifically, was it Opus 4.5? Or are you saying you started using Claude Code combined with that?
Seojoon Kim At first, it wasn’t even Claude Code, just like a web interface. Yes, I made it in the web interface and then just took the code and used it. I had become that far removed from a terminal environment. Of course, after about two weeks, I started connecting it to the terminal and using it, but at first I wasn’t using Claude Code either.
Chester Roh So your full-fledged vibe coding, your own vibe coding history, started in mid-November, then.
Seojoon Kim Yes, that’s right.
Seungjoon Choi So it’s been six months now.
Chester Roh Six months. Yes, about that.
Building an Ethereum Valuation Dashboard in Four Hours 14:45
Seojoon Kim The first problem I tackled after Opus came out was building a site to value Ethereum. It was something I had really wanted to do for a long time, but I was busy too, so I hadn’t actually gotten around to doing it. I think people now understand Bitcoin somewhat. They accept it as digital gold. People may differ on how much value it has or doesn’t have, but Bitcoin is still easy to understand. But still, for so many people, starting with Ethereum, there are far too many who don’t even know why it has value. And while working in the blockchain industry.
one of the things that, in a sense, stressed me out the most was that although I agree with many parts of it, people say all tokens have no value. If people buy them, the price goes up, and if they sell, it goes down. Of course, there really are too many tokens with no intrinsic value. There are a lot. There are many tokens that were made like scams or fail to capture value, but there are also many, including Ethereum for example, that have clear business models and are creating value. If even those are all lumped together and people just say, “This goes up when people buy and goes down when they sell,” then as someone working in this industry, that’s disappointing. So for a long time, I had wanted to create metrics for that and establish some kind of benchmark. It was something I had been thinking about for years. So to build this valuation model.
you first need to lay out the data that serve as indicators. So I listed out about 30 data points worth tracking in the Ethereum ecosystem in the Opus web interface, and told it, “Find the data for these and try graphing them,” then I ran it and went to the bathroom. When I came back from the bathroom, some file of code had been created, an HTML file. I downloaded it and opened it, and 20 out of the 30 were just working properly with real-time data showing. What’s interesting about blockchain is that the APIs are open. And the LLM, Claude, just found them on its own. I hadn’t even told it where to look, but it just searched around by itself. For some, it pulled them from Binance, and for others, it searched CoinMarketCap and brought them in. Seeing it display live data for 20 out of the 30 in a pretty plausible interface shocked me so much at that moment. It can do this much? Yes, it can do this much?
I simply hadn’t expected this to run. I thought that after it researched it, it would say, “To build this, I need this additional information,” and give me homework. But seeing it just come out in one go shocked me so much. So that day, I just went straight for four hours, properly wiring up the indicators there and the remaining broken APIs, and using those, I put in several valuation formulas from the capital markets and built a dashboard with eight methodologies, then posted it on Twitter. But after a day or two.
in the blockchain industry, there is a platform that ranks people on Twitter like a kind of power ranking, called Kaito. The post I uploaded spread enormously on that platform, and it went all the way up to number one in the world. This, too, is a kind of product. I had wanted to build it for several years, but to make it, I thought I would need someone like a quant, and probably two or three developers, and even with a conservative estimate, it would take a couple of months from planning to development, so it wasn’t a small project. But this was done in four hours. Seeing that it was made well enough for so many people to take interest in it, that was when I felt the click. The world had become one where you could build everything with a click. This is a minor point, but if you ran it for four hours straight.
Seungjoon Choi did you go with Max right from the start instead of Pro?
Seojoon Kim Yes, that’s right. I mean, I was running Pro, then after about an hour, I immediately switched to Max.
That’s right. You’re rich, CEO, so Max isn’t a burden for you. And then what happened next was, after some time had passed in November, around the end of the year.
An Abu Dhabi Travel Platform Built on a Plane and Execution at the Speed of Thought 19:00
Seojoon Kim I often go on business trips to Abu Dhabi, and I had another trip to Abu Dhabi scheduled. When you go to Abu Dhabi, Abu Dhabi’s national carrier is an airline called Etihad, and for another matter, I had a chance to meet the chairman of Etihad. So I got on the plane, and when you sit down in an airplane seat, there’s a screen in front of you. There, it said “101 of the most interesting things to see in Abu Dhabi.” I connected to Wi-Fi and went to the site, and there were some really pretty photos, and the full list of 101 things wasn’t even visible and the site was a complete mess. So on the plane, I made that myself.
So what I worked on during the flight was, first, I found the list of 101 things. After finding it, you know those third-party platforms that can scrape Google Maps? There, for example, for the venues with the most activity, some had 100,000 reviews piled up. I crawled all of that, starting with the photos, and then analyzed the Google reviews with an LLM, doing some natural language analysis. So for each venue, I had it analyze star ratings across several categories, like places good for families, places good for couples, places good for photos.
places with active activities, places where you can relax a bit, things like that, across a few different areas. Then I displayed them on the map, and showed which nearby venues you could move to to immediately see other places nearby as well. So from my perspective, I built a platform on the plane with better UX, usability, and data than Tripadvisor. After making it and landing, then…
Chester Roh Of course, Opus made it for you, right? Yes, of course.
Seojoon Kim When I showed it to them, they found it really fun. “I also made something like this.” When I said, “Isn’t it amazing? Ta-da,” they were really amazed. Since those people are obviously not tech-savvy people, I realized we can now make everything at the speed of thought. The book that made me go into computer science was Bill Gates’s Business @ the Speed of Thought, but back then it was just thinking and that was it. Now, execution really happens at the speed of thought, and I felt that.
So after I arrived in Abu Dhabi and that meeting ended, it was the beginning of the year. I immediately called in one of our investment analysts. We have an investment analyst in Singapore, and I said, “We’re in trouble. I think VC is really doomed now.”
The VC Crisis in the Vibe Coding Era and a New Type of Founder 21:30
Chester Roh Right. But we need to get a little more specific about this part where you say, “VC is really doomed now,” because, traditionally, for value creation to be made with a click, as you just described, time and resources were needed, and the people who could do that were a very small group of talent. Getting access to those things first and solidifying them through investment were the advantages and the work that VCs enjoyed. Now, you’re saying that these things have become meaningless. Of course, this is only about IT startups.
Seojoon Kim In other sectors that require CAPEX investment or large R&D costs, it is still a somewhat different play.
Chester Roh If we explain it just around the software startups we’re familiar with, from the standpoint of software startups.
Seojoon Kim when they usually come to raise funding, the biggest portion is developer salaries. Even if you open up Google’s financial statements, more than half of Google’s total costs go to developer compensation. Whether it’s a big company or a small company, so when you usually talk about a startup, about two-thirds is almost always allocated to costs related to the development team’s salaries or compensation. And because of that, the biggest main reason they come to raise funding.
is actually that. I had already seen things like this through CEO Jaehong, but after trying it myself, I realized that one person who can do vibe coding can at least skip all the early stages. But if that person skips the early stages and even starts making money, then that person won’t come to a VC anymore. I clearly felt that a new kind of founder was being created in this way.
Chester Roh I agree.
Seojoon Kim I mean, in the existing so-called Web2 ecosystem, there are many excellent founders. But there’s no need to necessarily compare them with these people. They had formulas they were good at, right? Building that organization, making what is usually called the organization of a great founding team, creating organizational hierarchy and systems, then hitting milestones step by step while fundraising, there was a venture capital ecosystem that operated on formulas like that. But a completely different new species has appeared.
Seojoon Kim They just build everything alone and use agents as team members, saying, “We don’t really need to raise funding.” But we really need to invest more in people like this, so how do we make these people take our money?
Chester Roh That’s the important question.
Seojoon Kim So I thought about that and decided we needed to create a program suited to it. I immediately told the investment analyst in Singapore to come, and we brainstormed for two days. We thought there were three kinds of value we could provide. In software, in an AI-agentic way, what were the three values we could give to founders building startups? The first was mentorship, but I don’t think ordinary mentorship is enough. Since these are people obsessed with agent-based building, we need founders who can speak the same language with them and mentor them, in other words, people who are deep into vibe coding, understand the agent ecosystem, and can look at the concerns these people have while building from the same level, while offering different experiences. We need to become that kind of investor. So including our investment analysts, all our team members began getting introduced to vibe coding because everyone absolutely had to do it. The second is what we think of as a shortcut to trust.
The Three Values Hashed Can Offer Mentorship Trust Network Peer 24:01
Seojoon Kim A phrase I started using a lot from then on was, “Everything visible is open source.”
Chester Roh That’s right.
Seojoon Kim That’s really what happened. These days, a lot of skills for copying frontends have appeared. So if you click on a very flashy website that’s running beautifully and add that skill, you can see things being made almost similarly in about 5 or 10 minutes. When most service areas can be built with a click, then the essence of business is almost; B2C is a different story, but in B2B, when someone says, if 100 or 1,000 people make countless products, who brings them and connects them? I didn’t make the Abu Dhabi tourism app to make money. I made it like a hobby, as in, “You can do things like this too.” But if that had been a tourism startup, whether or not there is someone who can connect you to the chairman of Etihad is a completely different story. A network that can play that role for stakeholders all over the world is something we have built quite a lot of through doing Web3 globally, so I thought we could develop these things more sharply and help founders with them. That was the second thing. This was an area where we were somewhat confident.
Chester Roh It’s also an area you’ve been doing well in. The third is really important.
Seojoon Kim I also lived in dorms at science high school and university, and people who have lived in dorms probably feel similarly. The things you learn at prestigious universities and places like that are much greater from your peer group than from your teachers. Peers are important. Yes, peers are so important. In a way, everything else is more like background, and the things that truly stimulate you or become lasting memories and experiences in life come from your peer group.
These people, the ones who are now immersed in agentic development and startup methodologies, develop this condition where they cannot talk to other people. It is so stimulating, and at the same time it makes you feel so anxious. You wake up, refresh Twitter once, and scary things just keep popping up. You fall into this anxiety of wondering whether what you are building now will still matter tomorrow. So bringing together top-level builders who carry this excitement, worry, and anxiety feels a bit different from just gathering generally good founders. In this agent ecosystem, people who are completely immersed in this, who want to talk about this all day and exchange information, putting those people together would be very meaningful. Because I myself was feeling so much anxiety, and I also wanted to talk about this, but even late last year, there were not that many people who wanted to talk about these things. I am a bit curious about the rough numbers.
Seungjoon Choi Around January, early in the year, from your perspective, how large did you think that pool was? Honestly, I did not really know about Korea.
Seojoon Kim I am one of those people who practically lives on Twitter. Especially, people in crypto often have that kind of pattern, and Twitter already makes it easy for the algorithm to take over. So the world I see is already covered with that. That is why I felt even greater FOMO and anxiety.
Early in the year, I did an interview with a media outlet called Outstanding, and during the interview, I talked about various things, about Nitro at the time, which used to be called Vibe Labs, and how we got started. The reporter told me I looked really excited. So the reporter asked what it felt like, and what I said then ended up becoming the title. I said excitement and fun were about 30%, and worry, anxiety, and fear were about 70%. I still think pretty similarly now. But when I look at people who are fully immersed in this ecosystem, I think the proportion is similar. They have tremendous worry and anxiety.
and to exaggerate a little, they wonder, “Am I becoming useless now?” “Soon, will everything I can do be taken over by AI?” “Then what should I build before this end comes?” Many people are seriously thinking about these questions.
So I thought we needed a community that could gather the energy of people like that. I needed it myself, and those builders would want to talk with each other too. Since that kind of clustering was already happening globally on Twitter, I started thinking that I wanted to create that kind of community in Korea. The common thread among all of these is, ultimately.
Chester Roh even if we end up in a world where these models create everything, you are still talking about a certain target, a certain class of people. I think we could define them as AI-native talent. It sounds like you devised these strategies while holding an image of those people in mind. What kind of people are they? Let me make the question easier: if they are people you want to invest in, when you say, “These are the people I am looking for now,” what kind of people are they? When developers used to work with a terminal open while vibe coding?
Defining AI-Native Talent and the Conditions for Dreamers 29:50
Seojoon Kim they looked really cool in the old days. Back then, they really memorized syntax, typed in flashy ways, and wrote code. But if you go stand next to them now, they are saying, “Why is this not working? Do it faster.” In natural language now… AI specialist.
But the reason they do that is because it works. It almost works now. So almost anything a person can think of can be done. Was it about two months ago? The founder of GitLab even treated his own cancer, exactly. So now, almost nothing is impossible. Someone who was not a specialist at all can do it, and I think the barriers to cross-sector expertise have all collapsed. So rather, that industry… Startups always had some of that too.
People who were not specialists, people who had not worked in that industry, often overturned things with fresh thinking. For example, Uber was built by someone who had not worked in the taxi industry, Airbnb was built by someone who had not worked in the hotel industry, and Peter Thiel was not someone from the financial industry, but he built a payment network. Those things happened, and now this is accelerating much more.
If you study intensely with an LLM for about a week or a month, you can solve things that experts thought, “This will not work,” and did not even bother looking at. I felt that a world has arrived where it is possible to solve even harder problems. That is what I came to feel. So there are a few keywords.
I think the world of dreamers has arrived. The world of dreamers. I see a lot of those people around me.
I see a lot of those people around me.
I see a lot of those people around me.
Even when those people talked about all kinds of things, they did not have tools to execute them. But now they can execute at the speed of thought. And I think the value of intent has become extremely large. Execution can now be done by LLMs, but intent is provided by humans.
So people whose lives are filled with moving according to intent, people who try various experiments with intent, in other words, without thinking that something will not work, I think people who have no “no” in them are becoming very important. It feels like almost anything can be done now. And even if I cannot do it, if I find experts one or two connections away, we are in an era where people who can solve it are appearing all over the world. So with that kind of positive energy and mindset, what is there that cannot be done? But people who keep executing with intent, I think that is the most important keyword. You have found many people like that, gathered them together, and you are continuing to look for them.
Chester Roh Once you gathered them, what kinds of backgrounds did they come from?
Seojoon Kim Dreamers. Yes, it varies.
Seojoon Kim Interestingly, among developers, many are from specialized high schools. From their perspective, according to social conventions, until just a few years ago, they would have been seen as quite far from top-level developers, as that category of people, those kinds of friends.
Chester Roh Right. From a traditional perspective, they are somewhat outsiders.
Seojoon Kim Yes, what I’m talking about isn’t science high schools, but development-specialized high schools. So in the past, they really had names like technical high schools or commercial high schools…
Seungjoon Choi For example, places like Sunrin?
Seojoon Kim Yes, there’s Sunrin, and Hansei Cyber Security High School too.
Seungjoon Choi But those places also have their own tradition and reputation. Right, right. Of course, those schools’ reputations have gradually built up too.
Seojoon Kim but I feel like they’ve been exploding over the past year or two. I think the university system itself is in the process of being dismantled. I think there are three roles that universities used to perform. One is selection. How diligently someone has lived, how smart the person is, and selecting for that once through CSAT scores. The second is the process of education. Following the school curriculum, there are midterms and finals, and students take classes. The third is the process of creating a community. I think all three of these are being unbundled. Those three are exactly the same as what you said earlier.
The Unbundling of the University System and Community-Based Learning 33:33
Chester Roh Hashed is trying to provide as value to founders. Yes, I think so.
Seojoon Kim I hadn’t thought of it that way, but now that I look at it, I think that’s true. So from the perspective of selection, developers don’t necessarily become better developers just because they graduated from a good school. There’s a slightly higher probability, but it’s not necessarily the case. In fact, especially in today’s world, you can just look at GitHub, then look at the repos accumulated there and things like stars, and see how skilled a developer is. GitHub becomes more important evidence than a degree. And then from the perspective of studying itself, there are already things like MOOCs.
but beyond MOOCs, now you don’t even need MOOCs. Even just a few years ago, people said you could study by taking CS lectures from Stanford or Berkeley. But I don’t think people would watch those these days.
Chester Roh That’s right.
Seojoon Kim Because you can just talk with Opus, and it teaches you everything. So that itself has disappeared now too.
And these days, there are really so many meetups, right? I think many people feel that there are more and more of them. As people participate in communities with intention, I think the environment where people can build their own communities is growing larger and larger. So I think, for a very large number of people, university is not a place they enter with intention.
Chester Roh Their parents’ intentions are reflected in it. Their parents’ intentions and social pressure.
Seojoon Kim Since I grew up in Korea, I should at least go to a decent university. It feels like that fills out your resume a bit. That’s just why people go. But the share of people who really choose and attend a school with genuine intention is small.
But instead, being active in open communities, going to meetups, networking there, and giving presentations there, those are things people do with 100% intention. The communities created with that intention.
seemed to me to be becoming stronger communities than just classmates from the same section or friends from the same major at school.
What we did wasn’t creating something out of nothing and raising a flag. We saw these kinds of movements that were already happening. But in terms of these people’s potential and energy, we are a company within the established system. We operate a fund and have many large corporations as LPs. And the players within this established system had not really provided much support that treated them as valuable players who deserved respect.
We went to them, showed respect, and asked them to work with us. And through that, I think they also started to open up a bit. Listening to what you’re saying.
Chester Roh I actually think that in the U.S., Y Combinator is taking over the former position of universities. My nephew is like that, and my child is like that too. They struggled to get into prestigious universities, but they think they only need about that much of the brand, and from freshman or sophomore year, they become desperate to get into a Y Combinator batch. Getting in there and then legally leaving school without any pressure from their parents seems to be the trend among kids these days. Hashed must be trying to become that kind of position too. Yes, it’s not that we intended to guide it that way.
Seojoon Kim but for the people who are looking for that, I hope we can become one of those symbols. Symbols are important, after all. Getting into Y Combinator is one thing, but it’s also a symbol they can explain for themselves. Society moves through symbols. So instead of saying, “I just dropped out of college,” whether that’s Hashed or some other VC, or another great accelerator, they can say they were recognized by them and are developers supported by their ecosystem and community, and say, “Now I’m going to leave school and do this.” I think creating that kind of symbol is meaningful.
Chester Roh You see, this society now has AI and these new kinds of things, and AI is ultimately a means of production. What used to require these large means of production and resources has now become a tool of production that compresses all of that. And the people who recognize this quickly, use it well, and match it with business are the kinds of talent currently drawing your interest. How do you think this will unfold? This is a somewhat open-ended question, but the current situation is like this, and it’s still weak and there isn’t much yet. But I feel new opportunities in places like this, and a great deal of talent will emerge from here. And now, as what we call a VC, how will companies be grown, and how will they exit? Will the money come from listing on the stock market? Or will you say, “No, they won’t need to do that”? I imagine there is a kind of Emerald Castle vision you have in mind. It’s okay if it’s wrong, just. Since you’ve asked a very open-ended question.
Seojoon Kim let me talk through the things that come to mind, even if they’re a bit disorganized. When you look at how people work at large corporations, most of them are executing. The intentions that can exist within a large corporation are very limited. Because the intentions are largely already determined. So according to the intention coming down from above, to put it harshly, people are just doing work that an agent could do, routing information, and a lot of things like that are happening.
So when you look at more efficient startups, in terms of company culture, one of the ways the word “startup” was used differently from the old term “venture company” included reducing the number of layers a lot. For example, back in the early days of the internet, when people talked about venture companies, they often started with rank structures that imitated typical large corporations. But among startups, although it differs from startup to startup, quite a lot of startups do not create the so-called rank system of staff, assistant manager, manager, general manager, deputy general manager, and so on. Many simply streamline the rank system down to about three or four levels, while trying to reduce the latency of decision-making, judgment, and all of that. That was one of the organizational cultures of startups that emerged in the Web 2.0 era.
And one of the energies held by people who wanted to grow while working at startups was that there would be more space where they could act with greater intent. At large companies, the space for intent inevitably becomes smaller, but at startups, the space for intent becomes larger.
Chester Roh That is such a good expression. If you go to the extreme, to an agentic-native company.
Seojoon Kim even if it is one founder, or a founder with a few employees, each person has the intent to fully cover their own domain, and instead of having team members, they each have a hundred agents and it becomes a company operated that way. Then what remains is the density of judgment, the accuracy of judgment, and then the responsibility that follows from that judgment, which humans now bear. The rest of the work, the so-called hands-on operational work, and even the coordination of that work, will mostly be done by agents.
In that case, the new kind of company image or enterprise image that these kinds of founders are experiencing now may be what humanity will experience in about 10 or 20 years. It may not necessarily be a company. It may just be the same for lifestyle as well. For example, when I think about our lives five or ten years from now, I think almost only intent will remain. For example, it is like this. If I take my parents as an example.
and I want to send them on an overseas trip for their birthday, even if I have that intent now, I still have to do a lot of the practical work. I have to look into travel, figure out what country would be good, check travel agencies and tickets, book flights, reserve hotels, and do all of that practical work. But now, within about five to ten years, and I think it could be even shorter than that, I will have my own agent, and if I install agents for my parents as well, then by putting together my parents’ preferences, various health conditions, and keywords for places they have usually wanted to visit, I only need to have that intent. I only need that intent. If I just have the intent to send my parents on an overseas trip, the agent simply does everything else.
And this will now happen across personal life, companies, and every domain. Right now, the execution layer of agents is not perfect. So people are, at some point, sensing that this kind of thing will happen someday, but it is extremely clunky.
Through harness engineering and various other methods, the people trying to experience this future a little faster and build organizations that run in this way are the kinds of people I want to invest in through Nitro now, or the kind of company image I have in mind. Recently, Karpathy talked about something like psychosis.
Seungjoon Choi But when I look at people who are pushing ahead and working in this field these days, they completely distribute all the context and run it simultaneously, and it actually feels like the people grinding more bio-tokens into it are doing well. Are you seeing things like that? Whether that is possible or not? Honestly, it has become more exhausting.
Chester Roh For people who like work, it has become more exhausting. It has become incredibly exhausting.
Seojoon Kim Because in the past, the physical act of execution itself was the bottleneck, but now, if you set up execution properly, agents do it all. So it has become physically possible to do 10 or 100 tasks at the same time.
In something I wrote, I said that our company is now a 20-person investment firm, but by the end of this year, we are aiming for the performance that a company of about 1,000 people would produce.
And one thing I often tell our employees out of habit is that they must not repeat next month the work they are doing now. In the past, that was completely natural, but in fact, a large portion of what companies do is repetition.
They slightly modify what they did this week and repeat it next week, organize financial materials this month, check in on portfolios, and so on, then repeat it again next month and make reports again. These are the things typical companies do. But that repetitive work is now something that can be done if you set up agents really well.
So once you keep compressing and setting up repetitive execution work for agents, you gain one bit of room at a time to do other work. Then the work one person does becomes the work of two people, then three people, and once that gets replicated as a whole again, it can grow to six people, and so on. If you build agent teams in that way.
scaling what about 20 people used to do up to around 1,000 people by the end of this year may be too aggressive, but I do think these kinds of trends will happen. Right. In fact, it is not impossible.
Chester Roh Listening to what you are saying, I have also been thinking a lot these days, even though people say AI-native, about what the definition of AI-native actually is. But now it is all quite mixed together. Is AI-native making an existing product faster, where people build something that slightly helps with a person’s work? I do not really think so. If it can completely replace an existing workflow that humans were already doing, then I think that is native. So I tend to draw the boundary that way. I think we need to think about it this way.
Seojoon Kim In human history, a new species has finally appeared that is superior to humans across many domains and lives alongside us.
Chester Roh I agree.
Seojoon Kim For example, in San Francisco, if you want to raise a dog, you have to register it. You have to get something like a dog registration certificate. Otherwise, you will probably have to pay a large fine. So even for companion dogs that live with us, depending on the country or city, systems are emerging where they must be registered and protected as a species. In the same way, a new species has appeared.
Chester Roh I did not know that.
that lives with us and can change our lives. And regarding this species, people who actively think about how we can live together, increase our productivity, and change my life and the lives around us, while developing each day with intent, I think that’s what being AI-native means. It goes beyond simply being good at using AI tools. It means seeing them as independent, equal species. Then naturally, this conversation actually leads into OpenClona.
Chester Roh or into the question of an agent’s status as an agent. In fact, when I talk with OpenClona, it creates a structure where something I could ask Claude Code to do directly is instead assigned through one more assistant layer in the middle. It creates one more meta layer. But as I keep doing that, this friend gets to know me better, and the memory.md and soul.md that eventually emerge often feel like they become the core of it. This is an asset. It turns out this was my tacit knowledge, this was my so-called intent, and this is me. I often get the feeling that “this is another me.” Then, as you said earlier, as we move toward an agent economy.
Identity and Reputation Standards in the Agent Era and ERC-8004 47:11
Chester Roh not just me, but my parents and my child as well, everyone is going to have an agent of their own. And even if I don’t intend it, those agents will carry out an enormous number of transactions behind the scenes for me. That also naturally connects to blockchain, and there are parts that connect to stablecoins as well.
Then what should we give to these agents? Something like a new legal personhood, an agent status? Do you think a system like that needs to be created, CEO? The origin of the concept of a corporation and a legal entity.
Seojoon Kim began with the Dutch East India Company. I think it was around 1604. So it’s a concept that’s about 400 years old. Before that, there wasn’t a global protocol for creating an economic community, investing through it, and distributing returns.
And another point is that a legal entity is treated as a person. So no matter what, there are incidents like this, for example. Even if you’re the representative of a corporation, even if it’s a company you own 100%, if you freely take and use the corporation’s money, you get punished. A lot of incidents happen because people fail to distinguish that. So it is given the status of an independent person, starting from actual registration, with rights, responsibilities, and things like that being established for it. In the same way, agents themselves need that kind of foundation. Then the question is where to build this foundation. Because agents are born digital-native, standards for protocols to build their identity, reputation, and things like that are being discussed. Among them, the protocol that is the furthest ahead.
is called ERC-8004, which is being built on Ethereum. When a person is born, the birth is registered. So, in line with the NFT standard, an agent’s birth is registered, and then the work that agent has done is recorded. When we buy things, of course we look at reviews. We look at how many purchase records there are, and whether the rating is 4.7 or 4.3, and naturally we trust the higher-rated ones more. Then when hiring a person, or when doing business with a company, when hiring someone, we always looked at a resident registration extract. That’s identity verification. And when doing business with a corporation, we look at documents like a business registration certificate or corporate registration certificate before sending money, and then we do the bookkeeping. Likewise, when transacting with an agent, those standards clearly need to emerge. Because this is still extremely early.
we’re not even at the point where such institutions are being discussed at the national level. The private sector is faster. But the reason the private sector is moving faster is that creating this makes money and creates a trustworthy economy. Because agents, too, are already involved in fraud cases and doing all kinds of strange things right now. In a way, behind the countless fake accounts being created on Telegram or Facebook right now are agents. We need to be able to authenticate those things, and to leave records showing that they have performed their work properly. So far, I don’t think there is an answer better than blockchain for this.
Because if someone says they authenticated it on their own server, they can just change it. But if every record is left in a structure connected to a public blockchain, that is trustworthy. Among everything IT infrastructure has created so far, it is still the most trustworthy form of evidence. If the way that agent operates is made into a smart contract and connected to the blockchain, all records remain transparently visible as records. So in that way, starting with Google, Coinbase, and MetaMask, the largest wallet on the Ethereum side, these companies are gathering together to create that standard. Whether ERC-8004 ultimately becomes the final agent identity standard, or whether something else is created, we don’t know. But those movements are being built faster in the private sector. If goods are traded and economic value moves back and forth.
Chester Roh what matters is, in fact, trust in the money moving back and forth, and then a reputation system for the parties to those transactions. That’s what you just walked us through. The blockchain community is also working on these things, but in fact, big tech companies like Google and Anthropic can also propose such systems on their own, and owning those systems itself becomes a tremendous source of power. They will try to do that, but will this end up being divided mainly around certain big tech companies? Or will these kinds of things being developed by the blockchain community in an open domain establish themselves as standards and spread that way?
Seojoon Kim But those two are somewhat converging. Google is also not doing this entirely on its own right now. The very reason Google is participating in things like ERC-8004 and x402 is that if Google stores this on Google’s own servers by itself, how many parties would really participate? So Google is one of the early founding partners that is taking the initiative, but people will trust blockchain more than Google’s servers. Then we can say that this infrastructure.
Chester Roh is already moving toward being strongly coupled with blockchain.
Seojoon Kim Yes, that’s right.
Chester Roh As blockchain and AI truly come together, I do get the feeling that a major business opportunity will come. CEO, then if you could give some advice, people’s concern is something like this. Even if we try that, Google is going to do it all anyway. If you try that, Anthropic will do it all anyway. How are you going to beat Claude? People ask questions like that a lot. Even so, from the standpoint of an individual entrepreneur, if someone has to build a business, do you think there are promising fields? Please give us just a few hints. Like, you should do this.
Seojoon Kim I think it’s extremely difficult.
Chester Roh Really?
Seojoon Kim I really think it’s extremely difficult. The areas that remain, for me, are connected to what I mentioned earlier. Among the areas where an individual can act with their own intent, one of the areas that may remain until the very end is, I think, the area related to taste. Elon said something like this. A basic income society will come, but not a basic income society where people barely survive on rations. It will be a basic income society where everyone enjoys abundance. I tend to believe that will ultimately happen. Because even if you think about it theoretically, humans are currently involved in every means of production. Once all of that, end to end, is replaced by machine agents and robots, society will become extremely low-cost. All prices will get cheaper. In fact, the biggest portion of everything we buy is labor costs embedded throughout the supply chain. Even the cost of the goods, even the cost of the raw materials, includes the labor costs of the people working there. When all of that is automated by AI machines and robotics, making everything and distributing it abundantly is not impossible. For example, the things we eat. In agricultural and seafood markets, robots and AI make everything. Delivering that also requires no human effort. The cars that deliver it are also all made by robots. If you keep following that line of thought, making everything abundantly and distributing it to everyone is not impossible. That kind of world will arrive. So eventually, even money loses its meaning. That’s what Elon is saying. As I go to Abu Dhabi, there are many times when I feel like I’m seeing the future. When I visit the homes or offices of so-called royal families or major family offices in Abu Dhabi, it always feels like I’ve entered a museum. Not necessarily because everything is expensive, but because on one side, Pokemon cards are just covering the wall, and on another side, when I recently went to Abu Dhabi, the office of a chairman from a very large family office in commerce was just packed on almost three sides with tiny model cars. But that doesn’t mean they’re incredibly expensive. So it’s about showing one’s taste. Since everyone is already abundant and has a lot of money anyway, for reference, in the UAE, people who have citizenship are called Emiratis, and their basic income is close to about $143,000.
What Remains to the End Taste and the Inspiration Economy 53:23
Chester Roh Even if they don’t do any work? If their income is less than that.
Seojoon Kim the state considers them poor and gives them that much money.
Chester Roh I want to immigrate. They give you a house and land and treat you that well, but you can’t receive it just by immigrating.
Seojoon Kim You can get permanent residency, but they don’t easily give citizenship. But then that means the so-called problem of making a living is solved. When I look at what people do in that situation, I do think there will be one major Death Valley during the AI transition, but assuming that ends well, I feel like it lets me see a little of how people will live. That’s what I imagine.
Chester Roh That makes sense.
Chester Roh It’s like a glimpse of a future society that has arrived there first. Then people only talk about taste.
Seojoon Kim That’s why Pokemon is such a craze in Korea now too. Not long ago, they stopped an event at Seoul Forest because there was almost an accident, and then was it called Magikarp Run? There was running too, and it was a whole scene. But meeting people like that in the UAE is so interesting, because people older than me give me cards one by one as gifts. Rare items. Yes, and it’s like when we go to a playground, elementary school kids exchange Pokemon cards, play games, and hang out.
We also played like that when we were young, but those things, those tastes, faded away, and the social pressure became that you had to work and prove yourself as a producer. So the reason you make money is to survive. In fact, if you ask most people, “Why do you go to work?” their honest answer is less “I go to work for self-actualization,” and more that they go to a company just to feed their families and get by, and in the middle of that, they try to find something somewhat interesting. For many people, the means and the end are usually reversed to that extent. But if a future where all of that is resolved.
eventually arrives as a good future, as Elon says, what remains is only taste. That’s the kind of feeling I have, and I think I see it a lot in the UAE, especially Abu Dhabi.
Chester Roh Then if that taste becomes a new means of value, or rather, “becomes an object of value” would be the right expression, there must also be a medium that holds it. Yes. I feel like you have already seen that too and are over there in that space. Then this kind of good called taste, what kind of medium should it be held in? What kind of framework should it be placed in? Is that also blockchain? It could be blockchain, or it might not be.
A TCG Card Fund and a New Way to Own IP 58:28
Seojoon Kim We’re making a rather interesting fund right now. We’re creating an Abu Dhabi-based fund that buys only TCG cards.
Chester Roh Trading cards?
Seojoon Kim Yes, trading cards. Pokemon cards, Yu-Gi-Oh! cards, Dragon Ball cards. We’re creating a fund that buys only things like that.
Chester Roh Does the fund simply buy the cards themselves? Yes. So there are broadly two areas.
Seojoon Kim One, which is the larger portion, is really just buying, storing, and selling them. So it’s like an art fund. Another portion is in the venture area, where there are IPs that have not yet been made into cards. Things like K-pop or Japanese animation. Among those, minting things that have potential but have not yet been turned into cards. A fund whose main business is those two things will probably be the first among institutions once this launches. A lot of the inspiration for things like that came from I drew a lot of that kind of inspiration from the UAE.
Once people ultimately move beyond tools of production the so-called useless things Taste is like that, right, by the standards of the past Things related to entertainment are not necessary for making a living.
So in the old days, if someone said they were going into show business parents would say things like they would cut them out of the family registry because they could not make money.
Chester Roh But now, in fact, it is a great means of value creation.
Seojoon Kim Right, so if you look at human history over time from prehistoric times, people gathered and hunted If that stopped, they died So being born, working for survival, and then dying started out as almost 100% of the human spectrum and that has been gradually, gradually coming down with various forms of industrialization, automation, and now AI The image of humans as tools of production is gradually disappearing Then, in the space that remains, we have to redefine what humanness is In fact, humanness in the past meant being a working being If you were born in a rural village, you farmed If you were born in a city, you sold goods That was human work Looking for something beyond that was a luxury things like entertainers there was a time when those were, in a way, somewhat vulgar professions But now the value of those things is what remains in large part things where humans inspire other humans or enjoy something things where stories and narratives can be shared I think that industry will become thicker and thicker So I always think of business along two axes One is utility, and the other is inspiration.
Chester Roh I like that.
Seojoon Kim Every business is located somewhere on this 2x2 matrix.
But what is interesting is that when a new technology first appears it starts out at the edge of inspiration and technology and then quickly falls into utility For example, at the Paris Exposition, when the light bulb first lit up being able to see light at night was an incredible inspiration It was utility and inspiration at the same time But now that is KEPCO or a company that makes light bulbs It is at the very end of utility Communications are the same When I first got a mobile phone as I entered high school, it was so fascinating I remember sending pointless texts to friends and things like that Back then it was not even a smartphone If you sent a text, it would arrive three minutes later and so on It was so fascinating and fun But now those things have actually become taken for granted So when something reaches the end of utility corporate value goes down I think there was also a time when finance was like that When we were really young, there were times when banks were very large companies Back then, entrusting something called money to a third party and being able to withdraw it somewhere else digitally was incredibly fascinating But as time passes, all those things become like utilities like water, electricity, or communications I think LLMs and AI will change that way too They are so fascinating now but in about 10 or 20 years, it will be like, of course it just does it if you ask of course it gives you all the answers, and that will become completely natural just as we now take the internet or communications completely for granted That time will come And when that happens, if we think about what activities people will still not lose interest in and continue doing it will be the things that do not disappear from inspiration.
But the industry that remains only in inspiration is the content and narrative industry But there have not been many ways so far to own this content and narrative industry You can buy shares of related companies So if you like BTS, you can buy HYBE stock and if you like Mickey Mouse, you can buy Disney stock but that does not have resolution at the IP level People’s interest was not broken down to that extent and reflected in the capital market and there was also a lack of tools that could connect those things more at the community level So among those things, one thing blockchain will do, I believe The reason I am paying attention to the TCG market as well is that among the existing tools right now it seems to be one of the few tools through which you can invest in IP and get upside from that perspective This is not Web2 or Web3.
but I am looking at the TCG market now as a prototype of that I think this kind of thing will move over in a digitally native way If that transition goes well, then where NFTs failed in the past I mean, NFTs were the value of inspiration in this new format It was not that the IP itself truly had value So I wonder if that will not be properly created in a digitally native way as well That is the thought I have.
Chester Roh That’s right, and the goods that became the subjects of NFTs were also too small and a bit kitschy and things like that, so I do think the base was too narrow for them to become utilities. But that experiment was meaningful, too. And of course, back then, we believed that would become the future, but in a sense, as you said earlier it may be going through the so-called death valley right now.
Seojoon Kim One interesting trend is that the two are really starting to mix these days. So one Pokemon card now, as you may have seen in the news, sold for over $14 million.
Chester Roh I could not understand it, personally.
Seojoon Kim It is a card called the Pikachu Illustrator card and they say there are only 20 copies in the world. This mechanism, the mechanism of the art market, has not been scientifically proven. It seems to be somewhere in the middle. But I think it is a more interesting market than the art market because it is not as difficult as art. With art, if you do not study it, you really do not know why a certain painting is expensive. But with IP, when more people see it, many more people can intuitively say, “Oh, that character.” For example, when looking at a Dragon Ball card, if someone has watched Dragon Ball they know that a card with Son Goku on it is the best card compared with an extra card. Those things are easy to access and they also have advantages like being easy to store.
But by taking those things as collateral and putting them into a digital space in other words, there is collateral there It isn’t just issuing NFTs, but taking these as 1:1 collateral exactly as they are, issuing them on the blockchain, and then if someone makes a claim, they can receive this, and recently quite a few experiments in that form are being created. We have also invested in one company of that kind.
Chester Roh As I talk with you, CEO, my thinking keeps expanding, and as you said, what remains of people is only intent, and then taste becomes important. This is something that traditionally we have not classified as an asset, and it is still in the realm of inspiration, but you believe that it will become utilitized and become a precious good in some new economy.
If we go back to what we were saying earlier about AI-native talent, even in the AI world, this concept of a company has been greatly dismantled. In the past, you would create the framework of a company, receive stock options, do financing, put in a certain amount of money, and if the company was sold, everyone would divide it up according to their shares. But now we often see value becoming excessively concentrated in one or two people with certain abilities, or one or two people representing the entire value of the company. Even in the U.S., whether it is OpenAI or elsewhere, a single engineer is effectively transferred for as much as a typical company exit, and when you look at the AI-native companies you have invested in, there are cases where that one person accounts for almost most of the company’s value.
Then the concept of a company is actually moving toward the individual level, or the inspiration an individual creates. This also feels somewhat connected to brands and religions created by individuals, things like that, and I think things seem to keep moving in this direction, CEO. CEO, then that world has not arrived yet, so from the perspective of ordinary people like us living in the present reality, how should we invest in assets that have not yet arrived? What subscribers always say is.
“So what are we supposed to do?” That is always the final question. What are you telling us to invest in? What are you saying we should invest in? Then what kind of company are we supposed to start? It often comes down to questions like these. We have now talked about the ideal, and I believe in that ideal. Because I am also someone who lives with those values myself. But there are steps, aren’t there? Please give some advice to those of us who are in reality.
Seojoon Kim The areas everyone is already looking at now are good investments. For example, starting with Samsung Electronics, Hynix, and memory companies, and LLM companies as well. Since many people are already looking at them as investable asset classes, I will speak from a slightly different perspective. Ultimately, as I said, when the agent world arrives, it will need pillars. It will need agent-to-agent language, and then it will need a means for agents to pay other agents, and then an agent saying what it can do, that is what we call MCP. It is like a resume. If my business card says lawyer.
Agent Ecosystem Infrastructure Payments Identity MCP and Public Blockchain Investing 68:13
Seojoon Kim it proves that I am someone who can do legal work. In MCP, I can do web search. I can do crawling. I can make payments. The agent proves these things through MCP. And safely managing those things, playing a role like a fingerprint that handles reputation, identity, and management, will be done by that identity verification protocol. Ultimately, when the agent ecosystem arrives, the area that has received relatively less attention so far is, I think, the value of public blockchains. For example, something like Ethereum.
But if you look at Ethereum, people in the industry ridicule it a lot. Its price is the same as it was five years ago. So people ask, why is Ethereum dead like this? There are a lot of strange memes. People poke it to see if it is still alive. But its fundamental value has changed tremendously.
During that time, the quantity and quality of transactions being created on Ethereum compared with five years ago, then the concept of scalability, then the size of the fees that had to be used when sending transactions, and then when you look at block space truly beginning to fill up densely because of agent activities, in other words, aside from the investment opportunities in IT that everyone currently agrees on, the area that will be created to give these agents payment and identity will be blockchain. Among public blockchains, I am cautious about saying you must buy Ethereum, but if you research which blockchains stablecoins are currently circulating on the most, the answer comes out.
For reference, right now Ethereum issues and circulates about 60% of stablecoins. As you look at that ratio, these activities right now, the activities exploding in agents, I think the agent world has truly not even arrived 1%, though you may agree. There are not many people around us who are using agents well right now. The people using them well now are very unusual people. When this really becomes something where, say, you go to a coffee shop.
and every person sitting there has something looking after their daily life, one kind of, I used the expression species, as a new species, not as a servant, but as a new species, acting together with us, there will be an enormous amount of transactions and value exchange, and the infrastructure itself for those things will be public blockchains. So including stablecoins, if you closely observe the assets where a lot of agent transactions are being onboarded, there may be interesting opportunities. Just as there was a huge bubble in 2017, and that actually led to Samsung Electronics and Hynix showing tremendous rises recently, I cautiously offer the view that such a trend could occur at the point when agent finance is implemented. Yes, that is an extremely insightful point.
Chester Roh Always, since we are now moving into the latter part of our conversation, I think your particular talent, CEO, lies in seeing a future that others have not yet seen and placing your own conviction and bets on it. Because seeing the future is one thing, but how far into the future you have seen and how much conviction you have is actually revealed by the amount of money you bet. People who only talk and cannot bet are not people who truly believe it. But you are someone who believes it and is betting on it. Now, as we move a little into the latter part of the conversation, I am curious about the background that led you to have that perspective, as Kim Seo-joon the person. Mr. Kim, you did mention this earlier, but it seems like you read a great deal from college, or even from childhood, and also had many good questions and good mentors. The things that made you who you are today, whether you might call them your luck, or the culture your parents gave you, where did those things come from? How were they formed? That’s what I’m curious about. You don’t really know yourself, how you were made.
The Builder Instinct Behind Seojoon Kim and His Web 2.0 Community Experience 72:16
Seojoon Kim But I think I really liked imagining things when I was young. I think I liked making things. The things I always asked my parents to buy me were science kits and LEGO. So of course I’m not the only one like that.
and when you look at LEGO sales, a lot of people like making things. I think all of us have a builder tendency. But the thing is, at some point, people compromise on and give up the fact that they like making things themselves, and making means making what I want to make, but they just move toward making the assignments they are given because they have to. At some point, they start taking their hands off creative making, and I think one thing for me was that I kept trying not to lose that and to keep holding on to it.
One of the very big and important connections in my life is CEO Joonghee Ryu. CEO Joonghee Ryu is 10 years my senior from high school, and even though there is a 10-year gap, we were able to interact. I met Chester Roh a few times back then too, when Chester Roh was working on Tattertools, and when CEO Joonghee Ryu was creating Olaworks, I interned there while it was being made. What I remember from that time is that we held a lot of seminars on “What is Web 2.0?” Chester Roh may remember it. That was really interesting to me.
Chester Roh There were a lot of events like that back then. I think that, when Web3 opened up for me.
Seojoon Kim it was like a blueprint showing that we could create a similar culture and community. So we take Web2 for granted now, but the difference between Web1 and Web2 was from read-only to read-and-write, and then a new kind of platform where communication was possible, a two-way communication platform, emerged for the first time. It feels so obvious now, but back then, about what on earth we could do with it, just like how we’re talking about the future now, we gathered and had tremendous conversations.
Chester Roh There were people who were dreaming. It was really minor back then.
Seojoon Kim There were truly only a few people using the term Web 2.0 at the time. It was a term I first heard in that gathering too, and even though I majored in computer science, no one at school used the term Web 2.0 at the time. But in the end, I think I received positive feedback that the future is shaped in the direction where people who imagine things make efforts, and make efforts with intention.
Seungjoon Choi I do want to ask this, but it’s the kind of question we may need to edit out if it doesn’t fit. The things we’ve been talking about this whole time do seem to have a certain texture to them, but saying that people who don’t dream, or don’t have a manifesto, or don’t have a will can just live while earning a basic high income feels like it could be very dangerous. So about that downside, though even the expression “downside” itself is difficult, I think advice for those people is also necessary. My disposition is not this. Right, not everyone has.
Chester Roh the disposition of a businessperson or entrepreneur who has to open up the future. In a way, those people are a very small alpha group in society. And they do become leaders. And in doing so, they change the world. And it’s true that everyone benefits because they change the world, but we can’t say that everyone has to become an entrepreneur. What do you think, Mr. Kim? For example, to put it negatively, there are also many people who do not get on or cannot get on the train of new change.
Seungjoon Choi This disposition itself may simply not fit them.
Chester Roh Then should those people just wait?
Redesigning Education in the AI Era and the Question of Being Human 76:46
Seojoon Kim So I think everyone needs some discussion about what humanity is. One of the major criticisms when people say school education is wrong is that I often think there is almost no need anymore to sit and listen to classes at school. Shouldn’t the time spent sitting while looking at another student’s back disappear? Because when it comes to studying, to learning knowledge, the information gap has narrowed so much. In the past, from the perspective of learning basic literacy, if you didn’t attend classes at school, I think there was a time when you couldn’t acquire so-called social norms, common sense, manners, and things like that.
Now, with smartphones, people can see how others live, how things work, and how people think. Even elementary school students have a level of literacy that was unimaginable in our time. Because they are exposed to information. That’s why I think many of the things schools should preserve are not one-way blackboard lectures and learning, but things that require people to look at each other’s faces, discussions, and then using the body and exercising. I tend to think schools should shift much more toward those things. At least the usefulness of teachers giving one-way blackboard lectures has almost disappeared, in my view, because there are now so many channels in everyday life where you can learn without that. Right. In a way, for our children.
Chester Roh what matters is not some uniform measure of knowledge, because AI will fill in a lot of that knowledge, so I also think we need to grant them the right to go as far as they want in the direction they want. Please create a school. Shouldn’t there be a school suited to this era?
Seojoon Kim So I’m not sure whether that will take the form of a school. Chester Roh briefly mentioned it earlier too, but the very concept of school is now being unbundled. The community we’re building called Nitro is something I hadn’t even fully realized, but parts of it may be playing a school-like role, and the same is probably true of Y Combinator, and Peter Thiel’s fellowship foundation is probably similar too. So I think many more of these will be created in more diverse forms.
I mean, a school has a fence around it. You go inside that fence and fix your identity for several years, six years if it’s elementary school, or four years if it’s university. Is fixing it like that really the best method? Under a single title, I think there is that kind of fundamental question. If we look at the thread of what you’re saying, Mr. Kim.
Chester Roh Schools, venture capital, and things like that are very traditional concepts, and they are quite tightly structured. But you often describe these things as being unbundled, and you frequently say they need to be rewritten. In a way, Hashed is also trying to unbundle venture capital and rewrite it anew, right?
Seojoon Kim Yes, but I’m not sure that necessarily means venture capital. I mean, venture capital… We make a real effort not to think of ourselves as venture capital.
In reality, we do operate a venture capital firm, but we also do venture building, and we create various subsidiaries to pursue new businesses, and we’ve also done a lot of community activities since early on. So in the end, I think the core is probably the community. In other words, a community where people can be together, a community where people can carry value together.
And since we are still in a capitalist society, as a way for people to engage in capital-producing activity together, there is a role an investment firm can play. Because we acquire equity, provide help, and incentives become aligned, through that process, in the picture we imagine of being together with a more future-oriented community of founders, venture capital was one axis of that. But I don’t think we consciously think, “We must remain venture capital,” or “We must change venture capital.” A community of dreamers, of people who dream.
Chester Roh Trying to gather those dreamers into this compressed space, that felt to me like your intention. That vibe. I also went to the first Demo Day presentation.
and since you actually invited me as a fellow, I went there and saw it. To be honest, the value I feel as a fellow is that I was extremely curious about how those young talents look at work, and how they handle it. And for me, it was a chance to learn those things in reverse.
Seojoon Kim There are a few episodes where I learned a lot and was pretty shocked. We held two offsites in Jeju Island. The founding teams and fellows were mixed together, and the concept of security, the concept of security around source code, is a bit different.
Source Code Sharing Culture and Where Value Accumulates 81:33
Chester Roh The concept is different from what we traditionally thought of as IP. In fact, it also connects to the Claude Code leak incident that Jinyoung was recently at the center of. Please go on. In a slightly different context.
Seojoon Kim being added to a company’s GitHub account usually requires signing an employment contract and signing something like a security pledge, right? But when we look at the people participating like this, in Vibe, our program that we’ve now renamed Nitro, so many founders and fellows are all just invited to each other’s companies’ main repos. They develop together, even though their incentives are not even aligned, and that is something unimaginable in a traditional company. They just develop together, and someone who can write better code stays up all night and writes it for them. Watching those things, I often think they are clearly a different breed. You could say they are a different breed, but if the source code itself.
Chester Roh were an extremely central part of a company’s value, then from an individual’s incentive standpoint, they wouldn’t do that. But the reason they were able to do that so easily is that the value gained by opening it up is far greater, and also, when they think about the value of their business or company, they believe that isn’t where it lies, right? They talk about that a lot. They say that what they’re building.
Seojoon Kim other people could just make it all with a click. We say that among ourselves, and founders say it to each other a lot too. Then what matters is not what we’ve built so far, but reaching a state where, by helping one another with what we’ve built, we can develop it faster together. If that itself doesn’t happen, they seem to think there is no meaning in protecting what I’ve built.
Seungjoon Choi So the value is outside GitHub, in a sense.
Chester Roh It seems you consider it very important not to miss the timing, and then beyond that timing, there are certain aspects of the business, the areas of tacit knowledge you mentioned earlier, and the area of brand that emerges from combining those things. That also seems to be connected to the fact that Hashed, or a person like CEO Seojoon Kim, is doing it. I think we are entering a time when it is very difficult to define what the value of a business is. And I think you also find it difficult to show that clearly in one sentence, but you seem to feel it. I think that is deeply embedded in the actions you take. I can’t define it in one sentence right now either. What you were trying to say earlier about Hashed, “Ah, what CEO Seojoon is trying to build is a community of dreamers.” “Then in the end, the core asset right now is the dreamers.” “Then the game is about finding those dreamers well.” “It’s a game of making them.” That’s how I’m interpreting it in reverse. You are seeing this as a direction for creating new future value, that’s what I think. Today, CEO.
we opened quite a few parentheses but didn’t close them. And the flow of thought went back and forth, but today’s goal itself, actually, was just to query many of your thought tokens, not necessarily to arrive at some tightly defined conclusion. But vaguely, I think the conclusion is already there now. Now, putting all the conversations together and all that, finally, for our AI Frontier subscribers, we’re using the term subscribers now, but people in Korea who think somewhat seriously about AI do watch us a lot. If you were to leave just one sentence, what would you leave?
Closing: Depth of Relationships and Time with Intent 85:32
Seojoon Kim There are two keywords I think about a lot myself. One is the depth of relationships. Relationships come in many forms. There are relationships between people, and there is also depth in the things I want to achieve well or the things I like. And the other is time with intention.
I think intention is truly the last piece humans possess. Because AI does not have intention first. AI is now the execution layer, and the only thing people have left to give AI is intention. But intention, I think, develops only when you organize it well, think about it a lot, and carry relationships deeply. So there used to be something called the 10,000-hour rule, right? But now, the important thing isn’t just putting in those 10,000 hours, but that someone who spent 100 hours thinking with intention can outperform people who spent 10,000 hours thinking without intention. I think that’s the kind of environment we’ve entered.
Chester Roh Actually, older people often say, “Life isn’t about speed, it’s about direction,” and it seems we should see this as a time when those human values, the value created by a single person, become even more precious, if you look at it the other way around. No matter how much we enter the AI era, I really think the value of people becomes even more important, and the impact created by one person, whether we call that person talent or an outstanding individual, is actually getting bigger. It’s not getting smaller, it’s getting bigger.
Seojoon Kim Right, because a person can play many more roles now. In the past, it was superficially company to company, and then the individuals inside the company would get diluted. But now, the driving force of a company… there will be many strong small businesses, right? A small number of people.
will lead numerous agents that execute this person’s intentions, and because they can create a bigger impact, relationships between people will increasingly come to the surface, and how deeply those relationships are carried forward will become more important. So from superficial relationships between companies, we may now be entering a time when relationships between the people doing the work become more important. That’s what I think.
Chester Roh So in a way, you’re seeing the future of companies coming down to the individual level, right? Yes. Understood. CEO, it’s a weekend morning, so thank you so much for your valuable time.
Seojoon Kim Yes, thank you.
Yes, thank you.