EP 101
EP 101. “Mission Is the Moat”: How VIZCOM Built a $50M AI Design Startup
VIZCOM Interview Starting in San Francisco 0:00
Chester Roh Today, as we’re recording, is the afternoon of June 15, 2026. I’m in San Francisco right now, so the time is a little later than in Korea. Today, from San Francisco, far away in Seoul, we have Seungjoon, and someone who has appeared on our AI podcast a few times, CEO Kim Minseok, joining us to have an interesting talk. I came to the Bay Area about ten days ago because of CEO Minseok. After I came, CEO Minseok said that entrepreneurs here are doing so many interesting things right now, the atmosphere is incredibly heated, and then because I’m also very interested in the biology business, people on that side, at the frontier, are also doing interesting things, so he suggested, why not meet them all? Over the past ten days, we met so many people. Researchers at frontier labs, really young, dreaming founders who are 20, 21 years old, and investors, and even people at big companies in between. So I gained so many learnings, and I’d like to cover all of them, but there are also many people who can’t appear because of media regulations, so we’ll introduce them slowly. For the first in that series, we’ve invited a very interesting person.
Introduction to Jordan Taylor and the Purpose of Connecting with AI Frontier 1:13
Chester Roh He is Jordan Taylor from a company called VIZCOM. Jordan, please say hello.
Jordan Hello everyone. I’m Jordan. I’m the founder of VIZCOM.
Chester Roh Jordan does speak some Korean, but Korean isn’t exactly super comfortable for him, so today, actually, with a sense of purpose around what kind of contribution AI Frontier can make, I thought it would be good to connect Silicon Valley, Seoul, and China’s Frontier AI a bit, and as the first step, we’ll try connecting San Francisco and Seoul. So today’s conversation will probably be conducted in English, and in between, there will also be a little time for Jordan to show off his Korean. But we’ll start by speaking in whatever way is comfortable, English or Korean, so let’s slowly get started. Welcome, everybody.
Minseok Kim Thank you for inviting me as well. I’ve appeared a few times too; I’m Kim Minseok. Since coming to San Francisco, AI startups are definitely concentrated here, and on a very small piece of land, many AI startups, frontier labs, and investors investing in them are gathered, making it a very unusual space, so while staying here for over two months, I’ve met a lot of interesting friends, and I had shared some of those stories with CEO Chester Roh, and now we came together, met many interesting friends, and if possible, we’d like to introduce some of them on this podcast, and among them, one of the friends I most wanted to introduce first was Jordan.
Chester Roh That’s right. That’s right.
Minseok Kim We’ll probably unpack the reasons in various ways on today’s podcast, but Jordan speaks Korean about as well as most Korean Americans, understands Korean culture very well, and in terms of revenue and product, he’s growing his AI product company very well, so I thought it would be good to share this story with the Korean audience, and I strongly recommended Jordan. That’s right. Yeah.
Chester Roh Welcome to the AI Frontier, Jordan. Welcome.
Jordan Thank you for having me. I’m really excited today.
Chester Roh Jordan is the most interesting person among the people we’ve met here, and so it’s actually I asked Matthew who would be the most interesting and most impressive. Thank you. Jordan, I just want to ask you your background. Would you
Jordan Taylor’s Study Abroad in Korea and Industrial Design Background 3:48
Minseok Kim share a brief background, you know, how you started, how you became interested in industrial design, and how you studied in Korea?
Chester Roh Please introduce yourself. Of course, of course.
Jordan Should I do half Korean, half English? So yeah, I studied in Korea. When I studied at Hongik University, I studied industrial design. After that, I went back to the United States. I was working at Honda and doing car design, and that’s when I realized, like, okay, well, I should maybe take my talents elsewhere. I ended up going to NVIDIA right after Honda. So at this point, I realized at NVIDIA, the state of the art at the time were things like convolutional neural nets, GANs, and really early signs of images being able to be generated, right? This was a very novel technology at the time. So I saw that was happening, and
my background was in industrial design. So I saw an opportunity here, like, wait a minute, I think the way that these models are going to improve is going to fundamentally change how I do my own work, which was drawing cars, rendering cars, doing this by hand in Photoshop. This technology was very clearly kind of solving a pain point, which creatives were spending a lot of their time doing instead of actually being creative. So that was kind of where the core idea began, coming from the world of car design and seeing an opportunity there. And then I kind of just was,
you know, I made a Reddit post about this, and then it went really viral on Reddit, actually. And that’s how I built the conviction of, like, oh, I should just do this full time. And I had about eight months of savings at the time, and just kind of lived off that. And my co-founder, you guys saw downstairs, yeah, we’ve known each other since like fifth grade, and then we just started building VIZCOM then. I see, I see. So it was in 2020 when you
Chester Roh decided, you know, okay, let’s found the company with this idea. 2020, yeah?
Jordan 2021, yeah, right when COVID started.
Chester Roh Right, right. So it was actually when we had GPT-3, right? Yes, the language model was not that perfect, but you know, we had a lot of image playgrounds.
Jordan And then DALL-E came out in 2022 or 2021. 2021. Yeah, and then there was Disco Diffusion before that, and Stable Diffusion a little bit before that. But yeah, around that time.
How Reddit Reactions Cemented Confidence in Founding VIZCOM 6:30
Minseok Kim What was the Reddit posting like? You wrote, and what was the reaction? Because you said that you got a conviction for starting a company. Yeah, working at NVIDIA is a pretty good job. So what was the moment you felt like, “Okay, this is what customers want. I definitely should do that”? What was the moment for that? Yeah, so I just made an app by myself. But then
Jordan after that, I just went to Reddit and was like, “Everyone, I made this. I made a web app that can render your sketches,” and that was the post. And I was like, “Oh, I made this. I made this myself.” And then I went to sleep, and then
I literally woke up and had thousands of upvotes. People were like, “Oh, this is cool. Wow, how can I access this?” And I had work that morning, so I was like, “Oh, wait a minute. I gotta call off work today,” and then I got to working.
VIZCOM’s Role in Turning Ideas into Products 7:26
Chester Roh So actually, we have a lot of questions about, you know, your journey, your entrepreneurship journey. Yeah, yeah, yeah, but before that, let us know, just introduce what your company is doing, because the audience needs some understanding. So of course, let’s talk about your company’s business.
Jordan Sure, sure. So let’s see, where do I begin? People listening, the chair you’re even sitting in started out as a drawing or someone’s idea somewhere. And this is known as going from paper to production. So that process of taking something that starts out as a drawing or an idea and then bringing that to the real world, that is kind of what our tool is really focused around. It’s like, how do we enable designers who have ideas
and authorship, essentially allowing them to manifest their ideas, allowing them to actually hold it in their hands? So that’s kind of the tool suite and the problems that we’re trying to solve. And we found a lot of traction in areas like automotive,
footwear design, industrial design. These are kind of the core pillars in which our tools function. Yeah.
Chester Roh What kinds of artifacts are mainly being made? Is it mostly about car design, or shoes and bags? What are the most frequently generated images? Yeah, yeah. Cars, shoes, furniture, everyday objects, things that are just around your house.
Jordan But there’s a lot of intention behind these designs, because these are people that were already designers or aspiring to be professionals. Yeah. So can we ask that,
Chester Roh if you don’t mind, the revenue scale? How much investment have you secured? Yeah. So I can’t share exact revenue numbers,
Jordan but how much we’ve raised is around $50 million to date.
Chester Roh Around $57 million. Huge amount of money.
Jordan Oh, yeah. Yes, yes. I think you are helping a lot of industrial designers to amplify their capability inside
A Workflow That Accelerates the Iterative Process for Industrial Designers 9:24
Minseok Kim of their companies. Would you share how your product can help their workflow or their capabilities? Yeah. So let’s see. Before VIZCOM, right, you’re having to draw
Jordan something and manually go into Photoshop, and then understand how to take that rendering and place it in 3D software, and that was very time-consuming. And now what people are able to do, one way I like to think about it is, they’re able to fail faster. So essentially beforehand, you can imagine a programmer, they have what’s called a compiler to tell them if the code is working or not. But people that design for the physical world, they don’t really get a chance to understand if it’s something they want to commit towards or if it’s working until they get the physical prototype a month later. So we’re basically allowing designers to not only explore more ideas, but to understand which ideas not to explore, because beforehand it was really hard to get that insight. So our tools are essentially allowing them to do those things faster, which inherently allows them to then understand what not to commit towards. So that’s kind of one way. And then, let’s say in the footwear
context, you’re designing a shoe, you’re getting through your iterations, and a month later, two months later, you get the prototype. Now you can kind of go straight from, and I’ll share some visuals with you guys, but from paper to an actual 3D-printed shoe in your hand the same day. That kind of stuff is what really accelerates the ideation process, but also product development in general.
Yeah. So traditional industrial design takes a long time for iteration. Your product actually accelerates this process and helps industrial designers figure out multiple ways and finally find the best design for the company. Yeah. Yeah. So actually, we are living in the era of ChatGPT
VIZCOM’s Differentiation in the Frontier Model Era 11:21
Chester Roh and Claude, and Claude and ChatGPT are literally doing everything. Yeah. So if I ask for a drawing and ask it to transform it into a fine 3D sketch or something, they are actually also capable of that. So, why should we choose VIZCOM over the other frontier models? Yeah, no, that’s a good question. So the way I think about it is,
Jordan it’s more than just a single function, right? It’s this idea of owning the entire process of: I have an idea, and now I want to actually be able to hold it. Now at that point, and we were talking about this earlier, it’s like the mission is the moat at that point. Yes, the mission is the moat, meaning that our desire to really allow designers to go from end to end, that desire alone is what allows us to have such levels of focus. That is almost, I would say, too small for maybe a frontier lab, but too big to ignore. So that’s kind of this dichotomy that we
find ourselves in, where we’re very specially placed, where we’re super oriented around people that design for the physical world. Yeah, and have a very particular focus on our ICP. So that’s kind of how I like to look at it. Now, things can always change, but that’s how I see the future: our level of focus and that mission being the moat. Interesting, interesting.
Chester Roh I really like the sentence that the mission is the moat. Yeah. So even though the big frontier models are providing similar functions, the community, your service as a community, yeah, exactly, is highly meaningful. Exactly. The community, yes. Communities are hard to copy.
Jordan Communities, yeah, yeah, are really hard to copy. That actually
Chester Roh aligns with your first posting on Reddit. Yeah, exactly. Your mission actually originated in the Reddit community.
”Mission is moat” Born from a Personal Problem 13:28
Minseok Kim Oh yeah, I think “mission is moat” is definitely resonating with me as well. For early-stage startup founders, they might ask you, how did you formulate this kind of thinking? I think, as you have multiple experiences from the beginning, how did you formulate this kind of perspective, like “mission is moat”? Would you share some more? Yeah, yeah, I think it was very personal to me. I think one
Jordan thing to avoid, or what I tried to always avoid, was trying to do something for the sake of being a founder. I was never trying to start a company. That wasn’t my goal. My goal was, oh, I personally am just spending so much time, and it was annoying. So I felt like I was trying to solve something that was very personal. And coming from industrial design, I was building a tool for myself. So that’s kind of how I was able to build this perspective. It was like, I want to solve this problem for myself, and then that’s what led me to kind of going down this path. So yeah, I would say having a very big problem that’s close to you is what kind of allowed me to have this perspective. Yeah, following up on the discussion, as a founder, you say
Minseok Kim “mission is moat,” and how do you spread your thinking to other employees and have the same mission, and then work for this? How do you actually structure your company and workflow, and how do you spread your… So is there any sentence that describes your mission?
Chester Roh Do you have any mission statement?
Do you have any mission statement? Yeah, no, I always say this. This may sound crazy, but I always say, if God didn’t make it, hopefully one of our tools did. It’s basically saying, I want the world to essentially reflect the possibilities of our tool. So you should go to the store and say, “Oh yeah, this was made by VIZCOM, right?” And it should feel that way. So I think just getting people excited about
Jordan the physical world getting exciting again, I think, a lot of things are so digital. We see our screens all day, that it’s hard to find signal in the digital world. To me, the next frontier question is now, okay,
not how realistic, but how real can you actually make it? That becomes the new question, and signal that people are looking for. Yeah. So yeah, that’s kind of where I think.
VIZCOM’s Technical Journey from Pix2Pix to Stable Diffusion 15:59
Jordan that’s kind of where I think.
that’s kind of where I think.
Chester Roh When you just founded your company. So you said you started in 2022, yes? Looking back, we didn’t have ChatGPT yet and GPT-4 not yet. And maybe at the time, it was convolutional neural networks, yeah, and we had GANs or Stable Diffusion. Stable Diffusion was way later. So what type of AI technology have you started with?
Jordan We were using some of the convolutional neural nets you mentioned, particularly the paper Pix2Pix. Pix2Pix, actually from Park Taesung. He went to KAIST. He was a very smart guy. We sat next to each other or we sat kind of close. I would see him sometimes during lunch, but I was always nervous to say hi. So I would just wave to him, but “Park Taesung, hello. Later, maybe let’s eat together.” But somehow I couldn’t. So Pix2Pix, so yeah, Pix2Pix was the paper, and then, yeah, that was just super inspiring. That was crazy inspiring. I don’t know if you
guys know BERT. That was the biggest language model at the time. That was so futuristic, but now it’s so old. But that’s kind of where we were at the moment.
Chester Roh Yeah, right, right.
Seungjoon Choi Pix2Pix is kind of the precursor of ControlNet, right?
Jordan Yeah, exactly. It was trying to do what ControlNet did, but in a different paradigm, right? Pix2Pix was the first technology
Chester Roh you leveraged upon. And then next is, could you please tell the technology transition over time? Yeah, so Pix2Pix was the fundamental thing
Jordan that fit our mission the best, which is using drawings, because our users already know how to draw. So that made up for a lot of the gaps in the models. So then naturally, I think Disco Diffusion came out, because people started realizing you could do text- based diffusion, meaning that you could use text to help guide the diffusion as the images are being generated. So that was kind of huge. That was insane. It was like, wait a minute, you’re telling me I can use words to essentially create an image? That was wild. So I’m thinking, let’s see. That was
- Then Stable Diffusion came out, Stable Diffusion 1.5 first. That changed everything, and then we just kept going. So then I think Flux, a bunch of other models, and Stable Diffusion 2.5. It kept going and going, and then things started really accelerating. And it was like new models every week, right? But yeah, that was the kind of time. Yeah. Is your product still dependent on
Chester Roh Stable Diffusion, or are you just in the transition to the modern models?
Jordan It’s kind of a mix of all of them now. They all have their own special cases and uses, right? I think how old something is should never be the judge of if it’s useful. I think people get so caught in the newest thing all the time. “Oh, the newest thing must be better.” But I think sometimes the older thing has its place still in the modern day. Yeah, so we still use it in some areas where we think it makes the most sense.
Product Development Starting from Problems and Internal Dogfooding 19:31
Minseok Kim I’m curious about how you actually, you should not only focus on technology, but also eventually you should satisfy your customer with the greatest product. How do you keep the balance between technology and product, and how do you actually help your team members dogfood your product? Yeah, so that’s interesting. I think a lot of people fall into this trap
Jordan where you start from the technology and then you search for a problem, whereas you should start from the problem and then ask yourself, okay, what technology can fix this? So I think we always start from problems first, the pain points. What are we really trying to solve here? So we start there. That’s how we kind of avoid that.
And one thing we’ve done is hiring designers internally. So we have car designers, footwear designers that actually work here, and they are the ones testing our tools before it goes into production. So that’s how we’re able to kind of, I guess you could say, dogfood our products, by hiring our customers internally. Yeah, so there’s a car guy downstairs, and everyone’s using our tool inside. Yeah.
Designers’ Judgment Amid Design Debt and Choice Overload 20:48
Seungjoon Choi In that context, I’d like to give a question in the context of dog- fooding. Do you manage design debt? I mean, nowadays we talk about cognitive debt, because there is such a hugely vast area of exploring space, latent space, so we have to choose something or select something with taste, but that causes overwhelm, you know. So could you talk about that? Yeah, it’s interesting,
Jordan because now you can make so many things that it becomes a new problem. Okay, now what do I pick? Which one is the right one? You see this with code. You’re generating so much code now that it’s like, how do we review this properly? And design kind of has this, but I would say it’s not as bad, because it now becomes more about your eye, like how do you spot what matters, and that’s on the designer. So I think that’s a good thing.
It’s a good thing to lean on the designer to make that choice, because you don’t want to really I personally don’t think you want machines telling you what looks good all the time. There’s something very human about why food tastes a certain way or why music sounds a certain way. It’s like, that’s us, people. So we really try to lean on the designer to make that choice. So yeah, it’s interesting, and I think
giving the tools to not feel like such a slot machine, but feeling more like an actual tool, like I’m in control and there’s a reason why I’m doing a certain thing. Yeah, that kind of mitigates creating so much slop, like a lot of stuff, because more doesn’t always equal better. So we sometimes think it’s about the qualitative nature of the little bit of things you’re doing. So yeah, that’s kind of how. But my question is, have you developed
Seungjoon Choi some kind of scaffolding tool for designers in that context, for dogfooding? It’s such a new problem that we’re kind of exploring it right now. We’re thinking about it right now, so
Jordan it’s not as much of a problem as making sure the output matches their intention. That is usually the core problem that usually drives creating a bunch of things, because the only reason sometimes they’re making so many is because it’s not getting it right in the first five or ten tries. So that’s one way I’ve seen that being developed, but we’re still thinking about it. It’s such a new concept. Yeah.
Jordan Taylor’s View on Korean Design Education and Professional Ethics 23:40
Chester Roh Actually, the reason why Matthew and I chose you as our first guest in the Valley is that you have a strong connection between Korea and the States. So I’d like to ask a question about what brought you to Korea. So when I see what made you come to Korea, I would like to ask about the life in Korea. Yeah, yeah, yeah. That’s fun. Let’s see, when was it? Actually, in high school
Jordan when I was there, I heard about KAIST. From a KAIST friend I heard about it. KAIST, KAIST the school. The school.
Minseok Kim Oh, when you were in high school? [Chester Roh] You heard about my school? Oh, really? That’s really smart. So I heard about it from a friend.
Jordan But I was like, oh, KAIST? Where is KAIST? And I looked at KAIST, I was like, oh wow. It’s like Stanford, but in Korea. And I was like, oh, but I don’t think I could do that. But I always liked design. I was like, design school in Korea, oh, Hongik University. So at first, I did a little bit of design school in the US at CCS, and then I was like, oh, we had a lot of international students from Korea. So I was like, oh wow, they’re so good at design. I was like, wow, design, how are they so good at it? Why are they so good? And they were like, oh, we’re from Korea, and before coming here, we went to an academy. I was like, academy? What do they do at an academy? So I heard about this, and I was like, wow, Korea, I really have to go. So I just went to Korea, and then I stayed there. Where, Sinsa-dong,
when I lived there, I lived in Jamsil. I went to a university in Gangnam and studied industrial design studied it. Yes. But as time went by, I learned Korean. I had a lot of friends too, a lot, so even today I still have friends. Yes, but yeah, and then let’s see. When I lived in Korea, it was really fun. Always back then, I liked sundae the most. I always ate sundae. But sundae, sundae, sundae. But Korean food culture as a whole was all good. So as time went by, I got really accustomed to the Korean mindset
and understanding how talented engineers were, particularly in Korea, compared to just the working style and the work ethic, I think was very special to me. So then when I came back to the US, I was like, oh, okay, wow, there’s a gap here, noticing that. I was like, wow, why are they only staying in Korea? But then I noticed there’s Taeyoung. He was an international guy from KAIST, and I was like, oh, interesting. So I just kept going back and forth between Korea and the US and building, making friends, and yeah, it was just a very natural thing. Yeah.
The Confidence Korean Talent Needs to Step onto the Global Stage 26:46
Chester Roh So could you please elaborate a little bit? A bit more on the gap you just felt, between why the Korean guys behave like that. They should come, they should behave that way. That way is the right way, but you just mentioned the gap. Could you elaborate? To be more clear for the question, you mentioned that Korean people are smart, work hard,
Minseok Kim and have a really good work ethic. So what actually are the barriers they have internally? What do you think about that? Because you have many friends. Right, right. Yeah, I noticed this, and I understand it now after
Jordan going to Korea and being an international student, and I felt this way too. So I was like, “Oh, wow,” I understand how they’re feeling, and I think sometimes it comes down to having not only the information, but the confidence that they can actually do that. Because I know it may sound kind of crazy. It’s like, “What’s Silicon Valley?”
“What’s California? Why would I go all the way there?” “I can just do that here,” right? But I really do think it comes down to, one, being slightly open-minded, but understanding that “Oh, wow, what I’m capable of doing is actually super special in a different context.” Korea is so small, and the moment you come to California or somewhere even bigger, like the US, right, your
mindset is like, “Oh, wow, there’s a lot to go around here. There’s actually a lot; this is a very big place. There’s a lot of things to do here, and the way I approach things is actually very unique.” Like a lot of people, especially in the West, like Americans, they don’t really think this way. A lot of people in the US don’t even have a passport or even travel super far, so their perspective is also very narrow. So the moment you kind of come in with a new perspective that’s maybe not so traditional, it’s even more appreciated, whereas everyone in Korea is so, so smart, very, very smart, so the competition is crazy. It’s crazy, crazy. So I think, yeah, sometimes there’s just taking that leap, and some of my friends have done this, and you guys are even early examples of this, right? I really do think there’s this kind of mindset, really, the mindset. Yeah, but it’s changing fast, I think, these days.
Chester Roh Yeah, I think so. I think so, yeah. Most of the younger generation Koreans, they speak perfect English, and they are
Jordan Oh, yeah, yeah, okay.
Chester Roh more self-confident and more self-assertive than the generation like Matthew and me. I was trying to be younger. So I think that your context is right. As Koreans, we are really, really smart enough, and our discipline is really good, so maybe we should be more self-confident in ourselves and be brave in a global context.
Jordan Right. So yeah, I really do think that’s it, because I’ve had many friends from college who are now working in the US, and they knew just a little bit of English at first, but then they just got really good really fast over time, right? So yeah, I really encourage it. So when you meet Korean entrepreneurs
Hiring Korean Talent and VIZCOM’s Expansion into Asia 30:11
Chester Roh coming from Korea to San Francisco, please help them. Yeah, yeah, please contact me,
Jordan or I am very open to connecting, and I love these things. Yeah, this channel is famous among AI engineers
Minseok Kim and researchers. I know that your team is hiring, so if you are seeking very smart and hardworking Korean talent, it’s an opportunity for you to brag a little bit.
Jordan Yeah, no, totally. We’re definitely hiring right now. So if anyone’s a software engineer, or even in product design, doing design, or any kind of engineering, even business ops, right, like how to really build a sales team, let’s say in Korea or in Asia, we’re hiring for all these kinds of roles, so we’d love to connect. Feel free to reach out to me. Just my name at vizcom.com, and that’s it. Awesome. I think a lot of Korean talent
Minseok Kim is going to apply and send messages to Jordan soon. But if people go, where should they be? What’s the menu there?
The Mindset of Finding Answers for Yourself in the Challenge of SF 31:20
Seungjoon Choi If people were to go, just in general terms.
Minseok Kim But in general terms, I think it would be hard to come as tourists. In my view, both sides need to be helpful to each other here, so if you’re a developer who can be of some help to Jordan, or a designer, or interested in this field, then if you reach out to Jordan with a sincere story, maybe you could get a reply? VIZCOM can’t become a tourist destination. The reason we’re doing this first shoot with Jordan is also
Chester Roh because Jordan really has a great affection for Korea. So Jordan talked about why Korean entrepreneurs are so smart and solid, and engineers are good too, but why they don’t come out into a bigger market, and Jordan also said Jordan would help a lot. Actually, this is already our second time meeting. Yes, so as I listen to this now,
Seungjoon Choi you’re delivering a kind of message and making a provocation. It seems like you’re giving people the message to challenge themselves by going to SF like this, and when they go without knowing what to do, of course they’ll look for Jordan too, but in general terms, how to survive in SF was what I was curious about from a listener’s perspective.
Minseok Kim But I actually want to ask you back, Seungjoon.
Seungjoon Choi Yeah, no answer. Okay.
Minseok Kim Right. Because you shouldn’t try to find the answer, and that answer is something you have to make yourself, so I want to ask back about that. I don’t think it’s a good question. Right.
Seungjoon Choi Okay.
Okay.
Chester Roh do something, we are actually looking for some textbook. Ah, right, right, right. What’s the best practice? So let’s just follow it. So I think that Korean culture is like half in the mix of Japanese and half American. Yeah, I would say that we are definitely different from that in China. Yeah, but we are half Japanese and half American, I guess. Ah, right, right.
Jordan How should I put it? Like just the willingness to try new things. I think, yeah, it’s a little bit there for sure. I think. Yeah, yeah.
Chester Roh So as Minseok said, as Jordan said, if somebody is really, really eager to do something, you know, they can find a way, right? Right.
The Optimism of Bay Area Startup Culture 33:49
Minseok Kim What I felt on the way here is that not only Jordan, but there seems to be a lot of optimism. In Korea, if you say you’re going to do this, people say, what if it fails, why are you doing it this way, it probably won’t work, that’s the kind of thing people say, but here, how can we grow it even more? It’s okay if it fails. We can just try again, let’s do it until it works. There’s much more of this mindset, of growing the pie and sharing it. I think there are a lot of people with this kind of mindset, with a lot of optimism. I’d like to ask about size and dreams.
Seungjoon Choi I think Chester mentioned this last time, that here, people grow their dreams bigger. That’s right. So actually, you said at the beginning of our conversation, your company
The $50 Million Investment and the Process of Raising the First Round 34:27
Chester Roh was funded with 50 million US dollars, which is about $57 million. About $57 million is a huge amount of money in Korea, and I think that normally in the startup community in the Bay, as we are saying it, first is pre-seed, and then seed, and then Series A and Series B. And in Korea, we do have a similar structure and stages, but the numbers are totally different. Yeah, normally here, you guys have one more zero. Yeah, let’s just begin with that type of conversation. So actually, you started the company in 2020, but maybe five years already passed, a lot
of painful things, and definitely you didn’t secure the funds automatically. You needed a lot of effort on that. Yeah, yeah. So can you just tell us about from the beginning of your company and your first fund? You were funded by Nat Friedman.
Jordan Oh yeah, he was part of it, yes.
Chester Roh Exactly. He’s a kind of figure in the market. Yeah.
Seungjoon Choi He’s the person who made Copilot.
Chester Roh You were funded by Nat Friedman. Oh yeah, he was part of it, yes. Exactly. He’s a kind of figure in the market. Yeah.
Seungjoon Choi He’s the person who made Copilot.
Copilot. He made GitHub Copilot, and through something called AI Grant, he discovered many AI startups as well, and now he’s at Meta Superintelligence Lab
Chester Roh Yeah, yeah, MSL.
Jordan He’s based there now.
Chester Roh Highly, highly famous figure. Right. Could you please tell me about the story, how you came up with
your first funding?
Jordan Yeah. So it was interesting because of that Reddit post. We started growing community on Discord, and two people. So Basis Set Ventures, they were one investor, and then AI Grant, Nat Friedman, they were the other ones. So Basis Set actually led our very very, very first paycheck. 500 thousand. That was the first one. That was Basis Set. But then right after, Nat came with this thing called AI Grant, and I was like, huh, what is this? First, I thought it was a scam. I was like, AI Grant? AI Grant? Yeah, it sounds fake. It sounds fake. I was like, oh, okay, this is real. Okay, cool. So we got into that. But my first, so Chang from Basis Set,
that gave us the original 500,000, she sent me an email saying, hey, I saw your Discord group is growing. I’d love to meet you for coffee. Got it? So I brought my laptop, and I was like, hey, this is what VIZCOM does. We can render your drawings. And she was like, oh, this is great. And the next day, she was like, oh, I can start you off with this much. And I was like, oh, that’s cool. And then, yeah, the rest is history there. Yeah. You know, AI Grant is very unique.
The AI Grant Experience and Building a Startup Without Fear of Pivoting 37:24
Minseok Kim It feels like a YC for AI companies, and specifically for the first batch of AI Grant, there are a lot of great companies including VIZCOM What are some of your experiences from AI Grant?
Jordan Yeah, that was really good. It was such a great experience, especially the first batch. It just felt very genuine, like everyone was just trying to figure out their thing. Everyone was early in their idea, so just having that common ground of, “Oh, wow, we’re all so early, we’re all trying to figure this out.” That was amazing. It felt like you were truly at the frontier of something big happening, so that was really cool. So I think that environment was very
special, and it was really fun, really, really fun. Yeah, it was literally in this basement,
a random basement in San Francisco. Yeah, it was cool. What are the things you learned from other founders
Minseok Kim or the program? What was the most helpful for that stage?
Chester Roh Was it really helpful to secure the next round of funding, or… Yeah, I would say the one thing I learned
Jordan was not to be afraid to pivot your ideas. It’s okay to change your mind. We were with Michael and
Arvid, who are now at Cursor, but they were doing a CAD AI tool during the AI Grant batch, and then by the time they were done, they were like, “Oh, we’re gonna do a coding tool.” And I was like, “Oh, wow, that’s such a big change. Such a big change.” And then the rest is history, right? But that was just that moment I was like, “Oh, wow, looking back, what if they never changed their mind and they were so stubborn, like, ‘No, I have to make CAD’?” They were like, “No, it’s okay. We can change ideas if nothing is working.” That reminds me of Runway.
Seungjoon Choi Runway once started as a tool, but they changed it to develop
models. Yeah, yeah. Exactly, exactly. So since Seungjoon mentioned the model,
In-House Model Development and Its Potential Use for Korean Manufacturing Companies 39:31
Chester Roh are you planning to create your own model?
Jordan Yeah, totally. So we’re currently working on this. We’re trying to understand what it is we want to solve with it, what it actually means for us. Not just simply training another image model that’s a little bit better, but something foundationally new. It’s something not just in one modality, even. So we’re thinking about this right now. I can’t tell too much about it just yet, but this is why we’re hiring. Yeah, we’re hiring for exactly these kinds of things. Exactly, exactly. A bunch of AI prosumers, including designers and decision
Minseok Kim makers working in large companies in Korea. Would you just give a glimpse of your blueprint for your products you are working on, and then how you can actually help Korean manufacturing companies? Sure, sure.
Jordan We work with many enterprises, for example, Hyundai and Namyang, companies like this that are just building physical products. We help their designers get a lot more out of their time. So now you’re no longer having to, I would say, spend as much time wasting on prototyping. You can just do those prototypes within a few days versus months. We had this the other day: the CEO of GM talking about VIZCOM. She mentioned from weeks to hours, which is a very massive time shrink. So yeah, I would say, one, they have to have an internal
initiative themselves of, okay, what is our AI strategy as an organization, which a lot of companies are doing. So that’s kind of one way. And yeah, those are the kinds of companies I would say that have really found the most success with this kind: people that are looking to accelerate their designers or enable other folks to be part of the conversation of the day.
University Communities and Design Students’ View of AI 41:25
Minseok Kim You talked a lot about community. I heard that you are focusing on the user side and helping them to use your product well, which is obviously product-led growth and bottom-up strategy. How do industrial designers or students in Korean universities approach and use this?
Jordan Yeah, the university. So we give VIZCOM for free to all schools. And yeah, schools like ArtCenter, CCS, Hongik University, places like this. The reason we do that is because when you’re a student, your workflow is extremely malleable, meaning you’re trying a lot of things. So then once you get an internship or a job, that’s when you bring your tools with you, and then that’s when VIZCOM gets spread into an enterprise. So students are using this in a way where, of course, they’re cautious because you don’t want
to lose the fundamentals of pre-AI. I think our generation is very special, where we learned things pre-AI, so we kind of know certain things that the new generation just will never have to learn. There’s no point, which I think could be good and bad. So there’s this kind of interesting in-between right now where some students are very accepting of these things, but some are very against AI, which is just I think AI has had a really bad PR outside of Silicon Valley and tech, which people don’t realize, I think. Right. A lot of Western students are very against AI, actually. Yeah, yeah, yeah, actually.
Chester Roh I saw one tweet yesterday. It was at the Stanford graduation ceremony, people were walking away.
Jordan People trying to talk, yeah, these kinds of things are very common. Right, I think so. So, let’s shift gears a little bit
The Valuation Frenzy Around Bay Area AI Startups 43:16
Chester Roh toward the landscape of the Silicon Valley AI entrepreneur scene, because you know that SpaceX got an IPO yesterday. Oh, wow. And Anthropic, OpenAI, their valuations are insane. And Matt and I met a lot of entrepreneurs here, and they are asking numbers, the valuations and the amount of money they are raising are literally really, really high. Yeah, I think that kind of fever is going around in the Bay Area.
So I’d like to ask about the general landscape, the Bay Area entrepreneur scene. Do you think it is, you know, how do I find the right expression? But do you think it is good enough? What do you think about this in general?
Jordan Yeah, the feeling of this whole situation? Yeah. I think it’s like,
Chester Roh Do you think all the valuations are justifiable?
Jordan Oh, yeah, I think some of it definitely is, especially Anthropic. These guys have created intelligence as a service. It’s literally like electricity, but for your mind. It’s going to be really weird to not have it. When Fable 5 went away, I was so sad. I was like, wow, I could feel it, like, wow, they took away such a they took away my dishwasher or something I really need. sense, I think it’s extremely valuable, because if we can automate the economy, that is a huge, insane unlock. An example of this is Qatar, the country. 95
percent of their workforce is done by migrant workers, and then 6 percent is just regular citizens. But most of the people in Qatar, they retire by the age of 50 because they found that they can just have a lot of migrant workers provide for their pension, so they can retire. So I think, once you have agents doing a lot of the economy, a very similar thing will happen here. And it’s like, that ceiling is unlimited. It’s beyond
what the current valuation is. And this whole nature of people talking now about post- labor economics, this world where working becomes almost optional, I think that really justifies the valuation. And you start thinking about that future, like 2030, 2035, it’s like, okay, wow, yeah, almost undervalued, maybe, I would say. So I think people here really believe in this kind of future, that they’re the ones building it. Yeah, it’s a little bit scary, of course, but that’s just the nature of really big change. Yeah, yeah, yeah, so I really do believe it. Yeah, for the most part.
Chester Roh Yeah, I partly agree with your perspective, but as an old man, many, many people in my generation, you know, say, okay, that company’s revenue is not that high.
Jordan Yeah, oh, there’s definitely, yeah, okay, there’s definitely a huge category of startups that are what we call this kind of English slang, LARPing, meaning, oh, I’m pretending, more so just doing it because everyone’s doing it. They’re not actually interested in the problem, and there’s a lot of these kinds of companies, of course.
But those ones, of course, I think they come and go. I think every wave has its version of this, like mobile internet. Yeah, yeah, yeah. I think that there are going to be a lot
Chester Roh of some type of sifting-through phase. So the good companies will emerge, and the bad companies will go away.
Jordan A lot of noise, trying to find signal in noise. Can you give us some data points about the stage of the company,
Silicon Valley Investment Cases and Sustainability 47:10
Chester Roh pre-seed, seed, Series A, and Series B? So definitely, we are actually seeing some standard: if you get to this amount of revenue, you would be valued at this valuation, and you can raise this type of money. So can you give us some ballpark of the valuation scheme of Silicon Valley startups?
Jordan Yeah, so depending on the, it really depends. So do you mean just in general? Yeah, yeah, yeah. So in general, oh man, these days, it’s so random. It’s so hard to know. Some people can just, just to give us some cases. So, okay, okay. So someone, I’ve seen some companies raise, oh, recently,
oh my gosh, this is a crazy one. Someone raised nine million dollars. at $500 million post valuation. $9 million, that was, I think, the Mechanize guys, called Mechanize. Yeah, they raised $9 million at $500 million post. which is very good, like, it’s very low dilution and a really high valuation. So things like that are happening right now, for sure. Let’s see, I’ve seen some just purely idea, like $4 billion valuation, but it’s just the idea. But I think because of their background they can do these kinds of things.
But I don’t know. I don’t think these valuations and these things are really good indicators of if something will be good or bad. It truly comes down to the founder and the problem. But yeah, it just shows it’s a reflection of how crazy the current environment is. Right, yeah, right. Yeah, it feels like a lot of
Minseok Kim nowadays, not downs, ups and downs. Definitely nowadays, ups and downs. So how long do you think it is, this type of high stage,
Chester Roh sustainable? Maybe two or three years?
Jordan I think even longer. I think it’s gonna go longer than we expect. Yeah, and if anything, maybe this is just the new norm. I don’t know. I’m really bad at predicting the future.
Chester Roh Okay, okay. And you have met multiple industrial designers in different
The Role of Designers in the AI Era and the Importance of Taste 49:24
Minseok Kim industries, and you have seen how their role has been changing. What do you define as the future role and future role of designers, and how their job description is gonna be changed? Hmm, yeah, because of AI, right?
Jordan I think it’s very interesting because I think it really amplifies the importance of the human touch, right? Because I think with AI, it kind of gets rid of a lot of the tedious tasks that you’re repetitively doing. So the designer is now more so focusing on things like, why does this look good? Or how does this look good? It’s not as important. Really having a good sense of taste is extremely important. Now, taste can mean many different things, but really being able to spot the signal and the noise, and why the audience you’re trying to design for would find that valuable. So these more judgment-oriented tools, like skill sets, are becoming more and more important. And you might see someone
whose job on a design team is to curate mood boards, because now data curation in that sense becomes really important, because the models can learn from that. But knowing how to curate those images is becoming the skill. So it’s like these new skills that are kind of forming, and that’s also really interesting. So yeah, that’s kind of one way to think. Your product is going to help the designers to nurture their taste and automate many tedious
Minseok Kim works and focus on impact. Yes.
Jordan Exactly, exactly, exactly. Yeah. Maybe my question is a relevant question.
The Essence of the Creative Process That Remains Despite Automation 51:04
Seungjoon Choi So, well, I’ll ask the question in Korean first. Would that be okay? Okay. As I was looking through the materials, I felt there was a certain tension in Jordan’s work.
What I felt had tension was that in the iterative process, the creative process, there was a feeling that Jordan really believes in iteration, but the actual product compresses that. But I felt there was a bit of tension between those two, and I wanted to ask about that. So, given Jordan’s original background,
Jordan physically understands that this is extremely important in industrial design, but the situation with AI keeps delegating and automating things. But just now, Minseok talked about tedious jobs or dirty work, and delegating those, but in reality it may not be only that. In the current state of delegation, is there something Jordan wants to protect, or should the flow be changed completely, or should Jordan follow the tide? I wanted to ask whether Jordan might have those kinds of concerns. So let me just paraphrase his question. So his question is that, you know, the
Chester Roh designing thing is really, really laborious and it requires a lot of time, but what you are providing to the designers is actually compressing those processes to make it really, really quick. So it’s kind of in a contradiction in a creative process, yeah, in an iterative way. Yeah. But AI is really,
really doing the things in a compressed way. So his question is mainly about how you manage those two contradictions. Well, actually, my question is about, I think you are a true believer that
Seungjoon Choi creative process is iterative, but delegating that work seems like omitting something in that process.
Jordan That’s right. Yeah. So yeah, yeah, I know what you’re saying. Okay. So it’s like, we’re not trying to
iterate, we’re not trying to offload, I would say, the creative process. Being actually creative is still up to the person. It comes down to the drawing, how they’re communicating their thoughts. There’s just different avenues in which they’re able to express creativity. Creativity in the older definition maybe had certain practices attached to it, but I think now there are new practices that are being attached to the word creativity. So creativity no longer just means painting or writing. It means more things now because of the technology. So I think there’s just kind of more definitions being added to the word creativity now, and they’re being expressed through these kinds of tools like VIZCOM. Yeah, it’s
evolving, just how language evolves over time, like Korean and English. These change over time, even. It’s the same thing kind of happening with design. It actually reminded me of the talk we had yesterday. Even though we are not coding at all, but
Chester Roh engineering is still required. Yeah, even though we are coding in English, the engineering still remains. Yeah, yeah. Exactly, yeah. Exactly, yeah. Okay, okay. Thank you for your time, Jordan. And we’re gonna release you from this. Yes, thank you so much.
Jordan Thank you, of course, for hosting us. Yes, and Seungjoon, it was really nice to meet you. Maybe when I go to Korea, we will…
Seungjoon Choi Okay, okay. Sure. Then now we’ll let Jordan go,
Visiting ICML Seoul and Opportunities for Global AI Networking 54:50
Chester Roh and the three of us, with Minseok, will wrap things up. So originally, we were planning to do quite a number of these recordings, but of course the frontier researchers can’t talk, and people like venture capitalists or entrepreneurs too. In early July, there’s a conference called ICML.
Seungjoon Choi Yes, that’s right. So I think a lot of really great people will be visiting Korea. I think they’ll be there from around July 5th to about the 12th, and during that period, a great many parties will be held. Many companies, even companies coming from the U.S., are renting out numerous cafes and restaurants near COEX and holding small parties,
It’s a timing that would be a shame to miss. That’s right. AI engineers
Chester Roh and AI researchers in Korea, or people who have already started companies, and even if they have received a lot of funding in Korea, the valuation scheme in this ecosystem is so different, and companies that are especially interested in Korea seem to be very interested in chip companies, and software infrastructure related to software orchestration, things like GPU orchestration, VLMs, and SLMs, those kinds of things too. But relatively speaking, they don’t seem very interested in model creation. That’s how it seems. Yes, yes, yes.
Minseok Kim They see it as a solved problem. That’s right.
Chester Roh Of course, now that Fable 5 can no longer be used, it has become a national issue for everyone, so we are also talking a lot about sovereign models, but,
Seungjoon Choi When you say it’s a solved problem, do you mean it’s hard to catch up now?
Chester Roh That’s right. I think they see that as a done deal. The recipe has been made public,
Minseok Kim and in terms of the compute, data, and talent needed to follow that recipe, they seem to think there are only a few players who can enter that game board.
The Potential for AI Agent Services to Emerge in Korea 56:39
Minseok Kim who can enter that game board.
Chester Roh And another very large sector that remains is AI agent service. Okay, okay.
In other words, I think we’re still before the emergence of a complete consumer service. People think ChatGPT or Claude has finished everything, but I don’t think they see it that way at all. I think they see it as only just beginning, and in that sense, what I tell the local people here when I’m selling Korea is that, if that is born, the place where the original form of the consumer business is born will be Korea. I’ve been saying that a lot, so I think we can set up a lot of meetings with those kinds of people this time too.
Minseok Kim They do seem to be observing China and Korea a lot in practice. People who are somewhat sensitive in that direction.
Chester Roh So we are going to create an email address for AI Frontier, and if you contact us there, we have had a lot of meetings with many venture capital firms and companies that are coming this time. So we have good connections, and it would be good if we could connect people and see what kinds of things end up happening. Very interesting.
Seungjoon Choi Then Chester, have a safe trip back, we’ll see you in Korea, and Minseok, I hope you’ll keep pushing forward.
Chester Roh All right, then we will wrap up around here.
Thank you. Thank you. Goodbye, then.