From Art to AI: Yohei Nakajima's Unique Path

Today on The Next Next, our guest is Yohei Nakajima. Yohei is a venture capitalist, an innovator, and an artist, best known as the general partner at Untapped Capital and the creator of BabyAGI, an influential open-source AI project. Yohei also recently announced the Agent Fund, which is an angelless rolling fund dedicated to investing $50K to $200K checks across the AI agent landscape. Now, Yohei is an interesting guy. He's a hard person to put in a box. He's one part venture capitalist, one part hacker building BabyAGI, the first popular open-source autonomous agent with task planning capabilities, part artist. He's an artist and creator of Pixel Beasts, which is a collection of 10,000 unique NFT collectibles. And he was responsible for the artwork, generative code, and even designed Associated and even designed associated games and apparel. And meanwhile, there's a bunch of things he doesn't want to do in his day job, and he looks to automate anything he can that he doesn't want to do using agents. So just a very interesting combination. I don't know that I've met anyone like Yohei, and we have a fascinating discussion in this episode. I hope you enjoy it. But before we get started, I'm Jason Jacobs, and this is The Next Next. It's not really a show, it's more of a learning journey to explore how founders can build ambitious companies while being present for family and not compromising flexibility and control, and also how emerging AI tools can assist with that. Each week we bring on guests who are at the tip of the spear on redefining how ambitious companies get built, And selfishly, the goal is for this to help me better understand how to do that myself while bringing all of you along for the ride. Not sure where this is going to go, but it's going to be fun. Okay, Yohei, welcome to the show.

Thanks, Jason.

I am excited for this one. We— I feel like we, we have some Similarities just in terms of the public learning and also the, I think, portfolio of interests. And I haven't met before, but, but dare I say quirkiness. I'm certainly quirky. It's meant as a compliment.

Which is why we run in the same circles.

Yeah. But, but also just quite different approaches in terms of the how. It seems like you Are more of a public learner with code and building, and I am more of a public learner with conversations like this. But we're learning about a lot of the same stuff, and I am not quite sure what we're going to cover today, but I'm pretty sure it's going to be interesting.

Well, I think one of the things you emailed me that I think is like, we can never talk about is like, how is this GenAI revolution going to change the way companies start? And I think that's a conversation I've had a lot. And I think that's a fun place to start. I don't even know where to start. That's a broad topic.

Yeah. So why don't we do that? I think the one thing I'll ask before we do is just maybe just describe yourself and your portfolio of interest to the audience just to frame the discussion. And then let's jump into that topic.

I'm a, I run Untapped Capital, an early-stage generalist venture firm. Within early stage, I think I'm one of the earlier side. I consider myself on the earlier side of pre-seed, right? I don't even do what I would consider late pre-seed. The way in which I invest is very top-down. I'm not, I don't like to be reactive to the companies that come to me more so. I start with picking an industry I want to dig in on, identifying a niche I want to deeply study, trying to build conviction either through building or conversations, and then go look for companies that would make a lot of money if my conviction is correct. And then as part of that top-down learning journey, the one I'm best known, the approach I'm best known for is the build-in-public learning approach, which is where I essentially use AI to mock up prototypes as a way to learn about the technology and meet founders. As far as portfolio of interest, I tell people that I'm a generalist that at any given moment looks like a specialist. And sometimes I do deep dives that are a week long, a month long, 3 months long, sometimes a year long, and the deep dives overlap. But at any given moment, there's a few areas I'm particularly interested in, but I like to think that as I go deep on different areas, they stack. together over time and then create this like beautiful graph that's like very wide.

And I know we said we'd then jump into the topic, but I guess one other piece of context that would be useful to slip in there is just, as you discussed, you code, you've been spending a lot of time with AI, you invest. In what order did these things come about? And, and how did they begin intersecting?

I've always built as a hobby. I've never been an engineer by trade, but my dad was an engineer, so he bought me HTML, CSS books when I was in middle school. I've always coded on the side, but always as side projects, whether it was a personal blog, a tech event site for the LA ecosystem, internal tools when I was working at Techstars. So building as a way to make myself efficient has always been there. So that would— that came first. I started investing when I joined Techstars, so that came second. But the build in public as a way to tie the two together came by accident while I was building Untapped. Before Untapped, I was usually working for other organizations, so the cool stuff I built was technically trade secrets. So there was a lot of experiments I ran, for example, at Techstars or Scrum Ventures that I never published because I was using internal data and it wasn't me making the decision on what to share. But as soon as I started running my own fund, There was no one stopping me from just sharing everything I did. And I just started doing it, not really thinking of an objective, but then over time realized that it leads to deal flow.

Now, did you start running your own fund right when you left Techstars?

No, I went from Techstars to Scrum Ventures, which is a Bay Area-based fund with close ties to Japan. So I did that from like 2014 to 20— 2016 to 2020.

And then how did your own firm come about?

I always knew I wanted to eventually start my own firm when I joined Scrum. And during my time there, I thought about different theses and I always thought I would start a firm when I found somebody to start a partner with. But there was an idea that I just couldn't get out of my head to the point where I had to start talking to people about it. And then as things happen, you start talking to people, you get more excited, you get this validation. And at some point it was, I just had to do it, which is probably the best reason to start something.

What was the idea?

So at that point I was really, and I, and it's still true. I think I'm one of the more experienced people when it comes to doing outbound sourcing at the pre-seed stage. It's, if you look at pre-seed investor, primarily sourced through referrals, and there's for better or for worse, a lot of allocation chasing. At Techstars, I ran top of funnel outbound across all of the Techstars programs. So I built a lot of know-how on doing outbound. And so the thesis was really leaning in on that strength, but the thesis was not just about outbound. Outbound was the how. But the reason for doing that was this fundamental idea that there are strong founders outside of typical networks who just don't get a chance. And if you are really honed in on outbound, my assumption was I can find and invest in those founders and that can produce better returns for LPs.

And as you set out to build that firm, when and how did AI come onto the horizon for you? Because I know you've been talking about it and building with it since Before it was cool. And where did that come from?

August 2022 is when I started playing with OpenAI's API. It was about 6 months before ChatGPT came out. But prior to that, I was pretty heavily automating a lot of stuff with no-code. Very heavily built our whole fun backend on Airtable. I think my highest month I was running like 15,000 Zap tasks. And this was pre-AI. When I got into AI, I learned about it through the Web3 community. I played around in NFTs. So when Midjourney, Stable Diffusion, and DALL-E came out that summer, A lot of the Web3 community was talking about it. So that's how I got, I started playing around with the image generation first. But then once I had my DALL-E account, I was like, oh, DALL-E has a text thing, which was the, which was DaVinci 002 GPT-3. And so that's when I started playing with it. And the first thing I did was ask it a couple questions that founders would ask me, and it did an incredible job answering better than I did. And that's when I thought, wait, I should try to see how much I can use this to make my job easier.

And so was the initial application of AI more to look at what you were doing internally with Untappd?

Yeah, the first thing I built was a wrapper called Mini Yohei. It was like, you're a venture capital specialist, assuming the following question comes from a startup founder, answer as an expert venture capitalist. And at that point, like ChatGPT wasn't out yet, so the value there was pretty incredible. And so for a short amount of time, for about 6 months, I did have portfolio companies using my wrappers sitting in our founder portal, which was fascinating. I had blog post drafting, SEO keyword generation, and it wasn't all the founders using it, but some of them were. But again, it died out after ChatGPT came out. But for that short period of time, what was fascinating was I could see what kind of blog posts my portfolio founders were thinking about publishing. And to me as an investor, I was like, that's incredible insight that I'd never had. And I did go down this rabbit hole of what other data could I get access to if I built tools? that are helpful to founders.

And what were your tool building platforms of choice at the time?

At the time, I was still heavy on no-code. So I was running it largely through Zapier and just doing API calls to OpenAI through Zapier until I realized that it would be much easier if there was an OpenAI integration built into Zapier. Fun fact is I built that OpenAI integration And was supporting it as an unofficial Zapier integration for a month until Zapier said, we'll take over. And I just handed it to them.

Cool. And so during that time, had it— it sounds like you were starting to use AI for some internal experiments and starting to pick off different processes and automating them. Was it making its way into the companies that you were backing as well, or? When did that start?

That started around the same time. I'd say like pre-ChatGPT, it was a, it was definitely a much smaller circle of people who were talking about building on top of it. And so I felt like, especially in that first 6 months before ChatGPT came out, I feel like just tweeting about building on top of ChatGPT, I met tons of founders and some of them I had a chance to invest in.

Got it. And I guess now taking a step back, how do you— so if you look at the landscape today, everyone's just adding the word AI to whatever it is they were doing, whether they set out to work with AI or not. There's so much capital that's getting dumped into the category. How do you think about what's happening in the short term and how do you think about What's happening in the long term and how do you reconcile the two as well?

It's moving fast. The best thing I can do, especially at the earliest stage pre-seed, I believe that leaning in on gut instinct is important, right? We just don't have the data to make quantitative judgments about which companies are going to be successful. So the way I think about it is how can I feed my gut instinct the right information? And I think it has to be a diverse set of information. I want to learn from builders about their challenges in building. I want to learn from corporate executives about the challenges in incorporating AI. I want to talk to day-to-day employees about their challenges in incorporating AI, and especially given company culture. And then also with consumers about what their experience is like. Also playing with my kids with AI. If I touch all of these on a regular basis, my assumption is that I will have a good gut, a better gut instinct around both the short-term and long-term future. around different tools that come my way. But I'm not going to sit here and say I have a clear vision of the future. More so, what I want to do is look, be able to look at a company, talk to a founder, and have a good, feel confident in my gut instinct about whether or not this is a company I want to work with.

And where are you in that learning process so far?

It's ongoing. I think it's going to be continuously ongoing. As VCs, the tech landscape changes all the time. And what's important for us is to, is to learn how to learn and to learn how to adapt to constantly changing information. I don't know if that answers the question.

Here's how I look at it. I graduated college in 1998 and fell into the Early innings of the dot-com boom and then bust. And then there, of course, was mobile and there was the cloud. And now there's AI. The ground is always changing under our feet, but it does feel like there are certain key moments where the change happens in a more condensed and profound way. I see, yes. So how do you think about the change that we're in the early innings of? Going through relative to some of the other changes that we've experienced in our professional careers before?

Things are definitely moving fast, faster. No, that's a good question. Maybe I'll try to answer the first question because it does seem easy. In the short term, right? I think, I mean, moving from chatbots to agentic builds, right? Really happened last year, right? Last year was really kind of the boom of people starting to build and implement agents. I'd say last year tooling was still pretty nascent and we saw a lot of improvement there with Arcade and Composeo were some tools that made tooling easier. MCP blew up in the last 2 months. So that's a tooling piece I feel like started getting better. I think this year in the short term, there's a lot of improvement to be made on memory or memory context management, whatever you want to call it, especially when you get to the enterprise level. How do you handle permission of information? with an AI as if I'm asked a question and the answer's in a Slack channel I don't have access to, how do you make sure I don't get that answer? So that's mix of context, memory, auth permission. So that's a field I think will expand a lot in the next year. I think in the long run, one of the things we're seeing is more and more companies that are not just AI native, but more agent native. Like I'm being pitched companies that start companies, right? We're gonna use AI to build a whole bunch of mobile apps. We're gonna use AI to launch a whole bunch of e-commerce brands. I think those will be fascinating experiments and we'll start seeing a lot more of those. But yeah, what happens in the long run? It's, that's hard to tell. I guess this is the short answer. I can imagine a lot of different futures and I feel like I'm just constantly trying to, trying to balance the probability of those futures happening in my head.

I guess one thing that I've been spending time trying to unpack is there is the technological readiness and evolution, and then there is the societal readiness and evolution. And it seems, if you look at history, that the technological is typically ready much earlier Then society adapts. And so the technology being ready happens faster, but then the societal readiness happens slower. And what are you seeing as it relates to AI? And I ask because there's this contrast of it's gonna— these big bold proclamations from the people closest to the technology, but then you look at society and a lot of society is still like, who's that woman that's high up in government that called it like A1? Yeah. And so like, that's where society is. Like, that's a High-up government official, right? And doesn't even know that it's AI.

To some extent, that's always been true, right? That is the typical hype cycle is when a new technology is introduced, we talk about— the people who are working on it talk about what it can do. And then as people catch on to that, that turns into the story of what this technology can do. And then once it hits the mass media, like all that nuance around, oh, we're not ready yet, disappears. And so everyone gets excited about the technology, but then realizes it's not ready as they try it. And then you have this disillusion period. Also in implementation, there's a 3 steps forward, 2 steps back piece. Like in AI, for example, there are going to be major issues. Companies are going to make mistakes that cost a lot of money. People are going to make mistakes that are going to cost them a lot of money. And those will hit the news. And when that happens, people will take a pause and slow down a little bit. It's going to be a back and forth. If you think about internet, mobile, there's still plenty of people in the world without internet, plenty of people in the world without mobile. That's changing, but it's been, it's been a much longer time than probably people who were early into mobile would've expected. So I think to that extent, I think you're right that we'll probably see something similar in AI, which is actually today, especially with the world being significantly more interconnected, especially with Twitter and so on. I do think the, the average person is exposed to what's possible much faster. And without context, so you feel that like the gap between what's possible and what we're doing today feels bigger than it was before. Is my gut reaction to what you just said.

Unrelated but related, I feel like the people that that are constantly like this this is the worst time period to live through and all that. It's yeah, there's a lot of bad things happening, but it's not like there weren't bad things happening before. Like one difference is that information. flows so much more freely now than it did before that you're just aware of a lot more of the bad things that were already happening. And so you think that it's worse when it's not worse, and maybe it's even better, but you just— it's like you have an IV of all the bad things as they happen, which you didn't used to have.

And for any technology, there's going to be good and bad, right? If you ask people, if you just blanket ask, is social media good or bad for society, a lot of people would start by saying bad because that's what they're close to and that's what their day-to-day. But then if you go back to some of the geopolitical instances that were really enabled by social media, there was clearly good that gave people voices that they didn't have. So again, social media has done both good and bad. And similarly, I think AI will have that impact as well.

That's another thing I've been sorting through is on the one hand, it's, oh, if it can offload stuff that I don't want to do, that's great. Oh, if it can help companies save money and have lower cost structures and raise less capital like that as a founder, like, that's great. But then it's like, what happens to People who are doing these jobs today, and it's oh, they'll be free to go and because they won't need to work. It's or they won't have an income, and they won't be able to provide for their families, and they won't. And so, well, what determines what happens? Well, like government, and then it's government's the one that's still calling it A one and not AI. And so, and and if money makes government work right, then who are the ones with the money? It's the big powerful corporations. And who are the ones with the power? The big powerful corporations. And Is it in their interest to provide for all the people whose jobs are? It's like they don't give a shit. So I don't know. I get the whole like it's gonna cure diseases and democratize information and automate all these things so we can build big stuff. Like all that's maybe true, but like how are we not? How are the rails gonna stay on for society? Right? Those are the kinds of things that that I worry about. And then as the cup as a venture capitalist investing it, that's not my lane, bro. That's for other smart people to figure out. It's like. All right. Well, who's thinking about it?

Technically, I'd say that today VCs are much more involved in politics than they used to be.

Certainly in the, in this calendar year.

Yes. In this calendar year. Some of the stuff you talked about, right? Technology will advance and we will either do it well or not do it well. But a lot of what you're describing feels less like a technological problem and more like just like a cultural problem.

But that's the thing. It's what will be the impact of technology that you can't look at in a vacuum. You have to look like you have to look at as the technology landscape changes so quickly, will the how will the cultural landscape adapt? How will government adapt? How will policy adapt? Like how will society adapt? And you can't answer the technological implications without answering the societal ones. And I think what's scary now is that society just seems really polarized and erratic, right? And so you can't really predict how society is going to react. Because you can't predict what society you're dealing with, right? So then it's, well, then the technology is the least of our worries. I don't know.

If we use language models correctly, I feel like they are very well suited to helping 2 opposing sides communicate better.

You've talked about that, right? But they're also just as capable of, for example, distorting reality and providing what seemingly becomes a pervasive view when really it's just the algorithms Manipulating and controlling information.

And it goes back to the human design. It's what's the intention of the design? What's the intention behind these technologies we're building? This sounds super crunchy, but like, I think if you come at it with good intentions and build technology, if the world does that, the technology will do well. But if we cut— if the world in general, on average, is trying to figure out how to use technology to capture as much value for oneself, then that's the technology that we're going to end up building.

So then doesn't it matter who is building the AI that we're all utilizing and relying on and how they're building it and what framework they're using to make those decisions about how it behaves and responds to information?

I think so.

Yeah, but aren't we all flying blind to that? I, I don't know if we are. Maybe some people know, but I enter stuff in ChatGPT or Claude or whatever, and then it gives me an answer back. But it's like my mother-in-law's always asking me like, who's building the algorithms and what, what determines if like what their value system is and what value system the algorithm is programmed to, to have. And it's, I don't know.

One of the reasons that OpenAI doesn't have all its prior employees is because some of these questions are coming up from my understanding. I'm not close enough to know the actual details. But I do think that's part of the reason is that some people felt like this should be built ground up from, with a different philosophy, which is why you see people like Ilya building super safe intelligence.

And if I guess bringing this back around to the, well, if you look at what's happening with startups, so unless you call this podcast something that I'm building, I'm not really building anything right now. And on the one hand, it feels, how could you anchor anywhere given how quickly things are moving? Let me take a step back and not have a horse in the race and just try to immerse myself and understand where the world is going. So that's one lens, but the other is what better way to learn than to build? And until you get your hands dirty, you're not learning actually anything, right? So I kind of, I feel that tension today. And like, I wouldn't want to be starting from scratch, trying to figure out how to keep up with this tidal wave, but I also wouldn't want to be an incumbent with legacy infrastructure and people and processes and culture. And so it's, I wouldn't want to be anybody. It's just a difficult time to exist, trying to figure out like, how this plays out. Nobody's safe in a way, which is exciting, but it's also like, it's unsettling.

Yeah.

Yeah.

When people disagree on the future is when you can make the most amount of money as a VC if you are correct.

And do you, given that you've started this discussion saying how thesis-driven you are, would you come out on that? Did you? And you, but you also said that you don't have a clear view of How things play out from here. So as a thesis-driven investor, like, how are you operating in that environment?

So far, so much, I, again, I like to think of what are possible futures and what are the risks for each. I mean, from, as a pre-seed VC, there's a couple of different categories. One category I'm thinking about is like, how does the early financialization of startups change? On one hand, you could say that you could start a company for $500K and then take it all the way to Series A, right? Kind of skip the seed, right? Because you can do so much more with little money. If that becomes the prevalent way to start companies, I'm actually really well positioned as an early stage pre-seed. But the flip side, the opposite kind of version of that is you actually don't need any money to start. You can start with zero money because you can build so easily. And so that you can actually jump straight to the seed or jump straight to the Series A. And if that future is more prevalent, then as a pre-seed VC, I have an existential crisis. But that's one example of like futures I look at. And trying to balance the two. Realistically, it's going to probably be somewhere in between. We're going to see both, plenty of both, I think. But those, but that's one pattern I'm trying to observe and understand. And then when I see either of these patterns, I try to see if it can help feed my future understanding of which pattern is more likely to emerge in the future. When it comes to starting companies, for example, right? I think there's the, even just looking at agents, right? There's companies that sell agents, right? Use our agent. Or build agents for people. There's agent tooling, right? Like here's tools that agents need to make to work. And then there's the, we just use agents internally. We don't sell them, but we're going to build a big company with few people. Those are 3 very different ways to enter the agent market. In my approach, in this case, we actually launched an agent fund specifically to cut a whole bunch of small checks across the agent landscape because I felt like that was the best way to learn about. Which of these futures is the most likely?

But it's agent fund, meaning that it's actually like agents that are do— like doing the diligence and allocating the capital, right?

Nope.

Or no?

It's a semi-autonomous. We're moving toward that. It is a playground to do that. But the fund is more of a thesis-driven fund. It's an angelless rolling fund that we write $25K, $50K checks into a whole bunch of agented companies. And we've touched on all these buckets I just mentioned.

And is that a part of the Untapped?

It's part of the Untapped family. I'm running it with Alex and Adam from Agentopt in the Bay Area. So it's the 3 of us running Agent Fund as a separate fund from Untapped. And it's much less valuation sensitive. It's really more about building out that agent community, but also to learn, to have access to all these companies, like building agent tools, using agents internally. And by having relationships with many of these companies, my hope is that I will have a better understanding of the best way to monetize in the agentic future by just getting all this, having access to these companies and founders.

When you think about the state of AI today, when it does come to, let's take coding as an example, although this applies to any, to a lot of different sorts of expertise, do you see it more as an amplifier of existing expertise or as, or as an enabler of Expertise of some semblance of expertise in the untrained?

It's definitely both. I wouldn't even say, I think it's both. I think it just levels up everybody. And I think what you described, both of those are very impactful in different ways, right? It'll help top performers in any space perform much better, which is really going to push the space forward to some extent because, for example, AI researchers who really master agentic solutions are gonna be able to start having AI run AI experiments more for them, meaning their speed of iteration is gonna go faster, which means the evolution of the actual underlying AI that they're building goes faster. So that's the experts using AI to be even more powerful experts. But on the other hand, if you look at something like vibe coding, you see all these people who've never touched code or never touched gaming, for example, starting to play around with it. Do I think, for, let's pick vibe coding gaming as an example. Do I think these people are likely to build the next massive Nintendo game? Not necessarily, but on a whole, does it expose more people to the idea of building games? Are some of these people going to go into game building because of that? And will they, are they gonna have fresh new perspectives that existing game developers don't have? I think all of that's true. And I think there's a way in which an industry can grow because of a surge, like a mass surge, like a lot of people joining it, which is the lower level, all being able to vibe code. And then another way to push forward in industry is to get the experts to like move faster and push the top up. And I think both of those happen because of AI. That makes sense.

When it comes to building teams and companies, if you look at just venture-backed startups as we've come to know them, what are your predictions in terms of How that plays out and what changes there will be, if any, in the coming 2 years, 3 years, 5 years, 10 years than the previous similar time period looking backwards?

It's getting cheaper to start companies. You can run bigger and more massive companies with less people. And that's a trend that's continued over time, right? That's a continuous trend, right? We don't need to set up server farms, all that kind of stuff anymore. I'm assuming the power law dynamic will still exist, meaning that few companies will produce most of the value, especially if you look at the startup world. Yeah, I do think that still holds. I wouldn't say that's strongly held, but probably partially because I want it to hold too, but also brand trust, all those things. I do think the power law will still exist. It's hard to imagine a future where every company is small. That just doesn't make sense. Which makes me think there's probably a way to build a company in the future. If you can figure out a way to grow, grow a team in a way that makes sense in the world of AI, and I don't even know what that looks like. I can imagine you building a massive company, but I don't even— that's a very vague thought.

Do you— I'm bouncing around a bit, but do you think our kids will have an easier or a harder go of it than we've had in terms of, in the aggregate, quality of life, ability to have a nice home, provide for your family, get a good education, et cetera? Is that road going to get harder for everybody or will opportunities abound?

Before I answer that question, what do you think Is today easier or harder?

Than it was for our parents?

Yeah.

Obviously it depends on— Your generation. Your personal situation is different, but I think it's harder. I'm extremely fortunate and for a bunch of reasons, and not to say I don't have my own challenges, but I think we're in a very fortunate position. I thank my lucky stars every day, but I feel like in the aggregate, it like, More kids are graduating and living at home, and getting those entry jobs is harder to come by. There, I think home prices are through the roof. Inflation, like cost of living, groceries, utility bills, right? And and then there's a bunch of it's like there's no safe havens anymore, right? It's like you used to, like in our parents' generation, you'd go to a company and you'd be a lifer, right? And then that lifespan has gotten shorter, and people hop around, and nobody's safe, and layoffs, and quarter, and. Dollars above all else. But now it's even taken a step further with AI. It's, is my— not only is my job safe, but is my profession safe? Is the whole occupation going to be wiped out by machines? So I'm just thinking, if I look at my kids, for example, it's skate to where the puck is going, not where it is. But like, where's the puck going? And if I don't know for myself, and I'm 48 years old, I don't know how to predict like where it's going to be going for them. So yeah, so it's— so I think Instead of learn this skill or learn that skill, it's learn how to learn, be adaptable, be resilient, right? Control the things that you can control.

In that sense, like, yes, it's definitely going to be harder, especially if you think about choice as a difficult thing, right? If, and if you go back far enough, if you were born into a baker's home, you probably helped in the bakery growing up and you probably ran the bakery afterwards. And if that's the life path that you assume for yourself from the moment you're born, that everyone assumes on you, and that's the life path you grow. That sounds easy, right? In the sense that you don't have to make these like existential decisions on what should I do with my life? You just kind of assume that's what you're doing with your life. To that sense, that seems easier. The problem with that society was that there was no moving around, right? Is that if you didn't want to do that, it sucks. You still had to do that. And so today we're in a world of extreme choice. I can see every profession that exists. I can learn about it. Every profession is available to me, which is just too many choices. And I think in that sense, yes, it creates Challenges, right? It takes longer to pick a passion. You're floating around for longer. Yeah, that does sound difficult.

But it also—I think globalization made things more competitive for everybody, right? But then at the same time, it's start speaking over my pay grade here. But it seems like we're trying to put the genie back in the bottle here in this country. But it's like the genie's not going to go back in the bottle on the global stage. So by trying to put it back in the bottle in this country, we're just going to end up. Isolating ourselves and taking ourselves out of the global discussion. And yeah, I feel like we're in a globalized world whether we want to be or not. And so either we— and it's AI, it's— look, AI is coming. You can either try to resist and protest and— or you can like lean into the wave that's coming because you're not going to put the genie back in the bottle.

Yeah, I have a weird logical jump. So we just talked about how the career choice is possibly one of the challenges and difficult choices. Previously, we didn't have choices, but also, you like, also it wasn't really optimized, right? It was just like where you were born into. Step back a little bit. When you think about information, right? When the internet first occurred, everybody had access to it. Like people were just reading what was on there. And then over time, as everyone started joining the internet, as everyone started writing, the way in which we see information became a filter. I'm not looking at every piece of information on the internet. There's algorithms deciding what I look at. And that's great. Not given that it's not like I'm going to Yahoo and choosing which links to click. I can just go to my Twitter feed and it just shows me interesting stuff that I'm likely to like. And if that is the future we want, and if you think about the fact that a lot of companies are using AI for hiring decisions, a lot of people are applying to companies using AI, is the future we're going toward where AI is just assigning us jobs and we're happy with it because we don't have the choice? It sounds dystopian, but if you tie some of the logic together, it doesn't sound too far-fetched.

It's microskills. One of the things I'm thinking about building is— so I look at, I look at my kid. He, he loves sports and he, he plays one sport pretty seriously. And as much as you put in front of him, he'll just do. And since he loves the sport, his tail's wagging, he gets better, he has goals, helps him achieve them. But if he has to think too much about what should I do and how do I, then he's just, oh, like that's too much mental. Capacity, like, where's my video games? Or where's YouTube or whatever? So it's like, how do you spoon-feed it ongoing in between all the formal structured stuff that you cart your kids all around to? And can you make it really easy and fun at home to just do it without all the mental math and friction? And can you build a platform? And so it's a similar concept, but around youth sports skills training for people that are serious and have goals. Like, how do you just provide that continuity and kind of bake it into the fabric of their days in a way That is fun. I think that's what you're saying as well, is take out the friction or the mental baggage around— not in my example, it's youth sports skills training, but in your example, it's just like livelihood, go about life. I don't know, it's a slippery slope because at some point you, you want that control, but at another point, like, with so much choice, it's just, it's almost paralyzing.

Well,

yeah,

but one of the things you mentioned, exposing them to different things, is I guess the best we can do as parents, right? In terms of AI, I, when DALL-E, when I started playing with DALL-E, I had these DALL-E sessions with my kids where we'd each throw out a term and we'd have DALL-E generate one image that combined the 3. We started doing that with Suno. So we'd have music, like my daughter would pick the genre, one of my other daughter would pick like the topic of the song, and I would pick something out, like an instrument. Then we just ask Suno to create the music. And I liked that as an exercise 'cause it made, it was more of a fun game that we played, but I hope that it internalized to them like what AI can do. And more recently it's been, they've been playing vibe-coded games where I just build a game over the weekend and they play it. They make requests, and then I show them that request can be like made, and they can play the new game within like an hour.

Given that that AI touches everything, what do you think about having an investment category or thesis called AI? Does that even make sense? It reminds me of having an early on it made sense to have a mobile category. Mobile's just everything's mobile. Early on it made sense to have a category called climate tech, but it's. Climate touches everything, but it's everything, but it's also not a specific cat— it's not a vertical, right? And so it's, do you need a category or do you just need everything to decarbonize and get more resilient, et cetera? But it's baked into the fabric of each industry across. I guess using that analogy, how do you think about it as it relates to AI investing?

I think so. I do. I think of AI, yeah, I think, I don't think of it as like a category per se. It is obviously a category in some, some aspects, but I think if you're building a startup today and you're not thinking about the impact of GenAI on the industry, you're just missing a big piece. You couldn't necessarily say that about every industry, like mobility, for example. There's plenty of businesses you can build without thinking about mobility. It's some, you can largely put it in the backend. But I think with GenAI, it's especially today, it's really important to think about how is GenAI going to impact my industry and how do I position myself well? Do companies need to have AI internally? I don't necessarily think so. But I do think they need to have understand where they fit into in the future of their industry, assuming Gen AI makes impact.

Everyone talks about how AI is going to enable us to work less or maybe not at all. But when I look at the founders who are in the trenches building in these categories, what I hear from them consistently is that I have two jobs, one and my team has two jobs. One job is to do all the stuff that we need to do day in and day out, and the other is. On top of how quickly everything is changing under our feet, which means that people are working more than they ever have before. So how do you reconcile those things?

It's not realistic. There was an early article on Every about how I was a VC who was trying to automate his job away. And then someone asked me that last week. One of my LPs was like, so how are you doing? It's not working at all. The more time I have, the more stuff I do. It's just, if you're a productive person who likes to do things, Yes, I've automated a lot of things, but any time I'm given spare, I find ways to use it. So realistically, my throughput has gotten higher, but I'm not— through automating my job, I'm not automating it away. I'm like automating, I'm like compressing it and just adding more on top.

As the pace of innovation accelerates, what do you think the implications are for how we build companies? Do you think people are going to Have to work. Will it be the same amount of work? Will they work harder? Would they work less? Will it be more stressful? Will it be less stressful?

I think initially it'll be more stressful. I mean, my personal experience is that when you increase your throughput, it's more stressful. I think I said that everyone's going to become a manager, right? When you're a manager, you have to— you can't— you're not just worrying about what you do, but worrying about everyone else does. And those are— those used to be just people, but now it's also agents as well.

What do you think the implications are in terms of team construction? You mentioned that you're thesis-driven, but how do you think about founding team composition and how has that evolved since you started investing?

Today, I'm definitely looking for, I am definitely attracted to companies who are thinking about how to build a lean team from ground up. If you thought about your team from ground up, like how would you build it? We're seeing a lot more like we want to build a $100 million ARR company with 10 people or less. We're seeing more of those types of pitches. And obviously as a VC, that's very attractive. So that's one thing we are already looking at. But it doesn't have to be that, right? There's probably still, there's always going to be more traditional ways to build companies.

For the people that say that, sure, it sounds great. Everyone wants that, but what are they actually doing to make that possible? And what are the elements of their plan that would make that a credible plan versus just a pipe dream?

I think a big part of it is like actually running a lot of experiments and learning, right? Because we don't know how to run a company that's fully AI-run. But if you start running experiments, you see what works, you see what sticks, and then you can hire around those capabilities. And what I mean by that, for example, we use a tool called Workello for AI due diligence. I can give it an industry, a company name, it'll give me a 20, 40-page report in 30, less than 30 minutes. When I think about hiring for Untappd, I am not looking for research skills. I do not need that. VCs who were hiring 5 years ago were very much looking for those skill sets in analysts because you needed a person to do that. Now we don't need a person to do that. So when I'm thinking about hiring for my firm, I'm looking for people who can build relationships, people who can stay on top of news better, people who are going to be better at running, who I can trust to try out AI tools that might be a good fit for us versus people who can do raw research. So that's just one example. But I think that, I think similar patterns apply to companies.

So do you, how do you think about our current education system? Is it keeping pace with the changes that are happening under its feet?

I don't think it ever has. It's slow, always. I feel like one challenge is that with technology is that like people who are really strong at technology coming out of school probably didn't learn it in school.

So for, as a parent, what are the implications of that as you think about how to set up your own kids for success as they get older?

Just have them be familiar with it, right? It's weird, but like I was on AOL Instant Messenger, right, when I was a kid. And in a weird way, right, that kind of native, like I am able to build a relationship with somebody digitally, right? I have friends who I've never met in person because I've only spoken on Zoom. I like to think that my early exposure to maintaining relationships via an AIM chatbot chat led to me internalizing the idea that I can build a relationship virtually. I think similarly, right, just exposing them to AI, letting them play with it, encouraging them to understand like how it can be used in a way that resonates with them will help them internalize how this technology might impact them in the future. I think,

did you wrestle with the, the, I guess, the counter tension in the screen time dilemma? More time outdoors, more time away from the screen. That's something that you worry about.

Well, yeah, at a high level, yes. Personally, no, not as much. We live in the forest. We don't—we actually don't give them that much screen time. It's more about like when they do have screen time, I want them to do something interesting with it. But we live in a forest. We go outside. My wife is like anti-tech, so we balance each other out pretty well there.

It's just hard because you—yeah, you want them to be doing something productive with it when they use it. But it's—it's—yeah, but we all want to eat healthy food and. I don't know about you, but I can eat healthy food except when the junk food is around. When the junk food is around, I'm not going for the healthy food. I'm going for whatever the craving is.

You just said it. When the junk food's around. My fridge is full of organic vegetables. I don't have junk food in the house. It's hard for me to eat unhealthy.

But how do you make the internet only full of organic vegetables? To stick with that analogy.

Parent mode, right? Like when I give them a kid, like every app has a time limit set to it. And if it hits that time limit, they have to ask me for permission to use that app longer. Mm-hmm. And we, I just block things like YouTube. My kids are young enough that I have that parental control. It'll get harder when they're older, but for at 8, 5, and 3, the parent, the Apple parental controls are pretty great actually.

I feel like what, what the, the, like the insider tech crowd does to, when they think about preparing their kids for the future versus what everyone else does when they think about preparing their kids for the future is totally different because the tech crowd is the crowd that's building the power of this, right?

I'm using OpenAI Deep Research to have it come up with a list of apps I should let my kids use, going and downloading it, setting parental controls, and then being like, here, you can use it. This is all healthy apps.

We're a family of like teachers or nurses or firemen or elected officials or academics or whatever. Then you're not going to be building the picks and shovels to power the AI revolution. And so does everyone else just keep going through their day-to-day and just hope it works out? Or are there—

yeah, yeah, that is hakuna matata. It will work out. Some, when we were kids, some kids had cell phones, some kids didn't. Parents disagreed on whether middle schoolers should have cell phones. Ultimately, I don't think it mattered that much. Like, the people who had cell phones didn't ruin their lives just because they had access to cell phones. People who didn't have cell phones, it didn't ruin their lives because they didn't have cell phones.

You talked about how you were thesis-driven. So how does that manifest today in terms of What areas you're most excited about deploying capital in?

Again, it's constantly moving because it's constantly shifting. A couple of areas that I'm specifically looking at, I think LLM native IP is a fascinating area. It's going to be hard to find the right investment because it's one, it's similar kind of consumer. It's either like there or not there. And like by the time you, by the time it's there, the traction's pretty big. But I do think there'll be kids IP, adult IP that's native to GenAI. If you think about mobile gaming, social gaming, with every tech iteration, there's IP that's native to that technological generation, right? Angry Birds, Rovio, Clash of Clans, all of those, so on. So I suspect we'll see GenAI-native IP. And so that's the space I'm interested in. And then there's a lot of the enterprise side, right? Better memory, better permission auth. Those are a couple specific categories I'm looking at, right?

Uh-huh. But it sounds like less on the application side?

I look at applications for sure as well. On the application side, I do. I find myself attracted to like super niche applications more so. I do feel like it's like the more general you are, the more edge cases there are to solve, the harder it is to grow your brand. I get attracted to we're building a full OS with GenAI OS system for Carpet cleaners, like that kind of stuff, like always catches my attention. I'm like, why carpet cleaners? I just picked that randomly.

Just from a practical standpoint, as founders are out there building, how should they think about you in terms of check size, stage? Do you like to lead? How should they get in touch with you? Just some practical information so they can filter when it might be a fit for Untappd.

Um, I like to say that if you're building something cool, just share it publicly on Twitter and I'll find it. That's how I find cool stuff. But you can always DM me. That's probably the easiest way. I do look at every DM. I don't respond to most of them just because I get so many, but I do look at almost all of them. We do have, I'm in the process of rebuilding it, but we have a site called dealflowdigest.com where as a startup founder, you can submit your round information on the backend. It'll match you to relevant investors. It'll send them that information. They can request an intro. And if you accept it, it'll automatically make the intro. I look at every company submitted there. Again, similarly, I don't respond to every single one, but it's a good place to do an official pitch to me. And I joke that if you want to pitch me, you have to pitch my friends. It's probably like 300 friends on there that are all investors.

How do you determine Where to prioritize your time when you get up each day? It sounds like you've got a wide range of interests and that so much of what you do is just by gut and by what catches your attention and fancy. How much structure do you put around it or do you just let your days run free?

I have very structured unstructured time. So during the day, 9 to 5, I'm very structured. It's like emails and meetings and like to-do lists, action items. I have a to-do list at the beginning of the day. I just go through them one by one. Oh, look, I have an hour to answer emails. I'm gonna do emails. Super structured. After 5, I get off, I hang out with my kids, I put 'em to bed. From 9 to 12 is fully unstructured. I never touch any of the structured work in that time, and I just let myself completely freely read, build, do whatever I want and make it very unstructured. So I have 2 times, unstructured time at night. And very structured time during the day.

When does sleep come into the mix?

If I fall asleep by 12, I can wake up at 7. That gets good 7 hours of sleep. I do have one superpower, which is I can close my eyes and fall asleep in 5 seconds.

Oh, I can't do that.

Yeah. It helps.

I get in bed and then like my— because typically I'll put the computer down and then it's like, all right, brush my teeth, get ready for bed. And then I get in bed and then it's like my computer's still running on super speed and it's, what do I do now? I guess that's, I'm actively trying to address that. So any tips that you have to be more like you in that regard, let me know. My Whoop has not been kind to me lately.

Meditation. I just, I essentially just meditate myself to sleep. There was, uh, there was an app called PZIZ, P-Z-I-Z-Z, that was developed for people with sleep disorders, which I do not have, but I have been on and off using it since college. Actually, in college I was using it very heavily. It wasn't even an app. It was like an actual audio clip that I just used the same one over and over again. But it has a guided sleep app. And over time, just using it, like I noticed myself repeating those words that are said in that app when I'm going to sleep and it just knocked me out.

So we've talked about the future, macro future. What about the micro future, the Yohei future? How much of that is scripted in your head? You're gonna be a professional investor for as far as the eye can see. Is that how you think about yourself or like what does the future have in store for you?

Being the VC, being a VC, I'm really enjoying. I'm enjoying it. I love it. This is definitely my life's work. I suspect I'll be doing this for a long time, but I don't have, I don't have a scripted life. I think, I think it was like 10 years ago, I was thinking about applying to business school and they had me write an essay on where you see yourself in 10 years. And I spent the whole paper talking about how if I can imagine where I'm going to be in 10 years, then I fail. Like where I want to be in 10 years is somewhere unexpected, somewhere I couldn't imagine myself.

I feel like we, we have that in common too. Like different ways to get there for sure. But, but I've talked about this before, even on the show. Like when I go out for a run, I don't have a plan. Like I, I don't know if I'm gonna run 3 miles, is something gonna kick in and I'll do a 20-miler that day? I don't know. So my— frustrates my wife. She's like, when are you gonna be back? I'm like, I don't

know. I use the analogy of a classical versus jazz musician. Classical musicians, it's, you have a script you're going to play and it's about the execution of that script that you've planned for yourself. And with jazz, again, with jazz, it's much more about, can you just jam out with whoever jumps on stage? And they're both very beautiful pieces of beautiful art, but very different skill sets.

I wrestle with that with parenting too, right? Because it's like, I relish in living such an unstructured life, but then because you see how it's like murky terrain and so competitive and you want to give your kids every chance of success. I don't know. I find that we're pretty structured parents in terms of helping them get to the next level and achieve their goals and keep the train on the tracks when it's like sometimes the best thing to set them up for the long term is like living in that uncertainty. But at the same time, that's like a scary thing for a parent to let go of the reins. So it's, I don't have answers, but it's just something I find that I'm wrestling with constantly.

Same, same here. I think it's constant ebb and flow.

This has been such a wide-ranging discussion. Is there anything I didn't ask that you wish I did or anything you want to leave listeners with?

No, nothing in particular. If you're interested in AI agents, go check out Agent Fund. It's a rolling fund, so anybody can subscribe. We've invested in a handful of fun companies there, which we'll be announcing soon. But yeah, and if you're not playing around with AI, please do.

Yohei, thanks so much for coming on the show. This Definitely has my gears spinning in terms of how the world's gonna play out. I have more questions than answers, but I guess that's part of the fun.

I think so.

Wishing you every success and looking forward to watching your progress and staying in touch.

Same to you.

Thank you for tuning into The Next Next. If you enjoyed it, you can subscribe from your favorite podcast player. In addition to the podcast, which typically publishes weekly, There's also a weekly newsletter on Substack at thenextnext.substack.com. That's essentially for weekly accountability of the ground I'm covering, areas I'm tackling next, and where I could use some help as well. And it's a great area to foster discussion and dialogue around the topics that we cover on the show. Thanks for tuning in. See you next week. See you next week.