What Sets Unicorn Founders Apart | Jude Gomila: Heyzap, Golden & 20+ Unicorns

Talent alone is not good enough because they can be demotivated talent. They could be talent that's tinkering on something that's completely irrelevant, just like not necessarily esoteric stuff, but like it's just not useful for anybody else. And they optimize the wrong thing. So if you get a talent that wants to build something for other people at scale, that gets really interesting.

Hey guys, welcome to Vintar Podcast. today we are hosting Jude Gomila from Silicon Valley. Hey Jude, welcome. Good to see you again. We met 15 years ago briefly, but, you know, we, we had quite a journey in the past 15 years.

Yeah, it's great to see you, Yuriy. Thanks for having me on. 15 years, we haven't spoken to each other about 15 years, right? So that in itself is quite fun to watch the whole Silicon Valley. I'm right in the middle of Silicon Valley right now inside the mine with the silicon.

Yeah, good to see you. Back then you were working on HazeApp with Iman. Yeah, so it would be good to learn what happened in the past 15 years. And, uh, but maybe first you can introduce yourself and who are you?

Sure, I'm Jude Gomila. I'm originally from the UK and studied engineering over at Cambridge. Um, I was always interested in math, science, natural philosophy as a child. I was like constantly wondering how does this work? And math, math I had a bit of a knack for, even though it's like scrappy. I was like scrappy, but I kind of like was keen. And what happened is when I was a kid, I changed school from this awful school that got shut down because it was so, so bad it got shut down. And all the kids moved to this other school and we sat there and because there were so many kids that had been displaced into this school, I don't think the teachers completely knew what to do because they were overcapacity. So they just like handed a stack of books like this, and they're all math. It's just math exercises from the very start. And they didn't know where we were coming in from, like how good we were, bad we are. And they gave me the first book and it was so trivial, it was so boring, but we had to do every single question before we could get to the next one. I saw these other kids on like book 7, book 9, and I'm stuck on book 1. I'm like, this stuff is so— it's like 1 1 2, 1 2 3, 1 4, you know, you keep— and that was for weeks and weeks on end. And all I remember doing I swear to God, this one, all I remember doing was doing this math book for all my time from when I came to school to the end of school. And I don't even remember anything else that they taught us. I don't even know whether they taught, because it's just like almost like you're left in the room and you just sit there and you just do this math book. Anyway, I sat there and I, I treat it like a race. So I was like, okay, I need to get to like book 7, book 8, book 9. This is where the other kids are. And I know they're slowing down because it's getting harder. It's getting more and more difficult. So I'm like, okay, well, I just diligently focus. So focus hardcore. And so I eventually catch up and I was like so pleased. And then, and that was like, I don't know, we were like 10 years old or something. I think it gave me a good foundation of this, like knowing just real basics, but having done like every possible permutation. of all these little basic questions. So then I got kind of interested in it and I really did like the physics side and the biology side. And, you know, so I started basically working pretty hard from when I was 14 to get into all this, basically to try and know as much as possible. And then I got into Cambridge, which is cool. I thought that was cool. And then did engineering. And then I was like, oh my God, it kind of ruined it for me. So, it basically, they made it very dry, and I was like, oh, why is it so boring? We're not getting to do the big questions here, like the big questions, because I was still interested in physics, but I learned a lot. So, there was a ton of material. I learned a lot about engineering across all, most engineering fields. And so, from electronics to electrical to blah, blah, blah, blah, blah. And then I did manufacturing master's, and I thought, I came out of it thinking, oh my God, it's useless. Manufacturing has left the West. And why did I do that? I was like, I learned loads of stuff, but why did I do that? And everything was going to software. So, I was like, hmm, maybe I should have done computer science. And so, I started to learn it. I started to learn it manually. And I got very interested in, like, Gödel's incompleteness theorem, the computability problems, Turing's work. I went straight to core rather than necessarily other things. So, I'm pretty interested in all the products as well. So, along the side, I was reading that we had something called the Argos catalog in the UK. I don't know if you know this. It's like a Costco, but in a way, in a sense, but there's no store. You just go there and order something like Amazon before online. And I used to read that product catalog and look at every single product and look at every single spec on all products, including every permutation of watches, all the toys. I was obsessed with the toy section. And I would read it every year and like memorize it. So, I knew all the products available and I also didn't have any money, so I needed to make some money. And I was very interested in how to make cash and I wanted to buy more powerful computers. That was my motivation. I wanted to buy more powerful machines. So, I did a lot of entrepreneurial experiments pre-university and then during university, but it wasn't quite connected to Silicon Valley or tech. It was very disconnected from that stuff. It was more like making money from first principles or like trading somehow, like buying something and selling it. I was really interested in all the esoteric technologies as well. And I was on the internet early. And there are other people that we know like Imad and other people, James, our CTO of Heysapp. We were all part of an early internet crew messing around on the internet, like looking at all the crooks and crannies of the web and testing things out and testing out like, you know, voice over IP when it first happened and testing out MP3s when they first happened. getting into this stuff. So, we kind of knew a little bit about the web, but we weren't making any money. We weren't making the companies behind it. We didn't really know about that stuff. After doing experiments in the UK and realizing that at that time, I couldn't really raise money in the UK, it was 2007, 2008, it was so difficult to raise any money and angel investing didn't really exist. So, there was a lot of buzz happening from like over in San Francisco and Silicon Valley. So, there was this general draw and I got out to Silicon Valley to get into YC. So, that was my objective. I didn't have a visa. I didn't have— I didn't break any visa laws, but I just came over and I was like on vacation and I was like, I need to get into YC no matter what. So, I started hanging out. So, I think I went to like an early demo day and I wasn't in YC and I didn't have any money to invest. And I just wanted to get in and hang with this crew. And this crew, everyone's doing very cool stuff. And I was like, this is a great group of people to hang out with. So, finding that community was critical. So, I went off looking for a co-founder to try and build a company and experiment with a few different people. And then me and Emma were talking about it. So, like, why don't we— we're both looking for kind of doing a startup co-founder. We'd done experiments ourselves and even advised each other. on different things, but we'd never considered doing it together. We're both right in front of each other. We're like, why don't we just do it? Why don't we do Paysapp? So we built a spreadsheet, we went through all these different ideas, including even accounting software. And we landed on like, let's do a YouTube for games. And we applied. Paul Graham did our interview. It was a sunny Mountain View afternoon, sitting very casually outside on the curb, super casual. And he's like, I don't understand what you guys are doing, but I really like you. So you're in. It's like, this is great. It was like our first break, first, like, proper break in Silicon Valley. So that was good. So we went in and we raised $30K and we actually spent, we needed visas, right? So we spent $15K or $20K on our visa, $10K each. So we used $20K of the $30K. And so we had $10K left, but we had not started YC at this point. We had 3 months before the start of YC. So we ran out of money before we started YC because just frugally spending money on basic food and accommodation. So, and I think Iman talks about this on other podcasts where it did get pretty frugal, man. So I know the differences in pricing between like a white onion and a brown onion. And I know what time that Safeway, you know, brings out the cheaper food and marks things down. So we went through some pretty And we walked into Chinatown, walked back with sacks of rice on our back and all this stupid stuff, right? We were really wasting a lot of time for 2 Cambridge grads that could build stuff, like trying to just survive. So the crash happened, the 2008 and '09 crash happened, and we were like low on money. We were going negative into credit cards. There's only a certain level of debt you can go because you can only get a certain amount of credit. And so we were going super low on that and we were trying a lot of different things and we were hacking. We were working hardcore. We did 7 days a week. And then we crashed. We both like were getting physically ill from working too hard. Then we pulled it down to 6 and we just held 6 as well. And everyone else was going out and having some fun. We just like locked ourselves away. It's like, these guys do not leave their office. We're like hardcore.

Did you code yourself?

Yeah, I was doing the front end and I did the backend and I did design. And I didn't actually know CSS or anything, so I had to learn it. And I had about I don't know, 6 hours to get going because it was like tight timescales and there's no AI to help you. And we were trying to do some stuff that was like difficult to do and basically hacking hard. So anyway, it worked and we got things out and we shipped and we started working on it and we ran and raised. It was 2009 happened, right? So the crash happened, half the batch of our batch died. We had Airbnb like starting to take off and it was really cool. to watch Brian. They were just working down the street on like 3 guys, you know, even Brian came over, took pictures of our office with his photographer. So, it's good early, early SOMA times. We were also told by YC not to go to SOMA. Don't be in San Francisco. You gotta be in Mountain View. And there were kind of a few people that broke the rules, like Dropbox were down there and Drew kindly let us work in his office for free. So, we got to watch these 6 MIT people hanging out and working in a very unorthodox style. Like, into the office late, like 12 PM or 1 PM, but they stayed till 1 AM and then playing Guitar Hero for hours and hours and hours. And like, how can you use— but then they'll go and code for hours and hours and hours. So it's kind of strange, but it was working. So we hung out there, we hung out and discussed, if you remember Daniel Ha, and these guys gave us free office accommodation, but we were inside other companies that were working and like seeing what worked and seeing the quality of the people and hanging out and having these conversations. So, we raised money in really by kind of brute force in a way, but meeting smart people. So, smart people met smart people. I had to do so many meetings. Me and Emma were like, you know, you're supposed to have one person raising, and it ended up being like, this is so dire that the only way to do it is 2 people to raise because we have to meet every single person. Because people didn't even have any money and they'll take the meeting and they didn't have any cash. So, like, they'll string you along 5 meetings in, it's like, I don't actually have any money. I was like, Why did you take this meeting then?

Yeah, why do they meet? Yeah.

They had to, to seem like they— some of them just wanted to talk to someone. They just wanted to talk to young people with ideas. And I was like, why are you doing this meeting? This is insane. So anyway, we found the algorithm was really to work with the smartest investors out there. And we enjoyed our conversations with the best people. And they enjoyed the fact that we enjoyed it. And we got into better loops with those people. So we cracked Union Square Ventures. I remember flying out, it was like a last amount of credit left. to make the flight. And they didn't know, we couldn't say, oh, we're desperate here. And we got the, you know, it's a fierce like 3.5-hour full partner meeting, which is just 3 people. And it went on and on and on. And it was just like, I'm not giving in here. I'll be here all night. We can be here for 3 weeks if you want to. I don't care. We're gonna, we need, you know, we need to do this because this is life or death kind of thing. And for them it's not, it's just another deal. For us, it's like, this is our life. It's like, it's not, it's We're done here or we're not. This experiment's done here or not. So there was no like backing down to any questions, like absolute relentlessness in every single question, exhaustive answering, creative exhaustive answering. And eventually they gave us a term sheet and it was not a bad deal. So we took the deal and that really got us going. We had like, I can't remember, we raised like $650K or $500K. It was a lot of money back then, especially if you were coming out of ultra frugal mode. And we were very careful with the cash, and we started building up a team, company, and trying different experiments. And lots of disruptions were happening in the background, like Facebook Gaming exploded, and we went over that way. And we couldn't quite get in there with our kind of more structural play, because we weren't making games, we were trying to make software for it. And then mobile happened, and we got into that, and that leveled us up. The Heyzap app kind of worked, and we did the SDK, and that got us into a lot of games. And we weren't making any money still. We're like, oh, we need to make money. We're quite far into the company at this point. So we released the ad SDK. And that worked. He like went to $1 mil a month very quickly. And we were like, wow, you make one move and you can suddenly change the business so quickly. So Heyzac went for a lot of pivots, a lot of experiments. And it's really us two learning even how to do a startup, how to do a company, how to navigate this fast disruption. And we had many, many shots on the goal. That was maybe the main learning that like we didn't just do one thing and give up. It was like, no, we tried a lot of different things and one thing eventually worked. And then we modified it again, and it got us to a point where, like, we had acquisition offers coming to the company. Funnily enough, we had an acquisition offer from Zynga in 2000, only one year into the company, for what would've been a very good deal, but we rejected it. We were like, no, we're worth way more than that. And sometimes it is worth taking something that's ahead of the game. So acquisition offers started coming in. We started getting a little bit bored of gaming just 'cause we'd been in it for a while. We're like, we could be building other things. We could be building Like, all these other things are happening around us. Along the way, I'm starting to invest in things. So, I'm like, at first, like, advising other startups that want to get into it, and they're giving me equity. Then I'm realizing all these problems with advisory shares is kind of a mess. And I started to just do investments. So, I'm going to all the YC Demo Days. I'm like plotting out my fantasy football.

What year did you start investing?

2010. So, about 1 year after getting funding. So, I used my salary. I still stayed in frugal mode. And I used my salary to invest in startups. And so I went to the Demo Days and I started noting like what was going to work, what's not going to work, and trying to understand the room as well. Like what other people were looking for when these demos were occurring and which ones were going to be big and which ones weren't going to be big. And it almost in a way, you could see this like 2 tiers of Silicon Valley. You had these young founders and you had these other people who had done it before and became investors. So one knew it was kind of a model that a lot of people were going to go through, like they were going to start doing investing. And you get to talk to other founders and learn faster. So I started doing investing in 2010. I started backing startups around me that were interesting. First with small checks, and I just kept pretty consistent. So I started increasing the check sizes, and then Naval was on our board. So we learned a lot from Naval and Alba Venga for about 5 years or something every month. So we did 2 hours every month with these guys.

What are small checks like?

Started with like $5K. $5K, $10K. Yeah. And at the time there was this like threshold of $20K, like $20K was the minimum number. And, but the SAFE note was kind of happening. It's like, well, why is it $20K? It doesn't really matter. It could be like $5K. Yeah, I'll take you for $5K because if you're giving specific advice. So $10Ks and $5Ks allowed like me to get into, you know, just into the door, right? And then it was like larger amounts. So I started increasing that. AngelList started happening and Naval invited me to various different new products he had, like the Syndicate. The original syndicate was an SPV. They separated that later. The syndicate is really just the marketing group and the SPV was the vehicle. But at the time it was called syndicates and people still call it syndicates 10 years on from the name change or something, I think. So we did this and I think Uber ran one in the early days and I ran a few and it was quite amazing. It was like, wow, $800,000 just came into this SPV with like a little write-up. And there weren't that many players on the GP side of it. So I tested this out and you had to invest your own check alongside it. It was separate. Then they combined it. So they went through these evolutionary models and I kind of went along with it while doing my startup. Paysafe got acquired in 2016 and I went straight into another startup. I was like, well, this is not enough. We sold 45 and Amar clearly as well was like, it's not enough. We need to go again. So I started working on Golden and I really like stepped back up for a year during the earnout and just thought, what is going to be important to me? Like, what do I actually intrinsically care about to become closer to the action of what's in my mind? And I was thinking about, I want to know about things. I want knowledge. I want to bring it back to the roots of like my passion that got me into university. And I want to know more about like these things, these strange things that exist that I don't know about. So, I was like, and I would go on Wikipedia and read about limits of computation and And look through all these gaps and see, wow, they're missing a ton of stuff. It's just missing a ton of stuff. Why? So I started working on that problem and I got really into it, really obsessed with it. And we raised pretty easily because we had had an exit. It was fairly easy to do the raise, to do something ambitious. It's pre-LLMs as well. So something was wrong with knowledge at this time. Like Wikipedia was sucking and we needed to do something. So let's pause there just in case you've got any questions before we go on for the golden story.

Yeah. Sounds like you guys split after Haze App exit, right? Like you didn't continue working together.

Yeah, we did 9 and a half years together. I mean, some of the experimentation is by experimenting with different pairings of co-founders. So even if you did a good job together, you might do a better job with just testing the water with a different team. And so we both invested in each other's seed rounds. So I did first check in Mercury. I think I did first check in Golden effectively. And we're pretty close, right? We're friends. I think it's okay to remix. With different people to try, like you're basically doing this big search space of all the different combinations and you can kind of get the band back together in different places. It's interesting because a lot of alumni went and started a bunch of different companies that went on to be reasonable. So the alumni of Heyzap were doing really well. And we had, there was like what to do with all this different talent. It was kind of like not 50-50, but it was like some people went to Mercury, some people went to Golden. And some people went and did their own things and set up companies that got VC funding. So the HeyTap crew were a pretty good crew.

So what did you learn from the first journey, from the first venture?

So much. Making mistakes. Hardest thing was doing something that you really believed in that was actually a fundamental mistake in startup law. So, you know, having a rule that you may have made that you thought was working, but it was actually the thing that's holding you back. So much. I mean, the number one thing was that the best, you have to get the absolute best talent. And you can't always judge it right in the interview. So, you have to be prepared to fire quickly. And within the first 30 days, and maybe even the first day, you would know, because we always had this rule of, like, ship something in the first day. And if you couldn't get through the mess, or, like, you know, you couldn't get something out that's meaningful, you probably weren't gonna make it. So, because the best people just find a way to get something useful out. I was like, oh, I can just improve that button. Like, the most minuscule features get to deployment in the first And how fast they did that was like proportional to how good they were going to be in the end. So we learned a lot of things and the main thing was like, just get the best talent possible. We did a lot of things right. You know, we tried, we just kind of like, the other interesting thing was maybe it is better to give in actually sometimes. We never were giving in, but we eventually get there. But like in that time, you know, could we have tried 3 startups? I don't know. It's kind of very stochastic, right? There's a lot of probability of like rolling the dice of these different combinations of people and market and timings as well. And assumptions that you go into it. So, yeah, that was a lesson, which is, and it's hard to say, like, should we have changed the strategy? Should we try 3 startups in that time or not? I don't know, because we can't rerun this, simulate. It's not like simulatable. We're not gonna know, right?

And it's a very different idea from what you started, right? What you sold eventually.

Quite different, yeah. We became a mediation ad network over time, and we really were seeking distribution each time, and then we tried to monetize the final distribution. So, each distribution attempt that we got, got us to another distribution attempt. And eventually that got us to a business model. And things are moving very quickly that the main other thing as well is that quality, we were learning what quality was along the way. So we didn't come in super high enough quality. So now it's like, okay, well, now we know what the quality bar has to be like really high. And that was something that we learned, like, and you can see it from Mercury, right? The quality is like there. So we learned quality and talent. We did a lot of other things that were right, like shipping fast, being creative, trying a lot of interesting things. Being cerebral about things, being fairly frugal about things, and actually architecting our exit. Yeah, that was a good experience. And then with Golden, we raised and we got Marc Andreessen on the board, which is great.

Did you have co-founders?

I didn't co-found it. So that was a wild thing. So that was another thing as well, like that I would always now recommend co-founding with people because it's just, you can never really take any time off and you have to just co-found. I had made a, what is a funny sequence of events actually, because initial conditions can really set you off and like can a mistake very early on that becomes like harder to undo. I had gone out to interview various different co-founders and the people I really wanted were all like just not available. They had made commitments to other people and they were co-founding with other people. So it's really hard to find a great person who is kind of available for co-foundership because they're gonna be doing something all the time. So that search alone can take a long time. And there's a little bit of luck in that, or you could solo it and bring someone in. Later on. So yeah, I didn't co-found it. And that made it like, I think that was like, it would have been better to co-found in general, because you can take a vacation, right? And I think actually, the learning there is that you can't do everything in your own head. You have to actually talk about it with another person, and like untangle certain things, and also splitting responsibilities of building and putting on CTO hat at the same time as CEO, which is like quite difficult.

So you were both CEO and CTO?

I'm PMing and raising. Yeah, it's PMing, raising, everything. It's like too many things, right? So, I tried to like build out the responsibility, push it down. And it's not quite the same as having a second person or a third person. So, I think the magic formula still stands of like 2 or 3 people is it. There are some businesses that have been built by individuals that has become like very, very giant, and they almost never give out any equity as well. They're like old school. I think for tech, it's moving so quickly. It's so complicated that it's better to have co-founders.

With the current tools that are available, would you build the whole company by yourself?

I don't think so, because I think part of it is just having a good time with your friends and sitting there and just hanging out and building the company and talking about other things as well, not just work all the time. So I think if you just do it by yourself, I mean, you could do it. And by the way, you could build a semi-lifestyle business that is generating a few mil ARR. It's compact. I think for ambitious things, you should bring a larger team together for sure. If you just want to make some quick money for yourself, go solo if it's quick. If you just need to make some quick money, if you want to try a more ambitious thing with a long-term mission, I think it's better to go 2, 3 people.

So what happened with Golden? What is the journey like?

So we built a pretty complicated system. So we built a knowledge graph. And so there's this like giant knowledge graph underneath it that we were like sucking all the companies. LLMs had not happened at this point, right? So The best in class was Crunchbase, Wikipedia, PitchBook, and we felt like we could take it together and build like a larger knowledge graph. So we went and did, built a pretty reasonable knowledge graph with UI to browse it. And then we started getting like a reasonable amount of distribution. So we were ranking on Google and still rank actually on Google for first position for various search terms. So we found this like pretty workable, like our thesis for basically distribution works. So we would print out information text on a subject. And add all these other components to it, structured data. And then it would, for a page that Wikipedia didn't have, it would suddenly rank at the top of Google. And so we were getting like number 1 positions. We're like, we can achieve a number 1 position on Google almost whenever we want for a canonical topic. Say for example, like surreal numbers or WeWork office space or, well, not quite that, but like canonical objects would like rank number 1 position. We basically kept doing that and building up this distribution and traffic. And then we added on a SaaS model to it, which was okay. It was generating like some revenue, not absolutely explosive revenue. COVID happened. So then COVID happened, which blew out everybody's like office culture and kind of swang things. And that was good and bad, as in for the company. It was good in that we could reset. It was an opportunity to reset certain components of the company on going back. It was bad in that it took out our physical culture. That we had in the office where people were like discussing ideas, hanging out in the evenings. And it all went back to like more clinical culture, which was like trickier to run, right, for a screen. So we ended up getting some good contracts from the government and I start, well, LLMs happen. So then the first like model comes out and they spent $50K or something to train it. And we were like, this is exciting. We should do this. Half the team didn't want to do it. And we were like, LLMs aren't going to work. What I should have done on that, in retrospect, learning-wise, is I should have gone with the intuition of the founder and just said, okay, well, if you don't like it, you can leave the team and we're gonna cut half the team out who don't believe in this new technology. I did not embrace it early enough because we even proposed it in a board meeting. Hey, why don't we spend $1 million of our funding and do like a more, much more serious model? So, I think when you do see something really promising coming along, it is okay To change direction. You can't do it on every single shiny thing coming through, right? Because you will never get anything done. So that was a big lesson. So we actually ended up getting more into AI towards the end of the company. And there was this parallel, there was like leaning on AI. We mostly got really, really good at like how to construct knowledge graphs, how to store information and knowledge on things, and how to kind of verify facts. Well, what we were trending towards facts. Crypto also happened. So we had crypto was starting to peak and we started to think, hey, could we solve this in a crypto-like manner? Could we build incentives for people to try and put knowledge forward into The Graph? And we built a protocol around this. A lot of people came to the Discord channel. It got pretty nuts, like 35,000 people in the Discord channel. And that was like an interesting difficulty of like trying to please the community, but also trying to build the right thing. And that tension there was tricky.

Did you launch your own token?

This is the funny thing. So at the time, I don't know if you remember like LBRY Coin and all these other cases going through where we hired a general counsel for crypto, crypto specialist, and we started looking into doing our own token, preparing for it. And he made the recommendation, do not launch a token. You cannot do this. SEC will not like it. So we tried to fight this thing of like, oh, we're being told that we can't do a token. And that's like part of the whole incentive structure of doing this protocol. And we read the case law and it was not favorable at all. So it wasn't looking good at all. This is pre-Trump era for crypto tokens. And actually it was deemed to be like, if you launch this, you're going to have a problem, you're going to have a headache. So we actually decided not to do it because of the rules at the time. And then everything changed and it's like the floodgates are back open again. And by the way, they could switch again in the next election cycle or the next election cycle. It could be that people that launch stuff are going to lose all their stuff. And it's a nightmare. It's actually a nightmare for this kind of stuff. fluctuation of rules. So we decided not to do the token. And that was like, it's funny because that was like the non-risky approach of an entrepreneur. An entrepreneur, nothing to lose, just do it. At this point, you know, I've got my GCs telling me you've got personal liability if you launch a token. And I was like, I don't want any personal liability eating into my portfolio. I've invested in like 16 unicorns from seed, right? And I've got something to lose now. I've got stuff to lose. So I took a prudent decision not to lose my own personal estate, not to put my estate and the farm at risk. So then we shifted to more more AI direction. And we started to become, we went more skunkworks inside the company. And then we, that effectively got us bought because we started to get kind of better at prompt. We got really ahead in prompting engineering and RAG. We got good at RAG to LLMs. And we got interested in like LLMs and knowledge inside LLMs. Like how are you going to verify knowledge inside LLMs or introspect it? And I'd read up on like a lot of the papers. This is about 2 years ago. I read up a lot of the papers on the current thinking of storing information and knowledge, not just in knowledge graphs, but into the actual LLMs themselves? And how do you verify it? So, I started thinking hard about this stuff. And that's where, I guess, for other parts we could go into, is like, you know, the interest of LLMs and knowledge and how we're gonna solve intelligence. It's become like a thing for me. Because it's still the consistent mission of Golden's, like, try and like get a system that can know as much of the universe as possible. And like, where would you pick that up? You pick it up from LLMs and not necessarily even LLMs. Like it may not even be a large language model. It doesn't necessarily have to have the language part because it could be a large model. That fascinates me. And I'm still, I've kind of gone back to research mode. So we sold the company last one year ago and I've gone back into thinking like solving like curiosities on my mind, like consciousness, the brain, how does this stuff work? How does compute work from like scratch again?

How did you know that this is the right time to exit? Maybe it's too early.

I think our shipping speeds, when the shipping speed slowed down and we had built a bigger team, but the shipping speed came down and that we were still, we were pretty early on some parts of AI and other companies needed AI. And we had an asset that in the future an LLM would get better and better and have all that data inside the LLM. So there was like good timing on the asset value. Our shipping speed was like slowing down because of more scale. And we had a lot of constraints building up. In terms of distribution also becomes a constraint. So, you know, you can't just like wipe the code base and be like, we're starting again. And like, we're firing like 80% of the team. We're going back down to 3 people. We could have done that. And we discussed doing that. And, or we could sell it. So the choice was like, hey, we can strip this down and like try something new and enter the knowledge game as it currently stands, which is very, very different, right? With GPT-3 and GPT-4, O3, except O3 wasn't out at the time. Or we can get the company bought. And I can free myself up to have no constraints again and then like rethink from scratch what to do now. So I chose to do that. And so we sold it. Yeah.

And it was, it became harder to recruit as well, right? Because so much competition around AI.

Yeah. So, so yes. And yeah, there were a lot of people recruiting was still good because people love the mission. But yeah, a lot of talent was going directly to like AI, direct AI, like foundational models or otherwise. Right. And some of the stuff like I'm under NDA. So I'd love to like go into certain parts of the story, which makes it spicier. And I'm thinking of all the NDAs that we signed in the final times of selling Golden. But I can tell you one thing. I spoke to pretty much most of the major AI players at the time, and I got a really good snapshot of like where they stood at the time with what they've got, plus the main players. And that's given me a reasonable starting position to be like, okay, I know where like FANG and beyond are roughly like, you never know all the secret projects, right? But like, and you get a feeling and then you're like, then you've got to get into the papers and see what's possible. And there's so many different startups. So I feel like I'm still a beginner on AI. I think everybody is, right? Everybody's a beginner. You could read all these papers and see all these different things, have all these different ideas, but there's so much happening that like, I think it's very few people that really have got like intrinsic feeling of like what, what is going on next. And there's so many things launching that, and we might still be in a real local minima, right? Like, there could be something that comes out after LLMs that might be very, you know, transformer paper came out, right? Like, we saw that, we're like, that's a big paper. But I'm seeing papers like that right now. And, you know, 3, 4 years, it could be something very different that takes off. So I think at this point, the future feels very tricky to predict. And I think a lot of people got lost and like, are we going to like lose our jobs or You know, what's going to happen, even like people that thought they were top of the game could like completely, like their role could be switched up a lot. I spent a lot of time to go back to first principles and try and like think about stuff that could be kind of quite futuristic and then come back a little bit to be like, where would you land? Where would you start that now? And try and build knowledge that is not being talked about so much today.

Do you want to build another startup or you transition to the investor seat?

I'm definitely ramping up my investing. I enjoy the building stuff as well. So I'm still building prototypes. And there is something very special about startup. When you look back at your history, like one focuses on the startup journey because it is a central thing that pulls everything together. So I do enjoy that. I'm pretty much mission-oriented. I will definitely ramp up investing because like I see so many good things that I need to put more money into the company. So that is definitively happening. I'm definitely exploring various, like incubating various companies. And, you know, if one like turns out to be the one, well, you'd be a fool not to jump on it, right? So I think I'm learning by building and learning by incubating and at the same time ramping investing. And that's pretty broad. It's like kind of a search space of what is like your golden ticket out to like the next level. And it's funny, if you sold your company twice and you made a bit of money, your needle becomes like quite difficult to move. So either you can set your sights on something very ambitious, which could increase failure rate or not. Or you can pull back to like, what do you intrinsically enjoy doing and seeing where the alignment on the subject matter, if that fits into a startup. And so, yeah, I've been learning a ton. So, I've been trying to catch up and of just learning where all the things we know, like humanity, like what are all these mysteries? So, I've been like searching through mysteries and trying to get to the bottom of them in my mind, coming up with like researching things that I feel are so, like, wacky that I don't even want to talk about them to anybody. So, and, you know, like, really trying to think about, I mean, AI is just quite incredible how you can sit there and have a conversation for half a day and learn so much and read through all these things and test it from different angles. And, you know, generally speaking, it's probably true or correct. And sometimes if it's, you can kind of verify by reading the citations, getting a feel for it. But I think we're in a time where I feel like I'm going for a new university degree where I've got no lecturer, but I've got an AI and I'm having a conversation with this and just trying different ideas out, running like, you know, 7 terminals with different things computing and waiting for them to come back to me.

Which tools do you use? What is your tech stack today?

I've rotated a bit, but the emerging kind of winner right now for me is ChatGPT. I know it's the classic answer that you hear all the time, but Perplexity is good. I'm liking Perplexity for fast web searches of today's information, and especially with the citations. I use Perplexity for quick answers on things that I want citations for. If I want very deep answers, I'm gonna go over to O3. And yeah, that, that works for me. So I navigate between the different models on ChatGPT depending on if I'm doing a mathematical question, if I'm doing something more writing-based, if I'm doing something more reasoning-based. So I'm kind of like, it's either I need maximum reasoning, I go this way. If I need maximum mathematical clarity, I go this way. If I need maximum writing perfection, I go this way. And sometimes there's also not reasoning, but the creativity of linking 2 disparate areas. Yeah. areas together and finding those 2 search spaces. And if I need something more practical, more like a guide on how to do XYZ and today's information, then troubleshooting with UI that might be changing, I'm gonna use Perplexity. Or if I'm trying to find out some today's stats, I'm gonna go more in that direction. So, I guess the dimensions of which tools you use could be depth of a particular type of optimization function. Sometimes you kind of want both. Like, I want really good reasoning plus mathematics, or I want really good reasoning plus biological reasoning or something. So my stuff is quite paper-oriented. So I want academic papers to feed into it, or it's very practical of like, I'm trying to pull off this, can you help me plan this out? And troubleshooting as well. But I do a reasonable amount of math in the editor. So that's also fun because that kind of needs a checker on the side that will check the math, like a rigorous proof checker. But I'm learning so much. I love it because it's it's faster than me reading a textbook. I can just go into an idea and have an idea, and every single experience can be very novel. Yeah.

Can we step back into investing? So you, you started investing in 2010. You are part of YC mafia, right? It's kind of— you have access to most of the YC grads, right? Kind of amazing ecosystem. So what did you learn about investing in early stage, right? Because you're mostly first check into 12, how many, 16 unicorns?

Yeah, probably 25. If we wait one year, 25 without doing any new checks, but 16 as of today. But there's ones that are creeping up on the bill and there's like trajectories of like other ones that just, yeah, take me up to 25 completely.

So what did you learn about these entrepreneurs? Is there a pattern?

There is a pattern. I wrote it down somewhere in this joint. Okay, so the first word that comes to mind is talent. So they gotta be talent, like super talent. They are super talented and they're super applied. So I think you can have a super talented person who's just sitting there doing like playing around, messing around, but they're like way more talented if they just push themselves and like sort out their direction. They could be incredible. So like this talent alone is not good enough because they can be demotivated talent. They could be talent that's tinkering on something that's completely irrelevant, just like not necessarily esoteric stuff, but like it's just not useful for anybody else. And they optimize the wrong thing. So, but if you get a talent that wants to build something for other people at scale, that gets really interesting. If they have this altruistic— some do want money. Like, there's like 2 ways in for like building something for many other people. One, like intrinsically want to build something better They, for other people, they empathize with that person all the time and they make the money indirectly. And maybe because they optimized on, you know, feeling the empathy for each user directly, they're actually better at making the product and therefore they make more money. And the other use case, the other person seems to be like, I want to make loads of money, I need loads of distribution. And it's indirect because they're focusing on the output in Kiefer Weisz's language in a way, not the input. Whereas the input's like almost the empathy model for that user. So, I look for, I really like the latter class. You can do well. I actually, you know what, all the really interesting unicorns are my favorites. They focus, the founder is obsessed with the user at large, and they really don't like, they kind of get disgusted by the current state of play, how inefficient things are, or poor quality things are. Some get quality-focused, some get functionality-focused, some get focused on time. Some get focused on how expensive things are. Like, I can make this cheaper. This is way too expensive for these people. So I think that combined with super talent gives you someone that could be a founder and build at scale. And some people are obsessed with the problem. So they're really obsessed with the product and the features and the little part, and they're not necessarily focused too much on other users. But because they are the primary user, they kind of map to a large set of people and they get obsessed with their personal interaction with the product. And I think that can work because that is a proxy to many other people if they're a more normal person. So that can work. So they're really, they're intrinsically like connected to the product in a super deep way. Yeah. It's like, where does the seat of the mental, their simulation of the mental model lie? Where does their obsession lie? And that's the other way. Like, what is this super talent obsessed by? Are they obsessed with like living in luxury and like being in a yacht and gambling at a casino and wearing an expensive watch? That's kind of the wrong obsession. Or they obsess with like, this pixel thing is annoying me. Like, I really don't like the color of the shading of that thing. Or some of them like the game. So there's some of these founders, like, they want to be— they just treat it like a game and they don't really— they just want to win the game. And it's like, they kind of distance themselves and skip this objectivity on the problem. And they're quite good as well to get. So there's different kinds of ways into this, but that's what I'm looking for. And you might say, what about the talent? How are you going to measure the talent? What's the talent? You know, like, how do you measure that subjective?

You meet a person for a few sessions. In most of them, it's a very competitive round as well, right? How do you get in and also experience that?

Yeah. So in terms of judging the talent, you can set a threshold. Is this the top 1% person I've met before? Like, how many standard deviations is this person out from other great people that you've met? And there's a certain threshold of like, it works. There's a certain threshold of it works to a certain market cap in a probabilistic way. In terms of getting— I don't really try to compete to get into rounds. So if I see like I'm competing, I'm already not interested. So I'm like, I shouldn't be competing because that seems like the wrong trade to be done. The trade to be done between me and the founder is I can add value to your company and there shouldn't be any— you could say, well, I can add value and this other investor can add value. And if there's serious value, then some competition can occur. But like the best thing to do would be the founder like somehow fit all these different valuable players into their cap table. And the most sophisticated one would be like, I'm going to make you all work and fit into this somehow, and we'll take it all. So I feel like if I feel like I'm competing or like rushing or trying to fight to get it, I'm just not interested at all. I'm more interested in being, well, one, as early as possible by thinking about the companies really early so that there's no competition at this point. So I don't want to enter the high competitive investing situation. And if it is in that way, then it's up to the founder. The better founders will find a way to make everything work out with talent. Talent the other way, hopefully, right? Hopefully they're taking investors that are talented. So main point is do not compete. Find a way that you become so— that the trade becomes so valuable that there's no competition for this particular trade that's about to go down between the investor and the founder. And then there wouldn't be any competition.

What value do you bring to the companies that you Yeah.

So I like seeing the shape of a company in the far future of what the company can grow up to be and trying to architect that with the founder and lock in to what they've got. And they may have a better shape in this area. I may have considered this. And there's all these building blocks of the adjacent squares and the markets to win, the markets not to touch. Some of them minefield, like, don't do that bit. Like, that's probably going to ruin your company. So there's us like figuring out what this company is going to look like and then running backwards. So I think that's good because it kind of sets us on a 10-year horizon. And by the way, I think with AI, Like, and disruption right now, it's a bit, it's the fog of war is higher, right? So we're both trying to look at that. And then there's practical things of like, the initial starting conditions matter a lot. So looking at some of the things we need to do to get moving very quickly, which might be, hey, you've got a great market position and great talent, but the quality is not quite there. Can you redo everything you built to like a standard? And let's just see if you can do that. So helping them calibrate on quality thresholds, helping them adjust. It might be even earlier where we've got to adjust like the positioning of the company, of what it needs to look like to work. Then there were pragmatic things like helping them with Series A introductions. That one's like almost the most boring mechanical thing, but it's still like a founder needs that. And introductions are still the way of doing it. I think strategic problems that are like niche are very interesting to me. Like, we've got this way to go or this way to go, which way would be better? And discussing that with them. So I think figuring out the gameplay against competitors is also interesting. So the competition's there, they've got this. If you do this, You completely trump them on this part. So, yeah, I think there's many different aspects. Yeah.

Do you find time to do this exercise with every portfolio company?

No, I think some don't need it. Like, some founders are so good that they kind of figured out really good answers by themselves, and they might need help on this one esoteric problem they can't, like, figure out. Some you invest in and you get buyer's remorse, and you're like, there's nothing I can do to help you win this game. Even if I, I'll basically be running your company for you. So, if you help them, you're kind of wasting time versus, like, all the other time, like, your opportunity cost of your time could be spent on even a company you've never spoken to yet, because there are other teams out there in the future who are gonna do big things. That's a shame. And, but you can help them, but it's, like, it's kind of like a lost cause. So, you've got, there's, like, a lost cause, and they can blow up to try and take all your time, right? Because, The worst lost cause situations, you will go and run their company for them, which is like a terrible idea. There's stuff in the middle. The stuff in the middle needs help. And then there are ones that are so valuable in your portfolio that if you just help a little bit more, you compound their network effect and their power law scaling even more. So, I think you can help the power law ones. You can help ones in the middle, jump into like the special companies. The ones down here, Mitigations, like, really doing it for, like, human reasons, so they protect themselves and they don't waste any more time. But it's like, you have to timebox it all as well. So, everything from my side is, like, minimal moves possible, and, like, staying away from certain problems, zones that I'm not gonna add too much value to, or it's the CEO's job. Yeah, it's quite ad hoc, really. It's not like, you know, it's pretty ad hoc. You've got, like, requests coming in, you've got ideas happening. And then it's like, it's happening. And sometimes it's down to the communication style of the founder. They might get a lot of help just because they happen to connect to you on WhatsApp. So a lot of it does have these like statistical random oddities in terms of like, it's not an optimal—

They get more attention because they're on direct channel with you.

It could be a style thing. It could be that you just connected to them better and like, it doesn't necessarily warrant it. It's not even necessarily— yeah. So there's all these different layers to it and it's just natural. Like, you know, we don't— I don't want to be too robotic about things. And, you know, you might— serendipity is allowed as well. You just never know. And, and you might be helping a failing founder who goes on to build a unicorn later. And I've seen— I've been involved in that before. And they, you know, they didn't do too well, they're failing, took 10 years, came back. Now they got the best MRR chart I've ever seen in my life. So surprising surprises do happen, right?

So where do you see yourself as an investor? Are you going to raise funds or you'll keep running like an angel plus syndicate?

So I think if you're raising funds in certain modalities, you can't say if you're raising funds. So I couldn't answer that question either way. So couldn't confirm or deny it. But I definitely, I have been deploying my money and other people's money with the rolling fund. And I want to scale up that. And I do think I can put a lot more dollars to work into the companies. And there's also interesting problems that need to be solved, like reality security. Like you could take, make a video that's pretty realistic now? And where is our safety on this? And watermarking doesn't work. And I've got a specific spec on this idea and I want to see it built. So these problems, this is where I'm like, I want to put time and money into this. And if no one's working on it and it's like the biggest thing, fine, I'll go do it myself if I really have to. And if not, then we— but some problems are intrinsically exciting to me because I'm learning a lot at the core. But yeah, I will be increasing dollars. I mean, to be honest, In a magical world, if you said, here's $1 billion per year, go invest it, I would go do that and just go deploy it. And it'd be like, you know, throw all these rules away from like all these classic GP-LP rules and all these other things they're looking for. And just like, just go, just go like find the best talent in the world. Literally, if they get their visas sorted out, pull them out from wherever they are and just build from the list of things that we need to, need to get done. Like, you know, here are like, you can imagine like a giant project of the world. I mean, it is happening, right? It's just a messy, it's a very messy organization. I'd like to see a bit more organization around the problems that we need to solve and alignment of like, this is a really good play. Because a lot, there's so many investors now, right? I don't feel like they're really talking to each other as a big group. I mean, you're setting up your giant network, which is a cool idea to create more alignment so that people aren't wasting, like backing 6 companies that do the same pet food subscription business as opposed to just like 1 or 2. And yeah, there's a lot of overlap in There's a lot of efficiency to be gained between all this money that's sitting in pension funds and like just stuck in cash offshore. And like 2 grads, like in a country where, you know, $20K per month can get them paid enough to build something really cool. I think that that's very interesting to me how it's like you've got offshore cash sitting around doing nothing at all. And you've got people that if they just have a little bit of cash, they can build something really cool. So we got— I want to see that gap closed. And You know, we could have like 100 million founders around the world. I mean, almost everybody, every kid could be like, you're going to be a founder. You know, that could be really cool. And if you want to do science, that's fine. You can do science. Or I just think we could do a lot better job here of ambition, especially in the West. You know, like what people wanted to be lawyers and doctors before, and we still need doctors. Not to be rude to the lawyers, but— What is that aspiration now for like the next generation of like, what do you want to be? You want to be a scientist? You want to be a So I think that in an AI world is an interesting question. Like, what are you going to teach your kids? And that's AI disruption-proof and career— like, what is the career for them? Like, you know, is every kid supposed to be like a foundational model researcher right now or like deep mathematician?

What are your thoughts? Who should they be?

Yeah, I mean, it's very tricky to know because you could say mathematics is going to get solved by AI. So if you say go deep mathematics, You know, maybe everything gets printed out by AI. Then it's like, what problems remain where like the human is going to be really useful? And, you know, with robotics, like I can imagine every single task in that I can imagine in robotics being automated. Sorry, in human labor being automated. I can imagine a machine for every single thing. And then it becomes a cost competition. It's like, you don't want to be minimum wage. You don't want to be a cost competition. And then I think of all this like analysis that people do Even decision-making and reasoning. And I think that's just getting better and better and better. So all these different, what were like blue, you know, white-collar jobs will, you know, can basically go away. The gold-collar jobs, say the knowledge work, is going away as well. So it squeezes everybody up to like higher and higher reasoning potential and designing systems that are not yet, not yet done. So I guess everyone at that point could be like trying to get really good at automation. So if we're going to get, if people are going to get automated, well, like maybe you get trained on how to automate things, which might accelerate it even more. But if you're not on the side of that, then you're on the other side. So maybe if a kid is like, yeah, we're going to teach you everything about automation, maintenance might be an interesting problem as well, like maintaining systems and dealing with scaling things up. And yeah, it's a tricky one. I don't know how this all plays out.

Even robots can maintain robots, right? It's kind of, in a way, it's all can be cyclical and you don't need humans.

Yeah. So there's a question of like, I mean, if the human brain is super Turing or operates on compute level that's above even quantum computation, there may be a whole class of stuff that we can get involved in. But that still seems very specialist. And it becomes an education problem as well, because if you need to educate everybody to be at a very high level to be able to have a job, then we got to be like, we got to start doing that education now. I'm not sure. I think we may have to, I mean, we should certainly create as much abundance as possible, lowest pricing that we can through automation. And I don't know. This is one of these big questions of like, lots of people have strong opinions. So there's going to be no jobs, there's going to be loads of jobs. And I think we're dealing with like a super complex system that's like, you don't know how this is going to— it's like looking at bacteria inside a jar and be like, I want to predict the shape of that bacteria once it grows. I think it's like a fool's errand to try to do too much prediction on it as opposed to like specific mitigation strategies or specific things we can do. Like every kid should be learning physics and math to a pretty high degree, in my opinion. Why not? It's kind of like a travesty if they don't get to see this cool stuff of what's under the hood, what's inside a processor. Let's do a processor today. And the teachers don't know either. So it's like, how do we get the teachers knowing this stuff? So that's my biggest gripe in a way of we can talk about all these job losses, but we don't talk about— once again, looking at the output issue, not the input issue, which is like, why don't we obsess about does a kid know math and physics to high degrees? and chemistry and biology, and can they keep up to date with it? And every single child, not like, oh, the kid's bad at math. It's like, no, bring them up. It's like, bring them up and get them really good and find out why they're bad at math. And like, you know, let's speed that up. So that's the bit that I think could be good to focus on to say, you know, you've got a billion people to like super educate to the highest possible degree. And that is maybe an insulation because everything else is going to be patched up. It's going to be redistribution of wealth. Or I mean, abundance is great because prices go down, but like if it's a competition situation, prices won't go down even if it could go down. So like the antidote here is like massive, massive volume education to the highest degree. Maybe the kids will do it themselves because they'll be super like native with AI and like doing it on the fly, different from how the older people do it. This is hard to know, but I am pretty optimistic in that I feel like the abundance will happen from all this stuff. We're going to learn a lot more physics and math and chemistry and biology is going to be solved. And so I'm like super, super positive about this. And space expansion is interesting as well because that opens up this like whole entire value chain of like undiscovered stuff. It might be that humanity hits a steady state, right, of population. It might just be like that is the natural maximum that this ant colony, you know, gets to. And that's, you know, we hit the 12 billion people and it's just in a steady state and then we optimize from there. I don't know.

What do you do for lifestyle? Kind of work lifestyle balance? How do you spend, you know, your free time?

Yeah, free time. So I like to learn about physics theories. I like to know about— so I actually spend each night, almost each night, watching lectures on physics because there's a lot of lectures out there and you could do 365. You could do 365. Like, there's thousands of lectures out there. So I watch like lectures on biology, or physics or math. Like, sometimes it's abstract math. I like content, so I watch, like, I will watch, like, all the movies, all the content to know that full dataset. And I actually, I spend time with my kids, like, doing teaching and building projects. And I have side projects as well. So, I feel like I have, like, 10 human life demands, you know, 10x, like, stuff to Every single minute. There's just like so much stuff to do. Exercise. Well, I think exercise is important for me, and I surf as well.

You have time for that? Okay, cool.

You have to make exercise work because it's like a basic, basic part of just maintaining your mind and body. Um, so I actually think you get better results if you exercise as well. Um, not just like health and longevity, like, you'll know, you know, you'll live longer, so you'll get to learn more things and see, witness all the cool discoveries. And you'll also be able to think better as well. So it's worth—

And great ideas come when you exercise or, you know, you're detached.

Yeah, walking and disconnecting allows you to free yourself from some of the constraints. So you get good ideas, hopefully good ideas. Yeah, I think just trying not to work all the time, basically, or at least the work should be your play ideally. And then it's like, there's not much difference.

Combine the two.

Yeah, but keep keeping busy, and that's not being idle with doom scrolling or any of that stuff. It's like, you know, don't just sit there, actually make stuff. I think making stuff should make up like 1/3 of your day all the time, no matter what. Yeah.

Great. Thanks for joining, Jude, and sharing your story. Lots of amazing insights. Not everyone is like you. You know, it's kind of It's quite a unique mindset, I would say.

Yeah, thanks, Yuriy. It's great to reconnect after 15 years or something. Yeah, I think the main takeaways are like, just learn as much as you can. Like, just be curious and just like keep reading about the mysteries that you're interested in and get to the bottom of them and make stuff for other people. And, you know, yourself, if you make it for yourself, there'll be people that will probably want it as well. All right, well, it's good to talk.