From $200 Million Revenue Founder to Frontier Lab with Henry Shi

0:00Hey, everybody. This is Ben Kesnoka, co-founder and partner at Village Global Global., and this is a network driven venture firm. tech app, and he scaled it to a billion in GMV, 200 million in revenue, 50 million users, and profitable. And then when things were going great at the end of 2024, it was, after all, super. He decided, he intuited that there was an inflection point going on. So he stepped back from operating and took some time to learn out loud and to really think about what might be working in AI. And he created artifacts, which we'll link in the show notes, which are really actionable today.

1:08And that's the AI crash course with 5 ,000 GitHub stars, plus the Lean AI leaderboard, and then the Lean AI company playbook. And we're going to talk about all of those. But then after about nine months, Henry came to us. He messaged us and he said, I know my next chapter. And there are a few things that spark joy for early stage investors, more than a very talented founder saying, I'm thinking about my starting something new. But Henry surprised us. And he used the F word, frontier lab. And he ended up joining Anthropic. And today, we're going to talk about all of that, what he saw, what he learned, what it's like at Anthropic, and what this means for all of us, whether founders, investors, or anyone building their career in the age of AI.

2:02So welcome, Henry. Thanks for having me. Wow, that was a very great, candid, kind intro. So thanks so much. And yeah, happy to be here. Happy to share everything I learned building the company, exploring the future, building AI, learning about AI, and how a company is going to be in this AI native era. So happy to have this conversation. Great. Great. Great. Well, I think you always have lived in the future. So let's take a step back to 2016. And can you share the origin story of super.com? Yeah, sure. That's great. So that was actually exactly 10 years ago in October 2015. I quit my job at Google.

2:40I went to Waterloo, studied computer science, been an entrepreneur for 10, 15 years, always loved building companies. I worked at Google for a year after graduating. Everyone, I guess I thought it was the dream job when you're a kid. You're like, wow, if I work at Google. In many ways, it was great. A lot of smart people, a lot of talented people, but also just a lot of bureaucracy, a lot of red tape. It was a massive company back then. And I realized that a big company just wasn't for me. It was hard to get stuff done. I had four different managers within a year. It took me eight months before I shipped anything to production.

3:11So really, I was like 22 at the time and I wanted to build. I wanted to get back into startup. So I my stock bested October 2015 and started a company. It was just my co-founder at the time, so we had no idea, no industry, no vertical, just people. We wanted to work together, and we're going to figure stuff out. So we iterated, talked to customers, the whole lean startup methodology, ended up in the conversational commerce space back when chatbots were all the rage, messenger chatbots, the first era. This was before chat, now it's coming back, but But the first area, and then very quickly learned, chatbots were okay.

3:48It's good for people like you and I. If you're busy, you want to spend time. You want to spend money to save time. But for most people, they have too much time. They'd rather spend time to save or make money. And chat wasn't necessarily great for that. So we focused on what people wanted, which is just saving money, good deals. And we said, let's get rid of the chat. Let's focus on what people want, good deals, good savings. And we focused on, at that time, travel savings. and then eventually expanded to other form savings and launch your own card, credit building, cash back, the whole gamut. But that's kind of how we got started originally.

4:18Amazing. Amazing. Okay, so now let's fast forward to December 2024. And you said after being a founder for eight years, you saw an inflection point. What did you see specifically? One thing is with my company, just seeing that journey going from zero to over 200 million in annual revenue profitably, having a leadership team, having an exec team. I've pretty much done everything within the company at that point, product engineering, design, strategy, growth, execution, OKRs, M &A. The whole gamut of things, 0 to 1, 1 to 10, 10 to 100 and beyond and back again. So I had a great experience. Honestly, if I were lucky and privileged to have been on that journey.

4:57But at the same time, kind of realizing that AI was coming. We were actually super early customers of Anthopic back, I think, in 2022. too. And I remember at that time I was telling my team, hey, like we had a shared Slack channel where they were doing custom work with us with researchers. And I remember telling my team, saying, hey, this company is going to change the world. We should take this very seriously, that they want to work with us. And I guess it was a little early for them. We were busy with our OKRs. Nothing really happened, unfortunately. But that kind of just made me realize that you had to really kind of be in it, to understand it, to prioritize it, to believe it.

5:35Otherwise, it's just going to fly by you. And by that time, the company's doing well. My co-founder was running the company, and we have a good team. So I figured it was an opportunity to take a step back, let the team shine, continue to grow. We had very well diversified revenue streams now, multiple use cases. So the company's going to keep going and scaling. And for me to be a little bit closer to some of my technical backgrounds or roots as well and spend some dedicated time exploring AI in this new wave. We give you a lot of credit because it takes courage to go untethered because you kind of like took off for a gap year to learn out loud and stuff like that.

6:14And I'm just curious, how did you approach that? Because many people travel, they do sauna and cold plunge. How did you structure your time to be so productive? You were so busy and productive. Yeah, yeah, thank you. And it was great meeting, getting to build a relationship with the village folks while being part of all your events and community and hosting things together. I think for me, I just love learning, and I get a lot of joy from that. So being on a beach, I think I get bored after a week, or not even a week, maybe a couple days. So I think I want to just stay busy, stay top of mind, getting the opportunity to explore and do things that I otherwise couldn't really explore.

6:54And I think when you're a free agent, you have a lot more freedom and latitude to do things. Whereas before I was like, hey, why is this fintech commerce guy doing an AI or whatever leaderboard, right? But as a free agent, you have the credibility of what you built, but also the flexibility and freedom to do other things. And it still makes a lot of sense. And you have the freedom to learn and to build in public and to learn in public. So I think that it's a combination of that. And then also in terms of productivity, I think it's just having the time as well, the freedom, flexibility. You have eight hours a day, actually probably more, 12 plus hours a day to just get into certain rabbit holes to learn.

7:41You didn't have to hit certain OKRs, and you had the flexibility. So I think that was a combination of credibility, flexibility, and time and community. Amazing. Okay, so you did and documented and still out there the AI crash course. So you're kind of a technical savant. Can the mere mortals learn AI in weeks like you did? Well, I wouldn't call myself a savant necessarily. I do have a background, but it's been four or five years since I really coded every single day. I was running my company for eight years. The first four years, I was running code, wrote the first line of code, and continued to do so.

8:20And the last four, it was very much about people, strategy, execution, goals, and just really steering the bigger ship. So I didn't have a lot of time to write code every day. So for me, part of it was getting back into some of the more technical details. I'd say it's definitely learnable, but it takes – and there are a lot of people who put out a lot of great resources to help. So what I did was just basically as I was learning, one of my friends recommended that, hey, it's actually good to learn in public because, one, it's an accountability mechanism to make you share what you learn. And you better be on top of what you're sharing.

8:59Otherwise, if you don't know what you're sharing, then you kind of look stupid. So if you're learning in public and sharing what you're learning, then it just forces you to truly internalize things. Okay, so then you also started tracking some great companies with the Lean AI leaderboard. So let's talk about the criteria for that and then some case studies. Yeah, so the Lean AI leaderboard is this leaderboard I built, I think, in March of this year, when around that time you were seeing all these tweets on Twitter around people saying, oh, Cursor went from 0 to 100 in X number of month with 20 people, Mercor, you know, 0 to 50.

9:36and all these crazy fast growing companies scaling very quickly with very lean teams. And for a long time, I had always believed, and many founders believed this, is the most fun stages of a company are early stages, like sub-50 people, when everyone's in a room just getting stuff done, super aligned, and you have that camaraderie. And then as you grow and scale the company, it just becomes a lot of people problems. He said, she said, and sorting out interpersonal dynamics. mix. So now with AI, if you can grow quickly and profitably with a small team, what does this new trend of company building look like?

10:11Do you still need traditional venture capital? Do you need to hire all these people? And how do you scale quickly and get real scale and revenue, right? So it was very fascinating to me. And I thought there must be more companies than just the five names everyone kept talking about, Cursor, Lovable, Mercore, et cetera. So I did this research to find all these companies who are super lean, growing quickly. And then I was asking my friend who, and kind of as a joke was like, what do you think investors care about? And he's like, you know, I don't think investors care about anything except for beating other investors on a leaderboard, like the Forbes Midas list and things like that.

10:43So I thought, okay, well, we have this list, we can gamify it, turn into a leaderboard and launch it. So I think I was actually on a ski trip. And then that night before, I was working on this thing. And I found a domain for trying to figure out what to call this thing. And I just lean in AI leaderboard.com and launched it. And it became really viral on social media and kind of got millions of impressions. And it just tracks companies who are going very quickly under five years old, over 5 million AR with less than 50 people. And the gold standard is over a million AR per employee, which is multiple times higher than traditional SaaS companies or even software companies.

11:22What did you learn about the types of companies? Like what causes, what enables that? A lot of these companies are just AI native AI first companies who've started, been founded very recently, and the founders are taking a very AI native approach. So instead of throwing bodies at the problem, they'll throw AI automation solutions at the problem. They'll build a lot of workflows in-house to automate things, just try to stay lean and nimble as much as possible. And I think a lot of people ask, what is defensible in this AI era when everything or many things are wrappers? And I think maybe, personally, I think one of the only defensible things is your speed of execution.

11:58Everything's moving so quickly, constantly adjusting, so you just have to keep testing, iterating, learning. And by having a small, nimble team, you can move a lot faster than having bigger teams. That's actually a competitive advantage. So you see a lot of companies purposely try to stay lean, try to stay nimble, really hire top senior talent who are generalists versus specialists because they have to learn and relearn new skills all the time. And then also, I think, you see a lot of PLG consumer growth companies, and I think oftentimes they're able to scale a lot faster. And so far, I don't think the enterprise sales, SDR, BDR, AM motion is fully solved with AI yet.

12:39So you still need human bodies for the traditional enterprises. But once you have AI agents and sales agents, I think we'll see more B2B and enterprise companies as well. So let's dig into the generalist versus specialist hiring. because that might seem at odds with being close to the customer. So if you have a lot of generalists, how do you get good customer discovery? I think being a generalist doesn't necessarily contradict with being close to the customer. In fact, if anything, because you have so few people, you have everyone interfacing with customers, right? Whether it's support or feedback or product.

13:14I think because there's so few people to solve all these problems, by nature of that, you're just wearing a lot more hats. You're doing a lot more things. You understand the business function end-to-end, and that gives you a better intuition about what customers want and their feedback and how to improve the product. And the journalists are not junior people. They oftentimes very senior people who just have a lot of experience and have different skill sets. That way you have someone with good experience who can still be a journalist and have to cover a lot of things, understand the business end-to-end, and I think that's what allows them to move very quickly.

13:50Yeah. And on that list, it strikes me as there were a lot of companies that target developers. And then some like Cal.com, which is a direct to consumer app for calorie counting. And so I guess that validates your point is not enterprise companies. It has to be maybe. Yeah, like prosumer, consumer, some infra companies, right? A lot of media companies like Imogen, Videogen. and there's some developer tooling for companies as well. But generally speaking, right now, because of, I'd say, the lack of AI sales automation or enterprise sales automation, I think you see a lot of these prosumer-consumer companies scaling very quickly, very leanly.

14:35But I think over time, you'll see more B2B enterprises companies as well. Let's talk about your journey again. So you're out there, you're doing your crash course, you're building in public, you think about the Lean AI leaderboard, and then you say, today, it's either entrepreneurship, VC, or a frontier lab. So walk us through the three options and what you chose to do. Yeah, I think historically, as a repeat, exited, more successful founder, I think many people felt compelled or maybe even trapped to the two paths of either being another founder or being an investor. And they feel like those two are the only two options available to them.

15:19And whether that's because of opportunity cost or lifestyle choices or preference or even pride in some cases, I think the common path are these two. And I think it's often very hard for founders to say, hey, I want to work for someone again. I want to be caught by the entrepreneurial bug. And I can certainly relate to that. However, as I was exploring, I realized that there's a third path, which I wrote about in my post and content, which is a frontier lab, which is actually quite a little bit different. And I'll explain why. Well, I explained the three options and I explained why the third. So on the venture path, I spent a lot of time on the other side of the table.

16:00I mean, I have a lot of experience as a founder, pitching investors, getting rejected 144 times. So very much an experience on the founder side. Their loss. Yeah. Yeah, and I can't say that it's been, and I think if you talk to most founders, it's one of the things they like least about starting a company, right? It's very much a song and dance. But what I didn't appreciate, or maybe I appreciate it more on the other side now, as I spend time on the other side of the table as a venture partner or sitting in certain meetings, meeting hundreds of investors, and then having entrepreneurs pitch me, is that it's also very much a sales job on the other side.

16:40And some of the stats really surprised me, but once you kind of see it, it kind of makes sense, which is most investors are certainly the top ones. The partners might make two or three investments a year, right? So they have to be super selective, and it's really about picking. And when you're super selective, you kind of have to go for deals that you feel like can return the fund, which is returning nowadays maybe$10 billion or more, right? And oftentimes that means converging on certain types of founders like ex-OpenAI, Stanford researchers who are doing world models for XYZ, right? Because that is the consensus of these big, big$10 billion returns.

17:20And what happens is, but these founders, because they fit that profile, they have many options. They have many term sheets already. So your job ends up becoming selling founders who don't really need your money to take your money, right? Because they already have 10 term sheets. It's like, how do you convince them to take your money versus someone else's? And then the irony of that is the founders who actually need your money, who you can actually help, you don't want to help because they have no term sheets. And if you're going to do two deals a year, are you really going to take that bet on this founder who's maybe unproven, unknown, and has no term sheets and no real consensus?

17:52Right now, there are definitely firms who take a really contrarian approach. But for the most part, I think it's very much, well, this founder fits the mold. They're going to be able to build a$10 billion company. Let's convince them to take our money. And for me, it just felt like very much a sales job, which is fine. But I wasn't sure of something that I wanted to do full-time right now. Maybe sometime in the future. But for now, I felt like I still had the builder energy and drive to do something. And I wanted to help more founders. And doing two deals a year is very hard because you can say no to 998 founders who pitch you.

18:26On the startup side, I also thought about it a lot. I was thinking, playing with a lot of different ideas and trying to test different things. But I think it came down to two things, which is one is there wasn't like a mission or maybe, yeah, just like a mission that I really wanted to spend the next five, ten years of my life solving. Like great companies take a long time to build, even Super.com. That was pretty much a ten-year journey. And so I knew the second time around that it was going to take a long, long time. and I want to make sure that whatever I dedicate myself to, it's something that I really, really believe strongly in the mission and the cause.

19:01And there wasn't that strong mission or cause just yet. And then two is I don't want to start up for the sake of doing a startup. I mean, you can do another B2B AI SaaS wrapper startup, and that's fine. And even these days, you can get it to 5, 10 million AR, and that's probably decent. But for me, my company is still going. I'm on the board. I'm fully vested. Companies continue to grow in scale, looking to go public in the next couple of years. So for me, it's like you can make an extra 10 million AR here, or I can just tell my team to do it, and they can do it and grow from 200 to 300, et cetera.

19:35And that's probably more impact and equity value. It's never been easier to go from zero to five, 10, even 20-plus million AR, but it's never been more unclear who's going to go from 10, 20, 50 million to 100 and beyond. maybe look at Jasper AI, for example, one of the darlings of the early Chachipitira, zero to 100 and now back down, and some are stabilized, but it's really hard to know who has staking power. And even some of these companies and the headlines you hear, like, are they margin positive? Are they margin negative? How much money are they losing through inference, through model costs?

20:08So it's really hard to know who's going to be around and who's going to get replaced and who's going to get eaten by the models. So it's very uncertain. This actually relates somewhat to this new type of funding called seed strapping, which I guess is a term that maybe I didn't coin, but popularized, which is this new form of starting a company, which is you take a little bit of seed capital and then you scale it and you use that to scale it to get an escape velocity. So a combination between bootstrapping and raising a seed fund. But it's unlike bootstrapping, you're not paying out of your own pocket in the early couple of years and you can play with someone else's money and get to growth.

20:48But unlike traditional venture, you don't have to keep raising serious A, B, C, D, E, F, G, get diluted to like 2 % and lose control and have to deal with all these like board dynamics, et cetera. So it's kind of a little bit of best of both worlds. Before AI, it wasn't really possible because it was really expensive to build product, to bring it to market, hire all these people. So you needed lots of, and you couldn't grow revenue so quickly and certainly not profitably. So you needed to keep taking capital. But I think if we look back, a little bit of irony is, do we really need to invest$500 million to make a B2B SaaS company?

21:23And now, especially with AI, you can certainly build all these things and get to scale. And because there's so much demand and willingness to pay for AI products, you can get to market much quicker and scale, make revenue and be profitable. So if you have that, then the question is, what is the role of traditional venture capital? So are there new forms of funding that can enable these types of seed-strived companies? I'm sure they may not be$10 billion companies, but I think for a founder, if you're like five people making$10 million a year every year, that's pretty good. And you're probably doing better than most venture-backed founders who are illiquid, who are stuck.

21:59And whether they can or can't raise, but it's just tough and you have total flexibility and control. So then are there new capital structures that can help scale these types of companies, whether it's like revenue-based financing or things like that. So these are some of the things I've been piloting with and I know we chat about at Thistle Village as well. Can you back founders in the earliest of stages in some sort of founder-friendly structure that has early DPI and returns? It's not uncapped, but you're getting returns, you can recycle that and then help many more founders than just the two founders a year because you have to return the whole fund versus here, you can help a lot more founders and help them get to seedstrap Seed-strapped escape velocity while still having good returns for the fund and early DPI.

22:44Right. And there might be optionality because some of those businesses might actually end up being huge decacorns or something like that and decide to raise following capital. So, yeah, we love that structure. Okay. But you rejected those two options of VC and founding. Yeah. So after going through, spending about half a year thinking through learning, building, experimenting, I sort of came to the thought that if I'm not going to be an investor or start a company right now and I'm going to do AI on the outside, may as well do it at the inside, at a frontier, on a good mission, make good impact with great people, great talent and city, and really take my experiences as an entrepreneur and builder and apply it to a good cause and push it forward.

23:26And I think historically, this opportunity probably didn't emerge for many founders because there was nothing of such consequence or scale or excitement. I mean, maybe Google in the early 2000s, maybe. But I mean, really, this is maybe one of the biggest inventions of our lifetime. And many people say maybe the humanity's last invention in terms of AGI. So, yeah, it certainly is very powerful. and I kind of saw the capabilities of these AIs developing exponentially. And I decided, you know what, if I'm going to do it on the outside, may as well do it with a good team on a good mission. So I got in touch with some of the folks there because we were super early customers.

24:11I had a friend from Motolu who was like an early researcher there. I've known Ben a little bit through the startup community. I was just chatting a little bit and they were like, hey, we have this zero to one prototyping role. It's maybe good for founders. And I was like, sure, happy to roll on my sleeves and do whatever. Like I didn't need to do X, Y, Z or manage my team. In fact, I'd probably prefer not to because having gone through that, just having managed hundreds of people, you really kind of appreciate the fun simplicity of just building. So we're chatting, and I think we were chatting through that process.

24:47And I think it was a little bit like I did pretty well. and then also they kind of realized I'd done a lot. So, hey, maybe you can just help Ben directly and help across the org and bring some of that founder energy and experience and just help execute across the org. So that's kind of the opportunity. But Anthropic is very much very disciplined and pretty fair in that process. So I think everyone more or less goes through an existing process and then through that process, eventually kind of maybe people figure out certain roles. But yeah, overall, it was pretty fair. and discipline process, and I'm pretty fortunate to be able to make an impact on the inside at the frontier.

25:26Well, and it's amazing because the talent density there is incredible. Sometimes the flip side of talent density is a lot of egos and a lot of independent thinking, which isn't always good for alignment, and I don't mean AI alignment. I mean, you know, human alignment. So can you tell us a little bit how do decisions get made, how do priorities get set? Yeah, yeah. But so coming in, I think one of the first things I realized is the mission is very much real. You definitely feel the alignment towards the mission, but really care about it and take it very seriously. And on the outside, I think there's a little bit of skepticism around like, are they just saying that?