Internet History Podcast | Chris Dixon of a16z

And, uh, at the time I joined, there were only a couple of investing partners and there's only one fund. Now we have this complex of funds. And, um, you know, I think they were probably looking, and I was a reasonably prominent angel investor blogger. They were probably looking at people who had been doing it for some time. I don't know exactly why, but I think in the end we just hit it off.

Thanks for having me, Brian. You and I have spoken many times, and I feel like we've gotten some of your beginning before, but let me go back to the very beginning and ask you about becoming a programmer. What were your first encounters with computers, and how did you know that this is what you wanted to do?

Yeah, I mean, I think this is a fairly common thing in tech, but as a kid I got a computer. This is back in the 1980s, and I just wanted to make video games and enjoyed programming. It was kind of a cultish, niche activity back then. I really enjoyed it. I think we'd say today I got into a flow state, just designing stuff and figuring it out. I was doing C programming and assembly language programming, making graphics and video games, and was really into it. Then I was in college in the early '90s, and at the time I felt like computing had gotten too corporate, sort of before the internet had really taken off. The computer industry moved from a hobbyist phase to Microsoft Office and Excel. So I actually studied philosophy in school and got into cognitive science, a little bit of what we call today AI, but more philosophy of mind, cognitive science, logic. I didn't really know what I was doing; I thought it was fun and didn't have a sharp career idea. But when I was in New York, if you know how to program and you're in New York and it's expensive, you end up doing various programming jobs. That eventually led me to two years post-school programming at a hedge fund, just to pay my student loans. But then the internet really started taking off, and I got into that. Fast forward, I was in New York and Boston for a while, and I ended up in the early 2000s starting an internet security company.

I'm happy to go into more detail.

Yeah, it was a company called Arbitrate and did options market making. I wasn't looking to work in finance; I was in New York programming, just trying to pay bills and pay down debt. It was the best paying thing I could find. It was a company we'd call quant trading, high-speed trading. I wrote programs that were finance-oriented, like Monte Carlo simulations, big simulations. I would do high-performance algorithms to make them execute quickly. There were about 20 people in our office in the R&D group, and I eventually became the lead of the finance part. I know a little bit about a bunch of things, so I knew enough finance to get by and was a decent programmer. It was interesting. I never had a real job before, just part-time computer jobs. I got a glimpse into Wall Street and how markets worked. It was intellectually challenging, but I realized Wall Street wasn't for me because it was too isolated. You're sitting in an R&D group in front of a screen, just trying to predict numbers. Once I saw the entrepreneurship world, which seemed more exciting, dynamic, and people-oriented, that appealed to me more.

Yeah, maybe to some extent. Although it's funny, because if I wrote a book about crypto called Read, Write, Own, I have one of the least finance-oriented takes on blockchains and crypto. I look at them as a new way to architect internet services. There's a financial aspect, but it's not my main focus. But I do think thinking about finance, money, and networks, and the intersection of how the information economy works, which has been my theme throughout my career, had common threads. Some things going on in finance pre-internet or early internet predicted key themes of the internet later on.

Yeah, I was there for about a year and a half. I was very lucky to get that job. The industry was very different back then. This was 2003 and 2004. Today people think of VC as a sizable industry, but back then no one had heard of venture capital. There were only about 10 to 20 operating firms. I was lucky to be in the New York office and work on interesting things, including the Series A of Skype. It was the beginning of the internet revival after the crash. For someone like me with no significant business experience outside of that quant job, it was a panoramic view of business, meeting interesting entrepreneurs. It was a great lesson. I realized partway through that I didn't have a great promotion path, and I was thinking about starting a company. They let me transition to being an entrepreneur-in-residence, working on the startup idea, and they ended up funding us. Today we hire junior people in a similar role at Andreessen Horowitz, and I tell them it's a great two to three year gig that gives a panoramic view and a deep education in startups and business.

Yeah, it's still there surprisingly, still around and actually a big business unit for McAfee. There's a category now with competitors, but security isn't front and center for most people. At the time, 2003, it was one of the worst times on the internet for security, specifically spyware and phishing. Phishing had really started in the early 2000s, and spyware, which is mostly gone now, involved downloading a screensaver and getting pop-ups. The insight was that past security issues were technical vulnerabilities, like worms or viruses exploiting bugs. But with spyware and phishing, the technology worked as designed; the problem was social engineering. All existing products defended against technical hacks, but the threat vector had changed to social engineering. The idea was a toolbar that would warn you with a big red box saying this is a phishing email or spyware site, like a trusted friend advising you. To build it, we had two co-founders from MIT who were more technical, and we built a web crawler that downloaded and analyzed software, testing websites and creating a classification system for good and bad sites. We spent about a year building that database, then released the product in 2005. Back then, few consumer internet products were venture-backed, so when we launched, people paid attention. We quickly got interest from big security companies like McAfee and Symantec. They visited, and eventually offered to acquire us. It was hard to break through because security products were sold through partnerships with PC makers like Dell or Comcast. The only way to make it work was to be part of one of these companies. It was also harder to raise money then, and there was a different ethos where VCs would often replace founders with professional CEOs. I had those concerns in my mind. We had nice offers, so we decided to sell to McAfee in 2006.

I was naive on the sale. I didn't really know what I was doing, but I had the benefit of two offers. I played a simple game theory thing, going back and forth, and got the offer to almost double. I thought I did a great job. Then I had dinner with the McAfee CEO after closing, and he told me he pre-authorized with the board twice what we sold for. So I hadn't even done a remotely good job optimizing. But it was a good outcome. At McAfee, the CEO who acquired us left a month after, and by the end of 18 months, we were on the third CEO. Management kept turning, which is common at big companies. Our expectations were thrown in disarray. We went in optimistic about transforming the culture, but encountered corporate inertia. I realized the key was to keep the product alive and integrate successfully, hoping they'd be happy. I'm happy it's still around and they felt it was a good acquisition. But it was clear that being innovative and staying on the cutting edge wasn't something we could do there long term. We stayed for our vesting period of 18 months, then left and started working on a new startup.

Yeah, we were interested in AI, and I joke now that I was doing AI too early, way too early. It was a machine learning company, started in 2008. I was involved in an advisory group to DARPA, an academic group discussing futuristic computer science ideas. Everyone was talking about machine learning, and I saw cool demos and got excited. Me and my co-founder from SiteAdvisor, Tom Pinkney, decided to figure out something cool with machine learning. We were eager to do something after McAfee, so we rented an office, recruited two engineers, and started playing around. In retrospect, with the first company we started with a problem and then found the technology. With the second, we started with a solution and tried to find the problem, which isn't a great approach. Also, the technology wasn't mature enough. Neural networks didn't work that well because we didn't have GPU power. They were out of favor. We used other methods. It was cool and worked, but not magical like today. We iterated and ended up with a recommendation product similar to Amazon's. In 2011, we were acquired by eBay to build that product for them. At the time I thought we never figured out the right product, but in retrospect we were just too early. Two years later, in 2013, Google had the cat video demo and deep learning burst onto the scene. It was 15 years too early. GPU computing power got much better. We had a good outcome, built cool stuff, but didn't get to the scale we wanted. It was just ahead of its time.

It's hard to say. You can use ChatGPT or any generalized foundation model, and they're incredible. A lot of what we did, like pasting in your favorite movies to get book recommendations, these models do a phenomenal job. One challenge is that generalized models are so good they subsume specific applications. The challenge today is how to make a product better than Claude, ChatGPT, Gemini, because these models are so powerful and can do many things.

Yeah. One of my angel investors, Ron Conway, a legendary Silicon Valley VC and good friend, was incredibly helpful to me. After I sold my first company, I thought what he was doing was interesting as a side thing. I was excited about startups and wanted to work with my friends starting companies. So I started angel investing, writing personal checks into friends' companies. I wrote my first one a week after the sale because I had money then. I did that for a couple of years. I wanted to be more aggressive. It was obvious to entrepreneurs at the time that there was a mismatch between what they were doing—building consumer internet startups on cloud services like AWS, needing $500k or a million dollars—and the venture market, which was writing $10 million checks built for the '90s capex-intensive model. The insight was there should be firms set up for that. I had two friends, Eric Paley and Dave Frankle, both entrepreneurs who sold their companies. We were angel investing together. Eric was getting job offers from VC firms, and we thought, why not start something together? So we started Founder Collective, the three of us. It's still going; I left after the first fund, but I'm still friends with them. We started it, and then I wanted to do a second startup, so I did it part-time while working on my startup full-time. Dave and Eric did it full-time. We raised money in 2008 and launched in 2009, investing that first fund from 2009 to 2012. In retrospect, that was a good time because the financial crisis meant many VC firms stopped investing, so there was a lack of supply. On the tech side, the iPhone and rise of mobile happened. The App Store launched in 2008, and the golden age of apps like Uber, Snapchat, etc., was 2009 to 2011. It turned out to be a good time. I think we did a pretty good job, but a lot of it was timing. In that first fund, we had Uber, Venmo, BuzzFeed, Trade Desk, and others. It was a great experience. Then I fast-forwarded and joined Andreessen Horowitz for bigger scale VC.

Financially, I don't know if I want to advise anyone to angel invest. It's very risky. I think you have to do it because you enjoy it, like a startup fan supporting startups, or it's a social thing with friends. It's very risky; you should consider it risk capital, not your life savings. For me, I still do it. If a friend is doing something, I want to support them whether it's a startup or a fund. The consistent thread is that all these things are talent businesses. If you know someone well and have faith in them, it's a good bet, as opposed to trying to predict tech trends. I learned that from Ron Conway; he's very people-focused. When I started SiteAdvisor, he liked me, but then he introduced me to security experts who gave negative feedback, and yet he still invested. I asked him why, and he said he just had faith in me and doesn't think about trends that much. There's wisdom in that because tech trends shift so much. People are much more consistent. With angel investing, it's become a more established category, with funds like Founder Collective, Lowercase, Baseline, First Round, and BetaWorks, which are part of New York's startup history.

Anniversary thing. So there were others but it was relatively small. I mean, we all knew each other. In fact, we were all often co-investors in each other's funds. It was just a very small group and yeah, like you said, it was kind of a social club. It was like an investing club or something. We'd often work together and collaborate and it's just a different thing. Now, it's much more institutionalized. That said, the tech world is much more successful and there are many more successful companies. So maybe it all balances out.