Latest / Cheeky Pint / Marc Andreessen and Charlie Songhurst on the past, present, and future of Silicon Valley

Transcript

Marc Andreessen and Charlie Songhurst on the past, present, and future of Silicon Valley · Cheeky Pint

• Marc (00:00:00): Do you mind if I start with a couple of questions? John (00:00:03): I mean, sure. Marc (00:00:05): Cheeky means what? Exactly. John (00:00:06): Okay. In the context of a cheeky pint, it is a pint you are not really meant to be having. Marc (00:00:11): And when it becomes established, it starts to attract establishment people, the social network. Right, exactly. And in that sense, the downturns as much of a pain in the butt as they are, are probably helpful. Charlie (00:00:21): You go back to banking, you go back to consultants. John (00:00:22): Yes. I'm sorry. Why is there more risk-taking on the West Coast versus the East coast? Charlie (00:00:27): Frontier FOMO leads to high trust, that sort of has a cynical truth to it. Marc (00:00:32): Category two errors are much, much worse by the way they torture you fucking decades because you read about the success cases that you've screwed up all the way up. And so you just learn the hard way. You have to be extremely open-minded. I have found people willing to tolerate any level of chronic pain in order to avoid Charlie (00:00:47): Acute pain. Marc (00:00:48): In order to avoid acute pain, people would much rather lose slowly over five years than have the conversation that involves a dramatic change to stop losing. Waiter (00:00:57): Alright, here you go. Alright, very good. Anyone need anything else? Marc (00:00:59): Finally, a legitimately Irish bartender. John (00:01:02): I have a scheduling issue with these because 5:00 PM clearly after work, pint’s acceptable. 4:00 PM I don't know, after work if you're a banker or whatever, 3:30 PM, now you're just drinking at the office. Marc Andreessen has been around the internet since the very beginning, really. He co-founded Netscape. He invented the image tag. He was there at the beginning and later he co-founded the venture capital giant, Andreessen Horowitz. So I'll be speaking to him along with our mutual friend Charlie Songhurst. Cheers. Marc (00:01:27): Cheers. Do you mind if I start with a couple of questions? John (00:01:33): I mean, sure. Marc (00:01:36): Well, there's just a couple of things. As a Midwestern American boy, there's just a couple of things that this is not my natural habitat. John (00:01:42): Okay. Marc (00:01:42): Cheeky means what? Exactly. John (00:01:43): Okay. In the context of a cheeky pint, it is a pint you’re not really meant to be having. And so if you were meant to be going home right after work and instead you stole away with a few coworkers, just off the books, aren't meant to be at the pub right now. That would be a cheeky pint. Marc (00:02:03): And then pint. The thing about pint that's just really puzzling is that everything else in Europe is like, it should be the cheeky deciliter. John I see. Marc Right. And so why is pint used with reference to alcohol but not with actual measurement? John (00:02:16): Because, I mean, Guinness and alcohol generally is part of a rich tradition. Guinness dates from the 1700s. It's part of why we have, it's the reason we have the canal system in Ireland. It was the largest company in Ireland at one point. Often the longest tenured institutions are universities and breweries and you look at the Belgians and things like that. And so I think tradition survives better in alcohol than it does in real sense. Marc (00:02:44): I have several more questions, but I will suspend them for the purpose of this conversation. John (00:02:48): I like this new format that we're inventing. So where I want to start is we have here various bits of Marc Andreessen memorabilia and a steal from one of my favorite pieces of Marc Andreessen content, your Miller Lite commercial. And it was only in a rewatch that I realized it was with Norm Macdonald. Marc Norm Macdonald. John What was it like meeting him? Marc (00:03:06): The one and only. In my experience, comedians are always a little bit interesting to meet because they're professionally funny and so their interest in being interpersonally funny is not that high. It's a lot of stress and pressure I think. John (00:03:19): I see. Marc (00:03:19): I mean he was very naturally funny, but John (00:03:22): He wasn't always on. Marc (00:03:23): He was not always on. And I mean, I will tell you in context you'll see the commercial. I just say in context it looks like we were in the coolest nightclub in the world. I will tell you it was in the middle of the day in what they call the Inland Empire in LA in some warehouse. And it was John (00:03:38): Like, not as cool as it looks. Marc (00:03:39): It was like 110 degrees outside. It was like 130 degrees inside. There was no air conditioning because it would screw up the sound. And then to create the smoky nightclub effect, they spray vegetable oil. John (00:03:56): Not water? Marc (00:03:58): Vegetable oil. It has to linger. The director was great and he was tremendously tolerant of me with no actual experience doing anything like that. But I think he did think he was Stanley Kubrick because we did like 150 takes. John (00:04:08): He was reading into his Miller Lite Act. Marc (00:04:09): And so, like, hour six of nearly passing out from the heat and choking on vegetable oil was not the most … Oh, and then the other great claim to fame is, it was a week later Miller fired their ad agency, which I would like to think I bear some responsibility for. John (00:04:31): Okay. So the thing I want to get into with you, about or spend a lot of time on, is the history of the Valley. One interesting place to start might be… If you think back to even 2021? As I think back to myself? Yeah. Is it apparent to the time? Marc (00:04:49): So my experience is no. And the nuance that I would put on that, I'll describe two reasons. Number one is there's an old line with respect to economists that also applies, I think, to investors and entrepreneurs, which is economists who predicted nine of the last two bubbles or nine of the last two crashes. And so it is extremely common. It's a difficult question because it's extremely common for people to call a bubble. When they're correct, they will then go around for years claiming that they're the one who called it. What you find is that those people generally were calling it continuously for the 20 years earlier. John (00:05:20): Peter Singer. Marc (00:05:21): Peter, for example, or earlier, there was a famous [magazine] Barron's. It's still around, but it used to be an extremely important investment publication. There was a columnist for Barron's, something Abelson. Alan Abelson. And literally, he wrote the same column for 40 years. ‘The end is here, it's all going to crash. It's all a giant bubble.’ And he wrote that, I think, continuously from, I forget the exact years, but from 1975 to 2015. And so, you have this kind of Cassandra thing where they kind of dine out on it. And so, I find generally that those kinds of people don't have predictive ability. And then I will tell you, look, the most sophisticated hedge fund managers in the world, generally, if you look in their backgrounds, at some point, if they thought they were in the macro business, they will have tried to make the trade based on what they view as obviously a bubble. And there were extremely sophisticated hedge fund investors. They went short tech stocks in the fall of ‘99 and then realized they were wrong and then went long tech stocks in Q1 of 2000. John (00:06:15): Druck has he talked about a publication? Marc (00:06:17): He's talked about it, but there are many others. Well, there's another guy who I won't name who's very active today, who's very smart. And I was talking to him on the phone about stuff and he just started laughing, and he said, he's like, “All I know is whenever I think the stock market is going to go up, it goes down and vice versa.” And this is like a guy who's, like, an investing legend. Charlie (00:06:32): What, when the bubbles started to burst, when was it obvious, in retrospect, that that was, is it 2000, 2001, is it only 2004 when you look back? When does …? Marc (00:06:43): No. So the sort of cliche, which is correct, is the market climbs a wall of worry. So what happens is when the market is rising every step of the way, there's some panic attack going on about ‘it's immediately going to collapse.’ And then what happens is there are drawdowns, I'm sure you guys will see the drawdown charts are really fascinating to see. Charlie (00:07:02): There's a big one in 1998 with the Asian crisis. Marc (00:07:05): So we all thought that was it. This is exactly what I said. So yes, there was a blowup in ‘98. There was an international crisis, and then there was a collapse of a big hedge fund at the time called … Charlie (00:07:14): LTC … Marc (00:07:15): Long-Term Capital Management. John (00:07:16): I read that book recently. It was really … Marc (00:07:17): Good. It's a fantastic book. It is a great lesson. And do not name your hedge fund, “long-term.” John (00:07:24): I thought the lesson was don't run 30 times leverage on the one trade. Marc (00:07:27): There is that, and also assume that academic superstars necessarily have a feel. Yes. But yeah, a lot of us said that was it. That's it for IPOs. It's over. That's it. The whole thing is going to cave in. So every step of the way, and then, conversely, we all got so used to it rising that there was a lot of speculation, I would say the median view among smart people. And when the Nasdaq first cracked. Sort of around March of 2000 was, oh, it was just another one of these momentary blips. And then the way I remember it—we'd have to look at the chart—but the way I remember it is fundamentally that, from 2000, 2005, there were five discrete moments where it fell apart. It kept cascading down. And my favorite version of the story is, we took our company, LoudCloud, public in September, 2000. And while we were on the road, we were on the road for three weeks. And while we were on the road, the NASDAQ fell in half. But that was just one of those things. So the answer to your question is, I'll put it this way, by 2003, 2004, you knew that it was really bad. Then what are the indicators, the indicator that everybody really knows it is, the longs all get fired, they lose their money, and then the PMs actually get terminated. And until that happens, there's still, I would say tremendous amounts of either uncertainty or, you could say, denial. Charlie (00:08:36): One of the great years for owning internet stocks was 2003 because you get the bottom and then you get this huge uplift, I think in eBay, Yahoo, maybe it's 2004. Marc (00:08:44): Sure. Charlie (00:08:45): But VC wasn't good through that entire period up until like ‘07…I would just say, look, you could maybe say this. You could say the entrepreneurial ecosystem got completely flattened by ‘03, ‘04. The idea of starting a company was ludicrous. Charlie (00:09:31): Got it. Marc (00:09:32): And so … Charlie (00:09:33): It, maybe, created too much fear in potential entrepreneurs. Marc (00:09:35): Yeah, that's right. And then look, the VCs panic. I say this, one of the cardinal sins you get in to venture is you're actually paying attention to what they're saying on TV … (00:09:45): And particularly on the financial news. And so it's in the NASDAQ cracks, it's very hard to keep yourself out of that psychology and to be enthusiastic about making an investment. But of course, if you're a VC, the rational thing to do if you're a VC, is to keep … so Fred Wilson's the guy who kind of really walked me through this originally, and he said, “Look …” his version of this would be, “Yeah, bubbles burst. It's all random and crazy and we never know what's going on in the whole thing. And you get wrapped up in psychology.” And so his rule of thumb always was you have absolutely, you have a disciplined mechanical process for the pace of investment and then also for the pace of exits, and you don't deviate from it. And a lot of that justification would be precisely so that you keep investing at the bottom. It's so funny, you see – this in the stock market. Everybody says, “Oh, buy low, sell high.” Everybody's an expert in bubbles. Everybody's read the books, the whole thing. But when the market has caved in, it is just, well, it's actually really funny. Like negativity. It's like just overwhelmingly, “You people are idiots. This whole thing is stupid. It's never going to recover.” There's 18 macro explanations … . And then, actually, at the real bottom, the other thing I found is people just completely stopped talking about it. Charlie (00:10:50): Yes … Marc (00:10:51): The idea of startups … Charlie (00:10:53): Crypto markets are a case study in this over the three cycles. Marc (00:10:55): It's just like it never even existed. It's just like, it's like the thing you would never bring up at a dinner party. And maybe to your point, that's what happened with the internet startups in 2003, 2004, which is you would not talk about it if you could possibly avoid it. Charlie (00:11:05): So in some ways, the social status of internet startups in oh three is similar to crypto and in like, 2020. Marc (00:11:11): Yeah. So the joke of that time was the two great kind of great VC trends. Startup trends of the late nineties were so-called internet companies, but B2C business to consumer and the B2B business to business. By 2003, the line was B2B meant back to banking, and B2C meant back to consulting. And so, oh, and this in turn is why you'll enjoy this a great deal. This in turn is why the employment decisions of graduating Harvard and Stanford Business School students are such a great indicator, possibly the best indicator of all of what's happening in the market. Because if they go into tech, the market's overblown and if they go into banking consulting, it's a great time to make VC investments. And that may be the best indicator I've seen the whole time because of the social status aspect. John (00:11:56): I think what you're describing is, you don't think you're capable of making macro calls. So you just have to decide what are sensible areas to be investing in over multi-decade time horizons, tech startups generally crypto, American Dynamism and pick your lane, and then you dollar cast average it to them. And then sometimes there'll be bubbles, like there'll be crypto 2021 moments. But that's fine because if you put the same dollars into these areas, I mean rough numbers, but consistently put dollars into these areas each year, the winners will more than make up for the years for everything that was hopelessly undervalued. Is that basically your framework? Marc (00:12:28): I would say that's mostly true. What I would modify is, it's actually not dollar cost averaging. If you're doing it in the stock market, it's dollar cost averaging, if you're doing it in venture , it’s not dollar cost averaging. The reason is, because, if you make the right venture investment, it doesn't matter how much money you put in, the upside is so great. And if you make the wrong venture investment, you lose all the money. Charlie (00:12:47): I think Andy Bechtolsheim’s 100K, I think, would be 30,000x. Marc (00:12:50): What's that, sorry? Charlie (00:12:51): Andy Bechtolsheim’s 100K into Google would've been 30,000x. That pays for a lot of other investments. Marc (00:12:59): In venture capital. It just turns out that the amount of money invested is, it has almost nothing to do with anything. And you, you're not trying, well, here's another thing. You never venture run a bargain shop like ever, ever. John (00:13:09): No, I agree with that. Marc What you need to do, I guess the way I would just modify what you said is just you need to keep investing. John (00:13:15): Yes. Marc (00:13:16): The danger is not investing too cheap or too dear. The danger is literally stopping. John (00:13:20): Sure, but sorry, we're pronouncing dollar cost averaging. It was the fixed amount of money that you deploy. Because I think the way people get into trouble is 2021 comes along and they raise some giant fund and that one has very poor returns. But if you invest 100 million each year, then you'll do pretty well. Marc (00:13:36): And you could also say this, the smartest LPs, so David s Swenson, who was considered to be the smartest portfolio manager for liquid portfolios, wrote a book where he goes through the following, and he talked about this a lot, which basically is for something like venture, you really got to look at it. You cannot rationally evaluate venture based in a single moment in time, a single fund, a single sector, any of that stuff. You have to basically look at it over a long enough period of time where you wash out the specific effects of— Charlie (00:14:00): Well, the proof that is, the inter-vintage volatility in any given VC is incredible. Marc (00:14:06): Right? That's right. Charlie (00:14:07): Which shows so much of it is just structural time. John (00:14:11): A top VC firm will have some 15x funds and some— Charlie (00:14:14): But it's incredibly stochastic because Google's founded in ‘99. So at the height of a bubble, Meta's founded 2004 at the bottom. Marc (00:14:21): Right. Charlie (00:14:22): There's no pattern that ties to macro, it just appears to be almost stochastic. You just can't predict. You've just got to keep doing it. Marc (00:14:28): Yeah, that's exactly right. And I would say that's sort of the core fundamental kind of truth of venture, which is really, it's something for people with a 20, 30, 40, 50 year time horizon, you have to get across. You have to get all the way across the cycles. Because what happens otherwise, if you're an LP, what happens otherwise is the minute you have a fund that's terrible, you pull out and that's precisely when you should have been going in the same behavior on the LP side that you see on the VC side. And so the smart LPs, what they all have in common is when they're making a decision invest in venture fund, they're making a decision to invest in that fund for the next five or six months. Charlie (00:14:56): So how much of an advantage for VC is having good LPs? Marc (00:14:59): Extremely, extremely, extremely, extremely. And again, this is very predictable. What happens is every time the market is hot, new LPs show up and pile in, and then when the market declines, they back out. And so the firms that have the VCs who understand the Switzer model are able to sustain over time and able to continue to invest in the downturn. The VCs, many new VC funds are raised in every bull market from basically tourist LPs. Those tourist LPs are extremely reliably prone to pull out. Charlie (00:15:27): So obviously that leads to the big question, which is how causal are the VCs themselves to the outcomes of the companies? It's the big, big question. Marc (00:15:36): I have a theory on it, but I have an indirect theory on it. But I definitely should not let the entrepreneur answer this question, but I just made an incredible strategic, incredible strategic mistake. This is where it all went south. You can see the look on his face already. John (00:15:52): One, presumably VC itself is very impactful because Stripe was just as a practical matter, not profitable for quite a few years. And I think that was the correct way to build Stripe. And so many companies, you build a bunch of tech and Stripe in particular, you build a bunch of tech and businesses start adopting it and they start growing. So you've two lagged curves. One, is you have to build all this stuff and then businesses start using it. And then those businesses grow themselves. And we just had Tobi from Shopify here. Shopify is now a massive business on Stripe, but they weren't when they started working with us in 2012. And so it's just the classic R and D thing, if you do work now for economic payoff later. And I think that tends to work well in tech. And then with specific VCs, it feels like the Silicon… I want to talk about the Silicon Valley high trust thing. VCs act as a very efficient matching algorithm between neophyte founders such as myself and experienced executives. And so you have this incredible talent engine, and I think in a weird way, people often miss… it's not about the money at some level, people miss that. It's about putting together a team in a very short order to go do this hard thing. And I think VCs are actually pretty instrumental in that. Charlie (00:17:08): I'll back in from the angel perspective, the single strongest correlation of how a company will perform, is how high status the VC does a Series A, is within the stack ranking of VCs, it is far more predictive than sadly than my own selection, or any other available I can find. It's almost deterministic. Marc (00:17:28): And look, some of that is because the top tier VCs can get the best deals, right? And some of that self-fulfilling prophecy. So here's my analysis having been on both sides of the table, mapping, mapping, what you said, my analysis basically is, that if you think about mechanically what's happening with a startup, a startup needs to basically get into a loop in which it's accruing more and more resources as it goes. And those resources are qualified executives, technical employees, future downstream financing, positive brand momentum, public perception, customers, revenue, throw weight in the government, all of these resources that you need to be able to succeed as a business. And so it is, there's a snowball rolling down the hill phenomenon, which is you're either a snowball rolling down the hill, picking up resources as you go, gaining size and scale and scope and power as you go, or you're not. And you're kind of stuck at the top of the hill as a snowflake and you're just not going anywhere. And so the question is, how do you get into this kind of aggregation of resources thing? Economists call this, what's the term for the things that are at the height of the power… preferential attachment. John (00:18:25): A sort of beating of companies. Marc (00:18:26): It's the Matthew principle, it’s a Matthew principle from the Bible, which is “He who has a lot will get more, and he who doesn't…” And so when a company gets momentum, you hear about momentum. When a company gets momentum, what it means is the next resource that you need is preferentially willing to attach to your thing as opposed to somebody else. That's the mechanical process that drives the power log… Charlie (00:18:44): That creates a chicken and egg question, which is does the product create the company or does the company gather enough resources to create a product? Marc (00:18:52): So that's part of it. But again, to create the product, it's often not just a process of, it's also like, okay, you got to create the engineers and then you've got to actually feel the product and give an example. You’ve got to have top-end security engineers. (00:19:05): There are only so many top-end security engineers. Where do they want to work? They want to work at the top companies. If you're a brand new startup, how do you convince them that you're going to be a top company? You raise money from a top tier VC. So that happens over and over again. The prosaic way that I put it is, my experience as a founder is a top tier VC is a bridge loan of credibility at a point in time when the startup maybe deserves it but just doesn't have it yet. And that credibility is harvested in the form of primarily personnel money and brand. And those three things turn out to be really important in the beginning. John (00:19:34): We're talking about the Silicon Valley ecosystem here, and you referenced Andy Bechtolsheim and his investment in Google. One thing that I find funny about that story is that's the case where he just wrote a hundred K check to them. He actually wrote a hundred K check to Google Inc. Even though they didn't have a company. And I think he got in his Porsche and drove off and is like, “Here you go.” But there were no terms, there was no nothing. And that obviously worked out really well for him. But that's not unusual. I've heard other stories. I think we even got some check like that where again, it was just like, tell me the terms later. And Silicon Valley is very high trust. How did that come about? Marc (00:20:07): Lemme tell you, that story is a great story. That is true. I will tell you there is another part of that story, which is the venture firms that turned down Google and the series A, which is the other side of things that maybe we should talk about, right? Because in retrospect, it all looks obvious at the time. It's not. John (00:20:21): Sure, but it wasn't obvious. Marc (00:20:22): Not obvious. Maybe that reinforces what you're saying, which is it's definitely not obvious. Look, I think it's just, quite frankly, you could have all kinds of theories about this, do all kinds of things. Talking about how wonderful everybody is. I think the practical reality is anybody who's been in the Valley for a while has had the experience typically, in the form of scar tissue, where there was some kid in a t-shirt with some crazy idea and you were like, “Okay, that's great.” John (00:20:40): Or, the opposite. Marc (00:20:41): Yeah. You pat 'em on the head and they go off on their way and then they turn around five years later it turns out, oops, that was Mark Zuckerberg, shit. I had my moment, I had my chance. The problem with missing, right? Remember it's category one— Charlie (00:20:52): It's basically, okay, it's FOMO leads to high trust. That sort of has a cynical truth to it that feels real. Marc (00:20:58): Yeah. If you sit around, yeah, it goes to category one versus category two error. Again, it goes back to the economics, which is Andy's hundred thousand dollars, if he got it stolen, he only loses a hundred thousand dollars. If he gets it right, he makes the 30,000 X return. And so, there's this thing, what you learn over time is the category two errors are much, much worse. And by the way, they torture you for fucking decades, right? Because you read about the success cases that you've screwed up all the way up. And so you just learn the hard way. You have to be extremely open-minded for people. Charlie (00:21:28): I have a confession here, which is what I tell entrepreneurs off to see VC. I say, “Look, don't try and convince them you're going to be successful. Just try and create a fear that there's this possibility for the next 20 years they might regret this as their sort of past personal billion that they missed.” Marc (00:21:43): When the company goes bankrupt, at least it ends, it's over. The pain is over. When you pass on the company that succeeds, the pain is forever. John (00:21:51): It's like the asymmetry of shorting. You're kind of shorting the entrepreneur. Marc (00:21:53): Oh yes, absolutely, a hundred percent. It's a horrible mistake. And so as a consequence, there's just this thing of… what it leads to is this incredible sense of possibility and incredible sense of optimism in a very positive way, which is like, you just need to be extremely open to the idea that you're going to run into the next big thing at any moment. And you really want to put, and say karmically, you want to really put yourself out there to be part of that. John (00:22:11): I think that's true, but I think that's maybe a different thing. You're describing that kind of success can come from anywhere. There's a big asymmetry in success where companies can 10,000x, whereas they can't go down by more than one X from their present position. But it seems like particularly the business culture and even moving outside, the fact that startups get really big is particularly high trust. So you have all of investing happens based on handshakes and people can just shake hands on, ’this is going to happen’ and trust that everything happens there. Even when it comes to when we buy companies, we generally agree with the founders at a high level of the terms. And there might be a single page or a two page term sheet, and obviously lots of due diligence will happen after that, but it won't be the kind of East Coast process, private equity process after that, where everyone's trying to pull a fast one that you can't trust the lawyers as fast as you can throw them. So there seems to me there's a particular kind of high trust relationship in how all the actors work with each other. Charlie (00:23:05): I was going to ask Marc why the East Coast and why Europe hasn't generated Silicon Valley, whereas you have Detroit, but then KU Japan copies it. And I think maybe he's actually already answered the question, which is maybe because I haven't had those 10,000 X returns, they haven't instilled the fear of FOMO. And it's the fear of FOMO that means you've got to take a trusting bet on a new person. And maybe that's the kernel that equates a high trust ecosystem. Marc (00:23:34): And maybe just add, I think maybe you're right that I'm being a little bit too cynical in my answer. It's also that you want your reputation to project into the future. And so, if you have a reputation, it's a fairly close knit community. If you have a reputation for being helpful and being positive and constructive and value add, then that plays well. Because then the person, you've done something nice for who's going to introduce you to other people in the future. Marc (00:23:56): It's the ultimate repeating game. And so there's that. And then look, I think the other side of it that you guys kind of alluded to, but I think is very important, which is, it's not zero sum. When I talk to my friends in Hollywood, which is not that far away and is its own entrepreneurial ecosystem. If you talk to anybody in Hollywood, they're like, oh my god, this is a shark tank. You're lucky if your friend's knife you in the chest. Generally, they just, it's in the back. It is this constant thing. And the reason is because there's just, at least my analysis, there's a fixed amount of money to be spent and made in movies, for example. And if my movie gets green lit, it means yours doesn't. And so even if we're close friends, we're going to undermine each other as much as possible. Whereas in tech, at least historically, you have this multiplicative kind of generative thing where it keeps expanding. Charlie (00:24:36): So why else? Why did nowhere else manage to get that ecosystem going? If you look at the history of this last 50 years, one of the stories that will come out is an utter uniqueness that tech almost became a Silicon Valley, or at least a West Coast monopoly. There's no precedent for that in any other industry. Marc (00:24:53): Well, I think we're back to that. Charlie (00:24:54): Yeah, exactly. Marc (00:24:56): You see this in data actually already, AI was reconsolidating into basically two places on earth and only one in the west. Charlie (00:25:02): No part of the industrial economy had that dynamic. What is it? Marc (00:25:05): So there have been a long parade of officials from other cities in the US and from other countries who have come to the valley in the last 30 years. I've met with many of them. They all asked that question I answered as follows, which is, “There are a set of things that you need all in combination.” And then usually at that point they get a stricken look on their face and they say, “Well, what if we can't do any of those things?” John (00:25:24): We build a really linear city. Marc (00:25:25): Exactly. Well, actually, it's surprising the number of people, and I don't want to bad mouth people always. People should try to make these things work and I'm proud of them for trying. But literally the number where it's like, “Wow, if we just built the right buildings, this would happen.” That's actually fairly common. And anybody who's been to Silicon Valley knows the key to it is— John (00:25:43): Go on El Camino Real. It's not the buildings! Marc (00:25:44): It's definitely not the buildings. So I think it's a formula and I think it's a list of things. It's like making a cake. They all have to be in the cake. And the best way I think I can describe it is, it's a set of things that have to do with stability and maturity and rule of law. So you need absolute contract law. You need liquid deep capital markets. You need expert specialists in all these different areas that really have real experience accounting and everything else. And so, there's a maturity and a depth, and that's stuff that developing market countries struggle with. But at the same time, you need the wild west and you need the spirit of adventure and the craziness and the willingness to take risks. And if somebody fails Charlie (00:26:23): And that's what the east coast missed. Marc (00:26:24): And that's what the east coast missed. And that's what Europe doesn't, right? At least when I talk to my friends in the East Coast or my friends in Europe, that's what they like, “Well, we can't do that, I can't take that kind of career risk. That's crazy.” And look, in a lot of countries and in a lot of cultures, if you take a risk like that and it doesn't work, it's a real problem. John (00:26:39): I'm sorry, why is there more risk taking on the West Coast versus the East Coast? Marc Ah, the frontier. Charlie (00:26:44): There's no established hierarchy. Marc (00:26:50): It's all in, what's his name? Winslow, the frontier guy from… Charlie (00:26:53): I was going to say it's Bonfire of the Vanities too. You would go join Goldman Sachs, you would join McKinsey, you would join existing institutions and go up them on the East Coast. Those just didn't exist on the West Coast. You effectively had a country of 50 to 70 million people. John (00:27:06): There was Wells Fargo. There's lots of institutions that you could join. Charlie (00:27:10): Yeah, but were they prestigious enough that they trapped young talent? Another way to say this is, why did Stanford do so much better than Harvard and MIT? Because obviously the input quality is the same. So there has to be something in the place they're sitting that creates a difference. Marc (00:27:27): I think there's a frontier spirit. I mean, I really do. John (00:27:31): But you're always skeptical of cultural explanations in other places. Marc (00:27:33): So there's clearly a talent aggregation effect that takes place inside the US. Look, most of the great people in Silicon Valley did not grow up in Silicon Valley. I tease, my wife grew up here in Palo Alto. I tease, I call her a townie, right? By the way, she has three more degrees than I do. So it's definitely not a status thing. But most people are imports. They get imported all through the entire rest of the country and around the rest of the world. And so it's definitely a selector, an attraction point for talent, and that's a big part of it. But look, I think if you just trace the history every step, it's not an accident that both Silicon Valley and Hollywood are the places that they are because the people involved went west as far as they could before they were literally stopped by the Pacific Ocean. (00:28:12): It was the ultimate selector in the build out of the country to the people who were the most oriented towards risk and to your point, independence, and doing their own thing. And that was true in the Gold Rush days in 1850, where San Francisco was ground zero for that. It's equally true today. Hollywood is the exact same thing. I mean, in Hollywood's case is actually funny because one of the reasons they wanted to get so far away is they were trying to evade Thomas Edison's patent enforcers, because Thomas Edison owned the patent for the film cameras. And the original Hollywood entrepreneurs had no desire at all to pay for that. And then Edison would hire the Pinkertons to come bust up the movie sets. Right? But you see what I'm saying? Rogue, renegade, iconoclastic. Charlie (00:28:50): But do you think the inverse is true? Do you think that certain people didn't move because it wasn't a fun city that had hit the scale of London or New York? Marc (00:28:58): A hundred percent. Look, we all have lots of friends in New York and London, and they're all just like, wow. My friends in New York, I dunno if you get two pints of this into them, they'll be like, they literally don't understand why anybody doesn't live in New York. John (00:29:10): Oh, I mean I think they'll tell you that at 9:00 AM on a Monday morning, you don’t need to get any drink into them. Marc (00:29:15): That is a very good point. John (00:29:16): It's a New Yorker cover. Charlie (00:29:18): So the frontier and a mining camp, you have to be willing to move to the mining camp? Marc (00:29:21): I think so. And then this gets into the danger. This is like the “back to banking, back to consulting” thing. The danger is,in a lot of ways is, it becomes established. And when it becomes established, it starts to attract establishment people. John The social network problem.. Marc Exactly. And in that sense, the downturns as much of a pain in the butt as they are, are probably helpful. Charlie (00:29:40): You go back to banking, you go back to consulting. Marc (00:29:42): Yes. And the only people who are left, and by the way, this was Silicon Valley when I arrived in ‘93, this had happened. And then this was Silicon Valley in 2004 as we discussed, which is you flush all the status seekers, you flush all the tourists. John (00:29:53): It's like fuel management for fire. Marc (00:29:55): Correct. Exactly. A hundred percent. You clear out the brush. Now look, how long can this last, I don't know. We're in a country that has at least certainly over the last 60 years has had a strong tendency towards stagnation. The thing that has kept this whole thing going I think, is just that there are these new platforms, these new paradigm shifts in technology. John (00:30:13): Everyone loves the defense company explanation for Silicon Valley. Marc (00:30:17): That's part of it. That's a big part of it. Yeah, so Steve Blank has done the best reconstruction of this. The typical Silicon Valley history goes back to the 1950s with HP and the 1960s with the chip companies. But the real history, I think he makes a very compelling case. The real history was actually defense tech startups in the late 1920s, 1930s. And you still see remnants of that if you lay drive around Sunnyvale, the NASA Ames. But this is the place where early, I forget the exact date, but early radar and early missile guidance systems and all that stuff, avionics, a lot of that was innovated here in the exact same way. And that was like a hundred years ago. Charlie (00:30:49): If you could go back, could you A/B test it? Is there any way you could have made Silicon Glen or whatever the Boston Corridor was called, successful? Marc Well, they did… Charlie And keep it successful versus the Valley. Marc (00:30:59): That's the problem. Charlie (00:31:00): Was there a point where it could have gone the other way? Or was it sort of inevitable from the fifties in 1970? In 1970, can it go both ways still? Marc (00:31:07): So when I arrived in the valley in ‘93, I think it would be fair to say the Valley and Boston were probably considered neck and neck. In Boston, and these are kind of forgotten now, but DEC, DEC digital, it was a huge, extremely important company. Ashton-Tate, the adventure of the word processor, I think was there. Lotus was there. And then you had later years, other great companies, EMC and others. And then there's a great book called Soul of a New Machine, which is one of the great all time startup books, which is about a supercomputer company in Boston in the late eighties. It was just extremely excellent, like literary book. And it really tells the story of a startup, but it also tells the story of Boston in that time and place. So a lot of leading edge super computing stuff was there. By the way, Thinking Machines was there, the original supercomputer company. John (00:31:53): The original. Marc (00:31:54): The original Thinking Machines, exactly. Yeah. So Danny Hillis, the company that's the forerunner of what we think of today is like large scale AI grid, cloud stuff was there. And look, MIT was there and was a tremendous generated, huge numbers of smart people. And so it worked really well for a long time. And then basically in the mid nineties it separated and then people in Boston will say that again, two pints in, they'll say that the final blow was probably when Mark Zuckerberg could not raise venture capital for Facebook and had to leave and come west. John (00:32:25): That was a meaningful signal. Marc (00:32:26): That was sort of the last, was the last moment. John (00:32:28): We call it the chapter marker. Marc (00:32:29): Yeah. I was just like, okay, if we couldn't do that one, and then by the way, in the counterfactual, had he stayed in Boston, maybe there would be an entirely new ecosystem there that doesn't exist today. Yeah. So I think basically it worked for a while. And again, this is why I locked in on frontier spirit. So what Boston has is all the stability aspects we were talking about, they just didn't have the same frontier spirit. And it just turned out that back to preferential attachment, it just turned out on the margin, the smartest people from MIT wanted to come here. And that was basically it. Charlie (00:32:59): If that's how to move ecosystems, sort of same question, but for companies: what's the company that could have been a trillion that you would have to change the least to make it a trillion? They get that one exec, they get that one lawsuit. It just goes differently. Marc (00:33:12): I mean there's many, many, many, I mean the all time story of that is a company called Digital Research, which should have been Microsoft. And there's a famous…. So the story roughly goes as follows, it's in the books, but it roughly goes as follows.So Bill Gates and Paul Allen had this little software company, originally in Albuquerque, down the street from Better Call Saul I imagine, which they moved to Seattle. They were very early, building programming tools for computers. And so when I first used Microsoft as a kid, it was Microsoft Basic, they were a compiler company or interpreter company, not an OS company. And then there was this PC wave with all these basically these sort of cat and dog kind of early PCs from like 76 to 82. And they basically sold the basic interpreter to all those companies. And that's how they got going. But they weren't in the operating system business. And then IBM decided famously to enter the PC business and then there was a network connection with Bill Gates's mother and the CEO of IBM. And they were on a board together and it resulted in the IBM team coming out and going up to Seattle and buying a license to Microsoft Basic, which is what everybody did in those days. And then the IBM team asked Bill Gates, what operating system should we use? And he's like, oh, well the standard operating system for PCs is called CPM, which at the time was true, it was the standard operating system for early business PCs. And they said, well who makes that? And he said, well, there's a company called Digital Research down in Santa Cruz and California. There's this guy Gary Kildall, you should go see him. And this was the synergistic relationship that he had with Digital Research at that time. So the story goes the IBM team, which is like 20 lawyers in blue suits, like get on a plane, go to Santa Cruz, they show up at the office to meet with Gary Kildall, discuss licensing CPM and Gary Kildall being a frontier-like person, decided not to come to the meeting, decided he'd rather go flying that day. John. John It’s a reasonable thing to want to do! Marc And instead had his wife who was the company's general counsel, negotiate the NDA. IBM was famous for its lawyers and the wife,the lawyer, was not about to sign the NDA and the day ended inconclusively. And the IBM team was like, all right, this is ridiculous. And they went back up to Seattle and they told Gates, if you can't find us an operating system, the deal for the interpreter is off. And Bill said, give me a few days. And Bill literally went down the street to an independent developer named Tim Patterson, licensed what at the time was called Q-DOS, Quick and Dirty Operating System, which is the true name of DOS, for $50,000 flat fee. Turned around and sold it to IBM that created MS-DOS. The kicker to the story is 30 years later, Gary Kildall was knifed to death in a bar fight. John Oh my God. Marc Sorry. I didn't want to bring the room down. Again, counterfactual and who knows? Who knows, who knows. John (00:35:52): No, it seems hard to argue that digital research would've become a trillion dollar company because Bill Gates had such a killer commercial instinct that there were, I mean obviously the IBM OS moment was the biggest moment. But there were several other moments in Microsoft's history where they steered things and it doesn't feel like if they get the IBM OS pick that, then it just, you magically become a giant company. Marc (00:36:15): Oh, no, no, definitely. You don't magically become the giant company. But again, this goes back to preferential attachment. Whoever got that, whoever got that IBM deal is a classic. It's impossible to remember how important IBM was at that time. Yes, IBM in the mid eighties was 80% of the market capitalization of the entire tech industry. They were the absolute gorilla. And by the way, the IBM PC, and then the clones ultimately that came out of it, completely standardized the industry. But all of the PC companies from before that went away. It was an extinction level event for everybody else. And so whoever got that deal, had he not gotten that deal, it's not even clear Microsoft would've stayed in business.. Having said that, he gets obviously credit for everything as well. Charlie (00:36:53): There is a trend where, if you go to the absolute cutting edge of tech, they're so sort of wilderness people, that they don’t have the conscientiousness. Marc (00:37:00): Correct. Right. Charlie (00:37:01): They go flying instead of turning up to the meetings. And it's almost like you get a second generation who go to the frontier, but are conscientious enough to institution build, and those become the super big companies. Marc (00:37:10): By the way, Dell's another classic case data from that same time. Dell computer was founded at the same time. There's like 400 IBM clone companies at that time that were actually the process going under, most of them just like vaporized. This is later, it was like five years later during the down cycle in the late eighties. And that was around the time that Michael Dell in his dorm room decided fit for the PC business. And that's exactly right. He was a version of that. He was a more systematic thinker than the Wildcatters who had been in the PC industry before that. Charlie (00:37:35): Is that how Oracle wins in databases? Because there's a ton of database companies back then? Marc (00:37:39): Yeah, I think Oracle was a somewhat different story. I think it might've been more of a story of just raw aggression. Charlie (00:37:43): Got it. Marc (00:37:43): Larry was always very into Japanese Samurai culture. Charlie (00:37:48): So moving forward in time, why did none of the pre-Google internet companies survive? Like Lycos, Excite, AltaVista, AOL, Yahoo, none of them. Marc (00:37:57): I think that you need to really rewind back to the differences between then and now. And I would just say a couple of things on that. One is the whole internet, boom bubble, whatever you call it, of that period was basically four years. It was basically four years in and out. For example, the companies you just mentioned, for the most part, my company got going in ‘94. Those companies really got going in ‘96. By 2000 like that, it was nuclear winter. And so it was a four year period. The business models either didn't exist or were brand new. We could spend a lot of time on that. But all the business models that you have today that have these big mega companies, those business models didn't exist. It was still mostly just packaged software in those days. And so it was really hard to build the kind of enduring business that you see today. And then I would say the third thing is the market was so small. So, the total market size in 1999 for internet anything was like 50 million people total, max maybe. Half of those people were on dial-up, which only barely counted. By the way, that was mostly AOL, which only barely had internet support, the way we understand it. They had a browser, but it wasn't what you're used to. And then the PCs were super slow, the modems were slow. And that was still the median internet experience in those days was you dial in for maybe an hour at night from your desk at home. And then businesses, by the way, were just, even businesses that had internet connectivity, were doing everything they could to prevent their employees from using it. Waiter (00:39:16): How are we doing over here? Alright, there we go. Alright, anyone need anything else? Marc (00:39:23): Finally, a real Irish. Finally, a legitimately Irish bartender. Waiter: I need a refill. Marc That would be fantastic. Thank you very much. Outstanding. So it was a very early crude time as compared to now. Charlie (00:39:36): So there's another question that leads to, which is normally you get bull and bear cases on crypto or defense or enterprise SaaS. AI seems unique in that there's very little in terms of articulate bear cases about why it matters. In fact, most of the bear cases go the other way, that it's going to destroy the world or something like this. Were there articulate bear cases on the internet during the bubble? Marc (00:39:56): Oh, I mean, yeah. Well the original bear case was just, “Nobody's ever going to make any money. This is ridiculous.” And then there was just a huge onslaught of, “This is just going to be cybercrime and porn and spam and fraud and abuse.” Charlie (00:40:06): You had the similar sort of equivalent to AI safety. Marc (00:40:08): Well, every new technology has a moral panic that's going to ruin society Charlie (00:40:10): Every consumer technology. Marc (00:40:11): And then look, it was just like, and then you just use the product and be like, this is a joke, it doesn't really work. Look at how long it takes the images to load. Is anybody really going to put their credit card in? So there was, I don't if bear case is the right term, but there was massive skepticism. John (00:40:25): Let's do the math on the bear case here for a second. I think the smartest bear case was that the internet's clearly a cool thing. You guys are getting way over your skis in terms of valuations here. And in particular, you're getting way over your skis in terms of the build out that's happening of the internet infrastructure, where the demand will take a while to catch up. And of course that was true where there was a fiber overbuild. And clearly there isn't an AI bubble in the sense that everyone really likes their tokens. The stuff that we are doing with AI or my personal Chat GPT usage. I really like that, you're not going to take that away from me. And so it's not a bubble in that regard, and it's sensibly priced and everything like that. It's a true tech ‘better, faster, cheaper story. However, there is a huge ramp up in AI data center build out. Oracle just had that 4x RPO beat that caused their stock to go up 40% and Larry Ellison to become the richest man in the world. Basically they're doing giant data center projects for AI companies. And one can imagine that there will be a data center bubble where people get too excited about the build out and we build capacity ahead of utilization and people finally, it's the last musical chair. People build that data center where actually no one wants to lease it. Do you think that is happening? Will happen? Is that a sensible framework? Marc (00:41:41): So I would say actually that is precisely what happened with the internet boom. John (00:41:43): Exactly. That's my analogy. Marc (00:41:45): That's right. And so for people who don't know this, what happened with the internet boom was there was this sort of internet software and services and Netscape and Amazon and these things. John (00:41:52): And by dollars, people confuse the dot-com boom. The internet stuff didn't matter. Charlie (00:41:56): It was a Telco bubble. Marc (00:41:57): It was almost entirely a Telco bubble and it was almost entirely a Telco crash. And you know that for two reasons. One is the sheer amounts of money involved were so much greater on the Telco side. And then the other is, Telco is where the debt came in. And to get a really monumental crash—depression, recession, depression—you need a credit bubble to pop. And the credit bubble was 100% I can tell you not on the tech companies, it was 100% on the telecom companies and it was massive and it was amazing. John (00:42:19): And some of them are dodgy stuff going on, like WorldCom. Marc (00:42:21): And then there was fraud. Right, exactly. And those stories are truly spectacular. My retrospective kind of explanation of what happened because, consistent with what you were saying, basically was there were a small number of people who were building the software and services and that was because it was just like they all had to be invented from scratch. And then there were just only a small number of people who even understood the software and how you could possibly apply it. There just weren't that many of us running around who did that. And so John Doerr had him saying “internet” and at some point “internet” became a cream that you rub on investors to get them excited. And when that happened, what happened was you had a much larger number of people who had a lot of knowledge about how to put buildings in the ground and how to fill those buildings with fiber. And the good news with being in the data center biz in those days— Charlie (00:43:02): Hold on, this gets to an interesting point, which is that when you get a boom, because the new people, there aren't enough people with the new skillset set to do it. Marc (00:43:09): That's correct. Charlie (00:43:09): That can never be the epicenter of the bubble. That's right. It's always where the 50-year-old with lots of capital are. That's where the epicenter is. So it was Telco in the internet bubble and all those Telco people sort of 50. And so now it's data centers. Marc (00:43:22): Well, and you need to play the way, exactly. And the way I would describe it is when the thing takes off, whatever the core thing, when the core thing takes off, there's just too much money. There's too much money that wants to come in and participate and it literally cannot participate. Charlie (00:43:32): But also it comes in the way, it knows how. Marc (00:43:34): It comes in, the way it knows how. And this is what you would find at the time, which was you would meet a lot in, I met a lot of these guys. A lot of these were Telco CEOs or people who start a lot of these new Telco companies, Global Crossing, all these new companies. But Global Crossing was one of the great kind, boom, boom, boom blow up kind of stories at the time. And the entrepreneur was this guy Gary Winnick, and he was actually a Drexel Burnham, bond guy from the eighties, Leonard Brown guy. And he figured out like, oh, we know how to put buildings in the ground. We know how to build fiber. You go to Cisco, you buy the devices, you rig up the fiber, Corning will sell you the fiber. And it's a known thing. And his expertise was going to the debt market, convince them to finance that and then he could go just hoover up capital. And in fact, he built tremendously valuable, tremendously important infrastructure. It's just that a bunch of that infrastructure was not actually filled up for 15 years. And in the meantime, much like luxury hotels traded hands three times, the people who own that infrastructure today are doing very well with it. Many of those companies went under. Charlie (00:44:28): It would be ironic if AI researchers are still underpaid, but there are too many GPUs per AI researcher. Marc (00:44:36): So this is the thing, and here's where you get into the question of whether you can have a reason by analogy and whether things are actually the same. And so then it's like, all right, is AI the new internet? And it's like, okay, if AI is the new internet, then you could maybe plausibly expect this kind of cycle. And for sure you do. You guys probably meet, I meet people all the time, which is like, I don't know how to invest in the software side of this, but I know how we're going to do a giant data center build. And this includes nation states doing this. And so you could say history is repeating itself. The counter argument to that is, I don't know that AI and the internet are even remotely comparable. Charlie (00:45:07): Another way to say it is if you could have sped up broadband by maybe five years, the internet bubble is in a bubble. It just seamlessly goes into 2007. You stilled 56 K modems in 2001. Marc (00:45:18): Correct. People forget. You remember this, but people forget or don't know, this: home internet broadband was not common until after 2005. And I was actually at AOL, I followed this very closely. I was, because we sold our company to AOL. I was at AOL on the executive staff in the board meetings in 1999. And the big question for AOL at that point was how to get from being the narrow band provider to being the broadband provider. We knew it would happen at some point, but it was unclear when, and ultimately the company couldn't figure it out. But the question those days was very much, and it was literally, it was cable modems or it was called ISDN. It was sort of proto broadband from the Telcos and it just wasn't happening. And in fact, it didn't happen at the scale in 2005. And then mobile broadband didn't really happen until like 2012. It was really bad. And people actually forget, the original iPhone from 2007 did not have mobile broadband Charlie (00:46:03): Or apps. Marc (00:46:04): Or apps. But it also, it was on the AT&Told, it was on the old AT&T. It was useless, too. So there was this just incredible lag for when an ordinary person could have the kind of experience that you can have today. And so, yeah, so one theory for why AI is different is actually no, the experience that you're having today just in ChatGPT, it's just so monumentally amazing. It's fully there and you have to watch it type the thing out, but the answer is spectacular. And so there's that. And then there's the other thing, which is just the metaphor. The problem with metaphors, which is, and one of the theories you could say on this is the internet was an interconnecting—it was a network technology. Whereas AI is a computing technology and maybe the only comp for AI that you can have is actually the creation of the computer. Right? Because it's literally, it's the first major reinvention of the fundamental model of what is a computer in 80 years, going from the Von Neumann architecture to the neural network. And if you trace the history back, they knew in the 1940s that these were the two paths. They knew what the neural network was in 1943. There was a big argument at the time of whether the computer should be based on fundamentally adding machines on cash registers or whether it should be based on brain architectures and just we had to wait 80 years for it to work. But now we have the computer industry V2, which is much more valuable and important because of all of the obvious things it can do that the sort of hyperliteral Von Neumann machines can't do. And so we've successfully unlocked computer industry V2, it's 10 or a hundred or a thousand or a million times more important and valuable. And all of your petty comparisons to bubbles in the 1990s just wash out because my God, look at what the thing could do. John (00:47:35): AI is funny because it is always the case that the hype cycle for technologies predates the technology being ready for that hype. And so Charlie and I often talk about the mobile internet hype. Yeah. That people were excited about, you'll buy cinema tickets on your mobile phone in the two thousands on a Nokia 3310, which is not actually how the mobile internet played out. And even the crypto excitement, the kinds of things people talk about with crypto of like, oh, you'll be able to make payments, whatever. We're finally getting to it in 2025 and any kind of meaningful volumes. But it’s a good 15 years from when people started being excited about it. AI is maybe the longest time lag from those things where, when was 2001: A Space Odyssey released? The books that it was based on in the 1950s and then 2001: A Space Odyssey was the sixties. Marc Sixty eight, yeah. John Yeah, exactly. And that was voice mode with tool use, like HAL 9000. And so I find it funny that we had such a specific vision that was pretty much right, but it took a long time for the tech to be, And there was various waves, Dragon Systems like the tech wasn't that good, but people were excited about it. Marc (00:48:41): Apparently there's, there's a book called Rise of the Machines that has the prehistory of AI. And I believe I remember correctly, there were actually debates about this in the 1930s. It actually predated even the sort of invention of the neural network. John (00:48:51): Okay, so roughly a hundred years later we're getting Marc (00:48:53): Yeah, yeah. They knew in, and I think Alan Turing and folks like that were involved in that at that time. There was a famous moment in the history on this. So Alan Turing, Claude Shannon, the inventor of information theory, two very important guys. During World War II, they're building the computer originally in World War II to beat the Nazis, crack the codes. And so Alan Turing and Claude Shannon are having lunch at the AT&T executive dining room in Basking Ridge, New Jersey, like 1943, and they're talking about exactly this topic. And Alan Turing starts to raise his voice, raise his voice, and finally he gets up in the middle of the AT&T dining room and says, “I'm not talking about building a genius computer brain. I'm talking about building a mediocre computer brain like the president of AT&T.” And so they knew, I think he knew that the path that they were on, the Von Neumann machine path, it was building this hyperliteral, you can almost say “hyper autistic” math savant in a box, which obviously was not going to be the thing that was going to be English language and write everything else that you were going to want to do. So he knew this is the wrong path, but he just didn't live in the time in which the technology was available to do what he wanted to do. And it just happens that we do. Charlie (00:49:56): What do you think as the emerging sort of heuristics of how the market works? So let me give an example from software. There's no inferior goods market for software. There's no cheap version of Excel or… Marc (00:50:07): There was at one point. Charlie (00:50:08): There was at one point, and it didn't succeed, which is the point. But in general, software's gone to one company, some horizontal, some vertical being the best. John (00:50:15): Because there's such a great deal. Charlie (00:50:16): Such a great deal because of central productivity... Is it the same in AI? Do we go with horizontal intelligence? Is there an inferior goods market where you end up with AI on device that's intelligent but not super intelligent, but you don't need it to ask the weather? Marc (00:50:30): The way I would think about it is if you think about, let's say this is a computer industry V2, what did you experience in computer industry V2? You had many different sizes and shapes of computers. And actually what happened at the time was the big ones got built first and then it literally was mainframe and then it was mini computer, and then it was sort of server and then it Charlie (00:50:50): It goes down to PC— Marc (00:50:51): Personal computer and then mobile phone and then embedded devices. And then by the way, and then it sort of multiplies out where cars and light bulbs and doorknobs and everything else. As you know, what you have is the consequences. The computer industry and specifically the chip industry is therefore in the form of a giant pyramid where at the top you have a small number of supercomputers and mainframes, and at the bottom you have billions and billions of embedded devices. And then you have everything else in the middle. And the reason you have that is because you have costume performance and fit implications for the specific devices. You don't want your light bulb to have to do a round trip to an IBM mainframe or something. It doesn't make any sense. You want 'em to have the embedded device so that it senses whatever you want, senses whether there's light in the room that's like a specific chip. And so I think the scenario in which you only have a few big AI models is the scenario in which not only are those models the smartest, but they're also the cheapest and the most power efficient and the fastest and easiest to adopt and use for every scenario. And I think that's highly unlikely just because if this is the breakthrough that we believe it to be, and it's the computer industry V2, you're going to want models in everything. You're going to want AI infused into everything. And then for a lot of those infusions, you don't need your doorknob to teach you quantum physics, but you do need it to be really good at knowing that it's you and not somebody else. And so you're going to have all of these kinds of hyper optimized use cases. And so my guess in the way we're batting is— Charlie (00:52:12): Everywhere, all at once. Marc (00:52:13): You're going to have that pyramid approach and then look, the economics are going to be a big part of that just because, I mean if only because the doorknob gets to run a local power and then the processing the doorknob needs to do is a tiny fraction of what you need to do when you ask GPT 5 a query. And so I think this is computer industry V2 in that way. John (00:52:31): And how do the markets play out? Is it just a normal battle price performance with proprietary players? How big a player is open source here? Can we, Charlie mentioned Oracle earlier. I feel like people today forget that the proprietary databases used to be the best databases all the way through the nineties, and you had to, step one of founding a tech company was write a check to Oracle and then you can do stuff after that. And then the open source databases, my people in Postgres became competitive in the 2000s. You don't like me reasoning by analogy too much here, but can you reason by analogy to the database world just how does the market structure pay out? Marc (00:53:06): Yeah, no, I think that's right. I think that's right. I think that's actually a good comp. Another one is operating systems. So when I was a kid, the world's best operating systems were specifically, I mean Windows, has its own trajectory and iOS. But what we used to describe as proper computing, on real computers like Unix computers, including supercomputers and workstations and advanced scientific applications, things like that. The best versions of Unix were proprietary for a very long time. And you had these really big companies like Deck and HP and others that had the IBM, that had their own versions of Unix, and they made a lot of money on those. And then Linux, same story, Linux came along, looked like a toy, and then 10 years later it was better than all the proprietary ones and all the proprietary ones died. That's my guess. Is that something like that? I definitely think we'll live in a world of a small number of big models that'll be incredibly valuable and incredibly widely used for many things. My guess is we are going to live in a world in which most aggregate AI is going to be executed probably on smaller form factors, and probably most of that is going to be open source. Charlie (00:54:08): So where is ground zero where the rate of change would be highest? Software development? Somewhere else? Marc (00:54:14): I mean, software development is a very good candidate for that just because you have people building for themselves I think. And you have this incredibly tight iterative loop and you see that with these new software, these AI tool companies. So that's a claim. And then by the way, the other advantage to software development is this is a really underrated thing with respect to AI adoption that a lot of the people in the field are missing as software development is not regulated. And so it's like impossible. Well, there is that they're trying, the enemies of progress and freedom are trying, and we are fighting them very hard, but it's like AI medicine actually can't move that fast, because it’s regulated and AI can't be a doctor. You can't get licensed. AI can't be a lawyer, it can't go make an argument at a court and so forth and so on. And so I think it's like, yeah, it's the unregulated fields populated by the same kinds of people who are building AI. John (00:55:00): Charlie had the interesting question of, are we overestimating the broad impact and underestimating the specific impact or what of at least for the next five years, as you say, AI and medicine or AI and law doesn't make that much progress because of some of the challenges, but software engineering is totally transformed. Marc (00:55:18): So the counter argument, I mean, I think there's a big argument in that direction. And by the way, actually, I wrote a big Substack piece. Maybe we can link to talking about how the employment shifts everybody's worried about are actually not going to happen anywhere near the velocity people think. Because a significant percentage of jobs in the US literally are licensed or unionized or civil service in a way where they literally cannot be replaced. And so I do think there is part of that. Having said that, I think things are going to pop in really interesting ways. And so for example, ChatGPT is in fact a better doctor than your doctor today with almost a hundred percent certainty. And just the fact that it can't literally be your doctor doesn't mean you're not going to ask all the doctor questions. And then you already have people online who are taking surreptitious camera phone footage of their own doctor asking ChatGPT during the appointment. John (00:55:59): I think the medicine use case is an interesting one because it turns out it was a space where most people were actually intelligence bottlenecked, which I mean it is like test-time compute. They were getting a very small fraction of their doctor's headspace. And if you put just more thought on the problems, you can get really good outcomes. Marc (00:56:16): And then by the way, medicine and law are also, you could also look at the self-driving car thing, which is, there's always this test for self-driving car. There's always this question of is the requirement perfection or is the requirement better than the median human driver? And if you apply that same question into law or medicine, it is just overwhelmingly clear that you're better off today with Dr. ChatGPT. Now in one sense, you can't live your life that way because it can't be your doctor. On the other hand, you can sit there all day long talking to it about your health. And by the way, I think there's going to be a lot of tension and a lot of drama in these different fields as that happens. Here's another argument that comes back around on this, which is the argument of like, “Oh, AI is horrible because it's going to lead to five companies controlling everything,” and it's going to be like, that's it. And the monopoly cartel fear. And there's a bunch of reasons to be suspicious of that, including things like open source. But the other reason to be suspicious is at least with downstream impact is, AI is already maybe the most democratically distributed technology in history. So whatever, 600 million people or whatever it is, the number now, is on ChatGPT in two years. And again, you compare that to internet adoption, it's like far faster. And of course the reason is because the internet exists today to be able to distribute it, but the world's most advanced AI is in an app that 600 million people have. It's not in the one that I have or that you have. It's the one that 600 million people have. And so this technology has already been hyper democratized, and so it's going to be in everybody's hands. And people get confused about this like, “Why would big companies do that?” And the reason is because the mass market's always the biggest market. You want to get to everybody if you're trying to build the most successful company and to be the company that is the most important. And look for sure, there are always concerns about aggregation of power or centralization of power, for sure. But there's this other thing which is what if this is just like the philosopher’s stone, the alchemy of sand into thought in literally everybody's hand right out of the gate. Charlie (00:58:05): You look back at the old companies, you look at the S&P 500 of some 1980, there's not that much change in the success based on tech. It's not like some bank gets better at tech than all the others and just goes past all the competitors. If what you're saying is true, you would say that old companies are going to adapt less well to this. Marc (00:58:25): 100% Charlie (00:58:25): And the level of change is going to be unprecedented. Marc (00:58:27): Yeah, I believe that to be the case. Well, so again, let's go back to the computer industry on this. This I think is a very interesting idea. So the computer industry started out by building the big thing, started out by building the mainframe. Thomas Watson Sr., who ran IBM in the 1950s, said he thought there was a world market for five computers. And it was literally one mainframe each for the three big insurance companies and then two for the Department of Defense. And that was it. And by the way, at that time, it was true that was the world market for those computers at that time. And then basically over 40 years, he went from mainframe to mini computer to client server to as you said, to PC and a phone. And so what happened, is over 40 years, the technology cascaded down into the mass market. And then today that culminated in the $10 Android smartphone in India. So that was that. AI, at least so far. And by the way, many other categories of new technology in the last 30 years, because smartphone is another example of this or have been the reverse, which is no, the individual it first the companies are deciding to go for the individual market first, the largest market, and those are the people who are the easiest to adopt. John (00:59:28): It's Andy Warhol, the president drinks the same Coke as you and I. Marc (00:59:31): Exactly. And then what happens is over time— what happens with smartphones, and this is what I believe is happening with AI, which is the individuals get it first, adopt it first. The small businesses get it second, adopted second, the big businesses get it third, and the government gets it forth, not because the governments and the big companies couldn't get it faster if they wanted to, but they can't absorb it. They have all their rules and then they have all their bureaucracy and they just simply can't absorb it. And again, it's like at the level of politics, structure of society, you could say this is a fight between the power of the individual versus the power of the state. Obviously there's fears of AI surveillance and all these things on the state, but the other side is every individual citizen being super empowered being a PhD in everything, including how to deal with the state. So everybody all of a sudden is a super lawyer. Okay? And then within business, it's the balance of power between small companies and big companies. And if you're just looking at speed of adoption, there's no question small companies are adopting faster. John (01:00:27): I was going to ask about that because Robert Solow said the computer age shows up everywhere except the productivity statistics. AI productivity is showing up everywhere except the hiring plans of your portfolio companies, which still seem to be hiring a lot of humans. What does the realization of significant AI productivity gains look like? Because presumably stodgy, large companies you believe will fight the gains at some level. They won't take as much AI productivity as they should. Marc (01:00:52): So I think the most basic question is the sort of fundamental question of, is this a centralizing power or is this a democratization of power? John (01:00:58): And so do you think it makes small companies more powerful in the battle against large companies? Marc (01:01:02): I think there's a really good chance of that. I don't know for sure and we'll see, Charlie (01:01:05): But it seems certain that it will make younger companies more successful against older companies. Marc (01:01:09): I would assume so. Charlie (01:01:12): Less bureaucratic companies Marc (01:01:13): But you have, let's take the employment of the jobs thing just because that's the one that gets all the headlines, which is just like, “Oh, all the jobs are going to go away. It's going to do everything.” So one version of it is like, okay, that is going to be the thing, and this is at least to the meme of five companies are going to own the world and you have whatever, three years to get out for the permanent underclass. The more conventional economic argument is the opposite argument, which is, this is going to deliver massive productivity improvements not just to companies, but also to individuals. When you put a technology in the hands of an individual that massively increases their productivity. And the way I think about that is AI just makes every individual a super PhD in every topic. That's the most dramatic increase in what economists call marginal productivity of the worker that has ever existed. And so as a consequence, every single one of those people is now capable of doing so much more than they were ever capable of doing before, whether they're doing that as a solo entrepreneur or whether they're doing that as somebody who works in an organization. And so, in that version of the world, you don't get the aggregating effects. You get some, but they're swamped by the democratization and the superpowers that every individual gets. And then 10 years from now, we'll do part two of this probably with the same glass of beer at the same room temperature. And we will be shocked by how much AI drove both employment growth and drove incomes because again, the conventional economic view is marginal productivity improvements. You want to hire more people at higher levels of productivity, they can do more, and then you pay them a lot more because they can. Charlie (01:02:40): A huge part of that is, when people think about this is intelligence but not imagination. If you go back to 1950, there's some movie where there is basically a single person is a cell in Excel. They're all sitting in a big room effectively doing accounting. If you described the sort of computing evolution, they would all say, I'm going to lose my job. But the jobs that emerge, video gaming, you couldn't imagine, you couldn't describe. So it's very hard, I think for people to overcome the jobs they can see existing ones disappearing, but they can't see the emergence of new categories. But we've always had the emergence of those new categories. And if you take things like sport, which I think is like 3-4% of GDP you can imagine that extending to 20% of GDP and whole new sports emerging. John (01:03:24): We have whole new sports emerging with eSports. And you can argue many of the existing—all sports have gotten way bigger over the past five years. Basketball is way bigger. F1 is obviously way bigger. They've all gotten much bigger. Marc (01:03:35): We're even bringing soccer to the US. John (01:03:36): Exactly. Marc (01:03:38): No, that's exactly right. And then the corollary to that, by the way, this is very difficult to talk about because people get very upset, but the correlator to that is those old jobs after the fact, you're just like, “I can't believe human beings were required r to do that.” Because literally as you're alluding to what happened, the original computer was a person sitting at the desk doing manual math all day long. Imagine if I showed up today and told you that's what your kids are going to be doing as a profession. You'd be like, sounds like torture. Charlie (01:04:04): Have you ever heard of Iain M. Banks, the science fiction author? The Culture series? Marc (01:04:08): No, I've actually never read that. No. Charlie (01:04:09): Okay. He tries hard to contemplate what a super advanced society with AI is like. And what's interesting is, everyone has stuff that looks like a job but is actually leisure. Marc (01:04:21): Right? Well, the best jobs in the world have that characteristic. Charlie (01:04:23): And then you have very complex status hierarchies as people aspire. And if you look at Gemeinschaftic societies like Formula One or something like that, you have a very clear sort of motivation states hierarchy for people within it that seems to fulfill a lot of human needs. John (01:04:37): Aren’t you describing being a VC?I've seen the activities in the conference. Marc (01:04:41): That's exactly what we like to say. It's a 9 3/4 profession. It's a country country club kind of thing. The other, by the way, great economic fallacy that I just see everywhere right now is this idea that AI is somehow going to be this hyper successful thing, hyper acceleration of productivity and dramatically change everything, destroy all the jobs, and yet somehow that's going to lead to people being eviscerated and being poor and not having anything. And the missing element there is that even if that serial plays out, which I think as I said, I don't think it's a centralization scenario, but even if it played out, the result would be hyper deflation of prices, which is the thing that people miss. And so the price in that environment with that level of productivity growth, the price business services will collapse and things that today cost a lot of money will all of a sudden all be cheap or free. Marc (01:05:22): Everything becomes oversupplied. Charlie (01:05:24): In Star Trek. There’s no—, GDP would be zero because the replicator does everything. Marc (01:05:30): And so things that cost a hundred dollars cost a penny, right? In that world, even real GDP looks like it's shrunk and everybody is much, much, much better off. And by the way, this is not the first time this, there have been periods of sustained deflation in the past. Charlie (01:05:44): When you see it within categories, look at the spend in CDs, music CDs versus music. Marc (01:05:50): I always talk a lot about the so-called second industrial revolution. So, the time in which our entire modern world was built, with everything from airplanes to freeways and everything else, 1880 to 1930, it's like that 50 year stretch. And for a lot of that period, they were in essentially a protracted deflationary depression. Because what happened was the technology for acquiring and processing raw materials was advancing so fast that there were gluts in all the different raw materials. And so it felt like the economy was caving in because prices were collapsing, economic activity was down, GDP was down. In reality, what happened was a massive surge in productivity growth and a massive surge in material prosperity. And over that period, both productivity growth and economic growth advanced something like three X of our time. But if you read the books at the time, they're obsessed with this problem of like, “Oh my God, there's this oversupply of iron. What are we ever possibly going to do with it? And it's destroying the economics of the iron production business.” John (01:06:43): Could you not have low productivity segments of the economy find ways to avoid the prices collapsing too much, such that you don't get this effect and people Charlie (01:06:59): Baumol’s cost disease. John (01:07:00): Yeah, we've gotten much better at healthcare over the past 50 years and yet… Marc (01:07:06): Yes, yes. So Baumol’s cost disease, but also just simply government. You see this today. So basically it's like today what happens if you chart, it's this famous chart. If you chart basically the prices of products across all these sectors, what you see, basically John (01:07:16): It's the deflationary economy and the inflationary economy. Marc (01:07:18): Yeah. There's two different economies and the deflationary economy is like everything electronic, everything, software, everything media. By the way, basically everything, all light manufacturing, clothes and everything. Well, not housing, so the price of clothes collapses, the price of housing, hyper rises, they deviate. On the other side, on the non-productive side, you've got housing, education and healthcare. And that sort of, I think explains a lot of the politics and sort of feeling of our society right now, which is just like everything that's optional and fun is getting super cheap and everything that's actually necessary to raise a family is like getting hyper expensive. And exactly to your point, these are two different economies. And then you look at, and this gets complicated, but if you look at housing and healthcare and education, what they all have in common is heavy government interference, specifically of the form of restricting supply. In all cases, the government basically restricts how many houses can get built. They restrict how many doctors can get licensed. They restrict how many education universities can get accredited, and then because restricted supply leads to prices skyrocketing. The voters get mad, and so then the politicians subsidize in all three of those markets. There's massive government subsidies, federal student loan programs, federal mortgage programs, federal healthcare programs, and if you insert basic economics, if you constrain supply, you cause prices to rise. And if you subsidize demand, you cause prices to rise. I think this is basically the state of the Western democracies over the last 50 years. Every step of the way as the price of the American dream—housing, education, and healthcare—as the prices of those rise, the pressure from the government to subsidize increases, which just drives the prices higher. And so you're in this ever escalating spiral. John (01:08:57): I'm presuming you're concerned about more of that. Everyone's making fun of the Boston City Council objecting to Waymo and maybe voting to preserve driving jobs and everything like that. And so we find more categories to turn into healthcare education. Marc (01:09:10): Yes, sinecure is fundamental, and then Baumol’s cost disease kicks in because now you have this different era. Then you have the hyper incomes being earned by people in the deflating sectors where there's massive productivity growth, and then people in healthcare get to command those wages, and then the whole thing compounds and gets worse. By default, this is what the governments are going to do. In fact, by default, it's exactly what they're doing today. And then there's a really tricky political economy thing to this, which is like the voters, it's very hard to tell the voters, “Don't vote for the guy who says he's going to subsidize housing more.” Right? John (01:09:41): Sorry. Are you worried about this as a political future? Marc (01:09:43): Yes, a hundred percent. Well, I think this is our political present. John (01:09:49): Sure, sure. I'm thinking an expanded version. Marc (01:09:50): Oh, an expanded version. 100%. I'll give you, well, the latest example of this latest example, remember the dock workers? Remember the dock workers strike? John (01:09:56): Oh, I do. Marc (01:09:56): Remember the guy with the gold chain, the whole thing, and we found out about the dock workers, and you dig into it and you're like, oh, the dock workers union. It's reason why John (01:10:04): European ports are way more productive than the US Marc (01:10:07): Automated reports. It turns out because you have unions and they have a tremendous amount of political stroke. One of the things that was discovered during that process that I didn't know is that in prior union agreements with the dock workers, they already had a one-to-one ratio of people sitting at home doing nothing to every productive dock worker as a consequence of the last whatever, 60 years of these things. So basically, there's a long history here that just never became visible in public, which is every time any kind of new automation shows up at the docks, the dock workers renegotiate the contract to preserve the jobs, which literally means people sitting at home. And that was before the most recent agreements. And that's just a micro example that seems to pick on. The much larger example is the civil service public sector unions, obviously. Right? Then here we're into teachers unions and nursing unions and all of these things. And then here we're into this fairly amazing bizarro world we've been in for the last 50 years where you have, especially around government, you have both civil service protections and union protections, right? So exactly. So by default, the political economy makes all of this worse and worse. By the way, this is why I think inflation doesn't mean what it used to. ( Inflation fifty or a hundred years ago used to mean the price of raw materials was so important in the economy that you felt it very directly. Now as you say, you've got this, you've got this, John (01:11:17): It's hard to talk about a single bundle. Marc (01:11:18): Yeah, because it is not the same thing. And this is the thing where you can't build a family off of the price of the iPhone. Just because everybody has infinite media on their iPhone for free does not mean that they feel good if they can't buy a house. John (01:11:30): I liked your, what was your inflation stat of if there's a hole in your drywall, it's cheaper to put a flat screen TV over it than it is to repair the drywall. Marc (01:11:40): 100%. Exactly. Charlie (01:11:42): Let me drag us back to AI a sec. You’re used to the 10x engineer, you going to have the 1,000x engineer with AI? Marc (01:11:49): Yeah, for sure. I think you already do in practice. And of course we have had for a long time, I mean we've had the 1,000x engineer for a long time. Charlie (01:11:56): Are we going to add another one? Marc (01:11:57): Yeah, it's becoming more visible. It's going to apply in more areas of software. And then look, the other thing is just the payoff to software is minimizing. The markets are so much larger now, this goes back to why would this time be different or they AI versus the internet? Which is just like, okay, this is the first time in human history that you've had 5 billion people connected on an interactive network. And if you are a provider of products and services that go into that market, if it worked, you may or not work, but if it works, it can get sort of infinitely large and actually really fast now. And so what is the upside? How many people are there in the world who are going to pay whatever it is, 20 bucks a month for the world's best AI? It's not all 5 billion, but it's a much larger number than you would've had 10 or 20 or 30 years ago. And maybe it's just simply market size. John (01:12:46): As you just heard from Mark, we're in the midst of a massive platform shift with AI. It's like the computer industry V2 and Stripe is the company building the economic infrastructure for AI. If you've used an AI product recently, you've almost certainly used Stripe. More than three quarters of the Forbes AI 50, including OpenAI,Anthropic, Lovable, ElevenLabs, Perplexity, Cursor, Midjourney—they all use Stripe to monetize their products and they're growing at a historic pace. We analyzed the top a hundred AI startups on Stripe and found they reached the million dollar revenue milestone four months faster than the SaaS companies that preceded them. AI companies choose Stripe for high converting checkout, instant global reach, and AI driven payments performance, all while monetizing in the ways that they need with subscriptions and usage-based billing. And then as AI agents begin to buy on our behalf, Stripe is building the tooling for agentic commerce so that any business can thrive in this next era. So come to Stripe to explore our MCP toolkit, agentic payment flows and everything your AI business needs. Charlie (01:13:44): You've been super early to crypto with a16z. It's probably the area where there's been the most concentration of VC performance, you, Paradigm, not that many others. Two questions. Why did, so few VCs focus on crypto, and then how important is stablecoins going to be? Marc (01:14:01): So sorry, the first question is why did they or why didn't they? Charlie (01:14:03): Why didn’t they? Marc (01:14:04): So what I've observed… I mean, so one is you could always just say the easy explanation, just they didn't understand it or they were focused on other things. What I've observed is that as technology has become more important, people's belief systems have a lot more to do with technology. So your worldview, the part of your brain that thinks about things like a larger and larger percentage of that is devoted to technology. And of course, if you're a VC, that's like a hundred percent, and then whatever you're spending your time on, you form whatever myths, legends, religion, cults. So it means the same question of why is the press so much more focused on technology than they were 15 years ago? Charlie (01:14:41): Centric was harder to be politicized about than AOL and eBay than it is today. Marc (01:14:44): Yeah, exactly. I think what we observed is a lot of VCs who were very logical and dispassionate on topics like SaaS for which there's no religion Charlie (01:14:53): It's hard to get politically motivated around Saas. Marc (01:14:54): For some reason, there was something about crypto where they just got locked in on the politics and it was just like, oh... So my theory of it, after a while, I just met so many people who would just…And it wasn't even that they were like, ”Oh, we don't think it's going to be valuable.” They were like, “It's evil. It's full on evil, it's a scam, it's a fraud, it's a this, it's a that.” If it works, it's evil. If it doesn't work, it's evil. One of my tentative conclusions, which is just money pisses people off. And so making money through tech is usually an indirect process. In this case, there was a more direct aspect. People have always built up all kinds of weird religious and political views around money. And literally what we experienced was people just got really upset and we can never understand it. Like what's the point of being a venture capitalist of all things? What's the point about being negatively upset about a new technology? John (01:15:43): And in particular, it feels like it requires high openness, where there's something about early crypto where it attracted folks like Balaji where there was all these grand pronouncements of like, oh, Bitcoin will supersede the nation state. It led to a lot of that kind of slightly cultish, very cyber punk. It really reminds me of the, what's the John Perry Barlow letter? Marc (01:16:08): Declaration of Independence of Cyberspace. John (01:16:10): Exactly, exactly. John Perry Barlow’s The Declaration of Independence of Cyberspace. There's a lot of that kind of vibe about crypto. And so it required one to be open-minded enough to think there could be something here. I think most people are not that high openness. Marc (01:16:23): And then it also got— John (01:16:24): But you're not that high openness. Marc (01:16:25): Building on that. Well, I dunno, but I'm not introspective, so I don't have to think about that. It also got right-coded. I think it got right wing coded. It got libertarian coded early, especially in the 2010s when everything got politicized, anything coded right, libertarian was bad. And then quite honestly, and maybe this will piss people off if I say it, but quite honestly, if you actually want to understand it, how it works, it actually is quite difficult. It is a complex technical thing, and I think maybe people don't understand the tag. Maybe people literally don't understand. I dealt with this a lot when I would deal with people who were causing us trouble in public, and I literally would try to explain it to them, and I just fundamentally couldn't, by the time we're using the phrase, “Byzantine general's problem,” you're done. It's never going to work. John (01:17:10): My observation is that we have a friend who talks about how crypto contains multitudes, and that's the important thing you have to internalize because the criticism you'll hear is sometimes something like, oh, “There's a lot of scams in crypto.” Speaker 5 (01:17:22): That's right. John (01:17:22): And it's like, okay, crypto is this big box, and within this big box are, there's a lot of scams happening. There's Vitalik types who are really interested in developing new protocols. There's people using it as a store of wealth, especially in emerging market countries. There's people who are just interested in speculative number go up games. There's people who are passionate about developing new payment systems and they're working on Bitcoin Lightning or something like that. It's this big box that contains so much different stuff. And there are strengths and there are weaknesses, or there are things that we might not like. Again, I don't like some of the rug pulling kind of scam aspects, but it's just a big box with a whole lot of different stuff in it. And people seem incapable of reasoning that way. Marc (01:18:06): They'd see what they want to see. And plus it was also like you want to see the tech guys taken down a notch, and this is some way that tech guys are manufacturing magic money. And then maybe another more focused way of what you're saying is, every new form of financial technology associated has been historically associated with some form of bubble and crash and sort of scams along with that. And the classic example of that that I think is illustrative for crypto, the invention of paper money. Charlie (01:18:31): John Law. Marc (01:18:31): John Law invented paper money in France about 360 years ago, and it immediately sparked what became the South Sea bubble. And actually he ended up, basically, his life did not go well after that. Charlie (01:18:41): He fled to Venice. Marc (01:18:42): Fled to Venice, and basically died poor. John (01:18:45): High-yield finance may be another example. Marc (01:18:46): What's that? John (01:18:46): High-yield finance. Michael Milken. Marc (01:18:48): Oh, yeah, junk. Junk ponds were completely discredited by the time Mike Milken was sent to jail. Yeah, junk bonds had been completely discredited because everybody, again, the moral story, everybody knew that it led to this massive bubble of all these, I mean deliberately high risk bonds. Who would ever do that? A decade later, that market was much larger than it ever had been in the eighties and was extremely well respected, and it played a huge role in the build out of everything since. And so new kinds of money lead to new kinds of scams. Charlie (01:19:16): Speaking of new kinds of money, how do you think about stablecoins? Marc (01:19:20): Yeah. Yeah. The stablecoins, I would say the way, I think primarily what I think is they've been super helpful to have stablecoins succeed because it's just an obvious incredible use case. It's worked incredibly well. They're being used all over the world for many different reasons. The numbers are now extremely large. I think it's great. It was originally, you guys probably know it was originally part of Vitalik’s early work. He had a very unfortunate—do you remember the original name for stable coins? Charlie (01:19:45): No. Marc (01:19:45): “Colored coins.” Oh, yes. Only an ESL speaker would pick that name. But the idea, the idea was a crypto token wrapping a real world asset. So that was part of the original thinking on all this stuff. It's worked incredibly well for dollars, I believe that will work incredibly well for many other kinds of assets. It's great. Now, having said that, the crypto purist natives are like, well, that's the main thing because it's a bridge technology of the old world. I think it's great, and I think it's fantastic that you have such a successful use case. Charlie (01:20:15): So fintech has generally not produced great companies or giant companies because it's been country by country demarcated. In fact, you end up with these very mediocre companies like Stripe, but they're fantastically managed. It seems like stablecoins could lead to the global scalability in fintech. That has been the prerequisite to making super valuable tech companies. Marc (01:20:40): I mean, we've had some fintech ones that we're very proud of, including Stripe. It just the level of one is regulation and then particularly Charlie (01:20:50): And payments is different because they did go global. They did, but not many companies Marc (01:20:53): They have regulatory constraints there as well. And then the last decade in particular, a lot of the western countries, they've been on a crusade against any kind of financial innovation, just on general principle. And so there's been these real regulatory government headwinds and then just look like dealing with the banks, dealing with the credit card companies, dealing with these... They're not psyched at the idea of some kid with some new idea. They're just not. And so even if you have regulatory clearance, can you actually implement the thing is an open question. And so I just think there's a lot of glue, a lot of stickiness. And then, I mean, look, you could also say, to be fair, it's a high hurdle to go to a consumer and to say, you should trust your money with some new companies. So there's a whole issue there. So yeah, so optimistic. Yeah, I agree with your optimistic point of view. Yeah, I mean, this was always part of the crypto philosophy, which was programmable money. If you have programmable money, then all of a sudden you could have financial services work a lot more like software. You could have a much higher rate of innovation. And you're right, maybe we're starting to get there. John (01:21:54): Was there someone who really got you into crypto? Marc (01:21:57): I would say the main person was my partner, Chris Dixon. It was very early and had figured it out. And then we were involved in Coinbase early on. And so Brian and Fred at Coinbase were super helpful in helping us understand it. John (01:22:07): How do you think— Marc (01:22:07): Sorry. And actually Balaji is at the head of that list, John (01:22:11): And how do you think Chris cracked it so early? Marc (01:22:14): Chris's entire life has been this pursuit of…it's just how he thinks. He's just born to do this, and it's in pursuit. He uses these terms. He used one term, he says, “What nerds do on nights and weekends” is one way to look at it. Second way to look at it is good ideas that look like bad ideas. And then his third most recent version of that is internet cults. It's like if it has a thriving subreddit, then something's going on. It's the other side of the people's negative emotion on this is the things that become movements early. The internet enables movements. John (01:22:49): Is there something that's… Yeah, this is the home brew computer club.. Charlie (01:22:53): John, how did you think about stablecoins for you? John (01:22:56): It's funny, when you were saying crypto is an internet cult, we find that it's very vibes-based in a funny way where there's always a thing of Stripe is pro crypto. We're super excited. Stripe is anti crypto, not going to make it. Stripe is pro crypto. Again, we've never been, that's how we've conceived of it. We just want to build things that people find useful. And the Bitcoin whitepaper dropped in 2008, and I want to say, and Stripe was founded in 2009. And so we've been watching all along. Charlie (01:23:25): Wasn't it February ‘09? John (01:23:27): Think it might've been ‘09. You're right. Yeah. But anyway, it was just before Stripe. (01:23:30): And so we've just been trying various things. Like we found it stellar in the early days. We tried Bitcoin support, original Bitcoin was a horrible payment method, you know what I mean? And the thing we have really noticed that's really striking is there's a level of consumer adoption and familiarity that allows for a mainstreaming. We just worked with Shopify to, they now offer stablecoin payments on all of their checkouts, or they're rolling that out on all their checkouts. That's just not a thing that would've made sense even three or four years ago. And so it's like we're talking about the internet stuff. Just at a certain point, Google and Facebook and all of these companies start to work, and if you try to launch Facebook in 1998, it doesn't work because there aren't enough internet connections. I think there weren't enough wallets for a lot of things to work. (01:24:14): Even you look at the stable point supply charts, we're growing at 40, 50% year over year. It's the grains of rice on the chess board. You don't need that many years of 40 to 50% year over year growth before it really works. But it is been really striking for us over the past 18 to 24 months, where we've been trying to make different things work at various points and are going to maybe shut off products that don't work. But now all of the products are really working all at once. I had some questions on the Andreessen Horowitz business. Why aren't you a hedge fund in that you, or why don't you do public investing? You don't have to be a hedge fund. You can just do long only. But aren't you in the business of predicting tech trends and evaluating companies? Charlie (01:24:51): After having done this conversation, we think you might be quite good at it. Marc (01:24:56): We've considered it. It's just you guys, if you spend time with public market investors, they just have a very different motion than what we do. And so they just Charlie (01:25:05): But is that tradition or is that fundamentally intrinsic to the ontology of the job? Marc (01:25:10): I think there would be a different way to run public money in a way that, for example, would've caught a lot of the Mega 7. I think that possibility exists, and literally you could just say it's as simple as apply the venture mindset to mega caps, and away you go. And obviously, obviously we now know venture scale returns when you get that right. I would just tell you, I tell one of the things that saves venture is that we're locked up and our investors are locked up. Charlie (01:25:33): It's a feature, not a bug. Marc (01:25:34): It's an incredible feature. And in traditional finance theory, they always tell you illiquidity, right? Illiquidity is a deficit. And it turns out Charlie (01:25:40): Which is true. But human nature is a bigger one. Marc (01:25:43): Human nature Charlie (01:25:44): Illiiquidity would be a feature if we were less messed up. Marc (01:25:47): It is so incredibly hard. And that gets sucked into the psychology of the moment. And I spend a lot of time at our firm trying to get people to not be sucked up for the psychology at the moment. So for example, it's just like an absolute ban on television news in the office. No, if it's on CNBC today, it does not matter to us. If it does matter to us, we made some horrible mistake eight years ago that we can't fix now anyway. And if it's anything else, we shouldn't be paying attention to it. Because the whole point of this is things that are going to take five or 10 years in the future to develop and people just need to get back to work. And I would bring that up just as like, okay, here's a very pragmatic challenge. You're running public money with a venture strategy. Alright, what's your lockup? Okay, now you got a quarterly lockup. Congratulations, big guy. The market rips your face off. All your investors redeem so much for your strategy. And so that's just really hard. And then people who have gone out to try to raise money on longer lockups are like, well, why would I do that? Illiquidity is a problem. Why would I lock up an Apple position? That's insane. And again, you can say nobody, the fact that nobody did that is illustrative of how difficult that is now. I dunno, maybe at some point we should. And then the other is just flat out opportunity cost, which is are you really going to spend the time dealing with that, that you could be spending meeting the next Mark Zuckerberg? John (01:26:52): You invest in companies that succeed and then go public. Marc (01:26:57): Can I tell the actual story? So we almost did this. We almost started the thing and we're like, all right, we have the venture mentality, we have the thing, but because of how the public markets work, we need a public market. I need somebody with some public markets background to even be able to raise the money. So we ran a long recruiting process and we got down to the final candidate and we met with him during COVID in, I'm going to say September ‘21, something around that time. And we said, look, just bring to dinner. Do the workup and bring your best idea. The one company that you would commit the portfolio to, would you like to take a guess for what it was? Charlie (01:27:29): No. Marc (01:27:29): Peloton. Oh my God. Which then proceeded to fault 99.9%. Charlie (01:27:36): So you missed a bullet. Marc (01:27:37): Right? And by the way, at the time, Peloton, and you remember at the time, you remember we used to talk about this at the time, remember Peloton was like, “Oh, this is a permanent… this isn't just a bike company. This is a movement, right? This is a cult and this is the brand and this is immediate.” And everybody had their theories, subscriptions, and recurring revenue. John (01:27:51): COVID where people overestimated the permanence of the behavior changes. Marc (01:27:54): Yes, exactly. Well, there was that, but there was also just these hardware companies, that kind of company. Fitness is a trend, fad driven business historically. And so anyway, it was just like that, just felt like a message from God. John (01:28:09): Go back to public market investing. So you invest in companies that then go off and succeed and go public, like Coinbase or Airbnb or all these sorts of companies. You then, because they're public, you get to distribute the stock. And so you distribute it to all the LPs. They get their shares, you get your shares. Do you hold the companies? Do you make a decision? Is it formulaic? Is it not formulaic? Are you secretly a public markets investor because you have to make these decisions? Marc (01:28:34): Yeah. So to be clear, there's two parts to that. The part, each of us as individuals does, do whatever we do with this stock. John Yes, but what do you do? Marc What do I do? John (01:28:42): But I'm basically saying, do you make active decisions or is it totally formulaic? Marc (01:28:45): Well, lemme tell you how we do it as a firm and then the individual. As a firm, we try to make it as mechanical as possible. We're trying to get out of the psychology of whatever's happening at that moment. You try to define a process upfront, but you do want to be discriminating. And so we have a magic box formula of things like quality. Are the founders still running the company, quality of the founders? Are they beating their numbers? What's the growth rate? What's the second derivative? John (01:29:12): What’s the service like in the pub? Marc (01:29:13): Exactly. Do they tolerate low performing bartenders? And then yeah, we have some schedule against that. There is a theory afoot and Sequoia is pursuing it that basically the venture venture firms and their LPs have left enormous amounts of money on the table by distributing too soon. And the best strategy over, if you backtest over 50 years, the best strategy, at least for the top firms probably would've been to hold everything in perpetuity. And so Sequoia notably is trying a strategy where they're trying to do more of that. I will tell you the LPs don't like that. The LPs… John (01:29:46): The LPs want their shares. Marc (01:29:47): Of money in and out, and they do have a positive argument that says, look, we're not paying you to manage public money. And by the way, they have their own needs. They have their own needs now more than ever. They're under real pressure in a lot of cases. And so if you ask an LP, they will tell you, yeah, we want you to try to shoot the lights out on as long dated horizon as possible. Having said that Charlie (01:30:10): We want our money back. Marc (01:30:12): Get us some money, please. Right. And so where this comes up is it is just the thing of, well, should we hold it for another three years and go for another doubling or should we burden hand on that anyway? So we try to run that mechanically on the individual side, I mean it really varies by the individual, just based on idiosyncratic life circumstances. Charlie (01:30:28): Off to big company world. For a couple of questions. How much should big companies focus on their competitors? Marc (01:30:35): I mean, so this is a real double-edged sword. So the easiest thing in the world is to focus on your competitors, right? Because got somebody to benchmark against, index against, and it's just been amazing how many other big companies start or stop their VR and AR programs based on whatever's doing at that moment. They seem to have outsource their thinking entirely to Meta. And so there is this dysfunctional version where you're kind of outsourcing your thought to the competitor and then there's the Peter critique of like you're giving into these Girardian kind of spirals. And I think there's something to that. Having said that, I mean, I see the other side of that all the time, which is the Andrew Grove side, which is only the paranoid survive. And isn't it great if you have an intellectual framework to be able to not think about your competition? Because that's a lot more fun if your competition's good, thinking about them is actually really painful. If you have this enlightened point of view that says you don't ever have to think about them, you're letting yourself off the hook. And so I think there's Charlie (01:31:27): Maybe the answer is whatever's most painful, thinking about them and not thinking about them is best. Marc (01:31:31): Well, and this gets to what I've experienced with big companies, and by the way, this includes in a lot of cases fast growing startups. They think a lot about their competitors for the purpose of trying to basically, essentially ultimately copycat with their competitors. You assume if your competitor is decent, you assume that for whatever it is they do, you assume they must have some analytical reason they're doing it. And so there's this natural tendency to try to build the analytical, analytical case to do the same thing. And so there's an overfocus in that way. Having said that, I can count the number of true competitive teardowns, I don't know, maybe on one hand that I've ever really seen because again, your pain point, the most painful thing in the world is to talk honestly about somebody who's beating you. John (01:32:12): Yeah, I always find that Jeff Bezos, we're not competitive focused. We're customer focused. Kind of a clever bit of misdirection because again, at Stripe, we think that our customers are very smart. And so if they're picking something else that is some signal of revealed preference that a well-informed person trying to do the best thing for them says this is better than Stripe. And so we do a lot of secret shopping. We do a lot of tearing down. We want to understand what's out there. And again, as you say, that shouldn't kind of define the roadmap. You should be able to come up with your own products, but if you're not coming at it from an informed place, something is horribly wrong. Marc (01:32:43): I think it is some combination of you need to be brutally honest with respect to what your actual issues are. And those actual issues include you're losing for reason X, Y, Z. Charlie (01:32:51): I mean, in some ways what they're saying is bureaucracies avoid pain, and so you need to steer them into pain. Marc (01:32:56): I would say it slightly differently, which is I have found people willing to tolerate any level of chronic pain in order to avoid acute pain. And so people would much rather lose slowly over five years than have the conversation that involves a dramatic change to stop losing. Charlie(01:33:10): Wow. Marc (01:33:11): And I've seen that over and over again. It's almost impossible to get people to do that. The level of aversion is incredibly high. John What founders or companies do you respect? Marc People seem fine, just bleeding out. I mean, it's just incredible. I mean you see it in other areas of…you see it in politics. I don't name names, but there are political parties, let's say in various places around the world where you just look at it and you're just like, I can't believe that you're willing to inflict the strategy on yourself with these results that are clearly not working. Charlie (01:33:39): Actually that leads to Marc (01:33:40): Yet they will not revisit their core assumptions. Charlie (01:33:42): If you look at companies that have died over the last 20 years, they do seem to have these very long, operatic deaths and they change a lesson. You would think Marc (01:33:51): Yes. Charlie (01:33:52): Do you think that's because of people that are prescient and see it just exit and so the people, you've sort of got a selection effect and the people that remain? Or is it just that it's too socially awkward to have a conversation that says we've… Marc (01:34:05): Most people would rather just put one foot in front of the other. Most people don't want to rock the boat. Most people don't want to be the skunk at the garden party. Most people don't want to call their own baby ugly. Most people don't want to. Yeah. I mean, most people, they don't want the reputation of being a troublemaker. It’s like this thing of, it's this very interesting signal. You have to decide whether you want to send as a leader, which is do you want people to bring you bad news? Because it's like if all people are doing you every day is bringing you bad news, number one, you're going to slit your own wrists because that fucking sucks. And then number two, you don't want people to just be complainers. And so maybe the most advanced version is, only bring me a problem if you're also bringing me the solution. But okay, now your life is a CEO is better, but what if there really is a problem and Charlie (01:34:44): They don't have the solution. They don't have a solution because it's beyond them. Marc (01:34:46): It's beyond them. And then they're the one that you're going to give the negative performance review to. So by the way, the other twist on the big company failing thing, which I think is really underrated, is the big companies that fail, the way the story gets written is they never figured it out. And the easy example of this is always Kodak for example, they never figured out digital photography. Well, you often find in the backstory is no, they actually figured it out and they did it too soon. Kodak had actually a very active digital camera program. Charlie (01:35:10): Then they got burned and then once burned, twice shy. Marc (01:35:13): By the way, Yahoo had mobile early Yahoo was all over mobile between 2002 and 2006. And then they got burned so hard on it that by the time the iPhone appeared it was too late. Charlie (01:35:25): Yeah, I think that if you did WAP, you were unlikely to succeed in the post iPhone world. Marc (01:35:29): Yeah, yeah. And quite frankly, I think a lot of the tech you mentioned the big tech companies, a lot of the big tech companies, they had internet fully deployed internally. They had TCP/ IP products. They actually knew it quite well. They were running it. It just was something that they were very used to that they didn't really think about in any way. And so yeah, there's the status quo bias thing. Charlie (01:35:47): So this is a good segue… John (01:35:49): Very intelligent sounding reasons as to why it won’t work from a recent attempt. Marc (01:35:53): People are really good. People are really good at the analytical explanation, as either as to why something won't work or conversely why something is going to work when it's clearly failing. John (01:36:01): But again, you just get to sound very convincing where it's like, that's a great point. We actually tried that 18 months ago. Charlie (01:36:07): Yes John (01:36:07): And no man steps in the same river twice. Charlie (01:36:11): That's a good segue into you've been on many boards, what makes a good one? Or maybe what makes a bad one? Marc (01:36:17): I mean, yeah, I mean step one is if it's a successful company, step two is if it's a good CEO. I mean the boards just can't do that. Just practically speaking, the boards just can't do that much. And even the old cliche is the hire, the fire, the CEO, and even that it's really fraught with peril. It's very easy for a board to blow that up. Charlie (01:36:38): I do remember your blog had a, “How do I hire a professional CEO?” And the answer is one sentence. You’re expecting a long article. And it's like if you need to do that, don’t. Sell your company. Marc (01:36:46): And that's probably an overstatement. And there have been some very successful professional CEOs over the years, John Chambers and Frank Sluman and others. But yeah, look, it's just really hard. It's just like is the company going to succeed or not? Is the CEO great or not? Is the company on the right side of history or not? That's honestly most of it. John (01:37:03): But even boards matter then? Marc (01:37:05): One of those things like you can't not have one, which is you don't want to run. If you are under another board, then you're as a CEO, legally liable for every screwed up thing that happens. You're much more likely to go to jail, you're much more likely for things to spin out of control. There are real requirements. Governance needs to be taken seriously. You're representing a lot of other people's money. So there's that. And then do you want to have absolute dictatorships with no examining nature inside ever? And then aspirationally, obviously the hope would be to be able to positively contribute. John (01:37:34): Yeah, you're given the governance explanation and you're saying that it's rare that founders are actually removed and our CEO are actually removed. And then even the cases where they are, maybe things are too far gone and everything, and sure, maybe that's true, but I feel like I would make a cultural pitch where, lemme try this on and you can react to it. We've found the Stripe board very useful because it's important to have to organize your thinking and have some accountability mechanism where you go on a quarterly basis and talk about things. And then we're doing this for the first time. And so there's lots of people on the Stripe board who have a different set of experience and come to us and advise us on various things. And we've gone and tried to pick the hall of fame of various industries who can then go behind on things. And I actually notice when I talked to way earlier stage founders, I think they underrate the value of a good board, where they are worried about the governance thing you say, where they don't want to give up a whole bunch of board seats and then have to do management of VC personality and everything, which is true, but they don't seem to take seriously. And again, maybe they just get this from investors, but they don't seem to take seriously the idea that you can put together a group who will meaningfully increase the odds of success of the company. I don’t know, is that just a particular thing to us, we needed more help than others or would you agree with that broadly as a cultural explanation where they're pretty useful culturally for management? Marc (01:38:58): Yeah, so what you just said is what we aspire to. So what we aspire to is that the boards that we're on are like that and that the CEOs that we work with want to have a board like that and that we're able to be a contributor to it. And so we aspire to that. I think there are many examples of that being true and hopefully on that, I've been an example of that myself. I think that's all true. Having said that, I guess a board cannot rescue a failing company. Well yeah, but there are a lot of people on a lot of boards and a lot of companies that are failing that are spending an enormous amount of time trying to rescue those companies. And so both in and outside of tech, and so just the higher order bit is still succeeding or failing and it's still quality of people versus not. Charlie (01:39:35): It ties into something you said Marc (01:39:36): The easiest thing in the world is to go on the board of a company that is going to succeed wildly no matter what you do and then to take credit for it after the fact. John (01:39:42): But presumably you believe, I mean that sounds fun, but… Marc (01:39:45): The hardest thing having been through it, the hardest thing in the world is to be on a team, on a board where you're struggling valiantly to keep the ship from going down and the ship is going down. Charlie (01:39:52): So that goes back to can you hire great CEOs or are those great CEOs? Someone wants to discover, to me, the people that have a reputation for great professional CEOs or actually great stock pickers. Marc (01:40:04): Yeah. Charlie (01:40:04): They understand tech deeply enough that they pick the company that's in a great position. Marc (01:40:09): Same thing for VC, same thing. Charlie (01:40:09): You can't hire them to turn around a phone company because they self psyched out of it. Marc (01:40:15): Know every once, I don't know, every once there's exceptions to everything. Every once in a while you get something. Charlie (01:40:18): Actually that is also a great book, which is a world full of heuristics in VC, single founders, multiple founders, but there's so many exceptions to each. Marc (01:40:28): You never back a married couple. You didn't back Cisco, right? Waiter: Last call, but you just made it. Marc: That was the saddest service. Is that the Irish, is that the Irish standard? Waiter (01:40:47): We specialize in economic infrastructure, not in hospitality service. Marc (01:40:54): I would just like to note that the result of that entire exercise is one less of my cups on the table. John (01:41:00): You're negative two drinks from that interaction. Marc (01:41:03): Net subtractive. John (01:41:05): You think people understudy the Elon method for running companies? Marc (01:41:10): A hundred percent, yes. John (01:41:12): Maybe just briefly describe that method and then why everyone is so incurious about it. Marc (01:41:20): And there's two reasons they're incurious about it. There was the original reason, they were incurious about it and now there's the new reason they're curious about it, which is Elon also generates emotion in people. Yeah, so look, you guys know that. How do you run a company? Well, there's been a hundred years of management books starting with Alfred Sloan's book, Alfred Sloan built General Motors. Alfred Sloan famously wrote a book that people like Andy Grove learned from that basically said, “Here's how you build a large multinational multi-product line industrial company.” And so there's this system and it involves somebody at the top of the company that's sort of overseeing this machine and they're getting reports and then respond to the reports. And then there's all these rules, both rules sort of inflicted from the outside and rules generated internally. And then there's Elon who just doesn't do any of that, just doesn't do any of that and has a completely different playbook. And the Elon playbook in a nutshell, as far as I can tell, I haven't worked for him directly, but from observing him and working with him as far as I can tell, it's basically number one, it's only engineers. Your company, people who matter in your company are the engineers, the people who understand the technical content of what you're doing for technology companies. And then you only ever talk to the engineers. You never ever talk to mid-level management. If you have it fine, if they need it for whatever to do their vacation policy or whatever, it's fine. But if you are the CEO to get the truth, you only talk to the line engineer. And so you just ruthlessly violate the chain of command at all times. And then your job as the CEO is every week to fix whatever is the most important bottleneck to the company's progress. And the way that you do that is you parachute in and you find the engineers that are working on that problem and you basically stay up with them all night until they finish, until they fix the problem. And then if there's no current major bottleneck, you spend your time instead doing engineering reviews, specifically engineering reviews, not product reviews, engineering reviews. And you get all the engineers together and you have them each present what they're doing for five minutes. And the result of that is every single engineer in the company, you know exactly what they're working on. If somebody's not good, you fire them on the spot. If somebody's great, you go all out to get them. Charlie (01:43:16): But what's the inverse of that? Because for 10 years after Steve Jobs, we had people doing sort of mimetic bad version, wearing turtlenecks. John Being an asshole. Charlie Exactly. I was trying to say that more diplomatically, but yes, being an asshole. What is the danger for entrepreneurs, what's the bad version of copying you on? Marc (01:43:39): Oh, the bad version is this is the critique. Actually my partner Ben Levi's critique, he's like Marc, the thing you don't get is as follows, which is that which is, that assumes you have somebody like Elon who can hold the entirety of every engineering topic in their head all at the same time. And so when you're sitting there with the 23-year-old engineer and you're working with them to redesign the database architecture or whatever, you actually are qualified to do that. Not just that one time, but every time. And then again, this goes right back to the last topic we just talked about, which is like, okay, how many of those people exist who can possibly do that? And we know the answer is one, I believe the answer is 10 or a hundred or a thousand. I don't know if it's a million. I tend to think we have more of those people than we think we do. I see a lot of founders who struggle with this because, so my observation for how founders kind of try to figure this out is in the beginning they sort of run everything. You just do everything. You just do everything, run everything you have to, and you have to have a unified vision. And you don't have this army of people anyway. And so you just do it. And then at some point your high value board comes to you and says, “You idiot, you're micromanaging. You need to bring in all these executives.” And then what happens is then you go the other way, you over delegate and then your high functioning board says, “You idiot. You're not involved enough in the details” and then you correct. And then what most of the successful founders I work with do is they end up with a hybrid model where they're deep in the details on some things, but they have a traditional system on the other hand. John (01:45:04): And do you think that works pretty well? Marc (01:45:05): I think for most of the founders we work with that have very successful outcomes, I think that generally is what they do. I think it works well, but it's not the Elon method. John (01:45:13): Sure. Marc (01:45:14): It's not the Elon method by the way, there's other aspects of the Elon John (01:45:17): I was going to say I feel like Marc (01:45:18): There more, there's other aspects. So another aspect of it is the function and purpose of the legal department is to file lawsuits. And I am not interested in all the rest of this stuff. You can go deal with it if you want to, whatever, whatever. But let's talk anybody who goes up against us, we are going to terrorize, we are going to declare war. And then of course as a consequence of declaring war, we're not always going to win all the wars, but we're going to establish massive deterrence. And so nobody will screw around with us. By the way, let me give you number three, which is becoming more and more salient. I think, and something we're trying to get our founders to do a lot more of. Number three is it's going to be a cult of personality and it's going to be a cult of personality, not just inside the company but outside the company. And we're not spend any money on marketing. We're not going to put any time into IR. What we're going to do is we're going to put on the show of all time and the company and the stock and the books and the videos and the products and the jobs are all a function of the cult of personality. John (01:46:09): I would add three things to that list too, and you can tell me if you think you agree. One is a focus on, and by the way, I thought the Walter Isaacson book got kind of a mixed reception, but I thought if you want to study the Elon method a bit, it was actually pretty useful for that. And so, the recent biography, one is picking sensible metrics for the business at any one moment in time. And so with SpaceX and as they're kind of building up the launch business dollars per kilo to orbit being the metric that we're going to optimize for. That's not totally obvious that it falls out. Even Tesla as they're wrapping up production, it's like deliveries per week. You could have focused on revenue, you could have focused on profitability, you could have focused on deliveries per year. The number of deliveries per week rolling off the factory line is itself an interesting choice of high level metric. So a big focus on, I think there's lot of this in Twitter as well, when he took it over, focus on what are the right metrics that we should be. And some of the criticism that's been levied at X is their focus on engagement minutes on the site has led to things like the ban on URLs, which I think a lot of people think is not de-boosting of URLs, which a lot of people think is pretty similar. So one is choosing the right metrics. The second is creating a sense of urgency. And people talk about this as inventing crises. But I would say the generous version is shortening the time horizons. And so it's funny, Elon was going around talking about it when he was sleeping on the floor of the factory in Nevada for Tesla that, “Tesla will go bankrupt if we don't do this and if we don't figure out model three production.” Tesla was a 200 billion company by market cap at that time. So it's like Tesla would go bankrupt or do a very, very non-dilutive equity raise, but creating a lot of urgency around this idea of fixing production and sleeping on the factory floor, which clearly shortens the timeline. And then the third is actually the businesses are really capital efficient. So I'm curious if you see this with hardware companies. I think sometimes hardware companies can be really indulgent with capital where they say venture capitalists will fund my vision of exploration for five or 10 years. And this is the risk now is people get into robotics and stuff like this that you get this self-indulgence and it's like, “I'll do my science project for ages and then I'll maybe figure out a product and figure out how to commercialize it.”. Charlie (01:48:18): I would say the other thing hardware founders do is they fall in love with the hardware and the product and they can almost get in, sort of redefine themselves as producers of science or beauty or product and sort of forget they're running a business. Or even worse, start to think of running the business as slightly unpleasant, beneath them. John (01:48:35): Exactly. Charlie (01:48:35): And maybe even not sort of intellectual enough. John (01:48:38): And so Elon's companies have always been very capital efficient and build a bad one and then build a good one. And so the boring company bought a commercial tunnel boring machine before they started developing their own. Tesla had the master plan where they build a low volume Roadster before they get to the high volume stuff. SpaceX just for what they do has never actually burnt that much capital lifetime and got grant money they were selling to the DOD, all this kind of stuff. So yeah, would you agree with those three? And do you think people can pick and choose? We can take some of those things without maybe the lawsuit department or something. Marc (01:49:08): Yeah, so I think that's all right. I would maybe add one more thing or kind of distill it out of a bunch of these, which is basically truth seeking at all cost. At least I find this to be the case with him. And I think this is really not, especially people who are mad at him, really don't understand this. He really, really genuinely wants to know ground truth and he really genuinely does not want to know anything that's not ground truth. And again, it goes back to our thing of how to confront bad news. He's absolutely ruthless and relentless in making sure that he actually understands what's going on. And you would think that that's common. And I've not found that to be common at all among people in business. Or you mentioned related to another thing, which is with Elon, “We're all going to die if we don't get this, we're all going to die.” Every other typical startup founder me when I was doing it, it's always like you're always trying to come across, you get front the company, you should be optimistic, brave face. Marc (01:49:56): It's going to be great. Really have faith. You should have faith. You shouldn't quit and go to another company, please stay with us. You're going to be great. John (01:50:04): Trying to weed out the nonbeliever or something? Marc (01:50:06): Apparently I think it's urgency, but literally it is just to be the guy who can show up there and just be like, yeah, if this doesn't happen or it's going bankrupt. I mean the number of other companies where that would happen that would just, okay, the talent would just bleed out. And then maybe I could add one more thing to this, which is he has what—you mentioned Steve—he has what Steve had, which is the people who worked for Elon and the people who worked for Steve, they often report after the fact that they did the best work of their lives. And they often report that they could have had difficult interactions along the way, or they could have had whatever, or by the way, maybe it didn't even end well. But literally they'll say like, “Wow, I got to work on the iPhone.” Charlie (01:50:48): There's a lot of very good SpaceX founders and they imbibe a work ethic that sort of reminds me about Goldman Sachs in the 1990s or something where they work incredibly hard and they work, they think from first principles and they're truth seeking. Marc (01:51:03): Yeah, that's right. Charlie (01:51:03): And they're risk taking both technically and they're risk seeking technically and risk avoiding in business. Marc (01:51:10): So then my version of your question is, I call this the question of the milli-Elon. It is like, okay, if Elon is a fault, Elon is a thousand milli-Elons John (01:51:20): You can microdose. Marc (01:51:22): Yeah, can you microdose, right? So can you operate at the level of a hundred milli-Elons or at 10 or at one, right? And a huge number of observers of Elon swear, it's a classic thing. He gets the classic feedback. Steve used to get this feedback. Lots of people get this feedback just, wow, you're great if you could just only just do 80%, if we could just get the 800 milli-Elon version and you could just not do the other 200 milli-Elon's, you'd be so much better. And literally, that's what I found with these guys is they've heard that a thousand times and it's a completely no up of a statement because there is no, for them, there's, there's no reduced version. And so if there's no reduced version of it for them, is a normal person going to be able to construct an optimally titrated dosage of milli-Elon's? Marc (01:52:06): And I aspirationally believe that you should be able to learn things and replicate, but it is a system. It's not just a set of practices, it's an entire worldview. John (01:52:21): I'm not sure. It's a whole system where if you don't have one thing, the whole thing falls apart. I feel like you can. Marc (01:52:27): Well, the other part of that though, that would be one. The other way of looking at that though, is the person capable of doing the partial version? John (01:52:35): No. Marc (01:52:37): You see what I'm saying. John (01:52:37): That I can buy. Marc (01:52:38): Are there people who can do the 300 milli-Elon version of it? Maybe. I wish I had met more of them by now. And then the other side of that is why is it understudied? And literally, I think this goes back to the same thing as why do people get mad about cryptocurrency is I think it’s— John It's tribalism. Marc Yeah, it's something, there was always something about him and how he operated that caused people to have an emotional response and then that is now magnified a thousand X or a million X and people are just not having it. Part of it is he's polarized the market very deliberately in the same way that I think a lot of great entrepreneurs do, which is unique. People tend to either love him or hate him. They either love the products, they hate the products. That's very helpful from a business standpoint, recruiting standpoint. It does create this cult-like thing. The thing you don't want in any market is lack of differentiation. He a hundred percent always has that. But as a consequence, I believe there are a lot of people who should be learning a lot more from him who cannot bring themselves to do it and to their own detriment. John (01:53:38): Can I talk about the media? So I feel like my framework is that there are often these new technologies that then cause an explosion in interesting media activity in new companies and things like that. And so there was the cable boom, and I'm excited for John Malone, this new book, but I sound interviewed him recently and he was talking about, they just thought up a lot of new channels when they had this pipe going to people's homes that could support a lot of programming that to invent new programming or to talk about creating Fox News. They were like, “Well, the existing channels seem a little bit to the left and conservative talk radio is really popular, so it seems like conservative news channels should work really well.” And it did. So that was cable and the internet came along and famously really worked from a media perspective. And in particular that was the big nail in the coffin for local newspapers where they were the main distribution outlet to people previously for information. And the internet went over the top. I feel like plausibly X is a big enough change to be a new media platform. A slightly trivial example, but TPBN is kind of a CNBC competitor, where I saw Matt from ElevenLabs, a great a16z company, and they did a fundraise and he went on TPBN to talk about it, but previously that would've been CNBC. but now TPBN is where he chose to go. And that's one example. There's lots of others. Is X that big a deal from a media perspective as to be kind of cable, the internet, then X or something? Marc (01:55:06): I think it is. And maybe the twist I would put on the TPBN or the cable thing is one of the things, actually, this is also what I'm about to say, a big deal in sports. There's also now the clip. And clips used to be weird, esoteric, and now clips are the main way that people can consume content. John (01:55:22): I see. So X and short form generally. Exactly. Marc (01:55:24): Yeah. So for example, a TPBN episode or for that matter, a sports game, now generates five or six or eight clips, or an interview or hopefully this discussion and then those clips go hyper viral if you're doing it right. But it's very common when you look at the analytics that the clips get like a thousand times the distribution than the actual program itself. And so is this, I think there's this art form, it is one of the reasons why a lot of historical television shows never figured out what to do with the internet. They didn't really understand that the internet native artifact was the clip. But the new media properties, the new media entrepreneurs, I think tend to really understand that. So yeah, I think that's true. Having said that, the impact of the internet is still mostly what it's been this whole time, which is a disintermediation mechanism. In the cable era, there were only 200 channels or whatever it was. In the internet there's a billion. So the overwhelming trend is still disintermediation, desegregation. John (01:56:16): Democratization is the other big trend to me in media right now. Marc (01:56:19): And Substack is a great example. Of course, Substack is a centralizing phenomenon. It is a singular platform and we have growth starts and are proud when they go. Everything's unbundling and bundling. Exactly. But it is not a rebundling in the form of a new magazine. And specifically the way that Substack thinks about it is they're not a publisher. There're a platform. And the distinction is they do not have editorial judgment. They're not trying to create bundles. John (01:56:42): The economics are different for the publishers. It's land reform for journalists. Marc (01:56:46): Yes, exactly. Exactly right. But again, there you would still say notwithstanding the success of Substack is like a centralized platform. It's overall effect is still disintermediation because specifically what it's doing is it's bleeding off many of the talented individual contributors at legacy media to have their own Substacks. John (01:57:01): It somehow feels to me like we're not done with the media changes. Marc (01:57:08): I think that's true for sure. Yes. John (01:57:09): Sorry, the media changes wrought by just this latest platform change of X and clips. The fact that again, TPBM of which I mentioned just because in our corner of the tech world, is from this year, last year. It’s a very new thing and we haven't seen all the last changes. Do you have any predictions? Marc (01:57:27): For sure. I'd expect to see more of those. Again, I would just say, look, what is the macro thing, the big macro thing happening, and I love what those guys are doing and I love what Substack are doing, but the big macro thing, if you just think about the world change, the big macro thing is TikTok, Instagram, and then short form video on X and a handful of other platforms. That's the macro thing. And so where the future of the macro culture goes, I mean look, I read Substack, but a thousand or 10,000 or a hundred thousand times more activity is happening on TikTok. And so the macro culture is going to be shaped I think much more by short form video, at least for the foreseeable future. And then as I'm sure is obvious now, but the role of AI production is about to really change things. John (01:58:15): And there also maybe the fact that there's a single global feed now, the fact that there's much less personalization in a way because so many things go to the top. And in a way, I really actually don't like the number of videos in my X feed these days. I'm sure they perform in the metrics or something like that. But if I wanted to scroll TikTok, I'd open TikTok and I don't want all the TikTok videos that get crammed. Do you guys get these in your feeds where you get just random TikTok videos from random accounts in your Twitter feed? Like no, I'm reading a newspaper here. I'm not trying to watch tv. Marc (01:58:49): Yeah, no, this was a big, I think the people who run these things have talked about this publicly, but all the old algorithms of things that your friends like, those are not as effective as just the macro. Charlie (01:59:00): We are more similar than we think. Marc (01:59:03): But also the nuances and interconnections are more subtle. Charlie (01:59:08): It's not the people, it's the people you don't know that you have connections with. Marc (01:59:11): Yeah, exactly. Right. You're probably more like a lot of other people you've never met than you are with the people you know. By the way, having said that, I believe the biggest, I think everything we just talked about is very important. I think the biggest thing that's happening is just we really, I think for the first time we're entering the true era of free speech. And I think that we started to get at that in the nineties and 2000s, and then there was a big reversion in the 2010s with the sort of censorship industrial complex that formed up and all the policies and all the government interference and so forth. And of course a lot of that continues on the part of the governments in particular, but in the US at least that project has failed and the platforms themselves are really liberalizing out. And then just the sheer volume and scope and variety of content and the number of ways that people have to get messages out in all kinds of ways in the hyper acceleration of culture where the censors don't even know what to ban. They don't even know what half the stuff means. We probably are living in the only true mass era of free speech in human history. And you're seeing things now, this is all the play, not real time, but you just see things now as just a normal user that you never would've seen 10 or 20 or 30 or 50 years ago. There's not even a chance. John (02:00:26): So does it lead to a political realignment? Marc (02:00:27): I believe it does. Yeah. So I think this is the big thing. I think Martin Gurri is the guy who has— you guys published his book. I think he really nailed it. And I think his thesis in his book came out in 2015, and I think a lot of people said either, “Wow, he predicted Trump.” Which is true to some extent, but that's not the big thing that he predicted. And then I think his prediction is in some ways so fundamental that it's easy to just kind of take it for granted and say, “Oh, of course that's what's going to happen.” But it's like, it's actually so fundamentally important, I can't stop thinking about it. Which is basically true transparency, true free speech. It's a fundamental solvent, basically dissolving all centralized institutional authority. And the reason for that is centralized institutional authority is never perfect, and it often has problems. (02:01:09): And in fact, it often has very deep and severe problems as we were just discussing. And the kind of show that a government agency or a big company could put on to claim that they're better than they are, that would've worked under centralized media, just simply collapses under conditions of true free peer-to-peer communication. There are just too many examples of too many things that go wrong for any institution for them to retain their credibility. And then Martin and I have this big debate about this. When I've talked to him about it, I'm like, “Wow, that's fantastic.” And he's like, “No, Marc, I didn't mean this was good. I never said this was good.” He said to me the following, he said, “Look, it is true that every major institution is much, much more broken than they have been putting on.” He said, “However, it is also true that we do not know how to run a society without large centralized institutions.” And so he said, “Those of you like me, who cheerlead the collapse of centralized institutions have not yet come up with an answer for what exists on the other side.” But anyway, point being, I think now we're really going to go through that. Charlie (02:02:03): The business version of this is you used to be able to push a bad product for a strong channel with strong marketing and sales. You just can't do that anymore. The product quality will out. It's deterministic. Marc (02:02:15): Yeah, that's right. And by the way, you get this phenomenon, you see this all over the place and you see this in, Gallup does this great poll of trust of trusting institutions. And the numbers are just all cratering and the crater, the declines are accelerating. You also see, and again, not to pick on specifics, but you also see this in these political parties. And you have a lot of this happen in Europe right now where these parties come in and they have whatever, 60% approval or whatever, and then six months later they have 15% approval. It's like, what the hell? Right? I mean, I'll give you an American example. Eric Adams in New York as the incumbent has 9%. And it's just like, how can you possibly have a system in which the ruler has a 9% approval rating? It's like, well, how did that happen? Well, it's all too transparent. Everything that's going wrong is too transparent. It can't be finessed. Charlie (02:03:04): The extreme version of this for good or ill is that the centralized state is an outcome of centralized media. The nation state is downstream from the newspaper. Marc (02:03:14): Right, exactly right. And so yeah, you can't hold it together. My counter arguments to Martin was basically, if you look at what the media landscape was like in colonial America, it was actually much more like what it's like now than it was in the 1950s, and you'd have 15 small newspapers in up the city like Philadelphia, and you'd have just enormous amounts of contention and name calling and all kinds of things. John (02:03:42): Anonymous bloggers. Marc (02:03:42): Anonymous bloggers, they had all that stuff. Benjamin Franklin literally wrote under 15 different pseudonyms and he would set them to be fighting. Charlie (02:03:48): Voltaire, the same. Marc (02:03:49): Fighting with each other, all these things. And it looks like we've lived this before. And he's like, yes. And it was a time of revolution. Correct. And so to me, and to me it's so fascinating. We're really in that now. I feel like that was still being held back as late as last year by the censorship apparatus. And now it's just like, okay, now it's all coming out. And maybe another way to think about this is the narrative for the last decade has been, the internet is a fountain of misinformation. And there is some truth to that. There's a lot of misinformation online. But the other thing is, according to the Martin Gurri thesis, the internet is an x-ray machine because every actually correct thing that all of these institutions are doing wrong is now being fully ventilated for the first time ever. And they cannot survive that. And that may ultimately include the governance themselves. John (02:04:40): We're describing one trend here, which is the move along the decentralization/centralization spectrum. And I think I'm not quite as enthusiastic as it seems pretty complex, that whole spectrum. But the other change to me again, seems to be the single global feed that's emerging. So take an example, the Astronomer CEO and that whole thing with the CEO with his HR lady being caught on video. That was just the front page of the internet for that day. Or two days. I was talking to someone who was saying they were talking to someone in China and they were joking about it, but it was just as prominent in China as well in the news there. And that didn't happen as much 10 or 20 years ago. And I don't even think it happened again, even when we had the internet and cable media because we didn't have the clip and the ability for the things to go as quickly. Charlie (02:05:28): I guess text is much more language barrier. But it's also language barriers prevent text from crossing borders, clips can cross borders. John (02:05:37): And just recommend algorithms for things right at the top, I think. So all these factors, do you just have thoughts on the implications of having a single global feed? Marc (02:05:46): Yeah, so this is right. This is kind of the monoculture, like global monoculture. John (02:05:50): Maybe. I mean— Marc (02:05:51): Well, Marshall McLuhan had this concept called the global village, and this is another one of these things where he said, “People think I meant it positively, and I actually didn't.” So he said electronic media formed the entire world into a global village. And he was talking about TV, but TV you could say had a certain version too, the early version of that, because it just spread a single video feed much more broadly. And what he said is, look, he said it used to be that every village was its own village. And so the things that happened in that village, if the wrong man kissed the wrong woman, it was a really big deal in that village, but it wasn't a big deal in the next village. Didn't even know about it. Now all of a sudden the entire world is becoming a single global village. And he said, here's the problem with that, is that villages are really fucked up. They're really dysfunctionally fucked up a lot of the time because everybody sees everybody else. They're tremendously judgmental. There's tremendous… the social relations have carried tremendous weight. If you end up getting sideways with the social relations of the village, you're in serious trouble. You might get exiled, you might die. They're prone to manias and panics, witch trials. They tend to go, they're hothouse environments. They tend to go crazy. And then specifically, I think the next version of that, I dunno if he said this, but other people said this is like, cosmopolitan societies are, they're written in nature, and they have the ability to have dispassionate communication. Villages are all about reality, right? It's all oral, it's all spoken. And again, it's this social hothouse I've spoken, and therefore highly emotionalized, de-intellectualized, highly emotionalized content. John (02:07:30): So Marshall McLuhan partly was writing about TV culture, but he was actually pressing into the clip culture. Marc (02:07:34): That's right. And then I think what he would say if he were here today is, I think he would say, “Yes, congratulations guys. You've got the global village.” He would say, “the Bible has the parable of the Tower of Babel being a disaster for a very specific reason. If you centralize everybody into a single giant village, you're going to have all the dysfunctionality, you're going to have the crazed panics and freakouts of a village basically happening all the time, which is in fact what we see.” I think our friend Tyler Cowen at this point thinks this is all very bad. McLuhan definitely thought it was bad. On the other hand, I grew up in a small town. It wasn't that great. The disconnected small town wasn't that great either. Is it really better to live in a world where there's only a few places where there's access to advanced thinking and cosmopolitanism? Or is it actually the fact that everybody on the planet can now be a full part of society and culture? John (02:08:32): I feel like there's a Marc Andreessen worldview that you've talked about enough that it's now a thing that exists beyond you. That's maybe just being dispositionally optimistic on technology generally and refusing to break any false nostalgia about the past. I was there in rural, small town Wisconsin. It wasn't good. Marc Yes, exactly. To cheeky pints. John (02:09:10): Exactly. Marc (02:09:11): There you go.