The Most Dangerous Investment Cycle In History | Paul Kedrosky

In Episode 505 https://hiddenforces.io/podcasts/the-most-dangerous-investment-cycle-in-history-paul-kedrosky of Hidden Forces, Demetri Kofinas speaks with venture capitalist, former Wall Street equity research analyst, and independent market and technology researcher Paul Kedrosky about why the AI investment boom is unlike any previous Capex cycle, how loose credit, data center real estate, competing AI utopian and collapse narratives, and a US-China technological arms race converge to make this moment more consequential and more dangerous than the dot-com bubble, the telecom buildout, or the railroads, and why the US economy may be headed toward a prolonged, Japan-style balance sheet recession and what that means for investors, businesses, and the public. The first hour covers Paul's framework for understanding AI as a general-purpose technology that substitutes for, rather than complements, human labor, and what that means for the scale and speed of layoffs, the repricing of white-collar work, and the social disruptions already underway. They also discuss the unprecedented customer concentration behind the frontier AI labs' revenues, how much of recent US GDP and stock market growth is attributable to AI-related capital expenditures, and why the fiscal space to respond with debt-financed stimulus will continue to narrow. The second hour examines how this disruption is showing up, or has yet to show up, in the physical world, from the demographic pressures pushing aging societies like China toward aggressive adoption of robotics and automation, to the technical dynamics that make warehouses and logistics far easier to automate than health care and skilled trades. They also discuss the probabilistic, multiplicative nature of AI-related risk, why meaningful regulation is likely to arrive only after a major failure, how investors can realistically diversify when even broad index funds are a concentrated bet on AI, and what Paul sees as the most likely trigger for an eventual repricing of AI-related assets. Subscribe to our premium content—including our premium feed, episode transcripts, and Intelligence Reports—by visiting HiddenForces.io/subscribe https://hiddenforces.io/subscribe. If you'd like to join the conversation and become a member of the Hidden Forces Genius community—with benefits like Q&A calls with guests, exclusive research and analysis, in-person events, and dinners—you can also sign up on our subscriber page at HiddenForces.io/subscribe https://hiddenforces.io/subscribe. If you enjoyed today's episode of Hidden Forces, please support the show by: Subscribing on Apple Podcasts https://podcasts.apple.com/us/podcast/hidden-forces/id1205359334, YouTube https://www.youtube.com/channel/UC8URhgYos5fjHqFSO4RSIEg, Spotify https://open.spotify.com/show/0TK6u251zGSlMgJKH6cOaP, Stitcher https://www.stitcher.com/podcast/demetri-kofinas/hidden-forces, SoundCloud https://soundcloud.com/hiddenforces, CastBox https://castbox.fm/channel/2629539, or via our RSS Feed http://hiddenforces.libsyn.com/rss Writing us a review on Apple Podcasts https://podcasts.apple.com/us/podcast/hidden-forces/id1205359334 & Spotify https://open.spotify.com/show/0TK6u251zGSlMgJKH6cOaP Join our mailing list at https://hiddenforces.io/newsletter/ Producer & Host: Demetri Kofinas Editor & Engineer: Stylianos Nicolaou Subscribe and support the podcast at https://hiddenforces.io https://hiddenforces.io/. Join the conversation on Facebook https://www.facebook.com/hiddenforcespod/, Instagram https://www.instagram.com/hiddenforcespod/, and Twitter https://twitter.com/hiddenforcespod at @hiddenforcespod Follow Demetri on Twitter at @Kofinas https://twitter.com/kofinas Episode Recorded on 10/21/2026

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Demetri Kofinas?

  What's up everybody? My name is Demetri Kofinas and you're listening to Hidden Forces, a podcast that inspires investors, entrepreneurs, and everyday citizens to challenge consensus narratives and learn how to think critically about the systems of power shaping our world. My guest in this episode of Hidden Forces is Paul Kedroski, venture capitalist, former Wall Street Equity Research Analyst, and one of the most widely followed independent researchers covering markets and technology, who publishes his analysis to institutional investors and the public. By the end of today's conversation, you will understand exactly why the AI investment boom is unlike any previous capital investment cycle in history,

• 0:43Demetri Kofinas?

  combining loose credit, a real estate dimension in the form of data centers, an unprecedented set of technological narratives oscillating between AI utopianism and civilizational collapse, and a technological arms race between the US and China for geopolitical primacy, all at the same time, and why that confluence makes this moment both more consequential and more dangerous to the economy and financial markets than the dot com bubble, the telecom buildout, or the railroads that came before it. Paul and I spent the first hour of our conversation today, mapping his framework for understanding artificial intelligence as a general purpose technology, unlike those that preceded it, one that more immediately

• 1:25Demetri Kofinas?

  substitutes for, rather than complements, human labor, and what that means for the scale and speed of layoffs, the repricing of white collar work, and the social disruptions already underway. We also discussed the unprecedented customer concentration behind the frontier AI labs revenues, how much of recent US GDP and stock market growth is actually attributable to AI-related capital expenditures, and why Paul believes the US economy is headed toward a prolonged Japan-style balance sheet recession, even as the fiscal space to respond with debt finance stimulus will continue to narrow. The second hour turns to how this disruption is showing up, or yet to show up, in the physical world,

• 2:07Demetri Kofinas?

  from the demographic pressures pushing aging societies like China's toward aggressive adoption of robotics and automation, to the very different technical dynamics that make warehouses and logistics far easier to automate than healthcare and skilled trades. We also discussed the probabilistic, multiplicative nature of AI-related risk, why Paul expects meaningful regulation to arrive only after a major failure, rather than an anticipation of one, how investors can realistically achieve diversification in a market where even broad index funds are a concentrated bet on AI, and what Paul sees as the most likely trigger for an eventual repricing of AI-related assets. Lastly, because this conversation deals with

• 2:51Demetri Kofinas?

  investing, nothing we say on this podcast can or should be viewed as financial advice. All opinions expressed by me and my guests are solely our own opinions and should not be relied upon as the basis for financial decisions. If you want access to all of this conversation, go to hiddenforces.io/subscribe and join our premium feed, which you can listen to on your mobile device using your favorite podcast app, just like you're listening right now. If you want to join in on the conversation and become a member of the Hidden Forces Genius Community, which includes Q&A calls with guests, discounted access to third-party research and analysis, and in-person events like our intimate dinners and weekend retreats,

• 3:34Demetri Kofinas?

  you can also do that on our subscriber page. And if you still have questions, feel free to send an email to [email protected] and I or someone from our team will get right back to you. And with that, please enjoy this incredibly timely, important, and valuable conversation with my guest, Paul Kedroski. Paul Kedroski, welcome to Hidden Forces.

• 4:02Speaker 01

  Hey, thanks for having me.

• 4:04Demetri Kofinas?

  I'm excited to have you on. This is your first appearance on the podcast.

• 4:07Speaker 01

  It's the first appearance on this podcast. Unfortunately, it's not the first appearance on any podcast.

• 4:11Demetri Kofinas?

  No, you are a prolific podcast guest, Paul. How long have you been independently? So actually, let's just start with a background for people that don't know you. Just give me your story here. Like, how did you get involved in this space? Did you come in from the finance side? Did you come in from the technology side? Like, what's your origin story? And how long have you been independently producing research and putting it out for the public?

• 4:32Speaker 01

  So I come at it from both the finance and technology side, I suppose. I mean, my day job, one of my day jobs is I run a series of venture funds with a partner of mine that we've been doing for almost 15 years. You know, so I come at it from a set of financial technology side, but way back, way back, I was actually an equity research analyst at one of the major brokerage firms. And so I, where I was coming at this largely from a financial standpoint, this by which I mean, you know, financial systems and how they function and how companies do what they do. And so I come at it from sort of a synthesis of those two perspectives, I guess. And way, way, way back in grad school, my PhD work, my grad school work was in sort of complexity and path dependency about how systems get

• 5:16Speaker 01

  into fragile states and what the consequences are. So in a sense, in a very loose sense, it's all kind of linked together, at least it is for me and my brain. And so, you know, that's how it all fits. In terms of independent research and so on, I was a university professor way back when at the University of British Columbia, University of California, San Diego. I still sit in as a fellow at MIT, at the Institute for the Digital Economy. I'm also a fellow at the Man Oxford Institute in Oxford. And so I still kind of maintain affiliations with universities, but I just found it was more interesting for me to do stuff semi-independently and just maintain affiliations with places that have interesting people doing interesting things. So I suppose that's, I've been doing that for the better part of 15 years plus in various forms.

• 6:00Demetri Kofinas?

  Yeah, it looks like you've done a great job creating an independent media company through, what is it, paulkedroski.com? Is that where?

• 6:06Speaker 01

  Yeah, I use that as a vehicle for a lot of my stuff, just because, you know, what's the old line from The Dark Knight? If you're good at something, never do it for free, I think is the line the Joker uses. So don't do it for free. So I do a lot of stuff like this, where I like to get out and talk to people, but I also have a number of institutional clients that I spend a lot of time with, but, you know, a small number, just because it eats your calendar if you're not careful, and mostly hedge funds and the large institutional investors.

• 6:33Demetri Kofinas?

  Well, I've really enjoyed your research. I've benefited from it enormously, so I can't recommend it enough for listeners. Were you an equity analyst during the 90s tech boom and bubble? Were you covering?

• 6:45Speaker 01

  I was indeed. Yeah, yeah. So that's, that was my scar tissue. And I think, you know, it's one of the ways I approach all of this stuff, not just in terms of systems, but is in thinking almost in terms of demographics. And we often repeat cycles in capital markets as people age out. And so a new cohort comes along who doesn't have any scar tissue from, you know, pick your favorite financial meltdown. It could have been the telecom bubble. It could have been the global financial crisis. It could have been the dot com period, whatever. And so, you know, having had that exposure, I think, was, for me anyways, was incredibly useful. I was a friend of mine, who's one of the largest short sellers in the world, was recently at a conference, and he put his hand up and asked, there was like 800 people in the room, and he asked, what percentage of you, and these are all fellow traders

• 7:29Speaker 01

  and hedge fund managers, were active during the dot com bubble? And it was less than 30%, which I think is really important, because that kind of, you know, we're all human, right? So that kind of cognitive and emotional scarring in terms of creating a, you know, receptivity to noticing things that look more fragile and dangerous, it helps if you've been through it. And the reverse is also true, that if you're completely inured to it and you don't care because you've never experienced it before, you often have an incredible run of luck in the markets, because you have no sense, like, you know what, I don't know what's going to go wrong. Nothing's ever gone wrong for me in my career. And well, how long's your career? Eight years? Okay, well, I get it then. So it can work both ways.

• 8:08Demetri Kofinas?

  Absolutely. So what kind of technology is AI?

• 8:10Speaker 01

  So for me, it's a general purpose technology, a GPT, which is to say, its best analogies are things like the canals back in the early 19th century and 18th century. It's like the railroads in the late 19th, electrification in the early 20th, the interstate highway system in the mid part of the 19th or 20th century in the United States. These are all general purpose technologies, which is a fancy way of saying that it's a general framework on top of which other things happen. And so for me, this is the way I think about it. And so that leads you to a bunch of metaphors in terms of thinking about it, like a utility, about its ubiquity, and about it as an underlying substrate on top of which a lot of things

• 8:53Speaker 01

  happen. So much like, as I said, much like railroads, electricity, the interstate highway system, even fiber in the late 20th century. These are all examples. So that's the broad metaphor. It's unfortunate that, you know, ChatGPT stole the GPT part and kind of repurposed it, but nevertheless, GPT has a much longer history than ChatGPT. So it's been something of a term of art within the industrial economics and economic research community of technology, which is this idea of a substrate on top of which things happen. So AI is a GPT.

• 9:25Demetri Kofinas?

  So how does it compare to other general purpose technologies, like the ones that you mentioned, in terms of how disruptive you expect it to be, at what scale, and how quickly and deeply those disruptions will diffuse throughout the economy and society?

• 9:38Speaker 01

  Yeah, so broadly, and I think that's a really important distinction to make is that as a GPT, it's very unusual because while it's a substrate, it's much more of a substitute substrate than necessarily a complementary substrate. So think about it like, let's take the example of the railroads. The railroads were really lethal for, say, horse-drawn carriages, as were highways, in the sense that they created a really effective competitor, but it didn't really change much in terms of humans. At the margin, there were people who changed occupations. There wasn't as much horse grooming going on and so on. But they were much more complementary in terms of introducing new ways of doing things, potentially increasing productivity. The difference with AI is that it's an incredibly

• 10:22Speaker 01

  broadly applicable technology that, for the most part, at least in its current incarnation, is really focused on symbolic work, what we sometimes call white-collar work. So it's and in those domains, it's at least as substitutive as it is additive, in the sense that it's very straightforward to see how it can replace tasks and occupations. And, you know, I consciously make that split, as well as making people potentially more productive. So in that sense, it's much more consequential in terms of its likely impact on humans, as opposed to saying, you know, I don't know, the emergence of railroads made it easier to move boxes around the country, but it didn't necessarily change that much in terms of individuals at scale. So the two things that are different, it's much more substitutive, and it's happening

• 11:06Speaker 01

  much faster. So the two together make it much more consequential in terms of the likely disruptive force that it has in the economy. And I think, while that's better understood than it once was, I still think that's part of the problem in terms of people. People have a bunch of rote mental models in terms of how they think about GPTs. You'll hear lines of patter like, "Technology always creates more jobs than it destroys." And this is based on a very small sample set of prior GPTs, and most of which had the characteristic of being very much complementary, not substituting. And so, yeah, that was sort of true, but it was true in a very small number of cases, because we only have, say, N of five, as

• 11:48Speaker 01

  an example, maybe, of prior, you know, wide-scale GPTs in a Western economy. So you have to be really careful about applying these blanket rules, like, it always does this, or never does that, because we just don't, we don't know. That's just not a, you can't say that from the data we have. And in particular, you can't say it, given that most of the analysis was ex-post. We know what happened afterwards, but we don't know what would have happened if in prior bubbles, people had been saying the same thing. If every time we had a prior bubble, people said, "Don't worry about it. Invest as much as you want," because they always create more jobs than they destroy, that turns the thing around, right? Suddenly, you take something descriptive and make it prescriptive. You're

• 12:31Speaker 01

  telling people, here's what to do, as opposed to, here's what happened in the past. So that's, for me, one of the key distinctions, is that it's much more of a substituting technology than a complementary one. We see that in the early data. There's new data on call centers, for example, from Ravello, that's really compelling here. Lots of other things out there that you can point to. Even Anthropic had some new data recently, that a report last week, talking about it's changing the nature of really, sort of, entry-level hiring, in terms of them doing less entry-level hiring. And you're seeing exactly what you would expect for a symbolic substituting technology being introduced at scale.

• 13:05Demetri Kofinas?

  So I'd like to double-click on your comments about speed and substitutability.

• 13:09Speaker 01

  Yeah.

• 13:10Demetri Kofinas?

  Does that mean that you concur with people like Andrew Yang, who are ringing the alarm bell, that we're going to see huge layoffs, and this is going to be socially disruptive in a way that previous technologies were not?

• 13:19Speaker 01

  Oh, absolutely. I think that's baked in here. And it's probably even more so because of the speed at which it happens. So you can separate the two aspects of it. One, that it's kind of a, you know, a heat-seeking missile aimed directly at white-collar work, and also because of the scale and speed at which it's happening, that's unprecedented. And you can think back about the railroads, which took like 25, 30 years to really reach full deployment, or think back about the Industrial Revolution, which was 30 or 40 years, in terms of, you know, the true scale of what happened. Or even the fiber moment was like five or six. All of these were much longer. This is happening with a much larger amount of spending and at a much higher speed. So we're not well-equipped to deal with that, in particular, whenever we keep applying this old metaphor of like, you know, it's

• 14:03Speaker 01

  sort of, I don't know how familiar you are with, there's a number of observers of this stuff. Like, one of the ones people like to point to is Carlotta Perez, who famously wrote this book, "Technology, Revolutions, and Financial Capital," which sort of describes how all of, some of these moments happen, and how we go through these wrenching changes. And then there's sort of this element of moment of disruption, and then there's losses and write-downs. And then on the other side, there's a new sort of golden age of technology. And the problem with these kinds of ways of thinking about what's happening and applying it to AI is, again, that's descriptive and ex-post. It's not that we went into the telecom bubble, saying, "It doesn't matter how much you spend, because it always works out," or that we went into the railroads build-out and saying, "It doesn't matter how many lines of railroad we build, because it always works out." This is just something, a conclusion

• 14:48Speaker 01

  we've come to after the fact. And so that's one of the reasons why this moment is more dangerous, is because we keep applying the wrong metaphor, these sorts of descriptive metaphors, as opposed to a sort of a proscriptive one, and saying, "Okay, how is this going to play out?" And the other reason why this one's really unusual, and this is more of an economic history perspective, is that most of the largest financial crises in Western economic history have had one of four factors driving them. So it's either been about loose credit, the easy availability of money, that tends to drive things. The GFC, the Global Financial Crisis, was in part about that. Real estate has been at the center of many crises in the US and elsewhere, including the GFC. So speculation about real estate.

• 15:32Speaker 01

  Technology, there's four of these, by the way, just in case you think there's 100. Technology is often at the center of this. A really good technology story can cause these, these moments, these crises. And then the other one is government policy, obviously. And government policy, where you, you know, the government's encouraging people to do things and creating an implicit backstop. So generally speaking, one or two of these are involved in almost every crises in Western economic history. This is the first one that has all four. We have a great technology story. We have a real estate angle. Data centers are implicitly commercial real estate, and a lot of speculation around that. We have very loose credit, which is changing a little bit in the last two weeks, three weeks, as we kind of have this debt market freak-out over Carweave and others. And we

• 16:15Speaker 01

  have a government angle, right? We've been told repeatedly, at least in the United States, that we're in an existential battle with China, that we must win. It doesn't matter the costs or the consequences. So the notion that, for the first time, one of these episodes sits at the intersection of all four forces is why this one in particular is both dangerous and misunderstood.

• 16:35Demetri Kofinas?

  So I want to, that I understand very clearly. I want to make sure I understand your first point about what also makes this different. Are you saying that there's a kind of reckless abandonment with which investments are going into the space? And is that because of, and is this a very Western phenomenon, because of the sort of AI, superintelligence, we are building a god narrative, and so it's all or nothing? Is that kind of explain what you're describing?

• 17:00Speaker 01

  Yeah, that, but there's also a race dynamic, right? There's an implicit idea that, "If I don't continue to spend, my competitors will." This is likely to be a market that's dominated by a small number of players, at least in terms of inference and the provision of tokens, the underlying sort of atomic units of AI. And so if you subscribe to that, then you immediately enter into this race dynamic, where I must continue spending, because I'm at risk of becoming an also-ran or even irrelevant. And so that's what sort of locks in place, at a deep level, why this competitive race dynamic exists. And you saw that in the recent calls from Dario Emmeide at Anthropic, asking for, you know, and this is a bizarre term, but pacing the frontier, which is just, you

• 17:42Speaker 01

  know, a fancy way of saying, "Let's try to slow things down." But you can't, because you get into this, you know, sort of prisoner's dilemma problem, that, "Yeah, okay, fine. What if we all agree, but China doesn't?" Well, then we're into this problem where China wins again, and we're back to this existential crisis. And so the race dynamic is baked into the very nature of what we're doing. And to your point, it's exacerbated, it's made worse by a kind of pseudo-religiosity on the West Coast of the United States with respect to technology, and this idea that there's no point in trying to slow anything down, because it's inexorable. It has its own force. It has its own morality and consequence. And slowing it down is, is anathema to this group of people. And so you've got these

• 18:24Speaker 01

  two phenomena operating and reinforcing one another. And then a capital markets that sees this happening, a race dynamic in this religiosity, who's more than happy to provide endless amounts of capital to try and keep the dynamic going, because they're making money from fees. They're not holding the debt. They're making money from an agency business of helping to create these special-purpose vehicles, off-balance-sheet financing, syndication, all these things. But they're not holding onto the instrument. So they make money from issuance, not from holding. And so you've got all of these things intersecting that creates the dynamic that we're seeing, which now means that, you know, we're now on a,

• 19:04Speaker 01

  as a percentage of GDP, as a capex moment, we're at, we've now exceeded all prior episodes in the last 150 years. It's larger than the telecom build-out inflation-adjusted dot-com, PCs, rail electrification, railroads, canals, take your pick. We're only behind interstate highways and World War II rearmament, so good for us.

• 19:24Demetri Kofinas?

  So I have some clarifying questions related to the race dynamics, in particular, when it comes to arguments that I've heard from folks like you, that this is actually a commodity business, and not necessarily a winner-take-all market for the frontier. Companies, but let's just put a pin in that real quick because I want to revisit or tie off a loose end related to the question about how disruptive this will be to society and the economy. Is there any way to responsibly model the amount of layoffs we're going to see or maybe the drop in entry level hiring or like, help me understand how disruptive this is from a numbers perspective? Can you put some numbers on it to help us understand what you're thinking here?

• 20:07Speaker 01

  Sure, there's lots of ways to get at that. So I'll just give you one or two, maybe. So if you approach it from the standpoint, just deterministically, let's just assume for the sake of argument that the current capex spending, the amount that's being spent on building out data centers is rational, which it isn't, but let's pretend it is. Okay. So if we pretend it is, and then we say, over the next four or five years, we'll probably get to 10, pushing on $10 trillion in data center spending. And if we apply a reasonable hurdle rate in terms of investor expected rate of return and then a margin on top of that, in terms of, so you can do the math fairly straightforwardly. It's a classic sort of return on capital argument and which a group at Columbia did recently, Bain has done as well.

• 20:48Speaker 01

  You get to numbers like four to six trillion dollars in revenues needed by, say, 20, early 2030s. These are all very squishy, but they're ballpark. And so then you say, okay, fine. And this is going to get to your question, sort of the backdoor way. You say, well, where's that going to come from? Right? So there's only two ways places it can come from. It either replaces existing spending or it's an all new spending. Okay. So when it's happening that fast, generally speaking, it's replacing existing spending, at least in large part. So then you could say, okay, so where is this coming from? Well, we've already said that this is sort of symbolic automation. Well, that's white collar work. So four to $5 trillion is about 40% of the US white collar workforce in terms of, if

• 21:28Speaker 01

  you look at white collar work as a TAM, a total available market, it's about 25% of the global market for white collar work, a little less, maybe 23. So you can start to say, okay, so in one way or another, a significant fraction of the US workforce is at the very least going to see their work repriced. It doesn't necessarily mean layoffs, right? So you have to be careful about making the distinction. But what AI is doing is repricing the value of cognitive work. And as it does that, as it reprices all of that, it changes the nature of people's relationship with companies. So it would be unsurprising to see, you know, numbers at least in the, you know, in the millions, in terms of the low millions, in terms of actual layoffs.

• 22:09Speaker 01

  But I think it's better to think of it in terms of the assumptions driving all of this require at least a quarter of white collar work to be repriced. And so then you can sort of draw your own conclusions. Does that mean fractional work, man, for these people? Does that mean massive layoffs? And then add a sting to the tail. Once you reprice that much white collar work, the rest doesn't stay the same price. Right? I can't have it. It doesn't work that way. You don't get to say, well, that corner is now cheaper, but the rest of white collar work, we still pay the same. No, everything gets repriced, just to a greater or lesser degree. So it's consequential, both in terms of layoffs and in terms of, you know, what we pay for this good that we call human work.

• 22:51Speaker 01

  And that's baked into the nature of what we're creating if you simply accept that these capex numbers require return on investment.

• 22:58Demetri Kofinas?

  Great. So we're going to have a chance to revisit that. Back to my question about unit economics. So I understand the race dynamics when it comes to the US-China competition, but what are the unit economics that underpin the business models of these hyperscalers that would explain this prisoner dilemma type race dynamic that you're describing in the commercial marketplace?

• 23:17Speaker 01

  Well, it's the same in the sense that if you believe that this is naturally a scale business, so there will be consolidation, which is implicit in having only a couple of frontier model companies, for example, then inevitably you want to be the large, one of the largest scale providers. And so then you're competing on the basis of the marginal cost of production in the limit. And the marginal cost of production is really just of inference, for example, is just, you know, heat, power and light. Right? I mean, at the margin, it's just what it costs me to keep the businesses doors open. If we're truly competing away all the margins and competing in a truly scale sense, right? So the response of the frontier model companies is the price will never decline to the marginal cost of production because we're so smart that we're creating all these new models

• 24:02Speaker 01

  for which we can charge premium prices. Right? So you'll see this constant flow of new models coming out of Anthropic or coming out of OpenAI. And the way to think about that is it's a defensive tactic to try and make sure that they can charge what the marketers call, you know, have a halo, have a halo price above the cost of actually producing tokens. And that's the game that's being played is you try to produce new models and justify how powerful they are by all kinds of talk about their exploits and other things. And it's a way of making sure that the cost of selling tokens doesn't collapse to the marginal cost of production.

• 24:35Demetri Kofinas?

  All right. But here's the question I have. How much money is projected to be made from these frontier use cases versus the vast majority of cases for which people will use AI? And also, doesn't open source throw a wrench in the, in these unit economics? In other words, most people?

• 24:54Speaker 01

  Oh, absolutely. So there's two or three different elements of this. I'm not going to try, I don't buy their arguments whatsoever. So one is that you can think about what's happening elsewhere, like China is a good example. China is trying to essentially become like almost an industrial manufacturer of tokens. So producing tokens at very close to the marginal cost of production that are, say, eight months behind US models in terms of their performance, which seems to be borne out by statistics and benchmarks from people like Epic and others. And they're producing these tokens at a vastly lower price. And that's important because, let me turn the story around. There's a really important misunderstanding of the nature of what happened in the last

• 25:36Speaker 01

  three years that kind of drives many misunderstandings of where we're going from here. And that is that, and this is an old tenet of being a venture capitalist, is that never trust your earliest adopters because they're really not representative of the mass market. Because by definition, they're early adopters, they're risk takers, they have extreme needs, they're willing to pay a premium price, they're willing to use risky technologies. That's not like normal people. Normal people aren't anything like that. The first four years of frontier model usage were dominated by whom? Software engineers, right? We know that from the data, something like 70 to 80% of frontier model usages by software engineers. Software engineers are very extreme cases for a whole bunch of reasons. They have relatively complex problems, but also just the nature of software.

• 26:21Speaker 01

  Think about how it works. I prompt a model with a relatively compressed prompt, say, you know, help me build a fitness app. That's like, what, like eight words, something like that? What comes out the other side? Might be a million lines of code. So it's highly expansive, right? Which is really unusual. Why is it unusual? It's unusual because think of the nature of white collar work. What's an awful lot of what you might use a model for is, I'm from finance world. So I'll say, you know, take this Goldman Sachs research report that's 30 pages long and tell me if there's anything there I care about. So it's compressive. It's the opposite of software engineering. An awful lot of the use cases, what normal people are using AI for, both inside and outside of large organizations, are for highly compressive uses. So why does that matter?

• 27:06Speaker 01

  It matters because it changes the nature of what you're extrapolating. If you're extrapolating from how these tools were used in the first four years on the basis of these early adopters called software engineers, you're making two huge errors. One, you believe that everyone cares about being at the absolute frontier performance, right? Because software engineers cared. Normal people don't care. They might not even notice. I've run the equivalent of the old Pepsi Coke taste test where I hide the model behind a harness and see if people can tell the difference. And normal people have no idea. They can't tell. It doesn't matter to them. And the second, obviously, big error is this one that white collar and usage of large language models is very different from how software engineers use them. So to take it back to your original point, that makes the margin pressure even more

• 27:51Speaker 01

  dramatic? Because if you can't charge a premium price because normal people don't care whether they're at the frontier, so you launching a really advanced model is irrelevant to them, the pricing pressure becomes even more extreme.

• 28:04Demetri Kofinas?

  So here's the thing I'm trying to understand, though. Are you suggesting that they don't understand this? Or that they're caught in this dynamic where they need to perpetuate a narrative in order to continue to get the funding that they need to make the investments to appear to justify their stock prices?

• 28:23Demetri Kofinas?

  Like, is it that they're caught in this race dynamic where they feel like it's a winner take all market? And therefore, they need to keep spending because whoever wins takes it all? Or they're caught in this narrative dynamic whereby their stock valuations are dependent on them doing things that suggest to the marketplace that they believe in this story?

• 28:46Speaker 01

  It's both. It's the honest answer, and I hear this all the time from inside. So, there's a story that's required, and this is why you continue to see them launching new models and talking about how this improves on, you know, pick one of the following 17 million benchmarks. And that's to maintain the momentum of the story and say, we're still making material progress. Your money is contributing to us producing new models at the frontier. And the untold part of it is, if you look at composite benchmarks, as opposed to just one-offing, because an awful lot of the benchmarks have been ingested by the models themselves. It's like they saw the test in advance. So if you look at composite benchmarks, which you'll see over the last three years, is

• 29:28Speaker 01

  while models are improving, it's becoming somewhat asymptotic, meaning that on an absolute level, they're not improving as much as they once were. So it'll, we went from like, say, around 18, 19% performance year over year on composite benchmarks three years ago, to more like 5% or 6%. And this is hugely important. So it's being missed largely in the conversation, and yet, to your point, it's really crucial to these companies because it's what allows them to continue to raise money and forces the race dynamic to continue, because you have to continue to show progress. Otherwise, the question is, and I say this all the time, that if it weren't true, the first company to stop training new models would actually be the winner, because their costs would collapse, right?

• 30:09Speaker 01

  You should, in a perverse way, the rational move somewhere in here is for an AI frontier model company to say, we're cutting all training costs, because the return on training has really declined sharply, at least in a dollars per unit improvement basis, and so now we're just going to provide inference. The data suggests someone should have done that already. It's not happening. And the answer, of course, is you cannot do that in this race dynamic without it having catastrophic consequences for your ability to raise money and keep the race going.

• 30:37Demetri Kofinas?

  All right, let's talk about concentration risk. And I'm speaking, actually, this could be applied widely, but I'm speaking just in terms of their customers.

• 30:44Speaker 01

  Yeah.

• 30:45Demetri Kofinas?

  How does this industry compare to, let's say, the internet when it comes to customer concentration? Where are the revenues coming from?

• 30:51Speaker 01

  It's wildly unusual, and I'll take you back to my equity analyst days. And one of the first things you look at when you're trying to evaluate a company, isn't just its growth rate, it's its customer profile. And it's not just, you know, do I know their customers? It's how many do they have? Because the single, one of the biggest risks you can have when you're following a public company as an investor or an analyst, is an air pocket in terms of their revenues or their earnings. And one of the causes of that can be, if they have a very concentrated list of customers, a small number of customers who account for most of their revenues, it doesn't take very much for suddenly, you know, an earnings per share forecast of $1.42 to get missed and it's a dollar instead, just because one customer held off on some purchases.

• 31:34Speaker 01

  So taking this over to the large language models, we have not seen yet, obviously, there's no S1 filing from Anthropic yet, but there's, you know, endless leaks about the nature of what their revenues look like. And we're told that something in excess of 70% of their revenues are eight companies. One alone, Meta, might be 17% to 18% of Anthropic's revenues. This is, for a company allegedly on a $50 to $70 billion run rate, this is unprecedented. I've never seen that in a company of this size, to be so dependent on a small number of customers. The only place where you sometimes see it is perhaps in the defense industry, where obviously your only customer is like four governments worldwide, so you might see it there. But that's very different, because you're in long-term contracts and you're relatively

• 32:18Speaker 01

  secure positions over a cycle. To see a commercial company operating in a private space with such customer concentration creates unprecedented risks as an organization, because you, in a sense, are be over the barrel in terms of your dependence on them and their ability to say to you, you know what, change this, do that, whatever else, because I can see how important I am to you. So with respect to its comparison to prior eras, there is no comparison. It doesn't, we never saw this level of concentration in any prior episode like this, whether it was the internet, fiber, GFC, railroad. And one of the most unusual elements of this moment is the extreme level of concentration and dependence on a small number of customers. And within that small number of customers, the dependence on a small group of people

• 33:02Speaker 01

  who happen to be doing a very anomalous activity called writing code.

• 33:06Demetri Kofinas?

  How much of economic growth do you think is a result of AI-related spending? And this second question I'm going to ask, it's a tag along, is very difficult to answer, but essentially, like, is there some way to estimate how much of a premium is built into the S and P, or some of these indices, as a result of AI spending? And the AI narrative?

• 33:27Speaker 01

  Yeah, yeah, yeah, yeah. So you can answer both pretty, relatively cleanly. So I've been tracking this idea. One of the things that first got me interested in this episode was sort of this anomalous continuing stubborn growth in the US GDP that didn't seem to be tied to orthodox economic statistics. And so when you dug under the hood, you found out fairly quickly that for at least the last eight, seven, eight quarters, something like, and it varied from quarter to quarter, 25% to 60% of GDP growth on a quarterly basis has been tied to AI capex, which is, again, wildly unusual, incredibly distorting, but obviously hugely consequential, because when you're that, such a high percentage of GDP growth, it masks the weakness in everything else.

• 34:11Speaker 01

  So something on the order of between a quarter and a little over half, depending on the quarter. And that's up there with prior episodes. It's probably on the order of the railroads, something on that order, in terms of its materiality with respect to GDP growth. And you can ask yourself, you know, get into all these numbers and so on, and it's like, well, why do I care? Why does that matter? Well, it matters because it masks that nothing else is growing, or even some things are shrinking. And the analogy I often make is that when you have a causal model of how GDP growth is working, and it's wrong, you make bad decisions. And I often joke that it's sort of like, you know, when the mailman comes to my house and my dogs bark, they bark and bark and bark, and the mailman leaves, and they think, you know, I did it, I made the mailman leave.

• 34:55Speaker 01

  And it's like, no, you have a bad causal model. Mailman always leaves, he's got other houses to go to. So, this is the same thing with GDP growth. Think of it in the context of the US experiment with respect to tariffs. If you believe that tariffs work, you look at GDP growth and say, oh, look, it had no impact. As a matter of fact, GDP was growing in a really sprightly way. And that's just a misunderstanding of the nature of what was driving GDP growth. And so this is important, not because we're trying to get at some kind of, you know, Guinness Book of World Records statistic on a thing that contributes the most, it's because GDP growth leads to wealth, and we're misunderstanding the cause. And the cause is almost monolithic. It's this great, you know, sort of spending hairball called AI capex. And so that's why it matters.

• 35:39Speaker 01

  It's because it's masking all kinds of other phenomena. And the same thing is true in inflation data. AI capex is badly masking what's happening in inflation, which is causing us to raise rates, when most of the economy is actually fairly weak, right? But nevertheless, this is obviously hugely important, this misunderstanding. And so then take it over to capital markets, you can sort of do the same kind of untangling and say, okay, how much of the S and P 500's growth over the last few years has been driven by AI-related stocks? You know, euphemistically, people talk about the MAG7, the magnificent seven stocks, which are, you know, largely AI-related. They were almost 45% of S and P 500 growth for the last five years. So, seven stocks. Seven stocks driving almost half of the S and P 500 performance.

• 36:24Speaker 01

  It gets worse, something like 55% properly understood of the S and P 500 is now tech, which means that if you think you're buying a diversified index fund, you're actually buying not just a tech bet, but a concentrated AI bet. So people are far less diversified than they might hope to be. Let's say I'm nervous about this moment, and I say, okay, honey, I'm going to buy an index fund. You're just buying AI again. And so, we know how much it's driving the S and P 500, and we know what fraction of the S and P 500 it is. And it's, both of them are at levels last seen, well, again, during the railroads episode in the 19th century and during the rail electrification. And as a side note, the Dow Jones, which the S and P didn't exist at that time, the

• 37:08Speaker 01

  Dow Jones did, right, as a major financial index? And the Dow Jones, in response in the late 19th century, to how the overwhelming share of railroads, of the Dow Jones, was on the order of half, created the Dow Jones transports, to, in a sense, to kind of like shuffle that stuff off to the side, so it wasn't causing so much trouble. So, you can see prior historical financial bubbles in the indices themselves, because we have vestigial remnants like fossilized tracks. The Dow Jones transports is a remnant of the, one of the last major bubbles. And so we are at the levels where the Dow pushed out the railroads, if you think about it in terms of what's happening with tech. And yet, there they are, dominating the index.

• 37:51Speaker 01

  And the only other sort of data point I'll add to that, why I care about that, is because more Americans own stocks than ever, so more Americans are exposed to the stock market. It's something like 4, 52%, something on that order of things are exposed to the stock market. And a slightly higher fraction of them have, in terms of their total financial wealth, is exposed to the stock market. So it's one thing to say I own a couple of stocks, but it's another thing to say, a significant fraction of my wealth is tied up in the stock market. Both of those are at near-record levels. So what that means is that most Americans are vastly overexposed to AI.

• 38:24Demetri Kofinas?

  Yeah, so this brilliant, wonderful answer. Also, I'm going to steal your analogy about the mailman and the barking dog. I haven't heard that before, but that is brilliant. We're going to have a chance to talk about what diversification looks like, whether it's possible, how possible is it, et cetera, likely in the second hour. But I've heard you say that the Fed is raising rates into a weak economy, as you said just now, adjusted for AI's capex spending, and that when that spending tapers off, the private sector will be left with excess capacity and loads of debt that it needs to service, which means that the economy will spend the next decade or more in a deflationary balance sheet recession. But I've also heard you say that sovereign governments don't have the fiscal space to provide demand-side stimulus. How can both of those things be true at the same time?

• 39:10Speaker 01

  They're not really in contradiction, so I'll start on either side. So the notion that we're used to this idea that on the other side of financial crisis, the government bails us all out. And that can take a lot of different forms. It took one form during the global financial crisis. It took a different form during the dot com bubble. In the case of the GFC, we had, in a sense, nationalized some of the banking system. And then, at the same time, we brought rates down to very low, interest rates down to very low levels. And so the obvious expectation would be if there is some kind of a crisis on the other side of this, that something similar would happen. And so there's two reasons why that will likely to be less effective than you might otherwise hope. And one is that the state capacity to continue to do these bailouts is very stretched.

• 39:53Speaker 01

  And you can see that because of the size of the US debt and deficit, because we're still, in a sense, working off the last two or three bailouts, I suppose. And that's one of the reasons why the US is being punished in sovereign debt markets in the recent jump in terms of the 10 year yield and this increasingly fraught ability to finance its own deficit.

• 40:11Demetri Kofinas?

  So you think, just to clarify something, so you think that the rise in long term yields is primarily a reflection of investors' concerns about US debt and deficits and not competition with AI for capex?

• 40:28Speaker 01

  No, it's both. It's very clearly both. So there's two things going on in debt markets at the same time, one of which is largely unprecedented, which is that the margin, and then certainly in midterm duration bonds, the US and other sovereigns, other countries are competing directly with the issuance of AI related debt. And that really matters, and this becomes wonky in a hurry, so I'll just give the short version, is because the nature of buyers in sovereign debt markets has changed a lot over the last, say, 15 years. 15, 20 years ago, the largest buyers of US sovereign debt were other sovereigns. So other countries were buying. And we used to talk about, well, what happens if China dumps all of our debt? What happens if Japan does? Blah, blah, blah, blah, blah.

• 41:11Speaker 01

  And the reason why they were large buyers was because they were trying to, in a sense, sort of inoculate themselves against their trade surpluses. So when you have a large trade surplus, you end up with a lot of dollars. You've got to do something with those dollars. You buy the biggest and most liquid things you can, which tends to be sovereign debt. So they had this incredible amount of, they were locked in buyers of sovereign debt who were, and this is the important part, who were yield insensitive. They didn't care what rates were. They just had a lot of dollars and had to put them somewhere. Now, flip to today, that's not the case anymore. More than half of the buyers of treasuries, in buyers in debt markets now, US sovereign debt markets, are yield sensitive. It's not sovereigns anymore. It's hedge funds. It's individuals.

• 41:55Speaker 01

  And so that changes the calculus dramatically. So whereas in the past, the issuance of AI capex might have not had a big consequence in terms of rates, now it has a huge consequence because buyers are rate sensitive. And they say, wait a minute, I can get 12% on some debt being issued by a hyperscaler with a AAA rating, or I can get 6% from the US government, who I don't trust very much not to deflate it away. Which one am I going to buy? And so that creates pressure at the margin in terms of what you might see in terms of the market based yield for the instrument.

• 42:26Demetri Kofinas?

  Okay, so let's go back to my question about

• 42:29Speaker 01

  Can I just?

• 42:30Demetri Kofinas?

  Yeah, go ahead.

• 42:32Speaker 01

  No, so I'll just, the last point in your question was, so then, let's take this all the way and get to the other side of this current episode and say, there's a series of railroad-like rolling crises, or a GFC, or who knows what. But nevertheless, so I've explained in part why the sovereign part is less likely to be material because it's already, in a sense, so obligated in terms of all of its continuing commitments and its current deficits and debt. So then you say to yourself, okay, fine, but we can still take rates to zero. We can still drop rates to very low levels. And we did that after the global financial crisis. We did that after the dot com episode. In each of those cases, we dropped rates very low. So the lesson of these so-called balance sheet moments like this, the best analogy

• 43:15Speaker 01

  being what happened to Japan after its real estate crisis in the 80s and 90s, companies don't respond in the way you might hope. Whenever they're highly leveraged, very indebted, and you lower rates, because they, rather than saying, what you want them to say is, oh, I'll take on more debt, which will drive more economic growth. Instead of that, they say, wait a minute, I'm already leveraged to the gills. I've got, in the case of Japan, I own properties around the world, and these are corporations saying this, so I don't care how low you lower rates. My project over the next 10 years is to reduce my indebtedness. Okay, apply that same model to what's happening now. We've got all of these very large companies that are taking on increasing amounts of

• 43:57Speaker 01

  debt in various forms, increasing amounts of external financing, long term lease commitments. You lower rates to zero, are they going to take on more debt? Absolutely not, because the project for them is exactly the same as the project for the Japanese companies, who are hugely indebted in real estate after their bubble burst in the 1990s. And so they'll spend that period trying to reduce indebtedness. So what does that mean? It means that your two major instruments of financial policy, monetary policy and fiscal policy, are much less effective than they would have been in any other circumstance, in part because of history, but also in part because of the nature of this episode.

• 44:33Demetri Kofinas?

  So Japan's a good example, though not, I mean, certainly it is for the purposes of looking at a balance sheet recession, but also, let's use Japan as an example, because again, I'm not convinced that we wouldn't have the, quote, fiscal space. I think Japan's debt to GDP during the great financial crisis, or right before, was close to 200%. And they engaged in fiscal, plenty of fiscal spending afterwards. So I don't know why, in a deflationary environment, sovereigns wouldn't have the balance sheet capacity to continue to spend. Obviously, if the economy came out of that, then it would run into huge issues, which is kind of what we've been dealing with now. But I just don't understand why, I guess I'm not convinced, that in the event that we did have a balance sheet recession, which I think is a real possibility, I think is

• 45:16Demetri Kofinas?

  an interesting thing to think about for those who are really concerned about inflation or runaway inflation. I'm not convinced that in that scenario, people should be concerned about owning government bonds.

• 45:28Speaker 01

  Well, what the market is telling you, even in advance of something like that happening, is that there's real nervousness in debt markets, such that they're requiring considerably higher yields, yields we haven't seen in decades. And there's no reason to think that the effect wouldn't continue were the US to continue to, because essentially, print money to operate on a fiscal policy level to try and do a bailout. So my point isn't that it's mathematically impossible, it's just that it's going to be much more consequential in terms of the impact on rates and yields. The US is no longer simply able to do this and say, we have so much balance sheet space and spending space that we can do this with impunity. That's just not the case, and the market's already saying that.

• 46:09Demetri Kofinas?

  Yeah, well, let's move on from this, but I just want to say, like, when I first heard you make your comments around the balance sheet recession, I actually thought they were very interesting, and it's given me a lot to think about. I just think that in such an environment where demand is collapsing, I would be less concerned about the risk to those owning US government bonds. And I'm someone who doesn't own any fixed income, because I'm concerned about the debt and deficit. So let's talk a little bit, let's go micro for a moment.

• 46:33Speaker 01

  Just one last point on that, just to make, I don't disagree with you. I think that it's inevitable that it'll happen. You just have to sort of think through, again, think through the dynamics. So the closest analogy in the US context was in the late 20s and the early 1930s, where while there was space to do something in terms of fiscal policy, we simultaneously had this incredibly hawkish moment where people said, you know, we can't afford to continue running deficits of this size, no matter the current economic situation, which was one of the precipitating factors that led to the Great Depression. So I'm not saying it's impossible. I just think that the likely fracture lines are already there in terms of the resistance to doing it, the consequences of that resistance, while it will cause probably a sharper fall

• 47:16Speaker 01

  before, all of a sudden, the policies change, because we have the same kinds of stressors that we had then.

• 47:22Demetri Kofinas?

  Yeah, I would take the under on that. I think we have a technocratic class of policymakers today who have learned very specific lessons from the 1920s and 30s. I think that's, that's less likely. But again, I think

• 47:33Speaker 01

  I couldn't, I couldn't disagree more strongly. By all means, continue thinking that.

• 47:37Demetri Kofinas?

  What specifically? You don't think that, so

• 47:40Speaker 01

  We have the opposite of a strong technocratic. We have a highly

• 47:43Demetri Kofinas?

  No, no, let me be very clear what I mean here. We have a technocracy, so central banking has gone through, not just central banking, but the entire economics profession has gone through a technocratic revolution since the 1920s, when people thought much more along the lines of political economy, as opposed to this kind of a formalized, mathematized, and quantified kind of economic modeling that provides neat solutions in theory, but which often do not correspond to how things actually behave in the physical world. And COVID, I think, is further evidence that the government is prepared to engage in massive amounts of quantitative easing, and stimulus, and fiscal handouts, and credit rationing, and anything else.

• 48:26Demetri Kofinas?

  So I think those two things, both the lesson that in the face of a depression, in the face of deflation, we need to print, which was obviously what Bernanke's philosophy was heading into the 2008 financial crisis, and the fact that we've actually seen evidence of that in the most recent crisis of COVID, I just think the infrastructure is in place. And if the government were to face a crisis, especially with these race dynamics between the US and China, they would have every incentive to print and to hand out money.

• 48:55Speaker 01

  They would have every incentive, but I think they'll fail at it, because I think in part, the things you're pointing to are better understood as kind of the death rattles of the technocratic system that you're praising. I mean, think about the consequences of what happened. They gave rise to crypto, because people were like, Money is essentially broken. I can no longer trust governments to do this kind of thing. It gravitates to the politicization of the Fed, where the entire Fed mission is increasingly under pressure.

• 49:21Demetri Kofinas?

  Absolutely, absolutely. But also, but the incentives don't operate the way that you seem to be describing, which is that the politicians aren't going to say, oh, my gosh, look what happened with crypto, look what happened with financial nihilism.

• 49:32Speaker 01

  No, of course not. But that's not what I'm saying at all. I'm saying it's much less articulate than that. It's more of a feeling that my constituents are no longer happy with this kind of heavy-handed insertion of the Fed into the economy, and the Fed is increasingly politicized, and we don't believe in this sort of technocratic view that we can somehow manage these things. And so I think there'll be a series of rolling mistakes. I think it's inevitable.

• 49:53Demetri Kofinas?

  Well, mistakes for sure. But here's a question for you. So in that scenario, you think the government would choose to enact austerity as opposed to hand out money in the event of a deflation?

• 50:06Speaker 01

  I absolutely do, yes.

• 50:06Demetri Kofinas?

  Fascinating. Well, I wasn't prepared to have this argument. So let's, maybe I need to wait a few minutes, have you answer some other questions, and I'll, and I'll think about it a bit more. But that's very interesting, Paul. I wish I had heard more conversations with you on this topic, because I'd be better prepared to discuss it. But let's go micro for a moment, and then maybe we can revisit this in the second hour. So one of the things that I really love about your report, you actually have, I haven't pulled it here in my, in my outline, you have a, a chart where you have something you call the complexity discount. And you essentially, to identify the types of businesses that you think are, are most likely to benefit from this general purpose technology.

• 50:40Speaker 01

  Right, right.

• 50:41Demetri Kofinas?

  And those that are most likely to be hurt by it. Help us understand your framework here, and what kinds of companies would fall in either bucket?

• 50:47Speaker 01

  Yeah. So it's just a different way of thinking about a lot of what's going on out there, and not getting wrapped up in the question of, you know, which occupations get replaced first. Another way of thinking about what's happening in the context of AI and large language models, is that they're causing the disappearance of this thing that has prevented people from using various products and services. And so sometimes when you, when you take away the complexity, let's think of an example like Photoshop, okay? Ridiculously complex product to use. When I take away the barriers to using it, do people use it more? Maybe, maybe not. It might actually just get eaten by people prompting a model and saying, you know, create me a picture of whatever, right? A squirrel playing golf. And I don't, still didn't want to do it inside of Photoshop, even if you make Photoshop

• 51:30Speaker 01

  easier to use. So sometimes there's complexity. So complexity reduces usage. But sometimes when you take away complexity, you realize the product really wasn't something I needed to use in the first place. Then there are other examples of where complexity actually reduces the amount of usage that a product could get that it would otherwise receive. And I gave the example of CloudFlare as an example. You can think of lots of others. But CloudFlare, for those of people who don't know, is started off largely as a kind of caching service, making it easier at scale for people to cache for companies to reliably be able to serve content over the internet, despite having unpredictable amounts of traffic and sometimes spikes in that traffic. So you could cache content all over the world. But it's since then launched a host of other products and services, many of which are

• 52:13Speaker 01

  wildly useful, but most of which are absolutely inscrutable to normal people. And so my example there is that it's a good example of how by insulating me, by which I mean anyone building anything on the internet, from the complexities of using that product, you actually discover there's a host of very useful things that you might otherwise be interested in using. There's these, you know, workers and blah, and blah, and blah. There's all kinds of different things that they offer. And so the disappearance of complexity is unlike, say, for example, with Photoshop, doesn't then have you come back and say, wait a minute, I don't need to use this at all. There's another way to do this. It's saying, oh, wait a minute, there's a whole bunch of other things I could be using with this. And so in their specific case, just to pick on them, that's created an incredible tailwind in

• 52:56Speaker 01

  their business. And so I think the stock has tripled or something over the last 18 months. Most analysts are telling, you know, really very positive stories, sometimes not very articulate in terms of pointing out to why it's all happening. They'll just say, oh, you know, AI is helping them or some such thing. But what's happening underneath, if you think about it sort of systemically, is that the complexity that sits in front of the usage of the product is largely falling away, because people are simply using things like, like MCPs, like connectors and other things to manipulate what CloudFlare is doing on their behalf. And there's lots of other examples like this. And I present this not to be some sort of like Uber CloudFlare bull, but as a way of helping people rethink

• 53:36Speaker 01

  exactly what's happening and not get wrapped up in, you know, whether or not I could vibe code a fitness app or something. I don't know. But think about it more in terms of how it's changing the nature of the relationship between people and the products and services that they use. And I think another example I gave was like, like Roblox is an interesting example of this, where you have this incredible platform that's, you know, one of the largest in the world in terms of, you know, sort of gaming content. And something like, I saw the statistic the other day, something like 30% of YouTube gaming content is Roblox related. It's some ridiculous statistic like that. But building things on top of this is incredibly annoying. I don't know this firsthand. I hear this from people all the

• 54:17Speaker 01

  time. And so you can imagine that when you reduce the complexity to producing applications that run on Roblox, it's not all of a sudden that you realize that Roblox is irrelevant. You realize that, wait a minute, there's this incredible platform sitting there that I could launch applications and run on top of, which is good for me and good for Roblox. And a kind of weird corollary to all of this is, you might say, well, the same phenomenon must exist in terms of building like iOS apps, like Apple iPhone apps and Android. And what's happened instead is we've realized that actually, we were fairly close to saturation in terms of people's ability to sort of think about and adopt new, new apps. And you see that in, in the data, there's a National Bureau of Economic Research report came out,

• 54:58Speaker 01

  I don't know, almost a year ago now, maybe eight months ago, showing that we all see the one side of this, which is this incredible spike in the number of submissions to the App Store. And people have seen this graph in many, many different forums. Co2, a hedge fund in New York likes to present it all the time. But what you don't generally see is the collapsing number of ratings per app. So what's happened is, we're launching lots of things, but people's ability or even interest or awareness of these things has really fallen off a cliff. And so, again, this is this split, right? What's complexity really hiding? Is it hiding a mature market? Is it hiding something that I actually don't need anyways? Or is it unlocking functionality and things that I might otherwise use? And I think, in a really pragmatic way, it's a useful way of thinking about how these

• 55:39Speaker 01

  tools are transforming our relationship with products and services. And you can take it lots of next steps, but that's sort of the starting point.

• 55:46Demetri Kofinas?

  No, I, I found this, again, you publish a lot of research, and there are different tiers, I think, also to your subscription. But this is a great example of, I think, how valuable I found your work to be. One last question before we move to the second hour. How would you describe the work that you do? And how is the work that you do going to be affected by AI? I mean, it's already being affected, but like, how disruptive do you think this technology will be to what it is that you do?

• 56:12Speaker 01

  It's an interesting question. So at a very sort of mundane level, it makes a bunch of simple things somewhat easier in terms of, you know, I like to, I think in, I'm a math guy, so I think in statistical terms a lot. So it's easier than ever to do, you know, basic modeling. And I think that's now largely a commodity, and a lot of it's just insulated from me. So I'll create models in a very, you know, straightforward way that doesn't require the same amount of work as it once does. But I always remind myself that the commoditization of analysis has consequences for anybody who pretends to do analysis. So the notion that you can have an edge from straightforward analysis is largely falling away. I just don't think there's much edge, and that's in financial markets terms, I mean edge, like alpha, in terms of actually doing anything like that,

• 56:56Speaker 01

  because it's so easily replicated. So you have to rethink, I think, a lot of where any edge comes from in terms of pulling together analysis, because so much of it's just commodified now. And so, which is great. I think it's fine. It's good. I, I never liked sort of the priesthood of doing Monte Carlo analysis. It was, never made much sense to me. So yeah, I think it's commodified almost the entire base of doing basic analysis. You're forced to be, more synthesis has to happen. You have to think across domains and bring things from other disciplines in a, in a more, you know, synthesizing way. And I think loosely you have to be more creative, but it's one of my objections to almost all of the lines of patter about AI, that, you know, humans will always be safe because we're more creative. And I think that's largely nonsense. And so I don't think it's just about creativity.

• 57:41Speaker 01

  I think, you know, synthesis, heuristics, and scar tissue help a lot, though.

• 57:45Demetri Kofinas?

  So I'm putting you on the spot here, but this is actually a question I've asked a lot of my guests. And it's a conversation I have with lots of people in person, because it's something that I ask myself every single day. Where do you think, you don't have to have a clear answer for this, but where do you think your edge comes from in this world or will come from? Or how you'll need to adapt in order to continue to be able to find monetizable value in what you do?

• 58:07Speaker 01

  The honest answer is I have no idea. I really don't. I sort of treat this as a very iterative process. And I have clients who have questions, and I'm always amazed at how interesting their questions are. And I try to like, think about what's behind those questions, and then use those in turn to build out sort of new foundational ideas and services. Because they say, okay, you know what, a lot of people are interested in this, and clearly there's not a lot of good answers out there about it. And so I'll sort of wrap something around it. But it's, a lot of it is driven in that kind of almost inductive way, that I think if I was to take it very deductively and say, okay, this is where I want to be in five years or 10 years, I just think that's nonsense. I just don't think it's realistic. So I take a much more, you know, inductive and iterative approach,

• 58:50Speaker 01

  where I see what seems to be, you know, for me anyways, generating some edge and some alpha. And I try and build around it, and I build around it until it stops working.

• 58:58Demetri Kofinas?

  I mean, I feel like the alpha that I get, let's take you as an example, and then I'm going to move to the second hour, and I'll give listeners an idea where we're going to go. The alpha, or the value I get from reading your material is twofold. One is the analysis. But it's not just the analysis. The analysis is useless if you can't help me learn it. So the speed and effectiveness with which I can learn from you is even more important. I mean, the, the analysis is table stakes. So I do wonder if that alpha will increasingly come from people's ability, in our work, let's say, to the extent that our work is somewhat similar, our ability to convey knowledge effectively to other

• 59:38Demetri Kofinas?

  people, our ability to teach others. And is there something that humans can do uniquely well in that regard? And essentially comes down to, you know, because it's become very common for people to say that intelligence is now free. And that may be true, but knowledge is expensive, or understanding, rather, is expensive. It takes time and effort to learn something. And so I wonder if there's some, some value there in between the analysis and the understanding that humans can still play an important role in.

• 1:00:11Speaker 01

  Yeah, I largely agree with that. I think the way that I think about it is I turn that around and say, what are models really, really good at? Well, they're good at next token prediction, which is a fancy way of saying they're essentially homeostasis engines. They're trying to drive people towards the center of a distribution. And the center of the distribution is often the right place to be, but the center of the distribution also misses every inflection and everything. And so what I try to really focus on, and I mean, actually running some experiments here now with respect to some machine learning models in, in sort of what you might call kind of tipping point detection in complex systems, is I try to focus on inflections. I think models are outstanding at generating the center of the distribution and genuinely terrible at the tails.

• 1:00:55Speaker 01

  And so I, I'm very much focused on the tails of the distribution, which is where I think most of my value lies.

• 1:01:00Demetri Kofinas?

  Great. So I'm going to move us to the second hour. I'm going to give listeners an idea of where we're going there. Of course, I do want to ask you a question that I posed early on, which is really, what does a diversified portfolio look like in this environment? Is it possible? And how close to one can you actually get? I want to revisit the conversation about knowledge work, and I want to introduce into that also a discussion about physical labor, and that means a conversation about robotics and industrial automation and what your views are on that. I'm also interested to understand what you think will drive the sell-off that will eventually come, you know, to pop this bubble as, as every single bubble has to eventually end, regardless of what, how you want to define a bubble, the boom in capex spending in AI.

• 1:01:42Demetri Kofinas?

  And I also, to the extent that we have time, like to discuss the spectrum of risks, in other words, like how seriously you take some of the catastrophic risks versus systemic financial and social risks that we've kind of touched on. And of course, I do want to revisit our discussion about a balance sheet recession and a government debt crisis and how exactly those two things can coexist. But again, we're going to move all of that to the second hour. For anyone new to the program, Hidden Forces is listener supported. We don't accept advertisers or commercial sponsors. The entire show is funded from top to bottom by listeners like you. If you want access to the second hour of today's conversation with Paul, head over to hiddenforces.io/subscribe and sign up to one of our three content tiers.

• 1:02:25Demetri Kofinas?

  All subscribers gain access to our premium feed, so you can listen to the rest of today's conversation on your mobile device, using your favorite podcast app, just like you're listening to this episode right now. Paul, stick around. We're going to move the second hour of our conversation onto the premium feed.

• 1:02:43Demetri Kofinas?

  If you want to listen in on the rest of today's conversation, head over to hiddenforces.io/subscribe and join our premium feed. If you want to join in on the conversation and become a member of the Hidden Forces Genius Community, you can also do that through our subscriber page. Today's episode was produced by me and edited by Stylianos Nicolaou. For more episodes, you can check out our website at hiddenforces.io. You can follow me on Twitter at Kofinas, and you can email me at [email protected]. As always, thanks for listening. We'll see you next time.