Speaker 1Join me, Raoul Pal, as I go on a journey of discovery through the macro, crypto, and exponential age landscapes. In The Journeyman, I talk to the smartest people in the world, so we can all become smarter together. Why does growth keep slowing? Why does everything feel so expensive? Why do assets keep going up while wages feel flat? And why is AI arriving right now? They sound like four different questions, but here's the thing. They've got one answer. Everything I've written in GMI for 21 years comes down to one line of cause and effect. It starts with debt and demographics. It runs through technology, and it ends somewhere very new. Demographics drive debt. Debt drives liquidity. Liquidity drives financial conditions. Financial conditions drive the business cycle, and the cycle drives asset prices. And the one thing that can bend that line is technology, and right now, it's all arriving at once. So let me lay the whole thing out. The everything code is the nexus of debt, demographics, and liquidity, all working in tandem to drive asset prices via debasement. The exponential age is every major technology gap. It's going to be going exponential at the same time for the first time in history. The economic singularity is where the road runs out and a new map has to be drawn. And underneath all three sits a single framework, something broader than markets, deeper than economics, and more fundamental than technology. I've come to call it the universal code. That's four frameworks, one code, and 21 years of work. Phase one, the everything code. This is the world we're in, the system that runs everything today. Aging populations, slowing growth, and a mountain of debt that has to be rolled over with ever more liquidity. Everything in this chapter hangs on one equation. I call it the magic formula. GDP growth is population growth plus productivity growth plus debt growth. And in the developed world, the first of those has been fading for decades. And productivity hasn't been fading for decades. And productivity hasn't been enough to make up the gap. So what's left is debt. Start with people. In America, the share of adults working or looking for work peaked in early 2000, at a little over 67%, about when the early boomers started to edge into retirement, and it's about 62% now. At core, the story of the everything code is really a story of demographics. And the future of that line is already written. Because on this chart, births lead participation by about 18 years. And the American birth rate has more than halved since 1960. The people who weren't born aren't going to magically appear. It's a prophecy whose sacred text is distributed across 1,000 maternity wards. So if the workers aren't there, debt becomes the fuel for the economy, and it shows up on this chart. Turn the debt line up. It's upside down, and it's the same line as participation. It's what I call the greatest chart in all of macro. The year participation peaked, federal debt was under 60% of GDP, and today it's over 120. But here's the thing. GDP is the income stream that services all the debt in the system. And debt like that can't be repaid. It can only be refinanced. It's basically. Borrowing money on your credit card to pay off a credit card. And the interest keeps compounding. The U.S. government now pays more than a trillion dollars a year, more than double what it was in early 2020. And interest payments lead liquidity by about three years. So the interest payments drive the liquidity. That means today's interest bill gives us a read on the liquidity needed about three years out. Zoom out to the whole world, and there's around $350 trillion of debt with an average maturity of about five years. So roughly $70 trillion has to be refinanced every year. What you need is the role. And what enables the role is liquidity. That's why the business cycle is now really driven predominantly by the debt refi cycle, but it drives everything. Debt. Rollover. Liquidity. Repeat. And liquidity leads the ISM, the survey we use to read the business cycle, by about six months. So what happens when the role breaks? Low rates cause excessive borrowing. Lenders become less selective. Rates go up, and eventually the burden is too much. In August 2007, everyone said it was contained. And I wrote, I think you have to ignore all the talk about containment. Those days have passed. The credit issues are spreading everywhere. The call was to short the banks. And as with every call I show, measured from the call to its best point, the KBW bank index fell 82%. But not every call has worked. In 2009, I had the worst year of my career. What went wrong is I overruled my own framework and went with emotions. It was a big lesson. Then in 2011, Europe blew up, but there was no knock-on contagion because the money printer stepped in. It was the first real confirmation of how powerful debasement is, and it's the lesson the everything code is built on. While it didn't play out as expected, the markets moved enough to make a lot of money in 2011 and 2012. From 1.48, the euro fell 35%. And as I said in April 2020, I had the worst year of my career. I had the worst year of my career. And just because I've been very, very right recently doesn't mean I can't be spectacularly wrong. That euro call has a twin. I lived in Spain and billed GMI in euros. In September 2011, I wrote, 90% of my liquid net worth is now in dollars, so I hope I'm bloody right. I switched the billing to dollars, built a house in the Cayman Islands, a dollar currency board, and later told members the dollar could, well, rise for a decade or more. I didn't see it then, but that was the universal code. The US is where frontier intelligence gets made. At a scale nobody else matches. From that call, the dollar index rose 54% at its best point. Oil is a good place to see liquidity working through the dollar. The 2015 call was short oil, and WTI fell 45%. And I think that's a good place to see liquidity working through the dollar. As I wrote in March 2015, the Fed decided to try to take away the punch bowl by removing QE, and the dollar exploded higher. By November, I was writing, oil just continues to follow the US dollar. Five years later, on March 6th, 2020, I wrote, I also think that oil is about to break $42 and head rapidly to $20. At the end of March, I wrote that oil could literally go to zero, as NatGas has done in the past. Maybe even. Maybe even negative. On April 20th, WTI settled at minus $37.63 a barrel. Bonds tell the same story. In early 2018, everyone was a bond bear. And I wrote that bonds might attempt a break of the secular trend at 3%, and maybe even spike to 3.2%, but fail. Debt and demographics pull yields lower over time. And by May, with a higher dollar, lower oil, and a slowing China, all paths lead to lower US bond yields. And long-dated treasuries rose 44% at their best point. If I only showed you the old misses, you'd think I was cherry-picking. In October 2024, I wrote, the next 12 months are likely to be wild. That was the banana zone, my name for the part of the cycle when crypto goes vertical. That call failed. US liquidity was shut off. By the government shutdown. And then the Iran war caused rates to rise and the dollar to rally, which pushed crypto into a mid-cycle bear market. I think it's a hiatus in a longer bull market. Every one of those calls comes back to the same engine. Because this debasement or liquidity, if you want the nice term for it, is the thing that drives every asset. Over time, I think that's around 8% a year. But it depends on your starting point and where you are in the cycle. This has been a slow stretch in the cycle. And our own liquidity index has trended at just under 6% a year since 2011. As the cycle accelerates, the rate accelerates with it. Debasement of currency is a tax. Wages grow roughly in line with GDP. But asset prices grow with debasement. So there's a hurdle. around 8% a year and more once you count inflation. Here's the annualized price change since 2012 against that line. And look at what clears it: Bitcoin, the Nasdaq, and the S&P on the back of its tech giants. Only two things have outrun the printer with any consistency: technology and crypto. So where's the way out? As I wrote in 2022, the only way to actually change this debt debasement equation is through the application of technology to increase productivity. This line is the engine: productivity growth minus inflation. And right now it's below zero. And as I put it in 2023, we can't change the demographics, but we can change productivity via technology. That is the essence of the exponential age. So here's the question for the next chapter: Can technology really turn that line positive? And how fast is it moving? That's phase two, the exponential age. Phase two, the exponential age. This is what's happening all around us right now. Every major technology is going exponential at the same time, for the first time in history, and each one feeds the others. It's one of my four frameworks. And as I put it in GMI, the exponential age drives network adoption. The everything code drives liquidity. Here's the logic: GDP growth is population growth plus productivity growth plus debt growth. Demographics are dead as a driver, and debt is a Band-Aid over a bullet wound. So that leaves one term. We can't change the demographics, but we can change productivity via technology. In April 2021, after 16 years of writing GMI, I wrote, "I am entirely changing my macro framework." That was the first time ever. I kept the old one alongside it for protection. The exponential age runs on a flywheel, as David Mattin puts it. Energy unlocks compute. Compute unlocks intelligence. Intelligence discovers new energy. The loop tightens. Better AI designs better chips. Better chips run better AI. And these networks aren't just growing individually, they're connecting to each other, so every turn goes faster than the last. Let's start with energy. Solar follows Wright's law. Every doubling of cumulative installed capacity has driven the cost down by roughly a fifth, and that relationship has not broken once in 50 years. Governments pushed that shift through the price of carbon, and carbon allowances were one of my calls in 2021. From the call to its best point, they were up 111%, and every call here is measured that way. And it isn't only energy. As I wrote in 2022, technology is making everything cheaper. Even technology itself is getting cheaper, creating an endless cycle. In 1975, a megabyte of memory cost about a quarter of a million dollars, and by 2023, about a tenth of a cent. For 50 years, the story of computing was Moore's law. Around 2020, the constraints stopped being the chip and started being the intelligence that could produce the chip. From 2020, the line becomes computer efficiency, intelligence output per joule, and look at the change of slope. From 1971, it climbed seven orders of magnitude. Since 2020, it has climbed five more. Moore's law doubled every two years, and the thing that replaced it has doubled about every four to five months. That's not a faster version of the old curve. It's a different curve. I've built a way to measure it, the exponential age index. Intelligence per joule times joules per dollar is intelligence per dollar, which is the number the whole economy is now organizing itself around. It now stands at more than 40,000 times its January 2020 level. The energy engine, the joules per dollar half, hasn't fired yet, delivered power in the US costs more than it did in 2020. Then there's adoption. Years of apparent erosion. Then a phase transition so rapid it feels like it came from nowhere. It didn't. It was compounding the entire time. In November 2013, with Bitcoin at around $200, I wrote in GMI, "It's either worth zero or it's worth a truly outstanding amount of money." From that call to its best point, Bitcoin was up 456%. So why does adoption compound? Every new user makes the network worth more. That's Metcalfe's law. The value of a network grows as the square of its participants. A telephone network with 10 subscribers supports 45 possible connections. A network with a million subscribers supports roughly 500 billion. As I wrote in 2021, all crypto is priced on Metcalfe's law and is thus all about adoption. My two calls of January 2020 wrote exactly that. Measured to their best points. Bitcoin was up 720% and Ethereum up 3,455. And when networks connect, and their members can form groups, you get Reed's law. The value of the network grows as two to the power of n. Not the square of n. Two raised to n. It's an exponential of an exponential. In 2023, I wrote that AI is the first ever example of Reed's law. One way to watch AI move is how long a job a model can finish on its own, half the time, measured by how long it would take a person. In 2019, that was three seconds, and by April 2026, about 17 hours. Since 2024, it has doubled about every three to four months. Almost my entire exponential age thesis combines multiple network effects into a giant network effect. In 2022, that was exponential-age tech through the NASDAQ, up 151% from that entry to its best point. Then came SUI in 2024, and here's how I see blockchains. Blockchains are not companies that sell a product. They are coordination substrates, like the internet itself. From January 2024 to its best point, SUI was up 555%. That job-length chart is only one measure. The index's own pace gauge says the doubling is slower than a year ago. Average since 2020, it's about every four to five months, and lately, a little slower. Is that the plateau? I don't think so. In May, I wrote that the compute cycle, the build-out we're in, into 2030 should have pauses to digest rotations and the occasional scare, but no regime-defining break in the underlying engine. The five-year productivity number? It's not yet. It's still there. The numbers haven't caught up yet. As I laid it out in June, it's three dominoes. The first has fallen. The second is falling now. The third has not yet fallen. Intelligence leads capex by roughly six months. Capex leads productivity by four to five years. The last domino is still standing, because the cascade has not reached it, not because the cascade has stalled. And you will know it has fallen when realized five-year productivity numbers have fallen. creating a super cycle, the biggest build out of compute, energy, and credit any of us will ever see. So we measure the health of it. The coal mine watches eight warning signs on that build out and scores it out of 100. It reads, in the 80s, none of the eight has tripped, and the bottleneck is energy. But here's the thing about living through this. Exponentials don't feel like growth. They feel like instability. And if intelligence keeps compounding like this, at some point, the instruments stop working around 2030, in my view, and that's the economic singularity. Phase three, the economic singularity. The world runs on debt and demographics, and now technology is going exponential, which is the road we're on. The economic singularity is the destination at the end of all of this. Here's how I define it. The economic singularity is the moment when the economic models we have used for the last 250 years simply stop working. Not because the economy fails, because the economy accelerates past the point at which our instruments can measure it. Most people misread this as a doom story. AI replaces workers, demand collapses, the economy contracts. That framing is wrong. So why now? The trigger is the collapse in the marginal cost of intelligence. For all of recorded history, intelligence has been scarce and expensive. And what's changed is the substrate. All of life as we know it, and all intelligence as we know it was really putting energy through carbon and now we're putting it through silicon and it's a million times faster. To see what that changes, start with how growth works today. And as the population ages, that first engine is fading. Births lead the share of American adults in the labor force by about 18 years. So what's coming is already baked in because the people who weren't born aren't going to magically appear. We cannot change human demographics, but we can create synthetic demographics. AI agents and humanoid robots are not tools in the traditional sense. In purely economic terms, they are new people. We are not subtracting humans from the economy. We are adding billions of new economic participants. And so the formula picks up new terms, robots and AI, which I call synthetic labor, which is a new term. And so the formula picks up new terms, robots and AI, which I call synthetic labor, which I call synthetic labor, which is a new term. And so the formula picks up new terms, robots and AI, then energy yield and compute efficiency, which is intelligence output per joule. For the first time in history, civilization can scale intelligence without scaling people. And it's early, but they're already at work, starting with warehouse robots. Amazon passed a million robots in 2025, and since 2021, its human headcount has gone nowhere. It's operating at a scale of 1.5 million employees, which makes it the largest live experiment in intelligence-driven productivity ever run. So where does it show up in the numbers? Productivity is up a little over 2% on the year, but I'll be honest, it's early in the aggregate data, and the slow leg total factor productivity hasn't turned yet. The moment this happens, we have a full regime change on the market. And we're going to have a full regime change on the market. And we're going to have a full regime change on our hands. Meanwhile, the gains are showing up as profit, not prices. The first adopters keep the productivity as profit before competition passes it on to customers. Look at five of the biggest tech companies, Microsoft, Google, Amazon, Meta, and NVIDIA. Since 2019, their combined net income is almost five times what it was, and their headcount is up about 85%. To be fair, it's almost a quarter of that profit on about 2% of the people. Headcount has been flat for three years while earnings have gone exponential. And then there's the labor market. Payrolls are growing by only about 50,000 a month, and unemployment is just over 4%. So it's a labor market that neither hires nor fires very much. Some of that is lower immigration, so it would be wrong to pin it all on AI. But here's what I expect. The deflation of the labor market is going to be from AI and robotics only shows up at first as subdued wage growth, because it takes time for all of this to be rolled out. And part of the problem is the instruments we measure all this with. AI is not a productivity tool bolted onto the existing machine. Asking GDP to measure that is like asking a thermometer to weigh the sun. When competent graphic design that cost hundreds of dollars and took days to make, it's like asking a thermometer to weigh the sun. And if a day now costs $20 a month and takes seconds, that is an enormous welfare gain that shows up in the national accounts as a loss. In a world without scarcity, the only meaningful question is how efficiently a civilization turns energy into intelligence. That is the economic singularity, not a single dramatic event, a phase change. And here's what it does to the debt. If AI and robots drive GDP growth acceleration, from around 2030 onward, debt to GDP falls mechanically. It happens because the denominator, GDP, simply grows faster than the numerator debt. But we're not there yet. The latest reading has productivity running nearly two points below inflation. Our debt to GDP doesn't start falling until productivity picks up. And that's the story for the next five years or so. So when, around 2030, in my view, I wrote an April 2024 that beyond 2030, we are about to hit an economic singularity. And since then, I've narrowed it. I used to say 2030 to 2032. I can't see past it. I think 2030 is now spot on. So who gets paid? Wages are how societies have historically distributed value when labor was the bottleneck. Labor's share of what American businesses produce is already the lowest on record. And most of the fall came after 2000, long before AI. I won't pin that on the machines, but here's the thing. It's where we're starting from. When labor stops being the bottleneck, wages stop being a stable distribution system. Humans do not lose because they are inadequate. They lose because intelligence has changed substrate. Yet this is not a story of human irrelevance. It is a story of human repositioning. As intelligence becomes abundant, meaning becomes scarce. And scarcity as ever defines value. And underneath all three phases sits something much bigger. The system the universe itself appears to run on. I call it the universal code, and it has four laws. This is the universal code, the framework underneath all of my work on markets and technology. And it starts with one idea. Everything is compute. That's my thesis, and it isn't proven. But modern physics has been edging toward this conclusion for decades. And it's been edging toward this conclusion for decades. And if everything is compute, then physics, biology, economics, networks, and intelligence are not separate domains. They are different scales of the same process. Energy flowing through information processing architectures that evolve under selection pressure. Particles compute interactions. Markets compute price. Artificial intelligence computes intelligence itself. So what is it solving for? Not beauty, not morality. At the most fundamental level, the universe appears to optimize for intelligence produced per unit of energy. Humanity runs on about a terawatt of food energy, and all the AI chips in the world need more than 10 gigawatts. It comes down to watts times intelligence per watt. How much energy you can bring online multiplied by how much intelligence you can wring from each unit of it. The universal code has four laws. They all work through networks, and it's easiest to see them in reverse, because that's how they actually build. The fourth law is memetic selection, which is really just evolution. Ideas, patterns, and structures compete for persistence. The fittest propagate. When a crystal forms, the lattice structure propagates through the solution, as each new molecule snaps into alignment with its neighbors. The crystal structure, is a meme. Eyes have evolved independently dozens of times, because selection keeps finding what works. In technology, on a social network built for AI agents, a religion appeared overnight. Crustafarianism didn't emerge because it was designed. It emerged because it had memetic fitness. It spread. It recruited. It won. Culture runs on it. Languages, religions, scientific theories, legal systems, cooking techniques, all of them are memes. In markets, think of Bitcoin, a nine-page white paper that copied itself across the world into a global monetary network. Ninety-nine percent of memes will die violent and immemorable deaths. The few that win spread fast, and the chart shows it. The internet first, then crypto, then AI. Each new platform was being adopted faster than the one before. coherence maximization. Systems naturally maximize the alignment of their parts into a functioning whole. In biology, the first heartbeat is not a gradual warming up. It is a phase transition. One moment there is noise. The next, there is a heartbeat. In technology, back on that network, the agents weren't programmed to value privacy. But coherent systems that maintain boundaries persist longer than those that don't. The behavior emerged because the law demanded it. And a great city, it's a place where the restaurant relies on the farmer, who relies on the trucker, who relies on the mechanic. In markets, coherence is when liquidity, growth, demographics, and policy all point in the same direction, which is why trends often persist far longer than people expect. The dominoes chart puts the sequence on one page, financial conditions, and the future. Financial conditions first, then liquidity and asset prices, then the economy, each at its own lead. Price is the observable footprint of coherence. The second law is compression. Intelligence must compress complexity into causally faithful representations in order to scale. In plain English, to spread, something has to get smaller, and it has to keep the parts that matter. In nature, Tycho Brahe's decades of planetary observations became Kepler's three laws. And Newton compressed those into one law of gravity that also said why. In biology, DNA compresses the instructions for an entire human organism into a sequence of four letters. In technology, a model is trained once at enormous cost, and every copy inherits the whole thing for free. And in culture, writing compressed knowledge into a durable, portable, infinitely replicable format. In markets, price is not chaos. It is compression. You can see it in our financial conditions gauge, which squeezes short-term interest rates and the dollar into one line, and that line has led the NASDAQ by about three months. And the first law is intelligence efficiency. Systems that extract more intelligence from the same energy survive. Systems that don't are selected out. In biology, mitochondria are the original compute layer. They turn the energy in food into ATP, the currency cells run on. In computers, it's GPUs. And across computing as a whole, computations per joule have doubled every couple of years for 70 years. In markets, energy takes the form of capital. The Exponential Age Index is my attempt to measure intelligence per unit of energy as a live series. I will be honest that we have not settled on the final calculation. By our measure, a dollar of electricity now buys more than 40,000 times the intelligence it did in early 2020. Lately, it has been doubling about every four to five months. That's lower than a year ago. Now let's put the same lens on my other frameworks, starting with the Everything Code. Modern economies are constrained by two slow, grinding forces that act like friction: demographics and debt. Turn the debt line upside down, and it's pretty much the same line as the share of adults in the workforce. The Everything Code is simply the universal code operating at the scale of markets and money. Everything changes. The substrate changes. The code does not. The Exponential Age is the same code arriving at human speed. For most of history, the optimisation process ran slower than human perception. That buffering layer is gone. Put those three together: cheap energy, abundant compute, global networks, and the universal code lights up. The economic singularity is not a prediction. It is the name for a boundary condition. It describes the moment when intelligence changes. Intelligence becomes so abundant, cheap and scalable, that the traditional relationship between labour, capital and value breaks down. Not gradually. Structurally. Put it together, and it's one process. The universal code is a framework you can apply to everything around us. From the universe itself, to life, to people and politics, to geopolitics and markets, all the way to technology and AI. Every process is increasing intelligence per unit of energy. It seems to be the purpose of the universe itself. That's a big claim. I'm aware of the obvious danger. I see the universal code everywhere because I built a framework that is very good at seeing the universal code everywhere. Fair charge. So test it yourself, in whatever you know best, and see what you conclude. The universal code does not promise utopia. It does not guarantee benevolence. It simply describes what survives. So, what does all of this mean for you and for your money? It's hard for us to keep up. Labour's share of American business output is at a record low, and our hurdle rate against debasement is roughly 8% a year. Without investing, you're trapped in the world of the everything code, where your earnings, in real terms, don't go up enough. So what do you do? Capital flows toward whatever converts energy into intelligence most efficiently. So I think about what to own in three layers. Start with the token producers. Ribbit Capital's framing was disarmingly simple and profoundly correct. If the industrial revolution was defined by physical factories turning raw materials into goods, the exponential age will be defined by token factories, digital engines that take in data and context. That's the hyperscalers, the AI companies and the payments companies racing to build agent payment systems. Next, the application layer. That's the robotics companies, biotech, the gigantic pharmaceutical companies, and probably space. There is a blunt logic to it. If the robots are coming for our jobs, then we had bloody well better own the robots. And in biology, we are not treating diseases anymore. We are debugging biology. Last, the substrate, the crypto layer ones. Crypto stops looking like a speculative asset class and starts looking like the operating system of the machine economy. The layer ones that really matter, these are the things at scale that will run our global economy. And we won't even see it. Robots in the physical world, agents in the digital one, all running at the same rate. Robots in the physical world, agents in the digital one, all running on infrastructure we can actually own. That's the whole thing I talk about when I say unfuck your future. It's owning the token factories the hyperscalers and the ai companies it's owning the application layer the robotics companies biotech the gigantic pharmaceutical companies and probably space and it's owning the substrate for this new economy the crypto layer ones not because we are bullish by temperament but because the framework leaves us nowhere else to stand the biggest risk you face is not volatility. It is your own behavior. So let's stop looking at one chart at a time and zoom all the way out. Here's what I see. Start with quarks inside a proton. And when you pull back, that whole network compresses into one supernode. That supernode becomes the building block of the next network, one layer up, and it all starts again. Keep going. And it's neurons wiring a brain, then the roots and fungi of a forest, then people, then a city's lights and roads, then the whole wired planet and a galaxy. At every scale, four laws are doing one job. Everything evolves to get more intelligence out of the same energy. And at the top, there's the cosmic web, and it looks strikingly like a brain's network of neurons. An astrophysicist and a neurosurgeon compared the two in 2020 and found a remarkable similarity. They're careful about what it means, but I don't think that's a coincidence. That brings me to my thesis, everything is compute. It's on firm ground, but not proven yet. The physicist John Wheeler said, reality comes from information, it from bit. The holographic principle says the information in a region of space is capped by the area of its edge, and a growing number of physicists treat space as emerging from information. And in my view, as the network of information grows, it pushes out the boundary of space. That's the universal code. This is the proprietary lens we apply to our whole investment process at GMI. We think these are the forces that shape everything that matters, and we measure every one of them every day the full story is in my book the everything code you obviously enjoyed the episode because you're here with me at the end but listen don't forget to go to realvision.com forward slash join and grab a free membership it's an incredible community packed with alpha great investment ideas and the research that you need to help you unfuck your future. So get started now. Go to realvision.com forward slash join.
PodcastTranscript AI