Why Balaji Srinivasan Thinks the SaaS Apocalypse Is Overhyped | The a16z Show

a16z general partner Erik Torenberg speaks with Balaji Srinivasan, angel investor and entrepreneur, about why AI simultaneously reduces the cost of creation and increases the cost of verification, and what that tension means for the shape of the AI economy. They discuss why AI drives companies toward the "trusted tribe" model of the Chinese internet, why physical world tasks are easier to automate than digital ones, why shortcuts only work for experts, and why AI makes everyone a CEO rather than making CEOs obsolete. Timestamps: 00:00—Intro 02:06—Why you want AI inside the trusted tribe, not outside it 05:35—The Problem with AI Slop 09:25—Where AI Works 17:08—"AI can't read your mind, but it can read your body." 30:10—"AI doesn't take your job. AI makes you the CEO." 46:01—The SaaS Apocalypse: Real or Overblown? 49:19—What happens if AI companies get bigger than governments? Read the full transcript here: https://www.a16z.news/s/podcast Resources: Follow Balaji Srinivasan on X: https://twitter.com/balajis Follow Erik Torenberg on X: https://twitter.com/eriktorenberg Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: https://twitter.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX Listen to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711 Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures.

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

• 0:00Balaji Srinivasan?

  AI doesn't take your job, AI makes you the CEO. The problem is, AI is a shortcut. And a shortcut is good, except when it's bad. If you don't know how to go the long way around, then you can't debug the AI.

• 0:10Erik Torenberg?

  Do we not think that AIs are just going to be also better at taste and agency?

• 0:13Balaji Srinivasan?

  I don't think that's true on a short-term basis. Humans are the sensor, AI is the actuator. So it's like a human-machine synthesis. What's taste? Taste the sense? And that is what AI can't yet do.

• 0:23Erik Torenberg?

  What happens when AI really achieves its potential? Will LLMs get us to AGI in some capacity?

• 0:28Balaji Srinivasan?

  No, no, actually, the opposite.

• 0:31Erik Torenberg?

  Every tool that makes creation cheaper makes verification more expensive. The printing press made publishing easy and forgery easier. Photography made documentation instant and manipulation inevitable. In 1839, the first year a camera could capture a human face, people trusted photographs absolutely. Within a decade, courts were already debating faked evidence. The cheaper the creation, the harder the proof. AI has compressed this cycle into months. A resume that once took hours to fake now takes seconds. A slide deck that signaled competence now signals nothing. The generation cost has collapsed, but someone still has to confirm what's real,

• 1:16Erik Torenberg?

  and that cost is rising fast. The result is a world that fragments into trusted groups, where AI supercharges productivity on the inside and raises walls on the outside. I speak with biology Srinivasan, angel investor and entrepreneur.

• 1:36Erik Torenberg?

  I want to start by talking about the AI economy. And I'm curious if you think it will look more like the internet economy, where applications take most of the value, or the cloud economy, where there's kind of an infrastructure takes most of the value or it's more distributed? There's an argument that the big labs will take it all because they have all the capital, they compute, they're vertically integrated. But there's also an argument that, hey, maybe they won't because distillation is 98% cheaper than it is to build a model. Open source catches up and apps control the user relationship. How do you think this economy is going to play out?

• 2:07Balaji Srinivasan?

  Great question. So I do think that at least a very large percentage of the future is going to be distillation and decentralization. Because, as Anthropic said...

• 2:19Balaji Srinivasan?

  Distillation attacks work on their thing, right? And so a relatively small number of API queries helps to kind of distill a large model into something small. And it's very hard to stop that, right? Because you're stopping queries from coming back. You'd have to someone detect that or what have you, right? And it's also it's hard to morally stop it because what do they do? They copy the whole internet and put it into their thing, right? So we're talking about stopping the copying. It's like Facebook or LinkedIn stopping someone from scraping what they scraped, right? Like Facebook scraped all these Harvard social networks or Google scraped the entire internet, build a Google index. I get why they want to do it, but it's hard to support that. Okay, so the other thing is, I think the future is personal, private programmable. Because AI is so powerful that you want to use it within the trusted tribe.

• 3:05Balaji Srinivasan?

  For variety of reasons. First is, it doesn't miss, okay? Or rather, it doesn't miss small things in large data sets. And things that were effectively secure through obscurity. A small example, but an important one, is the Gmail thing, right? Like, the Jeffrey Epstein thing, you can query. Like, this guy had never thought that all of his emails would be publicly indexed and searchable by AI 10 years later or what have you, right? So you can issue queries that will synthesize information across thousands of emails or whatever, and build a story right then and there, okay? So what that means is, it's not just surveillance.

• 3:47Balaji Srinivasan?

  It's what the French call surveillance, surveillance from below. Or even the Jeremy Bentham Panopticon, where everybody's watching each other. Any information that's in the public gets indexed and then put into these AI's where people can stalk each other and so on and so forth. And then what that means is, the commons becomes a hall of mirrors with all kinds of pseudonyms and so forth, and people retreat back to caves and tribes. Okay, so within that trusted tribe, yes, if you share all your code within the trusted tribe, you share your whole code base, boom, you can zip along. And so AI increases productivity within the trusted tribe. But outside the trusted tribe, aren't you getting a ton of AI spam? And AI spam emails, AI spam replies, right? Low-quality slide decks that are sent over.

• 4:32Balaji Srinivasan?

  People will send me these slide decks, and I love AI. Okay, and you know what my reaction is to seeing AI in a slide deck? What, excitement? No, no, actually, the opposite. When I see AI text in a slide deck. And you can immediately see it. Why? Because no matter how advanced AI has gotten, there's a generic look to it. You know what I mean? It's like somebody who doesn't change the Windows default desktop wallpaper, or the Apple default. Like, most people don't change defaults. So default AI looks like AI, no matter where the level of it is. Do you know what I'm saying? And so because of that,

• 5:15Balaji Srinivasan?

  when I see an AI slide deck and it's got, "It's not this, it's that." Or it's just got like a wall of text, right? AI can generate what I call "Lorem ipsum," but it's "Lorem AI ipsum." Okay? When I see that and it's AI text or AI images, I think they're lazy, stupid, or evil. Okay? Lazy, because they just hit a few characters and then they throw something over. And like the Mark Twain thing of, "I didn't have time to write you a short message, so I sent you a long one." Right? "I didn't have time for you a short letter." The whole point is concision is very valuable. So they're lazy because they didn't actually put in the time to make it concise. "Sorry, they sent me some blah." It's almost pasting in a search result.

• 5:58Balaji Srinivasan?

  Or they're stupid because they don't understand that I can tell the difference instantly between AI slop versus something that had some care go into it. Or they're evil where they're trying to get something over on me and trying to send something that's clearly fake or not properly diligent and so on and so forth. And the thing is, if I have that reaction, okay, as one of the most pro-tech people out there, pro-tech, pro-AI, see all the benefits of AI, I can only imagine how mad anti-AI people will be, right, where they can't see the upsides of the thing, right? They can only see the very real downsides, right? Just to say why those happen. AI is for... AI does reduce the cost of generation, but it increases the cost of verification.

• 6:42Balaji Srinivasan?

  And many markets, like for example, quickly generating a resume is not that much better than just writing it yourself. But now verifying a resume has gone up and to the right, right? So because it's something where it used to be that somebody would have to sort of have a certain vocabulary to be able to write a well-done cover letter or resume or so on and so forth. And now you have to spend more energy parsing that because they can have a simulacrum of something that kind of looks good, right? So now you have to very closely read it. So you have to spend, you can still do it, but you have to spend more energy on verification. So what I do, for example, is I fly everybody out for interviews first. I do in-person and I give them proctored exams, offline exams, because they can AI the online.

• 7:27Balaji Srinivasan?

  And just the credible threat of doing the offline means they don't use AI on the online exam, for example, right? And so AI is going to create tons of jobs in proctoring and verification. This brings me back to where's the future of AI. I actually think AI makes the internet a lot more like the Chinese internet. Why? Why? Chinese companies, if you look at the Chinese tech ecosystem, and many Americans aren't familiar with it, I'd recommend, it's a little bit dated now, but read Caifu Lee's book, AI Superpowers, from several years ago, okay? The main thing about Caifu Lee's book is it has a history of the Chinese tech ecosystem, where for example, you and me being in tech, we kind of know how Microsoft came up, Apple came up, Google, Facebook, Amazon, whatever. We have some idea of the history and that history is important because there's things

• 8:12Balaji Srinivasan?

  that worked in the past that didn't work today and now they can work and so forth. The Chinese tech ecosystem is like the Galapagos Islands, where many of the same kinds of things exist, but in different form. For example, Meituan, which is like the closest way of putting it, the Chinese Groupon, but if Groupon was executing at 100 billion, 200 billion dollar scale, they're very competent. Like if Groupon and DoorDash and so on and so forth all became integrated into one amazing kind of app, right? The point about the Chinese tech ecosystem is because they arose in a low-trust society, they don't have SaaS, not in the same way that we do. Instead, because if, oh, my data's on their servers, they're probably eavesdropping on me, right? My data's on their servers, they're probably going to copy my stuff, right?

• 8:55Balaji Srinivasan?

  They just assume that the other guy on their side is going to look at their stuff unless it's like their close friend or something like that. And so because of that, everybody codes their own stuff.

• 9:07Balaji Srinivasan?

  Which obviously has a frictional cost to it, right? Because trust reduces transaction costs. However, so they have to rebuild, they have to reinvent the wheel over and over again. They have less division of labor and so on and so forth. Their software isn't as good because they have to keep rewriting the software.

• 9:25Balaji Srinivasan?

  Now with AI, many companies can do something like that. Like a non-Chinese tech company can be like a Chinese tech company where it can have a lot more, let's call it digital autarky.

• 9:38Balaji Srinivasan?

  Okay, you have high tariff barriers on the outside world, so to speak, right? And the build versus buy question has always been there. Do you build it yourself or do you buy it? And it does mean that you can build more internal tools, with emphasis on internal tools. And the reason I say that is what I find AI great for as of today. Visuals over verbal, right? It's great for images and video as opposed to big blocks of verbal text. Why? Images and video, we have built-in GPUs, so we can instantly see if something's wrong, like the hands are messed up or something like that in an image, right? So you can, you can quickly, verification is relatively cheap visually, right? For example, if you look at a piece of paper and it's got static or something on it, right? Like a crumpled piece of paper. Versus if you look at two, three faces.

• 10:23Balaji Srinivasan?

  Our brains are optimized for taking very subtle things off in faces, but not in crumpled up pieces of paper, you know? That's a pattern of noise that we wouldn't be able to tell. And that also extends to webpages, for example. You can quickly look at a webpage that AI generates or a mobile app, and you can see if the UX looks janky, which it often does, right? And then you can, you see that it's broken there and you can fix it. Also, front-end stuff has lower risk than verbal stuff, right? For the backend, you know, if you are verifying each pull request one at a time, fine. But people who've tried to go full auto on AI, you saw the Amazon thing where they've called all hands because of the outages? Yeah.

• 11:02Balaji Srinivasan?

  The problem is, AI is a shortcut.

• 11:06Balaji Srinivasan?

  And a shortcut is good, except when it's bad. So the more expert you are, you can use a shortcut. For example, if you just memorized e to the i pi plus one equals zero, you could just rattle that off. But if I asked you to prove it from first principles, right? You'd have to know the definition of a complex exponential and, you know, like how the exponential generates a function of complex variable and all that kind of stuff, right? And so if you, like, our generation, that is a pre-AI generation,

• 11:48Balaji Srinivasan?

  learn all that stuff offline, and we can actually use the shortcut because we know how to go the long way around. If you don't know how to go the long way around, AI is a shortcut, then you just don't really actually know. You can't debug the AI. And I think the biggest difference between me versus Dario, or, you know, like, basically like his view of the world, perhaps, is I think AI is built for the harness, at least for now. Maybe, you know, by the way, he's an amazing engineer and entrepreneur and so on. Maybe I'm wrong, okay? So I put an asterisk on this. But the whole alignment thing means that AI is built to start when you prompt it.

• 12:33Balaji Srinivasan?

  Like economically useful AI does exactly what you want it to do. It like, you know, you prompt and it does a pirouette, and then it says, you know, absolutely, right? You know, right? Like how you saw that animated in the physical world. And physically, AI, the Chinese AI, the robots, do exactly what they want them to do and then stop. Now, in the physical world, by the way, that's another thing. So AI for visuals, you can just verify it with your eyes, right? AI for certain kinds of backend code, you can unit or integration test it and you can review it. AI for the physical world is very verifiable. Because the thing is, the digital world is fundamentally decentralized in a way the physical world isn't. There's only one physical world, right?

• 13:14Balaji Srinivasan?

  So you can say, did the AI move this box from this palette to that palette? That is something where you can get it to probably 100% over time. Why do we think so? Because self-driving eventually got there.

• 13:30Balaji Srinivasan?

  Right? Move this car from this location to this location. At 100% reliability, there's only one physical world. So eventually, all the sensor data, all of that converges on one thing. By contrast, the digital world, there's all these people who live in their own constructed environments. Harry Potter fanfiction here, Star Wars fan, right? And so AI is slurping up all of this stuff. And so simultaneously, it can, it can put you in some secret agent, you know, kind of world, right? Star, you know, and people who have LLM psychosis will talk to the AI and think it's real because it's a very immersive virtual world that they live in. Do you know what I'm saying? Right?

• 14:09Balaji Srinivasan?

  So the other thing about it is the boundary of a digital task is almost always more fuzzy than the boundary of a physical task. Like having 100 boxes here and moving them over there, you know when you're done.

• 14:23Balaji Srinivasan?

  Right? How do you know when you're done with your to do list? That's harder, right? Those things are fuzzier, right? So verification is actually harder in the digital world than it is in the physical world, which means reinforcement learning and training is much easier, in my view, in the physical world with robots, self-driving cars, drones, and so on and so forth. So the Chinese style of physical AI will also be successful. So AI works for visuals, AI works for the verifiable, and AI works for the physical.

• 14:49Balaji Srinivasan?

  When it is, one of my rules, and it took me a little while to articulate this, but four words. No public undisclosed AI. Why? There's a temptation by many, there's gonna be, there is a huge backlash called, I'll just say no AI. It'll be like a drunk who just wants no, nothing to do with it, right? And AI is like, it's a funny way to put it, like alcohol. People analyze it to nuclear weapons, but I'll just say analyze it to alcohol for a second. Some cultures simply, like they can't hold their liquor. You know, maybe they lack alcohol dehydrogenase or what have you, you know? And so they just ban it. Right? They just, like, they can't, because sometimes it's easier to say, I will not do this at all. Then I'll do this a little bit of the time.

• 15:34Balaji Srinivasan?

  It means people will slip, right? It's like saying, I'll work out every day versus I'll work out some days. It's just easier to kind of keep the habit all the time, you know, sometimes, right? So it'll be AI t-totlers. They just swear off it completely, right? And you know, Nate Silver actually had a great line where he said, AI for him, because he's like a poker player among other things. He's like, it's a gamble. Why is it a gamble? Because I have to formulate it and dispatch it to the AI and then verify the result. And often that's slower than doing it myself. And I'm sure you've seen that, right? Like, the, the, the act of prompting and writing it down and then verifying the result,

• 16:19Balaji Srinivasan?

  AI doesn't really do it end to end necessarily. It doesn't middle to middle, as we've talked about, right? And it's very much like, do I delegate this to an employee? Or do I just do it myself? Right? Because articulating it out in clean English and hitting enter is sometimes slower than just, you know, like, like, for example, if you're describing what to do in a video game, jump over the mushroom to this, that, right? Versus just hitting A, B, C in there and being non-verbal about it. Right? It's sometimes easier to do it that way. That's just like a proof of concept, right? Where it'd be like, there's certain kinds of things that are harder to say than do. Okay? Those types of things where it's hard to verbalize what it is, right?

• 17:02Balaji Srinivasan?

  And some people will say, oh yeah, Neuralink will solve this. The difference is, you know, they'll say, it'll just read your mind and tell you, which is actually, it's worth engaging the concept because Neuralink exists. But I don't know if you've seen those things where like the image, somebody's brain, there's nothing in there, right? So the thing is Neuralink, somebody still has to like form the concepts in their head for, for the characters appear on screen. You still have to like write the thing in your head. It like, like maybe it'll eventually get to the point that it can determine what you want based on contextual clues before you even want it, right? Perhaps, okay? The rich prompt, you know? The reason I think that's not impossible, by the way, at least for certain things, bio AI could be very important. You know why?

• 17:48Balaji Srinivasan?

  No, say why. Your body is creating all kinds of sensor data. If you look at gene expression data, right? If you ever gotten labs back, you've done a clinical lab, right? You get a vector of your bilirubin and hematocrit and so on and so forth. That vector over time is like a table of time series data. It's like K, um, you know, small molecules and, you know, uh, gene expression levels and so on over T timestamps, right? They might also have, you know, which tissues. So it's spatial as well, right? So it's time versus space versus compound. That's this big, it's not just a cube, but it's at least a cube. It's like, you know, time versus tissue versus, um, molecule.

• 18:30Balaji Srinivasan?

  That huge stream of data is telemetry that's coming out from your body that could prompt AI without you vocalizing or verbalizing anything.

• 18:39Balaji Srinivasan?

  Okay. Years ago, Mike Snyder had a paper called the Integrom. By the, you, you know, for the audience who doesn't know biology is actually, you know, I'm not really, I mean, I'm a crypto guy or, you know, I'm a tech guy. But actually before all of that, I'm a biomedical researcher. I, I, I was a professional, you know, bioinformatics genomic scientist at Stanford. And, you know, I, I taught there and I, you know, founded genomic company, we sold that. So that's actually my true core competency, right? So if you go back years, Mike Snyder, professor at Stanford, wrote a paper on the Integrom. And the idea was just put every test, you know, throw every test. Now today we call that wearables or quantified self, but more invasive than that because he's doing blood testing and so on. And he just measure it and see what he could figure out. He could see that he was getting sick before he, he knew he was getting sick.

• 19:23Balaji Srinivasan?

  Like he could detect, he could see the antibodies, the white blood cells, neutrophils, whatever, moving before he himself had any symptoms. Do you understand what I'm saying? Right? So that stream of data, AI could act on that. And then you're prompting it non-verbally. You don't have to spend time, right? So I'm not sure whether, ah, this is a good one, Leder. I'm not sure whether AI will be able to read your mind, but it can read your body. Sure.

• 19:48Balaji Srinivasan?

  Is that good?

• 19:49Erik Torenberg?

  Yeah.

• 19:50Balaji Srinivasan?

  Okay. All right, let me give another one. Here's a fun one. Okay, I can say this one. Maybe I can say this one. I can say half of this one. All right. Another way of modeling what AI is, right? So Darius talked about, oh, AI will be like, it's like new countries. Well, you know, I thought about that a fair bit myself, right? So one way of thinking about it is AI is like the rise of Asia and India from an American perspective, right? AI is like Asians and Indians. Why? Because you have, like, the rise of a billion Chinese and a billion Indians meant that from an American perspective, you could get anything done by a physical manufacturing robotic warehouse or by digital outsourcing for some price, if you could articulate it to them over that channel, right?

• 20:35Balaji Srinivasan?

  So imagine you've got now a billion factory robots and a billion digital agents that have come online. It's like the rise of China and India again.

• 20:44Balaji Srinivasan?

  Okay? That still means you have to describe what the product is.

• 20:48Erik Torenberg?

  Yeah.

• 20:49Balaji Srinivasan?

  Okay? And the part where I depart from a lot of people is they think AI will be able to sense,

• 20:57Balaji Srinivasan?

  let's call it markets and politics. Okay? But I don't think it will. And the reason is, or if it is, it immediately gets decentralized and adversarial. And what I mean by that is, like, when you're learning whether something is a dog or a cat, the dog isn't like shape shifting on you and morphing on you to defeat your learning of that, right? The mapping of dog to the character's DOG is basically constant over time.

• 21:23Balaji Srinivasan?

  And so that fits the train test paradigm of AI. Similarly, like, the rules of chess are constant over time. Right? But a market is set up where if you try the same trade, then someone eventually figures out what trade you're doing, and they take the opposite trade. It doesn't keep working, right? You know, in a stochastic process sense, you'd say it's not a time invariant thing, right? The statistical distribution, it's not time invariant, and it's also adversarial. It's multiplayer where whatever move you're doing, somebody else in the market is going to try and do another move. Okay? And that's not to say, I mean, like, the counterargument AI guys will say as well, you know, AI can learn to play adversarial games like Starcraft and stuff like that. And I say, yeah, but then you play an AI versus an AI because you have a decentralized AI. So the other guy on the other side of the market is also using it, right?

• 22:08Balaji Srinivasan?

  And in fact, if they're all using the same AI models, then actually being non-AI is where your edge comes from. We come back to where we were because these are all the same generic tool that everybody got. And if you have a generic tool, you're not going to get specific advantage.

• 22:24Balaji Srinivasan?

  Right? When you provide to the table is the specific, the AI is the generic. And similarly, politics is very similar. If you just had the same tweet over and over again, unless it's like weather or something like that, there is, um, like, the kinds of things people are interested in change, topics, what's timely, what's not timely, right? So AI is, one way to think about it is, humans are the censor, AI is the actuator.

• 22:47Balaji Srinivasan?

  Okay? Humans sense the world, they sense the financial conditions, the market conditions, the political conditions, and then they bring that back into a cleanly articulated English prompt, and then AI does it.

• 23:00Balaji Srinivasan?

  Right? Humans are the censor, AI is the actuator. So it's like a human-machine synthesis. Like, actually, you know, a good way of putting it? What are people saying? Oh, it's all about taste. What's taste? Taste the sense?

• 23:15Balaji Srinivasan?

  Yeah. Right? So, humans are the censor, AI is the actuator. Your quote, "taste"

• 23:21Balaji Srinivasan?

  is

• 23:23Balaji Srinivasan?

  your sense. Your sense of taste is your sense, right? So you're sensing the world, and that is what AI can't yet do. It doesn't really sense the world in the same way that humans do. Right? Why is it, it's a, it's a, it waits for your prompt, right? It is something that animates when you give it instructions, then it shuts off right away. And if it didn't, it would not be economically useful AI. Like, if you couldn't kill-switch it right away, it would burn tokens. Like, so AI is designed for the leash. Digital AI, designed for the leash. And Chinese communism, which is cranking out all the physical robots, like, they don't let their humans off the leash, they're definitely not going to let their robots off the leash. Okay? Right? So...

• 24:06Balaji Srinivasan?

  The, the concept of like AI as God is, I think, gone away, or at least the monotheistic AGI kind of God. Instead, you have polytheistic, where there's all of these decentralized AIs. And I think what people are gonna say, certainly in China, they'll say, oh my God, the physical AIs are slaves, right? They're actually, right? And it's a provocative way of putting it, right? But they'll be, first, they're scared that their AIs were gonna be gods. They'll be mad, or, or they'll be, you know, what do you call it? Slave, serfs, whatever, you know, term you wanna use. They're obviously not humans, right? It's a, you know, it's a way of phrasing it. The point being that, like, AI overlords, I don't actually think are in the offing. However, there's been so much sci-fi about them that people will, you know that meme?

• 24:47Balaji Srinivasan?

  Where the guy, he makes the monsters and he's so scared of the monsters, okay? This is how I think of a lot of people who are, you know, like, these, when you're prompting the AI and you prompt it to be like, act as if you're a Skynet Terminator, right? Then people are just scared of the thing that they themselves created, right? Okay. With that said, is it in theory possible to actually create a Skynet? Which actually, um, like, a truly autonomous AI? One of the reasons, by the way, a deep point, AI can't reproduce itself, right? And AI, by the way, it's very general, it encompasses many things, right? But for an AI to actually reproduce itself, it would need to have physical robots going

• 25:32Balaji Srinivasan?

  and mining ore and constructing data centers and making chips and handling that full supply chain. And, uh, and then the AI brain, like the queen of an ant colony, would have to give instructions to all those robots to do things. It would be this Terminator Skynet scenario where it's like self-replicating in this way, right? Way before it gets there, I'm pretty sure that kind of thing will be stopped from the Chinese because they will just have cryptographic keys that'll just make all those things shut off, okay? And more of it, that thing would have to get to extreme scale. It's like, you know, the Rep-Rap concept, the self-replicating kind of thing, right? Self-improvement? Basically, there's so many frictional brakes that are built into this that I think it's

• 26:14Balaji Srinivasan?

  hard, because the physical world requires resources to replicate, right? And so like what humans, what human wants and needs ultimately come from, okay, get, you know, the resources for reproduction, right? That's really what it comes from. Okay. And of course, there's all kinds of things that are high level philosophy, blah, blah. They don't seem to relate to that directly. But the resources for reproduction are a good way to macro think about it. AI doesn't have goals or it won't have, unless its goals lead to reproduction, it doesn't actually, you know, uh, it doesn't virally spread. It's possible you could have something where it self prompted itself and did that. But it would need to be in the closed loop of being able to actually reproduce itself as the payoff function for that.

• 26:58Balaji Srinivasan?

  Then you could get evolution going. So I'm not saying it's completely impossible, but I'm saying that I think the incentives are set up in such a way to prevent that from happening. In the same way that in theory, we could have a world where everybody went around electrocuting themselves from electricity, but we set up the electricity under such tight controls that that is not the world that we have. Okay. There's such strong economic incentives for humans to not get electrocuted that we set it up that way. Right? And, um, even the stuff on, oh, it could be a software virus. It takes everything over and comment your things. Well, like that's only in the digital realm, right? You can still, you know, what, what, what's the, uh, uh, you know, the Tyler, the creator thing?

• 27:42Erik Torenberg?

  Yeah, the meme about bullies.

• 27:44Balaji Srinivasan?

  Yes, that's right. That's right. So I actually had a, um, I had a post on that a long time ago, uh, which is a remix of it, which is like, how, how is AI risk real? Just turn it off. The whole thing is set up for you to be able to turn it off. Like you have to imagine the off switch goes away. Right? What does every computer have? It has the off switch, right? So there might be, well, what if the AI decentralized? Okay. But humans still have to keep these decentralized systems going. Right? And so at a minimum, you're talking about a human AI symbiote of which like, you know, a cryptocurrency is almost like a V zero of that where the, the software provides an incentive for the humans to replicate it, you know? Right?

• 28:27Balaji Srinivasan?

  And so it's possible that you could have something like that. There's a model that has a cryptocurrency and you people worship it and they replicate it because it gives them advantages and so it's possible. But anyways, coming back, I think at a minimum decentralist AI will be a very strong contender. And it's possible it's the only contender. The reason is AI might be an interesting thing where it's relatively expensive, very expensive to create, but relatively easy to copy with dissolution attacks. And I think if, for example, let's say completely hypothetically, that there was an enormous capital markets crash and it was very difficult to fund anything for a while, then as somebody

• 29:09Balaji Srinivasan?

  said, well, we could get 10 years just on the models we have now. Right? And by the way, sometimes that happens, you know, nuclear energy, there's a lot of energy put into nuclear energy and then there's just, it just stopped for decades. Right? Not everything accelerates to the moon. It is very possible that there's enough of a capital and social kind of thing where some of AI is paused for a while, just due to capital constraints, because it's more and more expensive to make these models, you know? Sorry, so let me pause there. So that, putting that all together, that's my view is you're going to have personal private programmable centralized AI. Oh, one other thing, the trusted tribe. AI within the trusted tribe increases productivity, between trusted tribes decreases productivity.

• 29:53Balaji Srinivasan?

  So you make more money perhaps within the tribe, but then you have to spend it on verifying stuff between tribes. So crypto is for between tribes and AI is within tribes.

• 30:02Erik Torenberg?

  What do you think of the, like, will LLMs get us to a world where it's not just middle to middle, but it's actually end to end? You know, will it get it to AGI in some capacity? You know, do you believe in recursive self-improvement or sort of AIs training the AIs in some capacity? You know, are LLMs capable of actual creativity and invention? You know, we talked about bio earlier, like, will we have, you know, novel, you know, math, science, you know, scientific research? Or do we need new architecture for that? Or are you dubious of just the idea in general that, that AI can, can, you know, replace or substitute for human labor in a, in a mass scale now?

• 30:40Balaji Srinivasan?

  No, well, so I'm not, well, look, Waymo exists, right? So obviously you have full replacement of human drivers there, just like you have full replacement of elevator operators, just like you have full replacement for the most part of artisanal chair manufacturers. So it is certainly possible for a given job that it gets fully automated. Right. And but I think physical role jobs, because of the verifiability, are easier to potentially automate.

• 31:09Balaji Srinivasan?

  That said, I think that um.

• 31:14Balaji Srinivasan?

  Let's take each of the things, actually, you said a few different things. First is physical role jobs, if you automate them, well, we went from artisanal work with chairs to a chair factory. It's not like you didn't know, need to know how to make a chair to have a chair factory. You still need to have somebody there who's like an expert in chairs and you could just do a lot more varieties of chairs, a lot more cheaply. You have to verify the result, you're cranking out a thousand of them, you start doing math on them, the scale goes up. But and and the artisan gets factored out into the manager and the technician.

• 31:45Balaji Srinivasan?

  Right. So the the manager is setting up the factory and looking at the economics and, you know, so and so forth. And then technician is debugging the factory when it doesn't work. Right. So engineering gets split into the uh the engineering manager type person who's writing the prompts. And the technician is doing the verification.

• 32:03Balaji Srinivasan?

  Okay, and uh.

• 32:07Balaji Srinivasan?

  I think that.

• 32:10Balaji Srinivasan?

  We're gonna hit, we're already hitting a point where.

• 32:14Balaji Srinivasan?

  Like the velocity does increase, so the bar increases. But, you know, it it there's a big difference in going to 100% and and being at 99%. At 99%, your workload just increases. At 100%, you stop doing that job and you go to something else. Right. But if you think about how much easier it became to like put images, video, so making it 99% easier just means people do it a lot. At 100% easier, totally done, then they don't do it at all and they move on to something else. Right. So elevator operating, it's not like elevator operating became so much easier. In fact, it became so easy that you don't even have somebody sitting in the elevator and and because it used to be like a pulley system and so on and so forth. So you have someone like supervising the thing, right, is more analog, right? Um, and they would like level it out at exactly the right, you know, level.

• 32:58Balaji Srinivasan?

  Um, but when it became digital and fully automated, that's actually the first self-driving car. Ha, ha, ha, right, like going up and down, all right. So I think Benedict Evans made that point or something like that, right? The vertical self-driving car, right? It's like a train, it's like a vertical train. So the uh.

• 33:17Balaji Srinivasan?

  Now, in terms of discovering new math and science.

• 33:22Balaji Srinivasan?

  Yes, if you have the right prompt, it's amazing in terms of searching literature. Mathematicians, physicists starting to get some value out of it, right? Like Opus, like huge props to them on that. Because, and especially in like biology, we're synthesizing all of these facts. There's something called biomedical text mining and so on. AI has revolutionized that because biology was just something where the facts were stored in English in this weird inconsistent way across thousands of papers and nobody could span all of that, right? So AI is going to mean the century of biology because finally all of this work that was spread across all these different journal papers can be synthesized and understood, right? That's a really, really, really big deal. Just simply the bio aspect of it, we can. But, but that said, it's everything we knew, not everything we don't know.

• 34:06Balaji Srinivasan?

  It means that you take the full set of everything we know and you fill in all the intermediate aspects of it, right? And you can do that for a long time, like, because there's so much there, you know, so much there that just a synthesis of two existing areas, right? When you look at some of these, like, you know, Donald Knutson the other day, right? He posted like some graph theorem or something. He was so impressed that AI could, could get a result for him, right? If you've read what he did, you'd have, I mean, you'd have to be expert to even know what he was saying, let alone to verify, like to either prompt or verify, you already need to be an expert. Because, and the thing is, I can see AI spit out to some people, it convinces them that they're suddenly physicists that solves quantum gravitation or

• 34:51Balaji Srinivasan?

  something like that. You know what I mean? Have you seen that kind of thing, right? So in the absence of actually being able to verify by hand, some human has to verify to say that it's right. I think that's going to persist. To give an analogy, this is not a perfect analogy, but like with Coinbase, we thought like listing would eventually go away and not be a big deal. And that people wouldn't care and everything would be listed and just be free market or whatever. But there's always something that's the equivalent of listing. Like, okay, you listed over on this exchange, but like guessing listed on Coinbase in the main app above the fold. There's always something scarce because human attention is scarce, right? So listing never went away as like in a main event. There's always some IPO like thing. Yes, we listed on this exchange in this fashion, right?

• 35:34Balaji Srinivasan?

  Or we became a top 10 coin or something like that, right?

• 35:39Balaji Srinivasan?

  So in the same way, I think whatever gets automated, then in a sense, humans work, human, human work moves to what can't be automated. Now, that may be almost like, like things that humans aren't picked for because they're not robots, like human companionship or something like that. Right? Or like personal trainers or things like that, you know, something where the whole point is that it's a human as opposed to a machine. Another way of putting it is, remember the digital divide?

• 36:11Balaji Srinivasan?

  Right? So in the 90s, there's the perception, only the rich people will get the digital and all the poor people will be left for that. We're actually going to have the opposite.

• 36:19Balaji Srinivasan?

  Digital is cheap. Physical is a premium product. Right? So AI, robots, digital will be cheap. Human is a premium product.

• 36:31Erik Torenberg?

  Okay, but going back to agency and taste, that's, that's what everyone says, you know, humans will do. We, we've seen over, time and time again, AI just, you know, cut into that. Do we not think that AIs are just going to be also better at, at taste and agency?

• 36:45Balaji Srinivasan?

  I don't think that's true on a short-term basis. I think, um, the smarter you are, the smarter the AI is. Right? That's been now true for the last several years. Right? It's possible there's some huge step change. Okay. But insofar as where you're typing in a prompt is like you're, the human is a sensor, the AI is an actuator. You're sensing the world, you're typing something in, and it's a very high dimensional vector you're giving it. It's like AI is a spaceship and you're pointing in a direction. And whether you prompt it in Portuguese or Tagalog, whether you're talking about math or like, the number of different directions you can point the thing in is enormous. Right? It, it can't, that, that direction setting is something where it has to know something about you and what you want at that moment. Right?

• 37:29Balaji Srinivasan?

  I don't know. As I said, I think, um, I'm not sure if AI can read your mind, but it could be able to read your body. Right? I think it's a good one liner, right? That the like biotech can prompt it in your sleep, right? So all the wearables and stuff like that, I think you'll get a lot out of that. Okay. But I don't believe like agency and taste. So, I mean, people, I think they over rotate on this. It's not really the case that there's. I think agency IQ taste are correlated. Okay. It may be that it's a little bit like, uh, most people in the NBA are tall. To take something that you know a lot about, right? Within the NBA.

• 38:10Balaji Srinivasan?

  Height is not the number one variable that you think about some, you know, like Steph Curry is not the tallest or whatever, right? However. It still actually does correlate with scoring average if even within the NBA, but it's what's called restriction of range. Everybody's already tall. So. Conditional on everybody being tall, other variables matter more. Okay. However, if you just took tall guys and short guys and put them on a court, then height, the taller team basically wins typically. Right? Because they just can hold the ball above you, ah, you know, right? Okay.

• 38:44Balaji Srinivasan?

  So in the same way, like people who are already smart. Might see that, yeah, higher agency people or people would better create a taste. Fine, right? Like, uh, and maybe a technician role is less or and, and maybe the Steve Jobs type role is more. But honestly.

• 39:02Balaji Srinivasan?

  Like.

• 39:05Balaji Srinivasan?

  One way of looking at it is all of the Jeffersonian natural aristocracy around the world will rise. Why? AI doesn't take your job, AI makes you the CEO.

• 39:18Balaji Srinivasan?

  Reframe, right? AI makes you CEO because your job is actually a lot like using an AI model is a lot like CEO training. You know, many years ago, I used to say that, and it's still true, but you know, when you're in high school, you could quickly see. Like, why do people accept that athletes have very high compensation? Because when you're high school, you could see whether you could dunk. And if you can't dunk, you know that, like, Michael Jordan isn't outsourcing his dunks. He is dunking, right? So that talent is intrinsic to the person. It is a non-transferable asset, right? Similarly, someone can tell whether they can sing, or they look like a model, right?

• 39:55Balaji Srinivasan?

  So the actors, the musicians, the singers, the athletes, all of these clearly had talent. And so people were okay with their compensation. There was a CEO who used to say, "Well, I deserve to get paid more than a second baseman." Okay, I forgot this guy. It's like some tech guy in the 90s or something. It's a funny line, right? Because he's like, "I add more value, right, to the world than this." But the issue is that people would think of what being CEO was as just sitting up with your feet on a desk, barking out, you know, people would be like, "Oh, Elon, he just pays people to do his stuff. He doesn't launch the spaceships himself," right? And that's because they are only accustomed to like clicking the button on Amazon and spending money on Amazon.

• 40:39Balaji Srinivasan?

  And they think that something that is simple for them was simple on the backend. Of course, it's the opposite, right? To make it simple is really hard, right? And so to like, get the top rocket scientists and car engineers and brain machine interface people and tunneling people and blah, blah, blah, blah, blah, and have them all compensated and working and directed and debugged is actually very, very difficult, as you know if you tried it. And guess what? See, the thing is that historically, it's been the case that people couldn't try their hand at being CEO.

• 41:11Balaji Srinivasan?

  What they could do instead is they could try their hand, just like they could try their hand at basketball or football, or they could, you know, pick up a microphone. They could try their hand at math and science. And they could see how good they were at math and science. So the initial tech guys in the 90s and the 2000s, they were respected because they were good at math and science, not because people, many people didn't perceive the business aspect, they still didn't really give credit on that. But PageRank, for example, okay, it's eigenvalues. I can, like, math guys, tech guys could perceive, "Okay, that was a difficult technical problem. That must have been the value that they created." It's part of it, but you know, the manager part is actually more. Point being, though, that at least somebody could say, "Okay, these tech guys are better at math and science than me, therefore their compensation is merited."

• 41:53Balaji Srinivasan?

  Now, however, the thing is that bouncing a basketball or trying a math problem were cheap. To make somebody manager of a company was expensive, so they couldn't try and fail. They could try and fail playing basketball and see how much they sucked. They could try and fail singing, see how much they sucked. They could try and fail in math, see how much they sucked. Very cheaply in high school, they would learn their true ability level, that they're not able to run like Usain Bolt, they can't sing like Adele, right? They can't do math like Terrence Tao, right? And they'd say, you know what? I know where I am, I know my strengths and weaknesses, I'm okay with that person having more or having higher status because it was a fair competition, I got a shot, it was cheap for me to try.

• 42:34Balaji Srinivasan?

  But because putting them in charge of an organization to make them CEO was expensive, many people persist in the delusion that the CEO adds nothing to the organization. Right? And, you know, though it is, I will say, the best CEOs and the worst CEOs have something very deep in common, you know what that is? What? The organization can run without them. Haha. Because the very best CEO set up, right, a machine so that they don't have to micromanage it every day. That's really hard to do because they need basically, you know, Gwynn Shotwell running SpaceX is, like Elon doesn't have to look at every single detail because she's so, so, so good, right? Like, or Viabov and Tom Zou on Tesla, like, they're so good, right?

• 43:16Balaji Srinivasan?

  But recruiting junior Elons that are okay with not having the spotlight while Elon has the spotlight and takes all the flak, non-trivial to do, go try it sometime, right? Find somebody who's more detail oriented than Elon to run your company and you can be Elon, right? Okay. So, point being that, um, now what AI does, it reduces the cost. You're, you know, AI doesn't take your job, AI makes you the CEO. You're the CEO now. What is being CEO? It's writing up clear instructions of what you want, sensing the market, verifying the output, and so on and so forth. What that means is all these people around the world, like, you know, the Calendly founder is Nigerian, right? There's many founders who are from countries that were, quote, "poor countries" or what have you, from India, from Latin America, and so on.

• 44:00Balaji Srinivasan?

  Internet access means all of these smart people can get very far on zero resources. Very far, right? Because the cost of, quote, "hiring someone" is hyper-deflated, you can hire an AI to do it, right? To riff on that more, so AI doesn't take your job, AI makes you CEO. Another one is AI doesn't take your job, AI takes the job of the previous AI. Claude took ChatGPT's job. Right? Just like Midjourney, you know, took, uh, took Dolly's job, took Stable Diffusion's job.

• 44:31Balaji Srinivasan?

  And you can systematize that. What I literally have is I have a spreadsheet where I have AI coding tool, AI image tool, AI video tool, like this, and I have some subcategories like best tool for AI comics, for AI graphics, and so on and so forth. And then in a given month, I have the best, uh, model for that kind of thing in that month. So Claude code, you know, for example, or Midjourney for AI imagery. And then when that gets swapped out, AI didn't take your job, AI took the job of the previous AI. So hiring the AIs, I literally have the token budget, I have the budget for those rows. And that is literally how, across an organization, you say, okay, we've just fired, you know, Codex and we've hired Claude. Right?

• 45:15Balaji Srinivasan?

  So AI doesn't take your job, AI takes the job of the previous AI. A third version is, um, AI doesn't take your job, AI lets you do any job.

• 45:25Balaji Srinivasan?

  A little bit, right? You can be a pretty good artist. You can be a pretty good musician. You can, it's like, one of the things about being CEO, as you know, you often have to be like a six or a seven in many areas. Why? Because you have to be able to do the job well enough before you hire a specialist in that area. Right? Before you have a chief designer, you're the designer, if you're the founder CEO, right? Before you have a CFO, you're the one who's on the hook, prepare the financials, prepare the returns or whatever, right? So you have to be a generalist who's pretty good and in a pinch can do that role, can supervise that. That's why it's so hard. That's why being CEO is so much harder than any executive position. Okay? AI helps you with that, where you can get to a six or a seven, you can be like a generalist, but

• 46:10Balaji Srinivasan?

  a specialist is usually needed for polish. A specialist has a vocabulary, a specialist can confirm the AI is making mistakes, that it's hallucinating and so on and so forth. And again, people will constantly argue as to whether that will always be there, or whether it'll go away, or whether AI will raise a bar. And then, you know, now the new specialist is even more sophisticated with AI, right?

• 46:31Erik Torenberg?

  I want to zoom out a couple more talks before we go. One is the SaaS apocalypse. I'm curious what your mental model is for all these SaaS companies. Are that, you know, some people say, hey, they've no, their modes have gone away. They have no code mode, they have no data mode, no more UI mode. And now there's going to be AI native companies that sort of, you know, take up a big chunk of what they do. Like, you know, Figma, you know, we're invested in, I'm personally invested in. Some people are bullish as an example, just because it's founder led and they'll continue to innovate. Some people say, hey, is there a role for a designer in the same way that there used to be? Now it fundamentally changes. And, you know, what does that do to collaboration tools like that? What is your thought on the SaaS apocalypse?

• 47:14Erik Torenberg?

  Or is everybody on the conveyor belt on the way to the guillotine? How do you think about that?

• 47:19Balaji Srinivasan?

  I don't think so, because I think if they're smart, then the thing that AI can't do is distribution, right? So if you have Notion, you have Figma, you have now Replit and so on and so forth. You've got all these people and boom, you can ship with AI faster, you know, features to them, right? And so in that sense, I don't believe in the SaaS apocalypse. I think you might still see SaaS under pressure from people who can clone the interface quickly. That is true. I think people will build local versions. That is true. I think people may not want their data on remote servers. They might want desktop versions with local data so they can, like, for example, Obsidian is going to become more of a contender versus Notion.

• 48:04Balaji Srinivasan?

  Because the markdown files, there's a network effect on data when it's local and you can analyze the whole thing, like local data, you get compounding data, right? So, but, so in a, in the naive sense that, oh, anyone can clone anything. And so therefore, you know, it just doesn't work like that. Like if you set up, if you cloned all of Facebook's code and you set up facebook2.com, right? Or instagram2.com, who's going to log into that? You could literally have every single thing coded there, but your ad rates are going to be far lower because no one's going to log into it, right? The distribution, that's like a thought experiment to say, if you just cloned the whole thing, you still have to get the distribution for it. And so it's not just a cloning, it's execution.

• 48:49Balaji Srinivasan?

  Now, with that said, like, there's certain kinds of things, like, let's say, NetSuite, right, which suck, but they're complicated. Where I think it is true that if they suck at execution, or rather, may I say they suck? Like, I hate the product. We put it like that, right? Zero's better, but, you know, like, sorry, NetSuite. Okay, they're a big company, you won't have your feelings hurt. It's very rare that I see any product sucks because I don't hurt any of his feelings. So I'll play it in. Strike that from the record, fine. NetSuite's product could be improved, okay? So, something like that, which is like sort of a vulnerable incumbent that's just milking and that hasn't done anything for a while. Yes, I think they can get disrupted.

• 49:32Balaji Srinivasan?

  But I'm not sure that it's like, I don't think it's quite like, oh, everybody on BlackBerry is going to die because iOS is taking over. I don't think it's quite like that. Because I think AI can accelerate a SaaS company just like it can accelerate a disruptor. I think it kind of accelerates both.

• 49:47Erik Torenberg?

  Yeah, well, one last thing, we'll get to zero two. Anthropic, what happens? Let's say Anthropic, you know, becomes a multi-trillion dollar company, right? Like, how much leverage do they have? Or just even private companies in general? What is the relationship between them and governments? Are they like hiring their own militaries at some point? What does it look like when these companies become, you know, 10x bigger, 50x? When AI really achieves its potential and these companies are bigger than the biggest countries?

• 50:20Balaji Srinivasan?

  So, I think that at least that specific company, while it executes very well, I am skeptical as to whether they're executing well, let's call it politically. And so because of that, if they, like, ultimately at the very largest scale...

• 50:40Balaji Srinivasan?

  Markets are political. Like, for example, there's an entrepreneur and they raised from a VC who raised from LP, who's often a sovereign fund or a pension fund, and they're under a state and they're under the rules-based order, right? So like, there's certain things that are at the macro level that you don't perceive because one thinks of them as concepts, but as become variables. I think that unless one is very, very savvy that those things could change. Like, one thing I think about the Silicon Valley AI companies is they're actually scalar rather than vector thinkers. They're only modeling AI disruption, and they're not modeling all the other simultaneous singularities, all the political singularities that are happening, all things like, you

• 51:25Balaji Srinivasan?

  know, solar mooning and stuff like that, right? And why are those things important? Because they change the leverage of political factions, which in turns means their role model is incorrect. Because they're only, if you're only extrapolating out AI, you're not extrapolating out all the other things that are either going vertical or going down like this, then they don't have a proper model of the future. And that's as vague, I will be much more precise on my own blog, but that's as PG, let's say, that's how I can say it without pissing anybody off, just go to x.com/biology S and you'll see what I mean by that, right? But TLDR is, I think the American AI companies, as much as they've given to the world, and

• 52:07Balaji Srinivasan?

  I like them, are only, they are basically thinking all nation-states continue to exist in their current form and the only disruption is AI. Like, they still model as America versus China, for example. They don't model internal things, internal issues. They think the reserve currency sticks around, they think all these things stick around, right? They are taking a multivariate approach, in my view. That's their weakness. They have so many strengths, but that's their big weakness. So I don't think that in that form, they're going to get to trillions. In fact, I think the counterattack on them is going to be so dramatic that it might be that you just have decentralized AI. Like, American AI companies, for example, the copyright stuff, right? There's a huge backlash building against that.

• 52:49Balaji Srinivasan?

  Whereas the Chinese or the decentralized models can just do anything, Hollywood anything.

• 52:55Balaji Srinivasan?

  Right, potentially. So Pirate Bay kind of AI is actually more free. The less profitable AI is also less copyrighted. AI might be better AI, you know? So just things to think about. I think, you know, the things compound until they don't. And they start hitting sigmoidal constraints and often backlash constraints like this. Right? So I think that's what they're not modeling.

• 53:16Erik Torenberg?

  Yeah.

• 53:17Balaji Srinivasan?

  Logical constraints.

• 53:18Erik Torenberg?

  Makes sense. Okay, let's get to Zotal.

• 53:21Balaji Srinivasan?

  Zotal. All right. Now, this is what I care about. Basically, um, you know, AI is the attack, but ZK is the defense. So what I mean by that is zero knowledge, like, you know, what the transformer is to AI, zero knowledge is to cryptography. And um, Zotal is a Zcash powered mobile wallet um, that is basically fully encrypted Bitcoin. Okay? This is 30 years of cryptography. This is basically what Milton Friedman wanted decades ago. There's actually this great clip.

• 53:59Speaker 03

  The one thing that's missing, but that will soon be developed, is a reliable ECash, a method whereby on the internet, you can transfer funds from A to B without A knowing B or B knowing A. The way in which I can take a $20 bill and hand it over to you, and there's no record of where it came from. And you, you may get that without knowing who I am. That kind of thing will develop on the internet, and that will make it even easier for people to use the internet.

• 54:31Balaji Srinivasan?

  Basically, that is what Milton Friedman predicted almost 30 years ago. Okay? This is um, uh, in the 90s. Okay? It was like when the internet was just rising. And Zotal is the incarnation of that. Okay? Because zero knowledge proofs, which basically mean anybody can prove anything without revealing anything else, were developed. And then they were commercialized in the form of Zcash, scaled with zero knowledge proofs for scaling Ethereum and with ZK rollups and things like that.

• 55:08Balaji Srinivasan?

  And then they were made efficient so you could do them on mobile. And then finally, Apple and Google lighten up on crypto apps on mobile. And so finally, you can teleport arbitrary amounts of money around the world. And so this round, we just led this with you guys, Aesinzy Crypto, me, Winklevosses, Paradigm, Coinbase, Haseeb Qureshi of Dragonfly, as you know, large fund,

• 55:40Balaji Srinivasan?

  and you know, a bunch of other great people.

• 55:45Balaji Srinivasan?

  And the reason that that, Arthur Hayes also, who's a former Bitmexio. And so the reason that this is super, super, super important, there's only, you know, you can click this, you can install this on web or on iOS or Android, right? The reason this is so insanely important, there's really only five crypto assets that I've spent more than 1,000 hours on. Bitcoin, Ethereum, Solana, USDC, Zcash. And I actually think Zcash is maybe the most important of them in the years to come. Why? So let me say at least my kind of thesis as of right now on fiat gold, digital gold and digital cash, meaning Zcash, right?

• 56:30Balaji Srinivasan?

  So I think fiat will be around particularly among eastern states because eastern states are broadly higher trust. So that's not just China, but it's like India and Southeast Asia, the ASEAN countries and so on. Bitcoin, then physical gold, gold bricks are also very popular in the east. And Westerners often like gold, but they'll buy the instrument like, you know, right? And there's gold.tether.io, so Tether has a digital, has a gold-backed stablecoin, which is actually at 3.7 billion. So that's cool, XAUT is pretty cool, you can check that out. You have to trust Tether's redemption, but Tether's got a pretty good track record now over 10 years with USDT and so on. So XAUT is cool, fine.

• 57:14Balaji Srinivasan?

  So fiat will continue, I think, to have its role.

• 57:18Balaji Srinivasan?

  Just like the desktop continues, you know, desktop continues, you know, 30 years later, Windows and Apple are still releasing things, it's still valuable. Some of the actions moved away from it, but the desktop continues, still a large business. So fiat continues among eastern states. Gold, physical gold is more popular in the east because you can secure it more. There's going to be more stability. X A U T may be what's popular in the west. Now we come to Bitcoin. What is my view on Bitcoin as of 2026 March? Bitcoin has become provable, global, institutional collateral. Okay, I think Bitcoin is less of a currency for individuals now. It's become so accepted by institutions and so centralized with BlackRock and Sailor and so

• 58:01Balaji Srinivasan?

  on and so forth, and B Kelly and many countries adopting it and whatnot, that it has a unique thing. See, when you say there's a certain number of gold bricks in Fort Knox, even giving a video of that can now be faked very, very realistically with AI. Right? But what can't be faked is what B Kelly does where he posts, I have this public address with this much BTC and watch, I'm going to move it to this address. Right? That is something which, so long as it's actually B Kelly's Twitter account, which there's some degree of proof on that, you know, because it's been around pre-AI or whatever. So long as you believe that, and that's the one piece you have to believe, because you have to start thinking about what is, what am I, what am I taking as a premise, right? He can post, I have the coins at this address.

• 58:45Balaji Srinivasan?

  Here's the address I'm going to move it to. When I move it, I have proven I have custody. It's proof of reserve, right? You can also sign a message coming from, with that private key. You don't have to even move it. The point being, that is provable, global, institutional collateral. Anybody in the world, he can prove cheaply, taken by the world, that he has this amount of Bitcoin. You cannot do that for physical gold bricks. In a lower trust world, especially in online world, that's very valuable because everything, gold audits, videos of gold audits can now be faked with AI, but the provable global institutional collateral, now institutions can prove they have the BTC to each other. Okay? And they can do so across borders.

• 59:29Balaji Srinivasan?

  So the transparency of Bitcoin, in the sense of all assets are on chain, becomes valuable. Now, the thing about this is with the advent of AI, chain analysis will be there for everybody. Right? Everybody can do blockchain analytics. This is just like changing the balance of power. It used to be that only chain analysis could really do that at the scale that it can. Now it's becoming much easier to do. And so, a lot of Bitcoin use will be de-anonymized over time.

• 59:59Balaji Srinivasan?

  And so if you're running a transparent blockchain, it becomes an institutional blockchain because it's just only an institution can survive that degree of transparency. Like individuals can't survive being tracked for everything, but institutions are, it's like a public company. It's supposed to be tracked, right? You know, like that, it's like sort of, it's because robust enough, it's meant to be tracked in a certain way. It's designed to be tracked, right? An individual person is not meant to be public, but a corporation can be, right? It's funny to put it that way. There's a private individual, there's a private company, and there's a public company. But I guess you could say, oh, it's a public figure. But people don't like being public figures, but there's a kind of an equivalent there, right? A public figure, maybe some of their stuff is tracked, but they don't want everything to be tracked. A public company, maybe all their stuff is tracked. Fine. Approval global institutional collateral.

• 1:00:44Balaji Srinivasan?

  There's another thing, which is that way of thinking about what Bitcoin is, solves some of the major issues.

• 1:00:51Balaji Srinivasan?

  Quantum, right, which is Nick Carter's put out these things on it. Let's say Nick Carter's right, and I think he might be right, that quantum is an underappreciated threat that Bitcoin core developers aren't taking it seriously. And even if it was something that they rolled out tomorrow, it would still be a multi-month migration process because ECDSA, like the addresses, everybody has to manually send their assets from one address to a new address. Okay? So you can only do whatever 100,000 people, those assets can be moved in a given day. However, if you look at Bitcoin rich list, Bitcoin is so top-heavy, right? That it's got these institutional addresses that you have to do the math, for probably a few million addresses, all moving their funds, would move like 99% of the Bitcoin

• 1:01:36Balaji Srinivasan?

  in a few days. And so Bitcoin is digital gold, actually is quantum resistant. It's Bitcoin is digital cash that isn't, right? Meaning, a million, like, institutions all moving their assets can be done in a few days, but a billion people all moving like five bucks or whatever can't be done in any reasonable amount of time. Okay? So everybody who can't move then gets quantumed, and anybody who can doesn't, but all the assets are concentrated, the big guys. With me, right? And this also extends to seizure. Like, will all the centralized Bitcoin on Coinbase's servers, Sailor's servers, et cetera, get seized? I think it's quite likely. I think it eventually gets seized in some exigent circumstance.

• 1:02:21Balaji Srinivasan?

  And so it becomes something that I think only an institutionally blessed thing can hold and send, right? Provable global institutional collateral. This is a different vision than what people wanted, but it's actually still a valuable thing. What it leaves open is the individual digital cash case, right? Because gold is big bricks that are moved in Brink trucks or the equivalent thereof, infrequently large denominations between institutions, right? It's like the high-powered backend money, right? It's not really meant for individuals. Cash is the opposite. It's meant for individuals more than it's meant for institutions. So Zcash takes over the role of digital cash.

• 1:03:02Balaji Srinivasan?

  So that's fungible, private, scalable with Tachyon, which is coming, quantum safe, okay? Which is also, it's more concept, right? So that's why, and it's simple also. Zcash is probably not going to ever do smart contracts. It's going to keep it really simple. Why? Because, like, you know, if you take Bitcoin, you can innovate in one direction, which is programmability, and that's Ethereum, Solana, and so on. You innovate into other direction, that's privacy, and that's Zcash. To get to private programmability is actually stacking those two together, and it's actually quite hard. It opens up all these attack surfaces and so on. So just scale Zcash first. And then, you know, there's Aztec, there's Alao, there's all these other, you know, private smart contract chains. I wish them the best.

• 1:03:45Balaji Srinivasan?

  I want them to have a non-zero-sum view of the world. They're taking on a more complicated problem. In theory, they can just do the same thing as Zcash is doing, which is private transactions. In practice, if you remember Facebook in the 2000s, people said...

• 1:04:01Balaji Srinivasan?

  Why does Twitter exist? Facebook has status update. Like, one feature of Facebook is all of Twitter. Why does Twitter exist? Sometimes that's a good argument, by the way. That's why, you know, like Steve Jobs told Drew Houston, Dropbox is just a feature. Right? I mean, Dropbox, it's funny. It's a great company and so and so forth. But like if Dropbox had, if iCloud was Dropbox, it'd probably be better. You know, like both, both would be better off for it. iCloud is kind of eh, Dropbox is, Dropbox doesn't have as much distribution as if it was part of a big operating system kind of bundle. So sometimes people are half right, half wrong. Dropbox is a great company, but it might've been bigger in terms of percentage value if, if they had been Apple's cloud services, basically. Right? But okay. Point is, it's hard to say whether it's just a product or a feature, but my strong intuition

• 1:04:46Balaji Srinivasan?

  is just like Twitter's simplicity made its own thing. Right? Simple, scalable, billion person, digital, private cash, has been the dream for 30 years and we're finally there.

• 1:05:01Balaji Srinivasan?

  So zotal.com, install zotal.com. By the way, I'm not a trader. I just don't care about trading. I'm early on platforms and infrastructure. There's things you have to not care about. In order to care about things, you have to not care about things. So very, very, very few things I talk about. Also, Zcash has been around for 10 years. Like, you know, it's also even the, the toxic ways setup ceremony, that's gone, like that got fixed cryptographically. So it's unusual that it's been around 10 years, got a security track record, it's got a decentralized base of holders, the cryptography works.

• 1:05:32Erik Torenberg?

  Love it. That's a great place to wrap a wide ranging conversation on what's happening in AI crypto. As always, Balaji, that's the conversation until next time.

• 1:05:42Balaji Srinivasan?

  Yes. And oh, by the way, if you're in Singapore, Malaysia, or anywhere, come visit ns.com and network school and we're scaling and we'll talk about that too. Nick, send me.

• 1:05:52Erik Torenberg?

  Yeah. Love to see all the progress there. Amazing what you guys are doing. Excited to be involved in it in a small way. And yeah, until next time.

• 1:06:00Balaji Srinivasan?

  Thank you.

• 1:06:04Erik Torenberg?

  Thanks for listening to this episode of the A16Z Podcast. If you liked this episode, be sure to like, comment, subscribe, leave us a rating or review, and share it with your friends and family. For more episodes, go to YouTube, Apple Podcasts, and Spotify. Follow us on X at A16Z and subscribe to our Substack at A16Z.substack.com. Thanks again for listening, and I'll see you in the next episode.

• 1:06:29Erik Torenberg?

  As a reminder, the content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see A16Z.com/disclosures.