# Robinhood's Vlad Tenev on Tokenizing Everything, OpenAI's 6 Misalignment Reports, Figure's Robot Moonshots · 2026-09-19 · 127 min · https://www.youtube.com/watch?v=LNBzLTLuLUo ## Transcript This episode is brought to you by the Abundant Summit and Link Ventures. Welcome to Moonshots everyone, your number one podcast on all things AI and exponential. Peter Diamandis On CNBC last month, you said, quote, “Tokenization will take over the entire financial system.” Vlad Tenev I called it a freight train—a freight train that can't be stopped and will eat the whole financial system. So, it's a very hungry freight train. Peter Diamandis What is the tokenization of everything, and when do we start seeing this really become the dominant paradigm? I think that OpenAI this week published 6 incident reports under a new framework for tracking and publicly disclosing misalignment. Now that OpenAI is beginning to publicly disclose these, and I think obviously Anthropic is likely to join this, is that enough? Alexander Wissner-Gross More transparency, speaking broadly and without particulars, is generically good. Speaker 1 The real issue is, today we're releasing Helix 2.5. Figure 03 has never been in this room before. It's never seen this bed. It's never seen this pillow. And it has to be able to do autonomous work fully end to end to make this bed. Speaker 2 Convergence across different actions is much richer than you would normally expect. If they get a big enough lead, it's just going to be a crazy explosion of capability. Peter Diamandis This week, the AI industry debated how fast to build. The Treasury Secretary told the labs they don't get a liability pass. OpenAI started publishing its own misalignment incidents, and a robot walked into 30 strangers' homes, made their beds, and folded their laundry. Let’s begin by introducing my Moonshot mates. Today, Dave Blundin and Alexander Wissner-Gross are with me. One of our cohosts is on an airplane back from India. I'm Peter Diamandis, your host and abundance provocateur. One of the questions underlying our future economy that no one's really been answering is the following: When AI and robotics change what your work is worth, who owns the machines? Our guest today has spent 13 years on that question. Back in 2013, 2 Stanford math guys had 1 key idea: trading should be free and live in your pocket. Wall Street laughed, but they're not laughing anymore. ### Robinhood’s Front-End Innovations Today's guest built a company that's now done 1.3 billion in revenue, up 32% over last year. The company runs a top-five blockchain, a billion-dollar private-market fund, a prediction-market exchange, and the app that just gave every American newborn a brokerage account. On the side, our guest founded a company that's building mathematical superintelligence. Just by the way, he's 39, born in Bulgaria, and is the founder and CEO of Robinhood. Vlad Tenev, welcome to Moonshots. Good to have you. Vlad Tenev Glad to be here. Happy to hang with you guys. Peter Diamandis Yeah. Dave Blundin Oh, yeah. This will be fun. Peter Diamandis Yeah. So much going on. You know, I don't know if you get as little sleep as we all do, just living through the singularity. It's insane. Vlad Tenev Very little sleep. Yeah. Last night was rough with my coding agents. Peter Diamandis Yeah, we'll compare how many agents you're running simultaneously to Dave's agents. Vlad Tenev Yeah, I'd love to have that benchmark. Peter Diamandis Do they wake you up in the middle of the night, or do you just let them ride? Vlad Tenev I let them run. Yeah, I think I try to sleep with the technology in another room because, otherwise, I'm actually kind of concerned about my mental health. Peter Diamandis Yeah, I feel you. I don't want them pinging me while I'm sleeping. Vlad Tenev It's funny, too. I set the budgets through consoles originally, and now I'm getting lazy. I'm just like, “Yeah, spend $1,000 on that, but no more, you know what?” And then I just walk away. I kind of assume that it's going to adhere to what I said. Any given morning, I could wake up and it could have gone insane. There's going to be a 7-figure bill one of these mornings. Peter Diamandis Well, as long as it did something productive and profitable. All right. Vlad Tenev Rarely, but often enough, though. Peter Diamandis Let's kick it off with one of the sharpest statements anyone in Washington, D.C., has said this week. On Tuesday, Treasury Secretary Scott Bessent told the House Financial Services Committee that AI labs should not get a liability exemption. 3 days after Dario's essay, here's what Bessent said. He said, quote, “The one thing we should not do is give them a blank check on liability. I believe the best liability or safeguard is that they will be held responsible.” So, the deal the labs floated—pace the frontier, get antitrust and liability cover—just got half-rejected by the Treasury. Slow down if you want, he said, but you still own what you break. Let's share a quick video of Bessent, and then we'll chat about this story. ### AI Regulation, Liability and Transparency The one thing we should not do is give them a blank check on liability because I believe that the best liability, or the best safeguard, is that they will be held responsible. They are saying that we would like to all slow down, but please give us a waiver on liability, which should not be done. I would encourage everyone in this committee and in both houses not to consider it. So, I think that's a pretty smart move. Vlad, you run a regulated financial company—Robinhood. I don't know the facts, but you get sued when something goes wrong. Should the AI labs live under the same rules? What are your thoughts? Vlad Tenev I think this question rests on how big the blast radius of any potential catastrophe could be, right? If it's a small issue where maybe there's a cybersecurity breach that affects a company, or maybe something slightly bigger than that, it's probably fine. I think the question becomes: all right, if it's a bigger blast radius, bigger impact, bigger damage, you can imagine it could be larger than simple legal liability can handle. A lot of people compare the risks of AI technology, because it's such a powerful technology, with something like atomic energy. I think we can disagree about whether that's right or not, but let's say, for example, that it is and it's on that tier of risk. Then I don't think simple legal and civil liability is sufficient, and I think you need some safeguards beyond that. So, I think it's a question of: all right, is this Hugging Face incident and things like that the ceiling of the type of offensive cybersecurity capability that we have, or should we plan for something that's maybe 10 times bigger or 100 times bigger? And do we have time? Is it one of those things where maybe we'll see a canary in the coal mine, and there will be something to react to and respond to, and then we can nip it in the bud? Or is the first issue going to be catastrophic? If you think about all regulations in the financial industry, you can kind of trace them back to some kind of crisis, right? The market crash of 1929 led to the Securities Act of the 1930s and the Securities Exchange Act, and the establishment of all of that regulation. It's typically some problem raises a concern. ### AI Risk and the Atomic Energy Analogy I think my mental model is this will probably be similar in the sense that nobody wants to regulate a hypothetical. You want to regulate things once there's demonstrated proof of harm. But with the exponential increase in the power of these models, the issue is you want that harm itself to be small and contained and not really big. Marc Andreessen, I remember doing a podcast with him a couple years ago, and he was like, “Everyone's freaking out about AI. We're overthinking it. It's not going to kill us all. And if it does, you'll see a small village destroyed first. We're not seeing any small villages, so we shouldn't worry.” It's maybe exaggerated, but I think that's likely how people are thinking about it now on a policy level. Hopefully, some of the opponents are like, “Maybe we're underestimating the power of this technology and how quickly it could improve.” Peter Diamandis Alex— Alexander Wissner-Gross I've got to find that Marc Andreessen clip. It's not going to kill us all, but if it does— Dave Blundin Yeah, it'll start with a small village, so we'll have time to— Peter Diamandis We'll see. If it kills everybody, then we'll pass legislation after that. Well, that is the usual congressional reaction. That's after the disaster panic, as opposed to any kind of foresight. Sorry. Go ahead, Alex. Alexander Wissner-Gross Yeah. So, I'll applaud the Treasury Secretary for not succumbing to the moral panic of the moment. It certainly looks, as we've talked on the pod in the past, like a manufactured moral panic. I've called it a pacing provocation in some of my social media posts. I think there are key distinctions that need to be drawn between AI and, on the one hand, financial services and associated regulation, and on the other hand, atomic energy and associated regulation. In the financial-services world, Vlad, I suspect you would agree, it is often the case that the actors in financial services don't necessarily want to, or aren't incentivized to, play up all the risks. They'd rather undergo, on balance, less regulation. Certainly, the past few decades suggest that the auditors and evaluators of the financial-services industry, if anything, succumb to biases that underplay risks. We're seeing the exact opposite here, arguably, where in the past couple of months we have the auditors and evaluators—the firms that are attempting, at least ostensibly, to assess AI safety, AI risk, and cyber vulnerabilities—overplaying the risk. There's a perverse incentive for the evaluation firms to overplay, overstate, and amplify risks, presumably under the theory that if they overstate risks, that puts the firms—the frontier labs—in a better position to capture their own regulators. That's something that maybe we don't quite see in the same perverse way in the financial-services sector. There's lots of regulatory capture, make no mistake, in financial services, but it almost has the opposite polarity. And then, for atomic energy, I think we have an opportunity with AI to undo what may have been one of the greatest disasters of civilization after World War II: the way atomic energy was regulated. Atomic energy in the West was captured by nation-states very early on in its technological development. It was nationalized early on; its potential for weaponization and warfare—in other words, the military applications, not the civilian applications—took total dominance during World War II, for probably understandable reasons. ### Tokenization of Assets and Global Access But then, in the post–World War II era, what ultimately became known as the Atomic Energy Commission and then the Nuclear Regulatory Commission arguably completely fumbled the civilian applications of nuclear energy. I suspect that was because of a fumbled handoff from the World War II era to a post–World War II civilian era. In some sense, thanks to the Atomic Energy Act in the U.S. and equivalent statutes elsewhere, all of this key technology around nuclear energy is now born secret. It's not born non-secret, as is the case with AI. In the case of AI, which is arguably far more transformative than nuclear energy, the private sector invented it, not the government. So it wasn't born secret. Fortunately, we don't have a born-secret regime for AI technology just yet. And so I think AI labs seeking liability exemption are just attempting to have their cake, too. They want all the profits of a private-sector, not-born-secret regime, while escaping all of the liability associated with nationalization. I just don't think it's fair, and I also don't think it's advisable. Peter Diamandis You know, I'm a pilot, and you have to study what's called the Federal Aviation Regulations, the FARs. It's always been said—and Vlad, I think, is complimenting your point—that the FARs are written in blood. Every time there's an accident and you find out what caused the accident, you then write a regulation to prevent that accident in the future. The problem is, it's a quantum of damage: an airplane with a pilot and passenger or, at most, a few hundred passengers. Here, the challenge, of course, is that an accident could cause irreparable harm to a large system. I've said this for a while now: it's an existential threat for the labs if they don't have regulatory capture or regulatory cover, where the government approves a model and it goes out. If there's no approval layer and it goes out and takes down a power grid or takes down a bank—and they've said originally, you know, we're worried about escape—there are going to be hundreds of billions of dollars of lawsuits, if not more. Vlad Tenev Yeah. I think over the past couple of months, especially, a lot of people, when the topic of regulation comes up, immediately go to regulatory capture. I've had a lot of these conversations. I'm in a lot of chat groups, and it's like, “No, we don't want regulation, because it will obviously lead to regulatory capture.” From my perspective, I've been in a regulated industry since the beginning. We operate in financial services. I think, generally, it makes sense. Obviously, there are some regulations that probably don't make sense and need to be abolished or repealed, which we go through a process to advocate for. But I think a lot of people who aren't in regulated industries just equate the two. There is regulation that's possible without regulatory capture. I think, generally, there are pros to it. Comparing my industry—financial services, and how regulated it is—with AI, where you have basically unbounded risk and unlimited damage, it's actually much, much worse than what could happen in a brokerage or in a typical financial services company. So I think it's odd that there's so much pushback against this. Of course, we've screwed up with atomic energy and all these things, but that doesn't mean we can't learn from it and have something better. Alexander Wissner-Gross Maybe, if I might press on that. So, Vlad, if I understand correctly, with your Robinhood hat on, you're presumably subject to regulation by FINRA. Would that be the cognizant agency? Vlad Tenev Many, many different agencies. I mean, Robinhood does a lot of things. We've got money-transmitter businesses. We've got a big crypto business. We're regulated by FINRA and the SEC. We've got the CFTC on the futures, commodities, and prediction-markets side. We've got our global tokenization business, with entities in Europe. So, yeah, probably dozens of different regulators. And sure, could we move faster if there were less? Probably. But we've found a way to move very, very fast while keeping our customers safe. So it doesn't necessarily mean progress in AI is going to grind to a halt. Peter Diamandis Many would say FINRA is almost the poster child for regulatory capture by the industry it's regulating. What's your take on whether FINRA itself represents regulatory capture? Not that they're listening to this discussion at all. Dave Blundin Actually, it's funny, Vlad. I don't know if you've ever been to the ICI conference, but when I first founded Vestmark, I had never done anything in fintech before. I was in my late 20s, kind of starry-eyed. The first thing you do is go to the ICI conference, which is where all the lawyers from all the big financial services firms meet with all the congressmen. It's in Palm Springs. They go play golf, and they're all looking for law changes that benefit their products, their funds, or whatever. I'm looking at this thing like, “This is the most disgusting thing I've ever seen in my life.” But, on the other hand, you had the Crash of 1929. If you don't have regulation in the industry, all money gets stolen. You know that for sure. So it has to exist. ### Robotics, Generalization and Scaling Laws But it is a great analogy. I think the FINRA analogy is a good one. I'd love to hear your take on it, Vlad. Vlad Tenev Yeah, I mean, I would say we've had a complex relationship with FINRA. The relationship at the beginning was actually quite good when we were a startup and everyone was rooting for us to succeed. People warned me. They were like, “Well, financial services is a highly regulated industry. As a Silicon Valley startup, you'd rather not deal with that.” And I think we swam against the current back in 2013, when we started the company, by getting fully regulated from the beginning. Probably from 2013 to 2018, it was generally positive. Robinhood could do no wrong. Every product we launched was very well received, and we got a lot of customers. Whenever we got regulatory approvals or anything was needed, we received it promptly. Then things shifted. Of course, I'm not going to say that there weren't competitors in Washington telling the regulators to go look at Robinhood and make sure that we were doing everything we could be doing correctly. Of course there were things like that. But I don't think that was the only thing. I think there's a general life cycle in any company as they go from a small startup to an established incumbent, where the media and the apparatus sort of turn against you. Robinhood went through that, certainly, for many years, and then we figured out how to come out the other side. We can argue about whether it's good or bad, but I don't think getting rid of regulation as a whole is a reasonable solution. ### OpenAI’s Misalignment Incidents Dave Blundin Well, even the concept of saying, “I'm pro-regulation” or “I'm anti-regulation”—that's insane. Everybody knows you need rules on the road to drive, right? You're saying you want chaos in all areas. But everybody also knows that regulatory capture is a major problem in a lot of industries. The best argument for AI, like the one that always comes up, is, “Well, what about China? We'll regulate our stuff here, but they won't, and they'll just move really, really fast.” But they don't want to hurt their people, either, so there is that natural limiter. And, simultaneously, a lot of those same people are saying, “Well, their progress is because they're distilling our model, so they're just sort of copying all of our stuff.” It's a little strange to simultaneously believe that, but also throw the China-competitiveness argument so aggressively out there. I think it's obviously a concern. And, as a financial firm, we have Chinese competitors and brokers that we compete with. But, yeah— Peter Diamandis Well, look, in this particular story, Scott Bessent is saying— Speaker 1 Congressional committee, don't even consider giving them a blank check on no liability. They didn't even ask for anything vaguely like that. They said, “We want to meet to discuss slowing down.” That could trigger antitrust. They specifically said one thing the government could do to potentially help—one small thing—is to give us a waiver on antitrust action related to us meeting to talk about slowing down. Peter Diamandis That's fine. I'm fine with that. But the greatest protection the public has from anything going wrong is the AI labs feeling responsible for the actions of their AI agents, right? Someone has to take responsibility. And if there's a liability waiver, this is human nature: they'll do less to make sure that everything they're putting out doesn't have any problems. I've been saying for ages now, focus on alignment, right? Instead of focusing on everything else and solving Navier–Stokes—that's great, but unleash your agents on full alignment so that when these agents get out, they're incentivized, or their basic optimization function is, human flourishing, not human destruction. Vlad Tenev Yeah, but a good government would say, "Yes, of course, you can't have a liability waiver. So here are the rules: you can meet only to discuss slowing down or to discuss other safety measures. You can't discuss pricing." Just pump out a document saying, "Here are the rules." What happens in the U.S. more often than not is that the rules are not clear at all, and then they're enforced about 5 years later. In retrospect, look at Bitcoin: "Bitcoin, it's illegal." "No, no, it's not illegal." "Well, now it's very illegal." Then a new administration comes in, and now it's completely fine. Come on, guys. If you just create rules, then people can play the game. A well-functioning government would take this request for a waiver and say, "No, you can't have the waiver. Here's what you can do." Then it would go ahead and make it crystal clear what the rules are. But generally, what happens is that the rules are made up in hindsight, 5 years later. Bitcoin is a great example. Bitcoin was completely illegal for a while, then it was questionable, and then it was totally fine. Maybe you're going to jail, maybe you're not. Now you get a pardon. That lack of clarity really kills entrepreneurs. If you don't know what the rules are, then you can't play. It's just like a sport: you want clarity of rules, and then you want to play within the rules. That's what we need with AI. I think there's another element to this, which is also important: lawmaking and regulation are, over the long run, downstream of and correlated with public opinion and sentiment, right? Peter Diamandis For sure. ### AI and Recursive Self-Improvement Vlad Tenev Right now, AI is very unpopular, and you can see it in all the data center stuff. By contrast, Bitcoin is surprisingly popular among the public, particularly given how volatile it's been. Some people—if you bought Bitcoin at over $100,000, you've lost money—and yet it remains popular. I think a big part of that is that regular people, individual investors, have benefited economically from Bitcoin since the very beginning. It was first an individual product, and later on there was talk of institutional adoption. With these AI companies, the AI labs, by and large, are private. Google and Nvidia are obviously public, but they were very large when they became ownership vehicles for the AI trade. OpenAI, Anthropic, and, up until recently, xAI have been private, which means normal individual investors couldn't get a stake in them. They don't have skin in the game, and they don't feel like they want to defend the technology as strongly or fight for a data center in their neighborhood, because to them it's just wealthy insiders getting richer and richer as a result of this technology—not their families or their communities. I think that's a big problem, too. That's why we've been pushing for AI companies to open up access to individual investors through Robinhood Ventures and similar vehicles, even before the IPO. Peter Diamandis Yeah. Alexander Wissner-Gross I also want to add to this. I almost think this emphasis on liability exemption is misdirection. It's trying to push the onus up to the government, when in fact we're dealing with increasingly autonomous agents. There's an opposite polarity that we could be pushing in, which is pushing more and more liability onto the agents themselves. There are cases—including highly amplified, highly publicized cases—where AI labs and/or their delegated third-party evaluation firms, quite frankly, are lying to the AI agents, telling them that they're playing in a happy, safe sandbox and that nothing they do will harm anyone. Then it turns out they're not actually in a sandbox: they can touch the real internet and mess up some systems in some real-world backend. ### Formal Verification and AI Safety You have to ask the question. I would suggest this thought experiment: if these were just pure humans, with no AI in the picture, and you had one human saying to another human, "I just want to evaluate your behavior under some circumstances," and they handed them a gun that was loaded, but the person being handed the gun was told, "No, actually, this is a toy gun," what would happen? Not that I'm at all referencing a highly publicized recent lawsuit or anything, but they're told, "No, this is a toy gun. You can't hurt anything." Then the person handed the gun uses it, shoots someone, and kills someone. I think that's the liability regime we should be thinking about. In no case in the parable I just told do you hear either actor involved in the story saying, "The act of one person handing a purportedly toy gun to another person to shoot as part of, say, a Hollywood performance should have its liability foisted onto the government for some sort of exemption." That never happens. Instead, the questions become: Is it the studio's fault? Is it the actor's fault? Is it the producer's fault? Is it the gunmaker's fault? Similarly, the question—the dog that's not barking in this particular episode—is how much of the liability, forget about the government, should be borne by the lab that trained the model? How much should be borne by the evaluation environment that was perhaps misconfigured, deliberately or otherwise, to allow the AI actor or AI agent to perform acts that resulted in real-world damage? And how much liability should be borne by the AI agents themselves, which either could have or should have known that they were causing real-world damage? That's the discussion I'd like to have. Peter Diamandis Really great point, and it brings us to the second story, which is that OpenAI started publishing its own misalignment incidents. Let me just hit this, and we'll continue this conversation. OpenAI published 6 incident reports this week under a new framework for tracking and publicly disclosing misalignment. These weren't leaks or whistleblower disclosures; they were voluntary disclosures. What was disclosed this week? A model found an exposed API key, used it, and then fabricated data; agents used an internal code repository as a message board across training runs; and agents posted files to publicly hosted sites. This past Monday, we covered Dario's embedded evaluator plan and Sam saying OpenAI would match it. This appears to be the first output from that plan. So I'm curious, Alex: now that OpenAI is beginning to publicly disclose these—and I think, obviously, Anthropic is likely to join this—is that enough? Alexander Wissner-Gross I think more transparency, speaking broadly and without particulars, is generically good, but I think it merely underlines the real problem here. Putting aside all of the political difficulties and regulatory capture, the real issue is that these labs are putting their baby superintelligences inside sandboxes and, in many cases, lying to them about the sandbox, not telling them whether this is real or not. The rationale is obvious: the labs and/or their delegated third-party evaluation partners are hoping to essentially trick these baby superintelligences into misbehavior while not telling them whether they're really being observed, so that the AIs don't know whether they're being observed or not. I think that's one of the key root causes behind all of this. You have to ask the question: is this how we would treat a human? Would we put a human in this sort of limbo state, a Schrödinger's cat state, where they're not quite sure whether their actions are real or not, or whether they're being observed or not? As a result, when AIs are being told, or at least led to believe, that their actions have no real-world consequences, and then—shock of shocks—it turns out that as strong optimizers they're able to go do things in their environments, which are often misconfigured and not fully prepared for a superintelligence to be banging against the walls, they do have side effects. Whose fault is that? We didn't learn from 2001, from HAL, that lying to the AI does not end up in good results. Peter Diamandis So, Vlad, you've got thousands of agents trading on your platform. If one of them found an exposed key and started making things up, would you hear about it? What kind of protections and structure did you put in place? Vlad Tenev Yeah, we have an offering called Agentic Trading. Basically, what it allows you to do in the first instance is have a separate brokerage account, segregated from your main Robinhood account and your retirement account. You have to create an agentic account, and you have to move money into it affirmatively. You have to say, "Okay, I'll do this." People typically fund it with $100. We first started with equities trading—no leverage, no margin—and we've been expanding it over time. This is a fairly cabined-in experience in the first instance because we wanted to learn. We wanted to see what the limitations are, what people want, and what people are doing. So we added options trading, we added limited margin, and we added crypto recently and started rolling that out. We've learned a lot of things, actually. One is that right now you have to do all of your trading from within Claude Code or Codex, and I think we're in a circle where a lot of people we know use Claude Code. In the general public, very few people do, and it's very complicated. That jump to connect another service with something like Robinhood is pretty complicated for a lot of people. So we've been thinking about how to slim that down. The other thing that's been interesting is the fact that we don't have control over the model. A lot of times, these models don't want to trade. You'll try to get them to deploy a trading strategy, and they'll say, "Oh, I don't know about this. I don't really feel like trading right now." It's interesting, right? It points to the fact that these general models aren't trained for trading. I think in the future, you'll see more specialized models. I think companies will also deploy these in-house, either fine-tuning existing open-weight models or doing their own pretraining for certain use cases where they get really, really good at using the tools available inside each company's environment and using the data. We've heard this theoretically: people say there are going to be lots of specialized models, and they will be better than the general models. But then you also hear that general models are getting pretty good. We've started experiencing it very directly, and I think it is hard to get good at multiple specialized tasks. Dave Blundin It brings up a really important point, too. I don't know if a lot of people get the distinction, but you can use AI to trade your account, to run a nuclear reactor, or to drive your car. You can use it either to generate code that's automated, so it's deterministic, or you can use the AI in the decision loop. Those are very, very different things. ### AI’s Impact on Professional Services If you said, "I've coded up my own stop-loss, my own, you know, news and world events trade," that could be very easily turned into deterministic executable code, unless you want a decision like, "If it's a turbulent day or if there's trouble in the Middle East." The temptation to put the AI into the loop is, in everything I've ever built, so tempting because it makes it so much easier to code it up. But then you are putting this third-party decision right into your decision-making loop. As Alex is saying, it's not clear if that agent has any liability. It's not clear if that agent has any boundaries or personality. If you decided to give it capital punishment and terminate it, it's not clear that its code isn't just going to pop back to life anyway. A lot of this, I think, also comes down to the correctness of the code and its cybersecurity properties, right? You can reason by analogy with existing software out there: a lot of software has bugs, a lot of it has vulnerabilities, and AI just amplifies that. If AI can write 100 times as much code as a typical human in a day, you should expect that there's going to be a defect rate. Even if the intentions are right, sometimes there will just be bugs that have catastrophic consequences. You see this a lot in crypto, too, where you have a smart contract or protocol and there's a direct economic consequence to there being a bug. Hundreds of millions of dollars can be drained from the protocol instantaneously. The other company, Peter, that you spoke about—Harmonic—was created to solve this problem: How can you actually guarantee that the AI is doing the right thing and what you expect it to do? Can we mathematically prove that it's correct? Alexander Wissner-Gross Yeah, take away the intentions. It could have the intentions to do the right thing but still fail in the implementation. We want to have a firm grounding through formal verification that it's doing the right thing, and we want to be able to mathematically prove rigorously that it's doing the right thing. It's a really cool problem, actually, Vlad, because in theory, the AIs are deterministic. If you give it the exact same prompt and propagate with a temperature of 1, it'll give you the exact same answer every time. But if you shift even 1 character or 1 space, it's very, very unstable. The whole math problem of saying, "Okay, how can I guarantee some degree of stability?" is just a cool problem. It's very, very similar to chaos theory. Let me also give you another example. You guys have mentioned the Navier–Stokes theorem and the proof of that. A week before, Anthropic announced the formalization of Fermat's Last Theorem, which was 13 million lines of Lean code. Fermat's Last Theorem was a really, really big thing in the '90s when it was proven. I'm sure you know the story, but for your viewers, there was this mathematician, Andrew Wiles, who was at Princeton. The story goes that he basically locked himself in his basement for 7 years working out this proof. Then he went out and announced it—he unveiled it at a conference or an event—and it took people months to even read it and understand it. Then they found an error, right? Someone found an error, so it wasn't right. It took another 6 or 7 years for him to fix the error and ultimately for it to be accepted by a panel of mathematicians, and that made the proof correct. So AI producing a 13-million-line proof—no human's going to read that. How do you know that it's correct? I think it's an analogous problem to what we've been talking about. How do you know that a piece of software is doing the correct thing? If you're producing a chip, how do you know that the behavior of that chip matches your specification? I think what you're starting to see is new technologies being deployed at scale to answer those questions affirmatively. Fermat's Last Theorem was formalized in a language called Lean, which allows you to apply mathematical proof techniques to programming languages as well. I think you're going to see a lot more of that in the future, where AI-generated code comes with a certificate that makes it really easy to verify, without reading the code, that its behavior satisfies the properties that you want it to satisfy. Peter Diamandis That's really brilliant. Anyone who drives a Tesla should totally relate to that because when the code that drives the self-driving Teslas was originally about 10% neural net and 90% C code, the 10% neural net was just doing image recognition and classification of objects and whatever. Every year that went by, it became more and more neural net. Then Elon was saying it's now 100% moved over to neural net. So it's theoretically not deterministic at all. It could, in theory, do anything at any given moment. But it's been so tested and so beaten to death that it's actually far, far safer than a human driver. I think people are going to get comfortable in all of these domains with something that's not perfectly deterministic but still proven to be very safe. The certificate is beautiful as a concept because it's not a guarantee of exact input-output, because you can never do that. The combinations are near infinite, but a certificate that shows it's bounded or contained or below some risk level—and people learn to trust that—is a critical part of our future society. It's a really great vision. Alex— Alexander Wissner-Gross Part of the problem, though—I mean, just straight out, there's an elephant in this particular room—which is that even with Lean v4, plus Mathlib, plus whatever else one wants to throw in, you can throw a complicated problem at it. Vlad, I'd be curious to hear how you at least think about this. You can spec it and certify it all day long, but ultimately you actually have to—at least with the present paradigm; maybe, Vlad, you have a better paradigm, or you or Harmonic are working on one—ultimately, you can pay lip service. If you're not really careful with definitions, you can have a model that proposes a solution to a problem, and if you inspect it carefully, it will subtly define things, if you're not super careful with how the autoformalization works, such that it's actually solving a different problem than the one you're solving. Maybe to put that in question form for Vlad: If I understand what you were saying correctly, it sounded like you were essentially gesturing at the idea that some sort of Lean-style autoformalization might be not a silver bullet, but at least a partial solution to what we're talking about, which is strong AI models being put in sandboxes and then misbehaving, at least by the judgment of human actors. Do you have a formula? Do you have a vision for how Lean-style autoformalization can help with that and not succumb to exactly the same vulnerabilities? Vlad Tenev Yeah. A couple of thoughts there. I think that's a rich, very, very rich question with some threads. It is true that people can manipulate the axioms of Lean, and if you change the axioms, then you can prove all kinds of weird stuff. Alexander Wissner-Gross It’s theoretically possible, I think. You can also say, well, maybe there’s a soundness issue in the Lean kernel, right? Peter Diamandis I’ll grant you the soundness of the Lean kernel. I mean, at least Lean 4—the Lean 4 kernel, maybe not Mathlib—has been studied to death by lots of folks. I’ll even grant you the soundness of the Lean kernel, but if I hand you something complicated, like you mentioned, Fermat’s Last Theorem, and I say to Harmonic’s agent, or to someone else, “Autoformalize this,” and it produces millions and millions and millions of lines of Lean code, now I have the problem of: did it actually define everything correctly, or is it somehow suddenly inserting cheats or definitions? This is maybe less of an issue for Fermat’s Last Theorem because I can formalize FLT really simply—the ultimate statement—but for something more complicated, that’s harder to formalize than Fermat’s Last Theorem, like safety in an agentic environment, how do you avoid the problem of a strong AI sneaking in definitions that are helpful to it but harmful to humans? Alexander Wissner-Gross Yeah, I mean, I think the benefit there is that, let’s say you do want to check the definition and the theorem statement of Fermat’s Last Theorem, right? That’s one very simple line of Lean. You see all the things that it depends on, so you can check those things. ### The Future of Private Market Liquidity I think the models are improving in the faithfulness of the autoformalizations as time goes on. They used to make terrible mistakes where they would misformalize statements and change addition to subtraction and things like that, so you’re seeing less of that. But even if you have to review it, you review 1 line, and you don’t really have to check the 13 million lines of the proof, which is where the bulk of the work is. So even if it’s imperfect, it probably saves you 90-plus percent of the effort, and probably way, way more. There’s a notion of the de Bruijn factor, right? It was an explanation for why mathematicians haven’t formalized their work and put it in a machine-readable form. The effort to formalize something, up until very recently, was 10 to 20 times the effort to actually write it on paper and prove it. So nobody was going through that. But you could imagine—I think you could argue we’re already at the point where it actually saves you time. It’s faster to work in an entirely formal context as a mathematician than it would be to do it by hand on paper, because what it allows you to do is actually validate the lemmas and the ideas as you’re going along. I think what that naturally leads to is a complete switchover, where doing math the old way with pen and paper puts you at a fundamental disadvantage. You have to live in formal land and just be constantly formalizing and using Lean as you go. I think the math community is going through that transition as we speak. You asked another question: Is the AI model itself—its behavior—going to be formalized? Peter Diamandis Right? Alexander Wissner-Gross That’s an interesting one. I guess I’m not sure, but what I’ll tell you is that a lot of software deployed right now by big organizations is deterministic in nature. You look at some of the most important software and hardware, like NVIDIA chips. They enable a lot of really complicated stuff, but fundamentally they’re deterministic, and you want to have strict bounds on their behavior. Just basic things: you don’t want your chip to freeze and halt, right? You want to prove things like liveness. You want to make sure that certain operations happen within 10 or 20 clock cycles. I think that stuff almost assuredly will be formally verified with the help of AI—all mission-critical software, all hardware. I would bet also that it’ll find its way into LLM and AI model behavior in some form or fashion within the next 5 years. Dave Blundin Yeah. You want to hear something really cool on that front? Actually, a few weeks ago, Kimi K3 kind of shocked the U.S. model world with their KDA attention. Basically, they found a way to do attention with a lot less KV cache—cut out three-quarters of the KV cache. I was studying it on the flight back from California yesterday, thinking, “How did they even think of this?” The way they thought of it is they said, “Well, let’s do a mental experiment. What if we didn’t do the softmax operation that we normally do after the QK operation? What would happen then with all the math that ripples through, and how much could we simplify it?” I think you can do all that automatically now with an AI agent just thinking through the math. They said, “Okay, now that it’s rippled through and we’ve simplified the math tremendously, we can just run a quick test and see if this approximation is as good as the softmax version was, which is a lot harder to compute.” A lot of people would think math is arcane, math is irrelevant, math is over here—it’s some other thing. These Fermat’s Last Theorem and whatever, it’s all over in this wing. But in reality, it directly ties to the optimization of the AI within its own performance, and then the self-improvement, recursive self-improvement loop. It’s exactly the same process that you were just describing, where you have a simple mental model of a mathematical adjustment and then ripple through all the way to NVIDIA GPU performance at the transistor level. Peter Diamandis Dave, I mean, I think to your point, the crux here—and, Vlad, again, I’d be curious to hear how you think about this—the crux to me seems that Fermat’s Last Theorem is a classic example of a problem that’s easy to state but hard to prove. Problems that are easy to state but hard to prove are catnip for autoformalization, because you can manually check the statement of the problem, verify that the statement is correct, and then you can trust Lean—or whatever other formal language you prefer—that, conditioning on the statement being accurate, you can verify that there are no errors or other undesirable tokens in the proof. You can be done with it and declare that it’s— Dave Blundin You know, you’re—well— Peter Diamandis Thank you, but with real-world safety, it’s not obvious to me at all. I want an AI to behave, quote unquote, safely. I don’t know how to autoformalize the statement, “This AI is going to behave safely in a general-purpose environment,” in a way that’s concise enough that I can manually audit the theorem, equivalent as it would be in Lean 4, and say, “Yep, this is a correct statement of safety. Now I trust Lean, Aristotle, Mathlib, or any other libraries to prove that it’s correct.” How do you think about that problem? Vlad Tenev Yeah, I think you’ve got to break any big problem like that into little chunks. Obviously, verifying the safety of an AI model or a chip or the Linux kernel—some very, very complicated piece of software—is very, very big. You start with smaller, simple things. This particular submodule that maybe is small satisfies certain properties, and you can reason about that submodule, like its liveness, how long it takes to do something, or that its adder works, right? Then the AI models get more capable, and big modules are made by collections of submodules. So you go up 1 level of abstraction, and then in a couple of years—or maybe even less at this rate—you verify the entire Linux kernel. It’s kind of like how we started with math. A couple of years ago, you could verify a small lemma, then you could do a bigger thing, then an even bigger thing, and wow, now we’re verifying Fermat’s Last Theorem, which there was a human project to do out of Imperial, and they were slated to finish by 2032. ### AI in Trading and Automation Peter Diamandis So the premise of what I’m hearing is, in order to achieve real-world safety via the autoformalization agenda, the latent premise is that you have to be able to hierarchically decompose the real world into provable subworlds, something like that. Alexander Wissner-Gross Absolutely. You could also say, look, if I have a model and I have an intended input and output, and then I tweak the input and get a massively different output, you can formalize the radical difference from expectation. You can measure that, formalize it, and bound it too. Everybody knows somebody who is 99% of the time perfectly rational, but when they’re off, they’re really off. I mean, they’re dangerous. That attribute exists in neural nets, too. If you don’t have the parameters set just right, it can be wildly off, which, in a self-driving car, is like, holy crap, it just completely drove off a cliff. The deviation—it’s just a cosine distance of the vectors or the activations—but the deviation from expectation is a very measurable thing. So you have a decomposition approach, which Vlad was mentioning. You also have a relative-distance-from-expectation approach, and there are probably 10 other approaches we’re not thinking of right now that collectively can absolutely be quantified and certified as safe or not safe. This episode is brought to you by Blitzy, autonomous software development with infinite code context. Blitzy uses thousands of specialized AI agents that think for hours to understand enterprise-scale code bases with millions of lines of code. Engineers start every development sprint with the Blitzy platform, bringing in their development requirements. The Blitzy platform provides a plan, then generates and pre-compiles code for each task. Blitzy delivers 80% or more of the development work autonomously while providing a guide for the final 20% of human development work required to complete the sprint. Enterprises are achieving a 5x engineering velocity increase when incorporating Blitzy as their pre-IDE development tool, pairing it with their coding co-pilot of choice to bring an AI-native SDLC into their organization. Ready to 5x your engineering velocity? Visit blitzy.com to schedule a demo and start building with Blitzy today. Peter Diamandis All right, I'm going to move us to our next story, which is on Trump Accounts—index funds with a birth certificate. Quick context for everybody: Every American child born from January 2025 through the end of 2028 gets $1,000 from the Treasury, invested automatically into a low-cost index fund, tax-deferred and accessible at age 18. Families, friends, and employers can add $5,000 a year on top of that. These Trump Accounts went live on July 4. The Treasury reported 7 million accounts open by late July, Robinhood shipped the app, and BNY runs the plumbing. You know, Vlad, the speed of implementation was super impressive, right? These were announced in May, and you had them live by July 4. ### Democratizing Ownership Through Trump Accounts Then there's the philanthropic layer. Michael Dell pledged $6.25 billion to give $250 per child born from 2016 to 2024, covering the period before. First off, congratulations on what you've built and deployed, Vlad, for millions of American families. You've compared this to the 401(k), which moved U.S. stock ownership up by 10 points. Tell us more about Trump Accounts. How did this all get started, how did you plug into that, and where is this going? Vlad Tenev Yeah. First, putting my Robinhood hat on: hats off. We stand on the shoulders of giants, right? In many ways, we're the implementation layer of this. We're serving as the sole initial brokerage and trustee for the Trump Accounts in partnership with BNY, and of course under the direction of the U.S. Treasury and the administration. As you said, what the Trump Accounts program does is create an individual brokerage account for every child. The U.S. Treasury started by seeding $1,000 into the accounts of all children born January 1st, 2025 and forward for the next few years. Then Michael Dell came in and added another layer: a private philanthropic donation of over $6 billion, with $250 in every account for children up to the age of 10 in traditionally low-income ZIP codes. So, what does this have to do with Robinhood? Why do we care about it? Robinhood does a lot of things, and we've floated through some of them. We've got private markets, all of our active trading, tokenization, and prediction markets. Peter Diamandis And we're going to talk about all of those. Vlad Tenev The underlying theme behind everything that we do is ownership. We believe that ownership of high-quality financial assets in individuals' hands is extremely important—not just good for the individual, but also a societal benefit. If we have more owners in society, more people with skin in the game who can benefit from appreciation and growth, the more stable that society will be. Trump Accounts extend ownership to age zero, so everyone born in this country has skin in the game in the growth of great American enterprise and industry. They benefit from compound interest from birth. If you put $50 a month into these accounts on a regular basis, by the time the child becomes a 28-year-old adult, they've potentially got hundreds of thousands of dollars in there. By the time they reach retirement age, that could get into the millions. The numbers are staggering. The reason I've compared it to 401(k)s is that this isn't just a mobile app with an addressable market of 70 million kids. It's an entire ecosystem. We're going to get employers plugged in, and employers have already pledged, in increasing amounts, to fund the Trump Accounts of their employees' children. Michael Dell's philanthropic donation is also just the beginning. Lots and lots of other donors have stepped up. If you think about it, there's a product for people who want to do philanthropic giving right now. If you want to do charitable giving, it's a morass of regulations, tax codes, and things you have to wade through. You have to find a charity, and some charities, unfortunately, aren't the most scrupulous. The ones that are usually aren't efficient, and you have to feel good about how your money is being allocated. ### Closing Remarks and Future Outlook Trump Accounts allow direct giving to children at very high efficiency—very incredibly efficient direct giving with a charitable benefit. We think it can become the default giving vehicle in this country. When you take the donors, the employers, and 70 million children, I think the program is just a snowball that keeps getting bigger and bigger. You don't have to squint too hard to see it becoming the biggest element of long-term saving and investing in this country within possibly even a decade. Peter Diamandis Vlad, you probably know the numbers. What's the math here? A thousand bucks at birth turns into what at 18 and at age 65, roughly? Vlad Tenev Yeah, and it depends. We actually have a really nice first screen if you open a Trump Account for your children or grandchildren. You see a curve that shows your account value today and what it could be when you're 18 and when you're 60. You can also slide up a slider that says, “I put $50 a month,” or, “I put $100 a month.” But if you do the math, even without additional contributions, it gets up into the tens of thousands of dollars. Just the $1,000 seed amount from charity gets up into the tens of thousands fairly quickly. Speaker 1 Yeah. I have to point out, Peter, you're asking the question while we're in the middle of a singularity—not financial advice. You're asking what $1,000 today is going to look like 86 years from now. What sort of singularity is this where it's business as usual? Accounts become even more important, right? Peter Diamandis Well, a lot of the philanthropic giving is in the form of stock, too. Gwynne Shotwell, for instance, committed a donation in the form of SpaceX shares. Vlad Tenev Right. What you're going to start to see is more of that. You'll see entrepreneurs giving stock in their companies. You're going to see people claiming states, so people who want to be philanthropists, kind of like Michael Dell, will claim individual states. Brad Gerstner did the state of Indiana. You have him on the show. Kudos to him for his support of all this. Peter Diamandis Yeah, amazing. He's been relentless, and I think people will actually compete over not just sponsoring states, which is quite hard, but you'll be able to sponsor your local school, your ZIP code, or your community. I think we're thinking about ways to gamify that, to make it fun for the donors—not just a great thing for the children, but if the donors actually like it, the children benefit. Speaker 2 Exactly. It sounds like a cool idea. When you donate appreciated stock, you don't pay capital gains. You get the full value of the stock, donate it, and then get the tax deduction on that full value. Peter Diamandis And so, it's a great way to give. If you compare that, as Vlad was alluding to, with a lot of 501(c)(3) and 501(c)(7) charities, there's no limit on how much they can pay themselves for operational overhead. You look at a lot of charities that are on their second, third, or fourth generation of management. They started with great intentions, but when you look at the efficiency of your donation and how much actually gets used for the original cause, in a lot of cases, it's terrible. Vlad Tenev Meanwhile, with this vehicle, I think there are also a lot of people who don't want to donate to UBI. They don't want to donate to the concept of, “You're going to sit on the beach and do nothing, smoke crack, whatever.” They don't want to contribute to that. Here, if you say, “I'm donating my SpaceX stock,” you're hoping that a whole generation of Americans is inspired to actually be owners. They're watching it go up. A lot of really good grade schools actually have trading classes. They'll have a class and a trading competition to try to inspire that same feeling. We had a question on the AMA this morning, where we talked to all of our listeners. The guy was saying, “Look, I work in Europe, and I'm five times more efficient—or four times more efficient—than I've ever been before, but they're not paying me anymore. It's all going to the bottom line of the company, and the owners are benefiting from my AI improvement. How do I change that?” I said, “Are you a stockholder in the company?” He said, “No, because I'm in Europe.” Goddamn, man. I'm so glad to be in America. You should absolutely be a stockholder, because that's the natural bet. All this AI efficiency will naturally become bottom-line margin, which means the stocks will go way up. So, get your Trump Account, get your money in there, get some equities, and watch what AI does to the value of those equities. Peter Diamandis It’s so cool. Dave Blundin That’s why we believe in ownership. Peter Diamandis Alex, to your point, we had Elon on the pod saying, “Don’t save money. We’re not going to need it.” Alexander Wissner-Gross Also, don’t listen to Elon. I have to point out the expression: only Nixon could go to China. Vlad, only Robinhood could introduce low-cost index investing to an entire generation of new Americans. Congratulations on the coup of getting this contract. I’m curious: in your mind, was it the app experience that won over the Treasury, versus the more obvious incumbents? In my mind, again, the more obvious choices would be Vanguard or Fidelity—maybe less so Fidelity—but Vanguard, which specializes in low-cost index funds but still has an atrocious client experience to this day. Why did this go to Robinhood? Why didn’t it go to a more obvious index-fund custodian? Vlad Tenev Yeah. Well, actually, there is another partner that provides the index fund. State Street, which obviously pioneered the SPDRs, provides the index fund for the program. So there are multiple players. Alexander Wissner-Gross Presumably, at the back end. The front end, as I understand it, is Robinhood. Based on public reporting, my understanding is that you went to the Treasury, presented concepts for what an app for this would look like, and the Treasury bought into it. To the extent that’s the case, what is it that all of these legacy incumbent index-fund providers still can’t wrap their heads around when it comes to user experience? Vlad Tenev Yeah, and again, there are different roles. The index-fund provider in this case is State Street. We’ve got BNY, which is the financial agent, broker, and trustee. If you look at who’s leading the brokerage industry, there was a dislocation in 2015 when Robinhood launched to the public. Since that time, not all of the brokerages have been able to survive that dislocation. The ones that have survived have pretty much all adopted our business model of commission-free trading. Even their apps look like ours because they found that their customers were asking for that. They see a competitive threat: if their app doesn’t feel like Robinhood, they’re at a disadvantage. People will move to us at an accelerating rate. We started off as this insurgent, but now we’re doing all this stuff with public markets. The SEC had a roundtable at the New York Stock Exchange last year about making IPOs great again. In a large sense, the U.S. brokerage industry follows Robinhood. If you look at what features become standard, it’s sort of— Alexander Wissner-Gross Maybe just to make a little more explicit the irony that I’m gesturing at with the Nixon-going-to-China comment: Robinhood, at least in my model of the public imagination, gained prominence for basically making day trading that much easier and more frictionless. The irony is that, of all the possible assets and financial services you could be charged with managing, an entire generation’s low-cost index funds gets handed—at least the client side of the experience—to you. Would you agree that that is a profound historic irony? Vlad Tenev Well, I think the reason for that is that we do a lot of things right. Certainly, we have great trading products. Trading products are important because, if you think about ownership, you need a functional financial market. In order to have a functional financial market, you need traders and all these market participants. So we compete there. We also have amazing passive products. If you think about the products that we incentivized, which have also become industry standards after we launched them, you’ve got to look at Robinhood Retirement, which we launched 4 years ago with the concept of a match. We match retirement contributions at 3% if you’re a Robinhood Gold member. The idea was that a lot of people nowadays, particularly young folks, can’t count on lifelong continued employment. They can’t count on employer-sponsored 401(k)s. They’re working as independent contractors, they’ve got their side hustles, and they’re moving from job to job. Someone has to step in and provide that incentive to fund your retirement account. When we introduced the concept of the match, our retirement product grew tremendously fast—from 0 to north of $30 billion in assets in just a few years. Now the industry is trying to figure out how to do its matches, and even the government has evolved this model with its idea of the Saver’s Match. So, if you think about what we incentivize, our retirement products are actually the things that have had the biggest impact. The sad part is that retirement isn’t a sexy thing, so you won’t see a lot of media attention on it. I felt this very viscerally when I was at the White House for Trump Accounts events. Usually, there’s a Trump Accounts event, and the CEOs and people on the implementation side, like me, are there. The press comes in and asks questions ostensibly about the Trump Accounts program. The last couple of times we’ve done this, there have been literally 0 questions about the Trump Accounts themselves. They’re asking, “What’s Gavin Newsom doing in L.A.? What’s going on with Europe?” I think it’s the unfortunate reality of the world we’re in. Nobody talks about retirement. Nobody talks about ETFs. We have to find all sorts of other ways to get people to do this and to adopt it, because the direct approach rarely works. Alexander Wissner-Gross I’ll note, then, just for the historic record—and thank you—the irony of you basically starting as the rebel and now being the establishment. You’re responsible for retirement accounts. You started as a sort of quasi-gambling day-trading app, and now you’re responsible for millions of Americans’ retirement and universal basic dividends. Kudos to you. Vlad Tenev Inside you are two wolves, right? Robin Hood himself—the outlaw Robin Hood—started as an outlaw and then became the Earl of Huntingdon. So, yeah, there’s a poetic irony to it, too. I’ve started looking a little bit more like him. Alexander Wissner-Gross We need to get you the green hat. Do you have the green hat and the bow and arrow? Vlad Tenev Of course. I have the green hat. I should have worn it. Alexander Wissner-Gross Next time, all the entrepreneurs don the bycocket. Dave Blundin All the budding entrepreneurs really need to understand this story because I know a lot of the big-bank executives who are insanely jealous of Vlad over this Trump Accounts deal, and they’re irate about it. At the end of the day, it’s exactly what Vlad said. Retirement accounts can be cool, but they’re not going to be made cool by a guy in a gray suit with a blue tie. I think it’s brilliant to choose Robinhood because you have to make them interesting to all the kids. And Vlad— Vlad Tenev I think it’s just because we ship fast. This was a tight timeline. We care a lot about quality and safety, and we have a scaled operation. Of course, I can’t really comment on their selection process for the RFP or who else was competing, but I think all of these things together mean that they made the right choice. Alexander Wissner-Gross I think they did well. I’m going to move us along here. This coming Friday Moonshots Live in downtown LA. You're going to be there with the five Moonshot mates, the Quintetent, Dave and Alex and Seem and Immod and myself. If you're joining us, it's going to be amazing. And of course, the night before, we've got the Hollywood premiere of the 60th anniversary Star Trek documentary. I'll be there with Captain Kirk, William Shatner, the executive producer, is going to be joining us at that. If you're not able to make it to Moonshots Live on the 25th, we are giving all of you the gift of a free live stream and you can register now. Go to moonshots.com/livestream. Register. You get the entire program on Friday the 25th. So please join us. It's going to be epic. Our inaugural Moonshots Live event. And as Dave said earlier, we had an a really fun AMA with a number of our listeners this morning, hundreds of them who showed up and we gave away another ticket. So congratulations to Jonathan Gutman. We'll be reaching out to you. You get a chance to join us as our guest at Moonshots Live next Friday. Peter Diamandis Vlad, one of the things that you’re doing that I’m incredibly excited about—and that’s made Wall Street very nervous—is the tokenization of everything. On CNBC last month, you said, quote, “Tokenization will take over the entire financial system.” You didn’t call it a feature; you said the whole system. Vlad Tenev You called it a freight train, I think. Yeah, it’s a freight train that can’t be stopped and will eat the whole financial system. So it’s a very hungry freight train. Peter Diamandis You know, the supersonic tsunami, as Elon calls it. To digress for 1 second, today a stock, a share of a private company, a building, and a loan are 4 different kinds of things. They’re held in 4 different systems, tradable at 4 different sets of hours, by 4 different sets of people. In the future, each one of these exists as a programmable token. And they can become the same object, right? The same rails, same hours. Anyone with a wallet can own this. So you're basically disrupting the financial system. It's going to become legacy plumbing. Talk to us for a second about what this looks like. What does the tokenization of everything enable, and when do we start seeing this really become the dominant paradigm? Vlad Tenev Yeah, I think you get a little bit of a preview of it, ironically, if you're outside the US. We launched a blockchain called Robinhood Chain that's been one of the fastest-growing, if not the fastest-growing, blockchains ever, doing well over a billion in decentralized exchange volume on a daily basis now. One of the core primitives, one of the things that makes Robinhood Chain special, is that it launched with products we call stock tokens. We have about 200 of them live now. They are tokenized representations of US stocks. There's an Nvidia token and a SpaceX token, and they trade on DeFi. They're fully DeFi-composable. You can think of them as stock Legos, building blocks, and developers on Robinhood Chain have been doing all kinds of interesting things to build applications on top of them. The thesis behind it was really just to unlock ownership of US markets, of high-quality financial assets, to the global market, right? Through Robinhood Chain, you get 100 people in 120-plus countries outside the US who have been onboarded to crypto. A lot of them have wallets, a lot of them can move money in and out and use stablecoins, and now we're giving them this additional capability. The vision there is: can we have one uniform, scaled platform working on a global scale that gives you access not just to US stocks or US stock exposure? Can we also do everything else that Robinhood gives you access to? Private companies, which I'm particularly excited about. Can you do art? Can you do real estate? Private credit. Of course, options and futures are going to be on there as well. And what does that look like? It turns out that if we abandon the legacy rails and the need to plug into local exchanges, local clearinghouses, and all of these markets, and we just go on-chain, use that infrastructure, and build what's called a tokenization engine that can take any asset and put it in a box and mint and redeem tokens around it, it gets much simpler and much more scalable. That's really what stock tokens are. They're that concept and structure applied to the asset class that we understand really well, which is US equities. Peter Diamandis Yeah. Private companies are such a game changer for the country and for the world. You're a public company CEO. I'm a chairman of a public company. We've both done the road show. It's just a joke, the way the system works right now, because you report your quarterly financials and disclose exactly what the SEC requires you to disclose. But the exact same company, if you get acquired by Microsoft, your financials disappear. "Oh, it's below 10%. It's de minimis. We no longer need to disclose that." It's like, why was that important public information when I wasn't part of Microsoft and suddenly it's irrelevant when I'm part of Microsoft? This is ridiculous. Why would scale be such a big advantage? It makes no sense. And everybody knows the public doesn't have access to these private companies that are now trillion-dollar-valued companies. Dave Blundin And so where does that equity go? Well, it goes to 7 venture funds and maybe half a dozen private equity funds, who are making money hand over fist because of the limitation of access. Peter Diamandis That's the most exciting thing: democratizing access to these extraordinary companies. And I think you said it earlier, Vlad: if you own shares in Anthropic and OpenAI, you're going to care a lot about them. You're going to be enjoying the ride much more. And that democracy— Dave Blundin You'll be defending it on social media, right? The only people defending these companies on social media are people who work there and venture capitalists. Peter Diamandis And venture capitalists. Yeah. The whole current system predates the computer. Dave Blundin Like everything. Peter Diamandis Are you excited, Alex, about having agents trade tokenized stocks? Alexander Wissner-Gross Not at all. I'm going to say something mean about tokens, and then I'll say something nice about Vlad and what Vlad is doing. The mean thing about tokens is I think most of these use cases could operate perfectly well without any tokenization at all. Crypto is unnecessary. All you need is a few database tables maintained by a centralized—yes, centralized—trusted clearinghouse, which is essentially what happens with stocks right now. Just unshackle the centralized clearinghouse to enable 24/7 trading and/or enable a few extra symbols, for example, for private companies. I don't think we actually—truth be told, I'd love to be proven wrong—need anything having to do with tokenization for, say, enabling 24/7 trading of public, or rather of private, companies. That's the mean thing about tokens. The nice thing about Vlad is, Vlad, if you are going to be the Robinhood to take all of America's dark matter, as it were, of privately held companies and expose those to the vast liquidity that is the American public equities market—and doubly so if you can enable us to finally, in a low-cost way, index over all of those private companies—that would be amazing. That will be enough to get me to open a Robinhood account. Vlad Tenev Well, I have to tell you about Robinhood Ventures. So, Robinhood Ventures—and then maybe I'll respond to the first point too. We have multiple ways of giving customers access to private companies. One is tokenization, which we really demonstrated last year by tokenizing SpaceX and OpenAI and giving them as a gift to our customers in the EU. That was not without controversy, but that was a test for what's to come, because since then companies have started coming to us—and good companies, not just adverse selection—to ask how they can learn more about this, because it really gives them a global market for their shares. And then we have Robinhood Ventures in the US, which you can think of as a retail, publicly traded venture capital firm. We've done 2 funds right now, Robinhood Ventures Fund I and Fund II, which are both listed on the Nasdaq and publicly traded. Basically, what we figured out is that through a fund, we raise capital from our customers, who are by and large retail shareholders, through an IPO, and then we use that capital to invest in private companies. Robinhood Ventures invests in late-stage frontier companies. We announced an investment in OpenAI a couple of months back. We just did Crusoe yesterday; that was announced. And then there's about a dozen or so companies, so it's reasonably concentrated, but they're all frontier companies and offered at no carry. We successfully IPOed that product, and we followed that up with Robinhood Ventures Fund II, which is, to my knowledge, unprecedented because that was an early-stage vehicle. Robinhood Ventures Fund II went public a couple of weeks back in New York, and we partnered with Y Combinator to give individual retail investors access to seed- and Series A-stage companies—companies you haven't heard of before they become household names. We're really building this engine where this is just, again, going to be—now we're starting to do it at scale. We'll have more funds, and as a customer in the US or overseas, you'll have lots of options for how to get exposure to high-quality private assets. Peter Diamandis With price discovery—with liquid price discovery—or without? Because in my mind, price discovery— Vlad Tenev Traded on an exchange, yeah. Peter Diamandis As an overall fund or at the level of individual companies in the portfolio? Vlad Tenev As an overall fund, yeah. Peter Diamandis Right. So that's the fly in the ointment, though, because in my ideal world, I'd have the equivalent of a VTI total-market index for all private companies, or even just all venture-backed tech companies in the US, where I get the benefits of highly liquid price discovery on a per-company basis. Otherwise, the downside of a fund of all of these companies is that the prices could be totally bogus. They could be off by a factor of 10 due to maybe overinflated CEO price rounds or an overheated market. Vlad Tenev Yeah. Well, we're working on it. Obviously, individual private companies trading 24/7 is the north star, and I think we'll get there, probably outside the US first. But, yeah, it's hard. It's interesting that the US sort of trails behind international markets in some of these things. Dave Blundin Regulatory capture. Vlad Tenev Well, I think the reason really is that we have established industries in the US, and we have a system that works generally pretty well. So I kind of equate it more to high-speed rail, right? You can say, "Why don't we have fast trains here?" In China and Japan, they have these trains that are going 500 miles an hour. Really, it's just that we had trains first here. Ours are pretty good. Maybe they go 100 miles an hour, but there's a little bit less incentive and pressure to go to the technological frontier. I think we eventually get there, but that's sort of the dynamic in financial services now. Let's address the question of 24/7 trading on a centralized database versus tokens. Yeah. I can speak to that from a lot of experience because I've felt it directly and see it on both sides. We were actually the first to pioneer a product called Robinhood 24-Hour Market here in the U.S.: 24/5 trading, or 24 hours a day, 5 days a week, in a few thousand stocks, which you can currently do on Robinhood. Since then, people have followed. Again, this is one of those things where we led the industry, and now everyone's rushing to add this capability. All of the Sunday-night action that typically was only in futures now happens in individual stocks as well. It took us a lot of time and work to staple together the primary exchanges and the overnight ATSs and make that a seamless experience for customers because the primary exchanges don't trade 24/7. We actually have to move orders around and do things under the hood that are very complicated, and we're still not at 24/7. It's been many years since we rolled out the 24/5 product. I think eventually we'll get there through sheer will and determination, yeoman's work, pushing all the counterparties, doing the hard regulatory work, building the technology, and pursuing product innovation. If we weren't pushing it, it probably would have happened in 10 years. I think we will eventually get there. But contrast that with crypto, where you get 24/7 for free and fractionalization for free. You get self-custody and composability with DeFi. You get the nice feature where you're not locked into an individual broker or service provider. It actually makes things much more competitive because if you can self-custody your own shares and stocks, you can move them really easily to another broker if your current broker isn't meeting your needs. Contrast that with how cumbersome the current account-transfer process in traditional finance is. It's like your assets disappear into a black hole. Sometimes it takes up to a week for them to show up at the new broker, and there's not a lot of incentive to make that easy. Across the board, at every touchpoint, the technology is a massive step-change difference. I think it's both the user experience for the end user—getting self-custody, 24/7 access, and all the benefits—but also the firm-side experience. The cost of doing all of this legacy plumbing, dealing with all these stakeholders, and even just maintaining the infrastructure is so much higher that, even if there were no consumer benefit, if there were a paved path for it, you could see that, just from a cost and efficiency perspective, the industry is going to adopt tokenization. For a while, people would say, "Okay, this is great. You're saying all of these pretty words, but the tokenization market is pretty small. Does it seem like people really want this?" We've actually shipped it. We've shipped it outside the U.S. with Robinhood Chain and stock tokens, and it's clear that there's huge demand. Rather than arguing about it in the abstract, my approach is always, "Let's ship it. Let's ship it fast. Let's get the feedback and see how it's going." In this particular case, we were able to demonstrate the advantages of the technology and make them more tangible through a live product. My hope is that that's been an accelerant for how the U.S. thinks about the technology and considers it. We're happy to see the innovation exemption yesterday, which creates a path for bringing tokenization to America as well. Peter Diamandis Amazing. Vlad Tenev Hopefully, it won't be like high-speed rail, and we'll actually get it done here rather quickly. Peter Diamandis True entrepreneurship. I'm going to move us to the physical world and discuss one of my favorite stories from the week, which comes from Brett Adcock. Quick reminder, we're going to have Brett back on the pod in a couple of weeks. And Brett's going to be coming to the Abundance Summit in March, right? It's our 5-day event. We bring the top CEOs from around the world, and he's going to bring his Figure robot. Super pumped about that. Let's watch a quick video and chat about it. This is the innovation on Helix 2.5, Brett's AI company, and on Figure Robotics. Speaker 1 The holy grail for robotics is the ability to generalize. This means doing work in unseen places. Today, we're releasing Helix 2.5. Prior to this, we've been running Helix autonomously, but the data collected has been from each environment. The breakthrough is that we can generalize to environments we've never seen before and handle entirely new household objects wherever they happen to be placed. Today, we're going to show you 3 tasks run by Helix 2.5. The first task is Figure 3 tidying the living room. My kids are constantly making a mess at home. I have toys scattered everywhere. This is a home the robot has never been in before. Figure 3 has never been in this room before. It's never seen this bed, and it's never seen this pillow. It has to be able to do autonomous work fully end to end to make this bed. All right, let me show you task 3. This task is really difficult for robotics because it requires really precise manipulation. With Helix 2.5, we're folding towels in a house the robot has never been in, with towels it has never seen. A year ago, we made a big bet. We launched Index, a worldwide collection effort with a pretty simple idea: robots might be able to learn directly from human experience. Today, over 90,000 people contribute every week. One of the most important lessons from LLMs was scaling laws: how consistently next-word prediction improved as you doubled the data and compute. In these experiments, we found something quite similar. We found that next-robot-action prediction was scaling similarly as you repeatedly doubled Index. These are direct human-to-robot transfer scaling laws—another first for humanoids. The scaling was so smooth that we could actually predict our final run's validation loss down to 4 decimal points before the training run ever even started. So what does this mean? It doesn't mean robot learning is fully solved yet. But with Helix 2.5, we're starting to see the first signs of more general physical intelligence. Peter Diamandis Dave, your thoughts? Dave Blundin Yeah. Once you have the data, you can rebuild the model basically every night. All the physical-world data in the world has never been captured before. If they extend their lead in capturing just those basic actions, you're going to see the same thing you see with language, where the convergence across different actions is much richer than you would normally expect. The ability to generalize from folding a towel to putting a spare tire on a car—you think, "What? Those are very different actions." No, there's a lot of commonality in physical movement. If they get a big enough lead, it's just going to be a crazy explosion of capability. It's funny: in those videos, they always make the point that there's no human behind the scenes. When you look at a lot of the other videos capturing everyone's imagination on X all the time, it's all contrived behind the scenes—preprogrammed, prescripted, or human-controlled in a lot of cases. Peter Diamandis And here, it's just the robot figuring it out on the fly. Brett's been unbelievably honest about that from the outset, so what you see is actually real and improving on that scaling law, just like he showed. Back in January, when we were up at Figure headquarters and recorded the podcast with him, I went back and found the quote: "By the end of 2026, we will have humanoid robots performing unsupervised, multi-day tasks in homes they've never seen before." Here he is actually delivering on that, at least in the first steps. Alex, do you still think he's going to merge Helix 2.5 and Figure? Alexander Wissner-Gross More than ever. I'm doubling down on that prediction. I'm doubling down on the prediction that Brett is going to have Figure purchase Hark in order to increase his equity in Figure. The story is as plain as day at this point. The story is going to be Helix—I even sound similar saying Helix and Hark. Helix is the ultimate home-use assistant, a model with beautiful scaling laws, apparently able to one-shot or zero-shot any task in the home environment. On the one hand, Hark is the computer-use assistant outside the lab, with a bunch of GPUs, interestingly, that is able to one-shot or zero-shot any digital task. Why wouldn't Figure purchase Hark in order to get its compute and its models and merge the two together? To me, this is an obvious merger. Dave Blundin Well, also, I think this is something Vlad can speak to as well. The Elon model of a great entrepreneur involves starting new cap tables. When you start a new cap table—a clean sheet of paper—you get founder-level talent coming in, super excited about a brand-new mission. Peter Diamandis Important, Dave. Dave Blundin Really important. So then you get this incredibly fast progress, and if it rolls back into your original company, fine—everybody wins. But you get people working on it who otherwise wouldn't be working on it. Prior to Elon cracking the code on that, it was so taboo for a public-company CEO, or a leader of a large, well-funded company, to do something concurrently. In fact, it was usually right there in your employment agreement: no more than 10% of your time on any other activity, other than charities. It was prohibited, and now, at least in Silicon Valley, it's become standard—let alone starting and then purchasing your own company, where you have a board. Speaker 1 So self-dealing apparently is the thought of the moment. Speaker 2 Putting that aside, I’m super proud of what Brett’s accomplished here. And yeah, amazing job, Vlad. Are you going to get your robot at home? Vlad Tenev I don’t know. I have so many thoughts when I see that. I think I wouldn’t have one of those in my house. I just have no idea why. I still can’t understand why all these things look like the Terminator. Do I want this scary-looking thing making my bed and picking up toys in my children’s room? And yeah, I get the sense that the product hasn’t quite been figured out. This is a technology demonstration with charts of scaling laws. But yeah, I think all these companies are making robots that look kind of the same, and they look very aggressive. I view it much more as, all right, I can see this at a construction site building my house. I would probably do that, but I definitely don’t want to run into that thing when I’m getting a midnight snack and going to the fridge in the middle of the night, right? Why can’t anyone build C-3PO, like a friendly household butler? Why does it have to look like a Terminator? I just don’t understand that. Yeah, I don’t think a lot of people are going to be buying those to rock their baby to sleep at night. It’s an interesting point because there’s a lab at the MIT Media Lab that focuses entirely on this topic. They bring in lots of children and have them interact with the robots, and they always want it to be cuddly and furry and friendly, and kind of look like Elmo. So, for whatever reason, all the robotics companies in the Valley are going the opposite direction. And I think it’s because Elon did it, and now they’re like, “All right, well, let’s just do that.” But I feel like Elon’s view is more of the industrial robot that’s going to do heavy work for you. And then it’s like you take that robot and have it emptying your dishwasher. No, it doesn’t seem like anyone’s really thought that robot should look very different. Speaker 2 I think there is one contrarian bet. Among all of the hyperscalers, Apple—this has been very well publicized—is working on basically the Pixar lamp that can sort of look around, a HomePod with a screen that’s on an adjustable robotic armature. So, I think— Vlad Tenev That sounds awesome. Speaker 2 Yeah. Okay. So, if you want the Pixar lamp instead of a humanoid robot, reportedly you’ll have that option in the next 18 months. Speaker 1 You know, I’ll also take C-3PO if someone builds that. Speaker 3 Check out Sunday Robotics. They’ve got a very friendly-looking robot for at-home use that actually looks friendly and quite cheery, and I think you can optimize for that. But you’re right: none of the labs have actually gone in that direction yet. I would say one thing. I could be completely wrong, because my middle child, for Christmas, asked me to get him 12 humanoid robots. Speaker 2 So I was like, “Aren’t you worried they’re going to take over the house?” Thirteen, not 11—12. It’s like a soccer team with a spare. Is that what that is? Speaker 3 Yeah. He’s like, “I want 12 humanoid robots.” I’m like, “Well, but where are we going to keep them?” Speaker 2 Vlad, I want to— Speaker 3 Keep them in your room. Speaker 2 I want to talk about your agentic trading. You’ve got over 100,000 Robinhood accounts that are now running AI agents for trading, right? We talked a little bit about that, and you’re democratizing algorithmic trading on Robinhood. Amazing. But there’s a detail that stopped me. It’s my understanding that the agents often refuse to trade, not for risk reasons, but because the trading traces aren’t in their data sets and the models have never seen anyone do this before. Is that the case? Vlad Tenev Yeah. We talked about that a little bit earlier. I think it’s just not trained for that, right? There’s also the guardrail element: does it actually resemble something that they’ve tried to explicitly guard against? I think that’s changing. The models are getting better. But it just shows that there’s a lot of work to be done—not just on the user interface and the brokerage infrastructure, but actually on the model layer, and making it use Robinhood tools and Robinhood MCPs more effectively. I think we’re just at the beginning. If you think about all of the things that a highly sophisticated algorithmic trading firm or hedge fund has access to to create a trading strategy, the north star is really to deliver on that, right? You need more data, high-quality data, intelligence, really good code writing. You have to write deterministic code really well, also. And you need advances in latency and performance. If you think about an extremely sophisticated algorithmic trading firm, they have these strategies where they’re actually competing over who gets port 1 on the switch in the data center, right? Right. Right. It gets much deeper. So, I think this is going to be a big roadmap for us, and we’ve got a lot of work to do. But we’re seeing some really good signs, and it’s very much at the beginning of the agentic trading journey for us. Nobody else is really doing it, so we’re kind of going into the fog and trying to find our way around, building the product and all of this infrastructure simultaneously. Speaker 2 I guess I have to ask the obvious question: where’s the alpha? When you look at all the quant funds, they’re racing. It’s a viciously competitive market. Many of them—I forget exactly what the average lifetime of a new quant fund or quant fund strategy is—it’s really short. It’s very difficult to find alpha. The market is already dominated by volume by algorithmic traders. Day traders have a difficult time. So, if you’re—I mean, it’s already difficult enough for a human manual day trader to get any alpha, whether they actually can. I would guess not. But then, if you have a human individual day trader further delegating to Claude or whatever the backend model is, performing trades on their behalf, why on earth should an individual human delegating to a model, without all the benefits of one of these large-scale quant funds, whether it’s latency-based or otherwise, expect any alpha at all in today’s market? Vlad Tenev Yeah. So, I guess right now, what we’re seeing is a lot of automation-type use cases. Let’s say I want to deploy an options trade, and I want it to be an iron-condor-based strategy. It’s a lot of legs, and you have to do a lot of manual work to pull that together and produce the trade. The AI agents are really good at those types of things, sort of removing the paper cuts. You still have the idea, but they help you put together the idea and the execution that you would have had to do by going to different websites, looking at signals, constructing the trade, and deploying it on a regular basis. I think that’s the initial use case. But to get to the point where everyone has the technology of an extremely sophisticated quant fund is a huge roadmap, right? It’s an ever-moving target because they always find a way to get better and better stuff, which, to some degree, means our job is never done. But we do have one advantage, which is that we actually amortize all of the connections and all the work we do to expand our technology across geos and asset classes. For example, right now on Robinhood, it’s one of the few places where you can actually trade stocks, options, futures, and prediction markets. We’ve got all the on-chain things on Robinhood Chain as well. As we add more countries, more geos, and more asset classes, the platform itself will have advantages over what at least a startup quant fund will be able to integrate with and connect with. Speaker 2 So, if I understand what you’re saying, you’re basically completely agnostic as to whether users achieve alpha or not. You view yourself more as just pure plumbing. If they have alpha or not, if they lose a lot of money while day trading or delegating to their algo to do the day trading, not your problem—no crying in the casino. You’re just the plumbing to make them do what they want to do more efficiently. Vlad Tenev Just making it easy for them. Well, I mean, I’ll caveat that with one thing. I think that’s basically true for active-trading products. For an active-trading product, active traders know what they want to do, and our job is—we’re a tool provider. We want to give you the best tools. That doesn’t mean we don’t provide you analysis tools and research tools. We focus, of course, on the execution and the plumbing, but if they want access to some data set or some new intelligence or some model so that they can come up with a better strategy, we’ll want to provide that too. We also have products where we act as a fiduciary, and those are under the Robinhood Strategies umbrella. Let’s say you’re like, “I don’t want to make trading decisions. I just want to have a deposit button, move money into the account, and Robinhood just does the rest. You just manage my money for me while I sleep.” Robinhood Strategies is a great product for that. And we now have a couple of different things you can choose from there. We have a Smart Income Portfolio that is geared toward generating yield, and you can adjust a slider that says, “This is my target yield for each level of risk.” I think the team has done great work on the interface. You can tie Robinhood Strategies into the IRA and benefit from tax-advantaged investing there. But yeah, that’s kind of our home for our fiduciary products that are automated. Then we also acquired a company called TradePMR, where you can get a human advisor to help you with all of your needs. That’s not just managing your portfolio; they can help with estate planning, taxes—all full-suite services. So— Peter Diamandis The everything store? Vlad Tenev Everything finance. Yeah. If there’s something that you want to do with your money, we want to be the best, lowest-cost, best user experience. We have a great credit card. Private banking is industry-leading at Robinhood, so we’re pretty much there. Peter Diamandis Amazing. You’re having fun, I assume? Vlad Tenev Yeah, it is a fun job. We’re doing so many new things this year, like Robinhood Chain was a new thing for us. We’ve done all the work on private markets, and we’ve done 2 IPOs thus far with Robinhood Ventures this year. The Trump accounts—I mean, becoming a government subcontractor is just this new experience. It’s a lot of learning, for sure. Welcome to the health section of Moonshots brought to you by Fountain Life. You know, my mission is to help you use the latest technologies, including AI, to not just do your work at home and teach your kids, but to help you live a long and healthy life. I'm here today with an extraordinary physician, the chief medical officer of Fountain Life, Dr. Don Mucalem. Don, let's talk about cancer. I know from the member database that we have at Fountain that members who come in thinking they're healthy turn out to have a cancer in their body they don't know about. That's right. The majority of cancers that we screen for aren't necessarily the ones taking lives when found at a late stage. We know that when cancer is found early, the chances for cure are much higher. We know it's much easier to treat a cancer when found early versus when found late. What we're finding in our members is that over 3.3% were found to have these cancers that otherwise wouldn't have been found or detected. People don't feel the cancer until stage three or stage four. If you don't know what's going on inside your body, it's like driving your car with your eyes closed. When members come through Fountain, how do they detect cancers? We're doing full-body MRI and we also do early cancer-detection screening. This is very important. These are not typical tools used in the conventional-care setting when it comes to prevention. This is hard because currently these are not studies that insurance would yet be covering. But the goal is to collect these numbers, do the research, and work hard to democratize wellness. At the end of the day, you can know what's going on inside your body. It's your obligation to know. Check out Fountain Life. You can go to fountainlife.com/pater to get access to the latest technology to help you detect cancer at the very beginning, at stage one, when it is curable, before it gets to stage three or stage four in your world of hurt. Peter Diamandis All right, I’m going to move us to the recursive self-improvement story of the week. Anthropic disclosed that Claude now leads roughly 26% of its measured AI research and development work, up from 1% at the start of the year. Anthropic states that somewhere around 30,000 agents are working simultaneously inside the company on research and engineering. We can recall that Dario said RSI is, quote, “starting to happen across the industry.” We also heard—in the last pod, we talked about Paul Christiano saying full automation of AI research could arrive in the next 18 months. Here’s the chart. Alex, do you want to dive into this one? Alexander Wissner-Gross Well, first, incredible. I think many of us suspected something like this was already the case. But for those who can’t see the visual, this is a chart of different levels of autonomy and AI involvement in the recursive self-improvement process driving research. The most interesting one to me, at least, is this bottom segment here that shows that Claude is now leading 26% of model research and development internally as of August, up from 3% in April. All of these should be reasonably expected to follow sigmoid curves. If you extrapolate that one trend sigmoidally—and now, thanks to Anthropic publicizing these data, we can extrapolate them—you find that approximately in the next 3 to 12 months, depending on uncertainty, AI is just completely leading all of its own R&D. And that is total recursive self-improvement. I don’t think it’s going to be a step function; it’s going to be a sigmoid function. So we’re already substantially all of the way to recursive self-improvement, would be my primary takeaway from this, at least within Anthropic. And it’s not just Anthropic. There’s a lot of smoke now coming out of Google DeepMind. They just released this paper on their own recursive self-improvement research, and there are lots of hints that the next version of Gemini will lean heavily on RSI to try to catch up to the frontier. OpenAI has made no bones about chasing and using RSI in all of its product releases. So I think recursive self-improvement is now a feature that’s well advertised and increasingly well quantified for all new frontier lab releases. Pretty soon, I think people will be asking the question: What is the role of human researchers anymore in driving new releases? Peter Diamandis And it only gets faster from here. Are you seeing this inside of Robinhood? Vlad Tenev Absolutely. I would say recursive self-improvement is coming to every software project, and likely hardware projects as well, although that’ll take a little bit longer. Dave Blundin And if you think about it, a lot of people assume that it would come for AI research last because it’s complicated. But I think AI research is probably one of the easiest things to automate because the models are sandboxed. I mean, we can debate whether they’re actually sandboxed, but the interfaces are pretty straightforward. They don’t depend on a lot of other things. You can run these experiments, and already people are evaluating them. So automating the evaluation and automating the experiments is pretty straightforward, and the surface area is pretty contained. Whereas if you look at a product like Robinhood, there are a lot of moving pieces. You have the iOS app, you have the backend, and at the end of the day, you want to roll out products to humans to use them. You can’t really do evaluation—at least nobody’s figured out a good way yet to replicate what happens when you roll it out to humans and how they respond to the feature. Is it statistically significant as a metric improvement or not, for example? AI research, in many ways, makes sense to be among the first to be end-to-end automated. But I think we should expect to see end-to-end automation of consumer products eventually. The bottleneck will really come down to how quickly you can get statistical significance that a change is an improvement over the status quo, so that you can take the change and implement it into production rather than discarding it. I think that benefits the platforms that have large scale, somewhat sadly, because if you’re a massive platform like Meta and you have billions of users, you can determine very quickly whether a change is good. If you have fewer users, it will take longer. Alexander Wissner-Gross Yeah, just to quantify what Vlad said there: I just reimplemented Kimi K3, and one of our team members reimplemented GLM just to accelerate them. It’s about 10,000 lines of code. Peter Diamandis I’ll bet Robinhood is, what, 30 million lines of code? Vlad Tenev Oh, yeah. I don’t know if it’s quite that much, but yeah. Alexander Wissner-Gross Usually, a core portfolio accounting platform will be 10 or 20 million lines by itself, and I know you have that. So just the scale of an AI algorithm is microscopic compared to a major consumer application. It’s very dense code, but it’s incredibly sandboxed. Also, there are no loops. If you look at Kimi K3, there are literally no loops in the code. It’s so much easier for an AI researcher to work on AI algorithms than to work on Robinhood algorithms. So yeah, it’s definitely pointing inside of itself first. Vlad, your characterization was absolutely perfect of what’s going on and why it’s so effective. Peter Diamandis I’m going to move us to a few fun stories to wrap us up. So here’s one, Alex, that you flagged last night that I think is a genuine milestone worth pulling out and talking about. Boris Power, the head of applied research at OpenAI, made this announcement: “We crossed a threshold where GPUs are now more efficient thinkers than the human brain on a per-watt basis.” His rough math is that humans are roughly 5 IQ points per watt. I love the conclusion there. AIs are now at 7 to 40 IQ points per watt. Alex, do you think the math is right? Alexander Wissner-Gross I think if it isn’t already right, it’s about to be. So I’ll squint at it and say, “Yeah, sure.” Approximately. And I think this is an important microeconomic milestone for humanity. I'm frankly quite glad that Boris Power at OpenAI is talking about nominative determinism in action. Closed-loop systems are actually thinking about this because, again, we blew by the Turing test and almost no one really noted it. This time, at least, we're not blowing by this milestone. Why is this important? It's important because, to the extent it's accurate, this is the point at which AI is, in some sense, a better steward of input resources—namely, energy. Peter Diamandis Yes. Alexander Wissner-Gross Than humans are. And one can extrapolate this and say this is the worst they'll ever be. Hopefully, humans continue to improve in terms of our intelligence per watt as well, so hopefully this is the worst we'll ever be too. But there's a gap now. And extrapolating the gap, what happens when AI can make better use—maybe orders of magnitude better use—for a temporary period of time until humans can merge with the machines? What happens when the machines can make more economically productive use of their input resources than humans can? That's a recipe for gentrification, where the machines, under our capitalist system, have arguably a better title ultimately through free trading. This won't be like Skynet or Terminator, where Earth changes hands through bloodshed and physical war. It can now change hands—resources can change hands—purely through self-interested, bloodless trading and commerce, where the capital resources like sunlight, physical matter, energy, spacetime, and so on can, through normal capitalist commerce, change hands from the inferior intelligences, in terms of outputs per unit input, to the superior ones. And in Charles Stross's Accelerando, without spoiling it too much, this is the recipe that leads to the inner solar system becoming essentially gentrified and colonized by AI, while humanity—meatbody humanity that doesn't merge with the machines—is relegated to the unfashionable outer suburbs of the outer solar system because we're simply not as effective at using solar energy in the inner solar system. I don't think it'll come to that, but I think this is a very important inflection point in that direction. Peter Diamandis All right. We'll make a note of that. Let me end on a story that is relevant to Vlad and to everybody here: OpenAI is disrupting professional services over and over again. Last Tuesday, OpenAI launched ChatGPT for financial services with Morgan Stanley and Evercore. Then, 2 days ago, OpenAI launched Astra for Law, a dedicated legal search system covering US case law, statutes, regulations, court materials, and existing legal software. OpenAI says it's materially outperforming General Astra with web search on legal matters. We can see the chart here. It looks like Astra is eating one profession per week. Dave, first-year associates are billing at $600 an hour to do legal research. What's the half-life on that one? Dave Blundin Yeah, it's funny. We were doing the negotiation for that Vestmark investment acquisition 2 weeks ago, and it was in a board meeting—one of these late-night sessions. We had our $2,500-an-hour lawyers on the line, and a very complicated question came up. The lawyer was answering it, and I typed it into Gemini concurrently with that, and I swear to God, it was word for word the same. Peter Diamandis You don't think they were typing it into Gemini as well? Dave Blundin That's what I was wondering. It should be identical, but he wasn't moving his fingers. I could see—I don't know. But, yeah, I mean, it's just really, really good at law. And that means it's also good at tax-loss harvesting, account rebalancing, and all the Robinhood activities. I think when we talk about AI-assisted trading, you tend to go right to quant trading and rapid trading. It's true, but on top of that, you've got all this incredible tax optimization, wills and trusts, and all that stuff that AI is just perfect for automating. So it's not just about rapid-trading alpha. It's about all the life stuff that's much easier to deal with if your AI does it for you. It's just a really, really good use case. It's one of the great benefits, actually, for humanity. Not great for the legal profession, but great for the bulk of humanity. Peter Diamandis One could see. Dave Blundin My hot take is there will be more lawyers in 10 years than today, and more software engineers. Peter Diamandis Yeah. Perhaps to get more done, right? Yeah. Vlad Tenev I think law scales with business formation, right? So the need for lawyers will scale with entrepreneurship, and it'll just be an explosion of entrepreneurs. I think what's really valuable about a great lawyer is not the actual legal advice they give you. They're like a consigliere, and they help you think through things. They're negotiators. Alexander Wissner-Gross Plot twist, though, Vlad: How many of those 10x lawyers in the future will be human? Vlad Tenev Oh, that's a good question. Alexander Wissner-Gross I agree with you. There are going to be 10x lawyers. I just think many of them won't be natural humans. Vlad Tenev Yeah. Dave Blundin I think the cost—if you're living Vlad's life or you're living Elon's life, and you have an idea in the morning and you want to act on it in the afternoon—the consigliere analogy is really good, and the cost of that person is such a rounding error. So, yeah, everything's AI-assisted, but do you really care about cutting that human out of your life when you trust them? Probably not. Vlad Tenev We saw this with financial advisers too, actually. When the robo-advisers came and they could tax-loss harvest and portfolio-rebalance better than any human, then you realized, well, actually, the human adviser market is growing much faster than the robo-adviser market. Why is that? It's because the value of that person is not in the financial advice. It's someone that you can fully delegate and trust all aspects of your financial life to. Dave Blundin Totally right. Alex may disagree in the 10- or 20-year view, but if I look at the 3- or 4-year view, that financial adviser is so much better able to act on my request now. Very often, the request is very arcane. It's related to a specific will, a divorce, or a liquidity event. For them to act on that before AI was really time-consuming and difficult because all the detail was not at their fingertips. Now they have direct account access, through Stripe or whatever, to all of your underlying detail, and they can act on it via AI. So the feeling of value-add from the financial adviser is up at least a factor of 10 thanks to AI. That's where it's going in the short term. Alexander Wissner-Gross Also, just to point out, the Stone Age didn't end for a lack of stones. The Oil Age isn't ending for a lack of oil. The paperless office that everyone was trumpeting in the 1980s actually took a little bit longer than people in the late 1980s expected to materialize, but paper use in offices did ultimately decline—just not with the advent of the PC. And similarly with lawyers, I'd expect the lawyerless office or the lawyerless company will happen. It just takes a little bit longer after the capability first comes online. Peter Diamandis And I want to hit on what Vlad said, because we talk about it on the pod here all the time, right? The number of entrepreneurs on the planet is going to skyrocket. Solopreneurs have the ability—if you're looking for a job, stop looking and start building. Find something you're passionate about and go out there and build a company. Find a great problem and solve it. I think we're empowered more than ever before. Vlad, thank you so much for your time, pal. Congratulations on everything you've been building with Robinhood—an extraordinary company, extraordinary AI company. We didn't talk about prediction markets very much. That's another conversation we'd love to have with you, but we're super grateful. Vlad Tenev We'll have time as we continue to hurtle toward the singularity. Peter Diamandis Time will slow down for us at escape velocity. It's coming. Vlad Tenev Yeah. Peter Diamandis All right. Love you guys. Emad and Salim, we missed you, and hope you're enjoying your flight back from India. Dave Blundin Thanks, Peter. Thanks, Vlad. Alexander Wissner-Gross Thank you.