Testing one, two, three. Testing one, two, three. Can you hear me knocking? It's very, very dark here.
Oh, so if I stay here, I cannot... Yeah. I can change, I can change the camera. Let me take a...
Oh, direct from the market. Okay. So...
Stand there. No, can see even if you don't stand here, also can see you. Yeah, also can see... I think people online also know as important the guy... not nice. Okay. They don't, too complicated for them.
Okay. Testing. Testing. Can you hear me? Testing. You stand against the light.
Testing because he's on the YouTube. So he can be, he was outside. Everybody to...
is extremely excited about what's going to happen after this. So, I'm going to keep my opening short and sweet. Um, I see a lot of new faces. Uh, so for those of you who are not familiar with the Singapore Fintech Association, allow me 10 seconds. Um, so we are the national industry body representing the Singapore fintech ecosystem. We bring together members — actually, Pix is right here — April, uh, fintechs, financial institutions, our regulator, the Monetary Authority of Singapore, um, as well as other ecosystem players to jointly discuss and surface issues that matter to our community. Um, together we help to build and develop Singapore's fintech sector and help to elevate Singapore's global competitiveness. So for anyone who is not yet a member and you're interested to learn more about SFA and the good work that we do, please reach out to any one of us. Um, Chikit is our treasurer sitting on the executive committee team. We have Tiff, Alex — I saw Alex somewhere. Oh, Alex Tway — um, as well as our secretariat team around the room. So who are all ready after the session if you're interested to learn more about us. So today, um, I'm really, really grateful and very honored that in partnership with the Ethereum Ecosystem Fund and sponsored by SFA member with an AI-native tech company building out here in Singapore, we are bringing to you a very special session with Vitalik Buterin. I think he needs no introduction. Everybody's extremely excited for him to come up. Um, but I just want to say, you know, we're all very honored to have him here with us. He is not just the inventor of Ethereum, but he has greatly influenced how most of us think about programmable money, um, decentralized applications, the future of the internet, and the future of the programmable economy in the age of artificial intelligence, or now known as super intelligence. Um, so we will, I will hand over time to Vitalik to share more after this. It will be followed by a panel discussion moderated by Chikit. As I mentioned, Chikit is a very valuable member of our executive committee, the co-founder and CEO of WHIP. If you're interested to learn more about WHIP, please also reach out to Chikit and his team after this. And then we will leave enough time for everyone to ask questions. I know everybody's dying — before Vitalik even came up to this room, I think there was a lot of questions already buzzing. So, we will have enough time today for you to get all your answers, all your questions answered. So now without further ado, please allow me to introduce Vitalik.
Okay. Um, do I have the... Cool. Thank you. Great. Uh, so today I wanted to, uh, talk about making finance actually secure and both the role and the challenges that both blockchains and, uh, other forms of cryptography are going to have in solving this, solving this very important problem going forward, right? So this is a big topic, right? And I think it's a big topic for a few reasons, right? So one of them is of course that finance is not just blockchains and, you know, there's traditional finance, um, and the kind of linking between traditional finance and blockchains, there is, uh, there are whole entire other layers of the stack that, uh, you have to think about. And then also blockchains are far from the only technology that matters, and that's also going to be one of my themes for today, right? I think, uh, you know, like often there is this pattern of saying like, oh, like if you want, like if you believe in blockchains, then you make things secure by putting them onto a blockchain. This does not always work. There's even times when putting things onto a blockchain makes them less secure. Um, but and then there are times when the best solution either does not involve blockchains at all or involves them only minimally, right? And so knowing the right way to combine different tools, including tools that have only existed for the last five years at least in a mature form, is going to be very important. So several hard problems stand in the way of, uh, finance being secure, right? So these are kind of easier problems and, uh, they're also actually problems that, uh, I mean, especially the first one, it affects blockchains and traditional finance equally, right? So quantum resistance is on everyone's mind. And another big one, I mean, especially on-chain, is of course scaling cost, right? And so if, uh, you've talked, like if you talk to someone who has gone in the weeds of basically saying like, hey, what if we try to like put this bond like on-chain or what if we put this, uh, like supply chain tracking contraption on-chain? Often the thing they worry about is scale and cost. Or if instead of putting it, just dumping it onto Ethereum directly, they come up with some kind of other and more complicated contraption which if you're not careful is often much worse. Often scale and cost are the reason why, right? So both of these, uh, are I think important and the reason why they're easier is that they are technical problems and they have technical solutions of a type where even if you don't necessarily understand every detail, like once you understand that it exists, you'll be able to kind of understand that it exists and move on, right? Other types of problems are more difficult. So first of all, I mean, in your sort of role — oops, this is just understanding the audience. Okay. So the room has clearly sorted itself very effectively from like less rolliness to more rolliness, which is, uh, good. But so rollups were like the kind of, the early version of rollups was the dominant scaling paradigm for Ethereum in the early 2020s. And there, you know, had optimistic rollups. There was zk rollups. There are like basically these contraptions that keep data off-chain but then at kind of publish or at least a limited amount of data and sort of proofs on-chain and effectively benefit from the blockchain security without incurring all of its inefficiency, right? And, uh, you know, these things had benefits. They enabled very cheap transactions but limited interoperability. So practically speaking, if you wanted to move, you know, a coin from Optimism to Arbitrum, you had to either wait a week or go through a centralized intermediary, which is not great. Um, limited ability to trust group systems and security councils, right? So a lot of these systems were secured by cryptography in theory, but they were secured by, uh, like basically multisig. So like, you know, like nine guys and like five or six of their keys could, uh, switch over the algorithm to something else in practice. And, uh, this is kind of justified as a way to make sure that if the code has a bug that they can still fix it. But the problem is of course that the security council itself, uh, becomes middle of a debug. Um, so they have these challenges, right? And, uh, in lean Ethereum, I think, uh, two solutions, uh, that we want to focus on here. One is there really is a spectrum here, right? So native rollups and EIP-8288. And I think the way that I would explain both of these is that so if you compare native rollups to pre-existing rollups that you might have heard of, native rollups basically instead of you having to write most of, like all of the code as a smart contract, like basically Ethereum itself exposes the functionality of verifying a proof of Ethereum execution. And so Ethereum, the Ethereum protocol would kind of pull that functionality in natively and so there would be just like much fewer lines of code to care about, right? So that's one. And then EIP-8288 is this, uh, proposal basically lets you have kind of aggregate signatures and proofs sort inside of this, you know, cloud in the network pool before they even get included into a block. And so think of it as basically an L2 that does not even need to have like block proposers in some sense, right? And so the way that I think about these newer technologies is basically they are sort of an evolution and a kind of a way of like really being much more clear and much more concrete about when and why we need to pay the costs of a blockchain, right? So blockchains, they're hard because they depend on the, or the thing that they do is ordering, right? Blockchains basically let you get consensus on which, which transaction came first. If you don't care about what happened first, you don't need a blockchain, right? Like there's other types of decentralized networks. Who here has used BitTorrent? Probably have. Yeah. So you have a decentralized network that's not a blockchain. They exist. Um, and if you care about ordering, right, if you care about, you know, the double spending problem, if you spend the same coin twice, you have to figure out like which one of those is first and valid, which one is second and invalid, then no, the blockchain can do that, right? But ordering is still a very tiny portion of the information and even the computation that goes into a block, right? And so the sort of the paradigm of the future, the paradigm that Ethereum is moving toward is basically if something requires like actually global ordering, like it requires order where it's order that is theoretically between kind of everyone's transactions, then you do it on a line. If you only need ordering within a kind of smaller subgroup, then that's what native rollups are for. And then if you don't need order, if you just have computation and you need to do that computation, you need that computation to be verified, then you use like basically the mele and, uh, like you have, uh, like these kind of cryptographic proofs that computation happened and you generate a proof of the proofs of the proofs. You generate this kind of, the network sort of collaboratively generates this one big proof and then you sort of forget about all of the, uh, original computation. Nobody needs to repeat it, right? And so essentially you like, you use different types of technology to do different jobs, right? And you use a blockchain for the specific thing that the blockchain fully requires and then it becomes more efficient if your requirements are slower. And then if your requirements become even lower, like, you know, basically more and more stuff actually happens off-chain, right? And so what's happening here is essentially we're moving from just talking about blockchains as envisioned to sat by Satoshi to combining blockchains with other forms of computation and then using cryptography, specifically using zero-knowledge proofs, to kind of verify things without everyone in the world having to download and having to re-execute them, right? So use the blockchain for what the blockchain is good at. Use these newer forms of cryptography for everything else and it becomes much more efficient, right? So that is like a bit of a sneak peek into some of the side of how we handle, how we're planning to handle scaling. Now I mentioned quantum resistance, um, what, and the way we're handling quantum resistance is basically like I forgot to tell you, but by the way, all this stuff is quantum safe. Um, so that's, uh, like the, if, if there's one thing that you need to know about quantum safety that I think is sort of not percolated enough through like a lot of kind of public culture, it's that being quantum safe is not that hard. There are, for every single thing for which we have quantum unsafe algorithms, we have quantum safe algorithms. You have to switch to them. You have to eat the cost of sometimes being five or ten times less efficient and then sometimes, you know, you, you have to like do another thing to get your efficiency back, right? And, uh, there basically it's hard work that engineers have to do. There is nothing fundamentally impossible or apocalyptic about quantum. AI is, uh, similar and different, right? So I'm sure who's here has heard stories about AI hacking into everything. Yeah. So, you know, like we are now in the era of AI where instead of worrying about abstract crazy sci-fi risks of doom, you're supposed to worry about realistic risks that affect regular human beings in the real world, such as autonomous agent swarms randomly deciding to hack into things. This is AI in 2026. Um, now AI is going to hack into everything that can be hacked into, right? So a lot of people have this paradigm where they assume like, oh, everything will have vulnerabilities. You cannot avoid everything having vulnerabilities. Like you just have to keep things hidden and, uh, basically kind of patch things up reactively, right? And like this is not going to work, right? AI is able to find exploits in hours that used to take months. Um, AI is able to almost instantaneously cook up complicated chains of actions that take advantage of multiple exploits at the same time. Um, AI is able to find exploits in closed-source software which is just a binary. AI is, AI is able to do crazy stuff. Attackers have AI. Um, and, uh, it's very sophisticated, right? And so the question is, what do we do? So I am actually, uh, very much an optimist about the long-term future of cybersecurity. I basically think that at some point this concept that, oh, you know, like everything is bugged and like actually everything is insecure will just stop being true and, uh, we actually will kind of get to a much better equilibrium. But the transition to get there is challenging. The transition to get there involves doing kind of new and different kind of things, right? So traditional model of security is kind of what we have on, uh, at the top here, right? Basically, you know, imagine you have these kind of different layers and then each of those different layers have their own bugs and you're sort of hoping that nobody manages to find like all of the bugs to like actually break into the system, right? It's like, okay, you know, you have your thing and then your thing is inside of a sandbox and then like maybe they can break into the sandbox, but then to figure out how to, you have to have the source code and maybe the source code is secret and then maybe the IP addresses are secret and, uh, this does not work anymore, right? Because AI is just powerful enough to just zoom through and find an attack and everything. So we will need to make software that is actually secure and software that is actually secure is, uh, something that actually has been already more and more of a trend. Uh, so there's a graph I have, uh, that basically it asks, uh, like if you use top-tier security practices to write code, how many bugs can on average do you see per, per thousand lines of code? And that number over the last 35 years or so has basically dropped from like one to 0.1 to 0, or from 10 to one to 0.1 and like to 0.01, right? It's actually been happening even before AI, right? Even before AI, you had things like strongly typed languages, right? You know, you basically switched from things like, uh, C++ to things like Rust. You had better memory management. You had like better practices. People have a better understanding of how to write code, much better testing practices. And AI and actually formal verification is, formal verification enabled by AI is the next step in this, right? Basically what this is is you write down a list of properties that kind of make up like what you think it means for your program to be secure, like, and you try to prove them as mathematical theorems, right? And the idea is, well, if you can write down a list of mathematical statements about a program that can define properties that the program, that the program should have and you can like try to define as many of them as possible, right, and keep thinking about them and, uh, you then just like send AI and prove all of them, right? And if AI can prove for theorem and stokes, it can prove a safety property over a program with 100,000 lines of code. Like this is basically the argument is right now the biggest challenge is basically, well, how do you like, are you really sure you've like gotten all of the different safety properties you care about, right? And like sometimes you don't, right? And the principle here is basically that you're expressing your intent in as many different ways as you can at the same time, right? Because any one way of formalizing your intent will always like be miss, be missing something really crucial, right? It's like, you know, if you tell like a magic superb like, "Hey, I don't want anyone to die of cancer anymore." And then the magic superb's like, "Oh, well, I'm just going to make a super virus and kill everyone from that first," right? So, you know, the sort of like standard, uh, like argument for why like being safe with super intelligent AI is really important, right? And you know, security is the same thing, right? And so what you want to do, like even strongly typed programming languages, they do this, right? Like you basically, you know, you have like a the int c equals a plus b and then at the same time you're expressing that c is a plus b and then you're also expressing that c is an integer, right? Like you actually have the execution logic, then you have the types, and then you also have the test cases, right? You have like three ways of expressing your intention and then you know, you run the test suite and you run the type checker and like you're basically mathematically checking all three against each other. With formal verification, you can basically like express intent in an almost unlimited number of ways, statements, and then compiling your program is proving the theorems. So this, uh, kind of approach is, uh, basically where things need to go. Um, and this sort of thing is, uh, actually like something that we, you know, like in Ethereum we've been thinking about for years. Like we actually started the AI formal verification effort back in 2024 and it's just only been accelerating since then. And, uh, the, um, and I think, uh, one of the challenges with Ethereum right is that sort of actually being secure is the only way for Ethereum to be secure, right? Like something like Ethereum, it cannot, like it does not even have the option of being sort of quote secure by hiding its source code because people need to see the source code in order to trust it. Um, something like Ethereum does not even have the option of KYCing like literally every single user so that if you know, like one of them breaks into it, like you can, uh, you know, go and send the police to arrest it, right? Like Ethereum does not have that option. It's an open protocol and so being actually secure meaning like actually proactive, you know, like you're like you actually have the wall of security is the only option that Ethereum has available to it, right? And, uh, so look, if we already have formal verification projects ongoing for the consensus algorithm, for EVM implementations, for many parts of the protocol, for lots of applications that are sitting on top of it. You know, Gro had even actually another thing you just need for verification for is to implement some of this advanced cryptography, right? Because some of these proof systems, they're logic against thousands of lines of code. You have to prove that it's, you have to prove that there aren't any bugs. Like if you don't, then there are bugs, right? And, uh, there's an effort called BRL that basically actually does that. Like cryptography is actually ideal for this because often the protocol is like a thousand lines of code and then the statement that is the thing that it's proving is often like 10 lines, right? And so this is something where Ethereum has been doing a lot, right? But security of the chain is only one aspect of security. There's also security of the user layer and then there's also other things that might be insufficient. So data, data privacy is probably is one of those, uh, things that I think the blockchain space did not properly think about for like over 10 years, but it's, uh, I think increasingly becoming understood as part of just being, being the core definition of, uh, security. So security of the chain historically and still for, and still now and still forever, it comes from decentralization. There's many thousands of nodes and, uh, the network can keep running even if almost all of them go offline and the network can keep running even if a large majority of them become malicious. Um, also in Ethereum, we've historically had multiple implementations of every single, like major piece of software. And so with multiple implement, basically even if one client has a bug, then, uh, like the network actually still keeps on going because the majority of the network is running other clients. This actually saved us in, uh, 2016 because, uh, in 2016 we had a situation for about one month where we had one like very smart and very determined attacker that kept on finding security flaws in, in the Ethereum clients and they never actually managed to find the flaw in both at the same time. And so like actually a lot of people just flipped from running one clients to running the other clients and the network kept on going. So that's a where it's, uh, security has come from. Our strategy is shifting more and more from being about redundancy, uh, to being about formal verification, right? So redundancy is like you write software doing the same thing five times and then as long as three of those five are correct, you're fine, right? Formal verification basically says you write one, you, you make one single artifact to make one single piece of code that you trust. And then if you want, you can have five different groups write their own implementations or write like lists of security properties that it has to satisfy. And then you have like these mathematical theorem provers that basically verify that the things that everyone wrote are implementing exactly the same function. And if you discover they're not, then like that's where you discover, you know, you need to like everyone needs to pay, you know, put their attention toward and actually see like, well actually, you know, like which side actually better reflects people's underlying intent. Um, so instead of five different software components that are covering for each other, you have one software architect or, or artifact that has five different types of groups that are covering for each other. So effectively the decentralization is still there, but it's kind of pushed to a level where you're paying the costs of it only during the software development process and only in an automated way, right? So in real time during execution the thing is hyper efficient. Security of the user layer. Um, so probably the most important shift that we've seen in the security of the user layer of questions over the last few years is that, uh, smart contracts are no longer the biggest problem. Actually the security of smart contracts has improved massively over the last decade. Like basically ever since the DAO got hacked, we've seen better standards, better audits, better tests, better formal verification, things have been better and better. If you do not do crazy things, if you act, if you behave conservatively on Ethereum, you will be fine. And if I'm wrong about this, then like, like 90% of my money disappears too because, because it's been a smart contract and, uh, it's been fine, right? So when you see numbers that have like, like this, when you see numbers that like reflect blockchains continuing to be unsafe, one thing that you have to remember is that the blockchain world is a crazy world which, uh, you can very easily tell by going to conferences such as Token 249 next week, which I'm not going to. And you know, a couple of years ago they had like 10-meter tall pictures of Justin Sun's face, right? Like, you know, you go there like basically there's like people who are just like voluntarily being completely crazy and then there's people who are just like doing reasonable normal things and following recommendations, right? If you are the second type, you're basically fine. If you're the first type, then like, yeah, like they don't do any of what I just said, right? I'm like, yeah, like they're going to like lose freaking $2 billion, right? And like they'll keep losing $2 billion because anytime there's an advance in the frontier, they don't spend that advance on safety, they're going to spend that advance on like being even crazier, right? So I think like that's like, like that's something that's really important to keep in mind, right? That sort of, you know, the crazy frontier and the level of safety you actually get if you're non-crazy are different. Um, so of course I recommend being non-crazy, but if you're non-crazy, there has been this shift in what is the main thing you have to worry about, right? So over here, so Safe is a non-crazy application, right? I mean, Safe has like, I mean it's actually a fail. All my money is in, right? And it's like a wallet. And, uh, Biden lost a huge amount of money. And the reason why they lost this money is that actually the smart contracts were fine. They lost money because the interface that they used to, to use the smart contracts got hacked. And this has been the sort of like dirty secret of, uh, a lot of like blockchain stuff is that sort of the core is more decentralized and safe, but then stuff that people build on top of it is like kind of more centralized and it's more centralized in ways that like actually really compromise all of its properties, right? So if you think about what it means to access a smart contract through a user interface that's hosted on a website, right? I mean this is even like true with banks, right? Like a bank is a website, right? There is a server. There is one single computer that is hosting the content. And if that, if that particular computer gets hacked by North Korea for the exact interval during which you are logging in, then North Korea controls what version of the application you receive and it's going to contain a button and that button is going to say, um, you know, like send and, uh, you have like text boxes and the text boxes, you know, you're going to put in, you know, your, your best friend's address, you're going to put in 100 right USDC and then actually the send button sends your entire balance to North Koreans, right? And like all you need in order to get, get caught by this is for the server to get hacked and give you a bad UI and for you to be too lazy to like check through what you're signing again in your wallet. So it's like actually not hard, right? So this is what I mean by the user layer. Better protecting the user layer is the frontier of safety that actually matters in 2026. So wallets and front ends are a big issue. Um, then, um, we have hot wallets, cold, cold wallets, uh, like basically other kinds of infrastructure and then also a big one is trusted intermediaries, right? So basically, yeah, like Liquid got hacked. When Liquid got hacked, you know, if there was a multisig and effectively individual participants in that multisig all got hacked, right? Like this happens very often. So this is like entirely centralized components. This is, I mean, 90% of this is one hack and that one hack was a centralized component, right? Basically, yeah, places other than the smart contract are like where the vulnerabilities are. And so we, the, the next like important step is to harden, like is to harden this one. Um, another thing that's on the horizon is that I think kind of agent-based access to blockchains and to smart contracts is only going to increase. And if you're using an agent, then like you're not even using UIs at all, right? And like I think agents done well can even be, can be very security increasing because, uh, like effectively you're just completely skipping the UI, right? At least the battery did not fall out. Um, so the agents done well can like completely skip the UI and they can like actually completely cut out a huge number of sources of vulnerabilities. The problem of course is that while the agent could make a mistake or maybe the agent could get convinced by like some kind of adversarial attack to try to, to do something bad with your money, right? And so you want like agent and something else. You want like, you want multiple levels of redundancy, right? And so thinking through security of the user lawyer in the context of agents which brings both opportunities and risks is, uh, something that's also very valuable. So security is a whole stack problem, right? So you have the blockchain protocol, then you have clients which are implementations of the blockchain protocol, then you have wallets, you have to think about security of your wallet, then you have application front ends and then, you know, the user, right? Even that, you know, like individual people are targets of social engineering and you have to remember that and then also you have to kind of go back down, right? So there is the operating system. So on every device that you touch money on, every device that you touch kind of authorization and authentication, the operating system is something you have to think about and then finally also hardware, right? So, you know, I see the new, the new meme is apparently that instead of talking about AI, we're talking about SI and, uh, I actually believe in talking a lot in talking about SI, but for me SI is element, element 14 on the periodic table. It's silicon and, uh, you know, the silicon that your stuff is running on is like really is the thing that underpins all the security, right? And SI implemented improperly can also leak your keys, it can leak your information, it can leak all kinds of things, right? And, uh, and so you basically have to kind of think about each and every one of these layers and these are things that apply I think both in the context of crypto finance and they also apply almost just as much in the context of traditional finance, right? So one of the things that I think that like one of the big opportunities that also comes from like blockchain-based finance is smart contracts, right? Basically the idea that you can define the logic by which your program like author, what, how your program authorizes what it can do as a piece of computer code, right? Like you can make an account that says you need four of these people to authorize a transaction. Bank accounts do not let you do that, right? Like even if you're an institution, you know, there's like single signature and double signature and it's this like weird thing you have to spend a lot of time to figure out how to get. Um, but now on-chain, like any structure of authorization that you want to make, like you can just go do it. It's a piece of code, right? And, uh, I mean these days with an agent, you just deploy it in 10 seconds. Now, of course, if you're handling money, you do not want to spend 10 seconds on figuring out how to handle your money, right? And so, like, what you want to do is you're not like, you want to spend the extra kind of space for effort that you have on making sure that everything that you're doing is actually secure, right? So here we're talking about basically reducing the need to trust centralized counterparties and replacing it by the need to trust math. In order to trust math, you need to ver, basically verify the math for math. And, uh, this is something that you can do, right? But this is something that like we actually need to think carefully about. There's, there's good ways of doing it, there's bad ways of doing it. Um, and, uh, there's kind of an entire discipline to this and, but then also if you do not care about getting into that discipline, there's like a growing set of these kind of safe standard patterns that, uh, you can go and use. So if we, if people are interested in self-custody, who here holds, holds their own coins and like holds them in their own wallet, not on an exchange? Okay. So this is, uh, also very relevant I think for me. Yeah, like I, I really think self-custody needs to improve and, uh, actually, uh, like this is, uh, like this is one of the key benefits of, uh, like digital assets, right? Is that you can basically define like the actual underlying logic for like the mathematical function that controls them, right? And the challenge is that I think good self-custody depends, defends you against two risks at the same time. The first risk is other people being evil. The second risk is you being, you being stupid. So, um, basically, you know, like the reason to care about this is that the institution that, you know, you put your trust in to secure your funds, sometimes it is FTX. We hear sort of FTX. Okay. Yeah. That's good. More like a lot of hands. Um, okay. I think like this, I think like this room is actually arranged from like left to right by like, like willingness to raise hands. Um, okay. How do I test this theory? Um, okay, raise your hand if you're closer to the white wall than the black wall. Okay. Only a few. Okay. So that's, um, great. Okay. Who here has heard of Mount Gox? Who here has climbed Mount Cox? Actually, no. It's, it's not an actual mountain. It's Magic the Gathering online exchange. Like, people literally put $460 billion into Magic the Gathering online exchange. I mean, 10 years after that, people like spent $180 million of transaction fees on Ethereum buying digital monkeys. So like I get it, but you know, money and monkeys, you know, the only thing between them is a K, right? Um, yeah. Uh, so this is one thing that you have to worry about, right? Um, the thing you have to worry about on the, uh, is the other thing is that humans are not infallible, right? And I think people who care about like holding their funds themselves because we're, uh, not like, you know, you're worried about counterparties, like tend to be very sort of self-sovereignty-minded, very individualistic, very sort of like I want to be in full control myself and, uh, often this ends up like sacrificing redundancy that's like actually very critical for security, right? So humans are not infallible. Who here has forgotten a password? Okay. Who here has lost a piece of paper? Who here has like failed, like lost, um, like lo, like had their grades in school suffer because you lost a piece of paper and you didn't hand in an assignment on time? Okay. So if you like store your coins in a way that some people tell you to and then you have a risk of suffering like 100,000 times more than that. So humans are not infallible. Our phones are not infallible. Who here has lost a phone? Who here has had a phone stolen? Who here has had a phone suddenly stopped working and like got bricked and never been able to recover it? Um, same thing to a laptop. Any of the same for a laptop. Okay. Any of the same for like a hardware wallet, including non-blockchain hardware wallets like UB Keys. Okay. Um, who here is, uh, not, um, honest about their failures with objects? Okay. Um, so other risks, your house burns down, you know, or potentially you live in a place and, uh, your place gets invaded, you know, that also part of the 2020s. And suddenly at that time, you know, you have to leave on an hour's notice and you never come back. Also, you know, like lots of people literally are interested in cryptocurrency because they're trying to hedge against political risk. Well, that's a political risk. So that's, um, that's one $5 wrench attack. So someone can, uh, like basically put a gun to your head and, and, and order you to like hand over every password you have and every object you have. And sometimes you forget your password. My basic, uh, kind of like re, I mean, maybe this is like specific to me as a nomad, but my basic test is that if let's say like for a form of self-custod to be secure, it has to be able to survive me suddenly losing all of my possessions without exception, like basically everything on my body, including backpack, including clothing. Like imagine, you know how like in a video game, you know, like maybe like you die and then I don't think this is how they actually work. Like usually your inventory survives death, right? But like, you know, imagine you had a video game where you die and then of course you respawn, but you respawn naked, right? Like you need your, you need your coins to survive that. Most people don't really think this far. And so this is why self-custody ends up often being less secure than it could be. So my solution to this sort of thing has always been multisig, right? Basically, yeah, one of the big lessons that, uh, I think you can probably take away from all of this is that decentralization solves everything and like that's not quite true, but it's like close enough to be true that like you might as well believe it, right? You know, we, yeah, like blockchains having many nodes protects the blockchain from being taken down. Software packages having many different people figuring about different statements to prove about them best guarantees their security. And then from your wallet, like you want to have multiple keys and you want those multiple keys to be of different types. So ideally, I think my only, some of them should be things held by you. Some of them can be things held by friends and, uh, some of them can be things held by institutions. Also, one of the frontiers in doing this is, uh, more complicated policies. So for example, you can have one key that's able to spend limited amounts that has limited authority and you need multiple keys, like you might need like say four or six that are able to spend as much as you want or that are able to replace the, uh, the primary key if you lose your primary key. Sometimes like you can, you can go even more complicated. You can say you need four out of six to do anything immediately, but you only need two out of six to do something after a one-year delay, right? So a lot of people kind of, you know, like do, you know, like do end-of-life planning in this way, protecting against kind of very exceptional situations in this way, right? Like the benefit of smart contracts is that you can choose much more custom policies and, uh, like this is a very important security benefit for someone, right? So all of the like basically, you know, you need in, you need to combine different kind of keys that are held in various different ways so, so that none of them all share the same risk. And this is a, a really fundamental thing that I think almost no one does. And this is something that like we are actually trying really hard to improve the user experience of. So that's like some of the issues in security on the user side, right? You know, there's like protecting your money when it's outside of an application. Vendors sort of protecting your money when it's inside of an application. We haven't even talked about oracles yet and there's like a whole rabbit hole of making better oracles and making oracle-minimized DeFi, like all of these things matter, right? So problem for protecting your data, I think, uh, privacy is part of security. Um, who here wants like Twitter to find out about everything in your last 100 days of purchase history? Who here would be okay with that? Okay, yeah, brave. Um, who here is brave? Okay. So I personally prefer Firefox. Um, okay, so how do you kind of secure privacy of, uh, financial data, right? I think, uh, so if you want to talk about the simple things, so zero-knowledge proofs, right? Basically zero-knowledge proofs, you know, they're the technology behind Zcash, they're the technology behind, you know, like things like privacy pools and Railgun, they're the technology behind all of these systems. And the core of it is basically that when you make a transfer, you're instead of revealing to, to the chain which coins you're spending, you're revealing a proof of the fact that you have coins that you can spend that you have not yet spent. So you're proving a mathematical statement that I am allowed to spend this without publicly revealing why you're allowed to spend it. And, uh, this is kind of the big cryptographic magic that's only become possible over the...
Over the last 10 years, it has really become trivial to build around, to develop, and to make quantum-safe — really, just in the last two years. And then EIP-8288 will basically make that kind of scalable and quantum-safe at the same time, which is one of the reasons why it's really worth watching.
Another thing is that many forms of selective disclosure are possible. So there are many reasons why you might want to still disclose aspects of your spending history and receiving history to different people. Obviously there are regulatory reasons to do that. You want to be able to prove sources of funds as an organization. Many organizations need to be able to prove things about what kinds of sources they're getting money from. You also want it for auditing — you want to give people inside your organization who are auditing the ability to see transactions without the ability to spend. So many different forms of selective disclosure that disclose many different aspects of your spending history on a very fine-grained level are possible, and programmable zero-knowledge groups are a really big part of this.
I think the best paradigm to think in, if you're thinking about privacy, is basically: think from a paradigm where at the beginning everything is opaque, and then ask what things do I need to reveal in order to operate safely, and to whom. And then you reveal those things to those people, and you don't reveal other things. So thinking explicitly about who has what data is, I think, one of the big lessons, and zero-knowledge proofs really make that possible. And zero-knowledge proofs are in the process of being made both quantum-safe and...
There are things to be careful about. One of the ways you can screw up with privacy is this issue of anonymity sets. The basic story is: imagine you have one of these systems, you deposit 10 coins, and then one day later you withdraw 8.739 coins, and then two days later you withdraw 1.261 coins. Guess what? 8.739 plus 1.261 equals — anyone want to take a guess? Okay, 10. Good. So it took you guys three seconds to figure that out. It will take an AI three microseconds. So basically, they are scanning for this stuff. And this is the simplest sort of example, but there are many more examples.
Basically, every action you take unintentionally reveals information about yourself, often in ways you did not even realize. There are amounts, there is timing, there are patterns, there is even network-layer information — there are a lot of different things. And all of them sort of reduce the set of people among whom you're hiding on-chain. Essentially, if you reveal too many individual bits of information, that can be enough to let theoretically pretty much anyone on the internet actually trace and identify your activity, even if you didn't mean to. So this is one of those things that's good to worry about.
One of my lessons from security is that the frontier is moving forward. The frontier is moving forward in two ways. One is that defense is getting better — things that before you could not defend against, now you can. At the same time, there are parts of the stack that are not as well defended, and the attackers are getting better at attacking them. So the frontier is moving forward in the sense that the area you needed to worry about before is often fine now. The area you were not even thinking about before is now where you need to worry.
For the privacy side, one of the frontiers is layers other than the chain — the network layer. We have no idea who is running the majority of nodes in the Ethereum network. It's an open protocol; anyone can spin up a node. So who can see your transaction as it passes through the network? This can also be used to identify your activity. Centralized services: if you talk to Ethereum through a browser wallet, then that browser wallet is often pinging dozens of different servers with all kinds of information, revealing things about what your activity is, about when you're online, about what information you're reading. All of these things are leaking information about you to people you never intended to authorize.
So short term, actually, I do recommend more people and more institutions run their own node. I have a node on my laptop. There is a bit of a dirty secret, which is that a lot of people don't realize how easy it is. And I think they don't realize how easy it is for two reasons. One is that there have been recent improvements in the technology, and over the next two years the improvements are only going to accelerate. These recent improvements basically mean that right now you can sync an Ethereum node within six hours — you were actually not able to do this five years ago when Ethereum was much smaller. Snapsync is much better. You can fit Ethereum inside half a terabyte, so lots of computers can fit it.
The other big reason is that running software is technically hard, but these days AI can help. AI can help you do things even if you're not technically skilled at all, and that's something always worth keeping in mind. So that's short term. Longer term, there's an entire privacy roadmap that includes things like private reads, private information retrieval, network-layer privacy — basically using cryptography, using these more advanced mathematical methods to do a better job of not leaking information.
The endgame in all of these cases is that every bit of information about yourself that you reveal should be an intentional choice. It should not be an accident. That is the endgame, and that is something the technology is really getting closer to actually making.
So, making finance actually secure. This has two parts. The first part is how blockchains can help, and the second part is what you need to do to make blockchains actually work for this. For how blockchains can help, the majority of the answer is eliminating counterparty risk — eliminating your dependence on single individual actors that the solvency of your assets depends on. And there's another smaller part of the answer that also matters a lot: enabling much more programmability. Saying, let's put our money into a wallet where four or six people can move it — this is not something banks easily offer. On a blockchain it's completely trivial; your agent can set it up and do the thinking to set it literally faster than the 12 seconds it takes for a transaction to go on-chain.
That's where blockchains help, and blockchains can help even going beyond basic financial use cases — they can help make a lot of different types of infrastructure more secure. The question is all of the challenges, and there are a lot of challenges. So many people worry about quantum resistance. Who here worries about quantum? Who here worries more about quantum than about AI? Okay. Who here worries more about AI than about quantum? Okay. Who here worries about both? Good. Who here worries about SI? Good. Silicon is good.
So quantum resistance, scale, and cost. Who here has ever paid more than $1 for a transaction fee on-chain? More than 10? More than 100? More than 800? So reducing transaction fees basically can only happen if you actually let the chain handle more. And so this is getting quantum resistance and scale and cost at the same time. All of these things are a big part of the Ethereum roadmap.
Then the AI cybersecurity apocalypse. If vulnerabilities exist, they will be found. And so the new bar for security — the thing that five years ago we thought was impossible — now becomes possible, and at the same time it becomes mandatory. Code that actually does not have critical vulnerabilities is basically what you have to get to. And this basically requires all of these very powerful mathematical methods, also enabled by super-powerful AIs. And at the same time it requires a big shift in mindset in how you think about Ethereum, how you think about technology, how you think about blockchains.
And then finally, security at the user layer. You cannot think about security without thinking about security of your data. So all of these things matter, and at the same time all of these things have solutions. These solutions are often a combination of better technology, better math, and at the same time better and newer types of thinking in how you actually use and apply these tools. And there is also a bigger challenge in packaging up these solutions in a way that is as easy as possible for more people, and people who understand all these things less, to be safe. And I think this is all a big part of what all of you are hopefully also going to be able to help work on. So thank you.
I'm okay to go straight to the panel.
Hi. Okay, can I invite... Okay, well, sir, I have got a drink first.
Okay. So, hi everyone. My name is Chaiik. I'm the CEO and co-founder of Wave. As earlier on, Holly, my president, has introduced me. Wave is actually a regulatory technology company headquartered in Singapore, founded about 12 years ago. And actually today, SFA celebrates its 10th anniversary, and next week we celebrate its 12th anniversary. So I'm very glad to actually be able to moderate this panel for all of you. Wave is actually an all-in-one AI-native regulatory compliance technology platform, and we cover both the Web2 and Web3 ecosystem. At the same time, I'm also the treasurer of SFA, so we do a lot of what we call voluntary work for the ecosystem.
So today I'm very glad that I'm joined by three gentlemen. Obviously, Vitalik needs no introduction. And in the middle here we've got Mr. Lasi. He's very, very experienced, deep in cybersecurity. Do you want to do a short intro about yourself?
Sure. Hello. Yeah, if I'm a domain expert, cybersecurity would be it. But well, this is one of the reasons I got interested in this space in the first place. Side is a wonderful security tool, and it's also a lot of security-oriented thinking. But I also manage the Ethereum ecosystem fund.
Yep. And on my right we've got the very brightly attired Mr. Douglas Gun. Douglas is the founder of Sarah CX, and he's a serial entrepreneur and investor since the age of, what, 12? Do you want to say a bit about yourself?
Yeah. So, hi everyone. I'm representing my team from Sarah. We went live three months ago. We are a non-custodial currency exchange for stablecoins on Ethereum. So if you combine stocks, bonds, commodities, and the crypto market, FX is actually two times bigger at $9.6 trillion per day. It's actually also the only market in the world where both sides of the exchange are actually valid. So the currencies are moving on-chain as stablecoins in ringgit, euro, dollars, Japanese yen, and so on. So anyone can be a liquidity provider on Serum. So 63 of them are currently averaging about 20% APY in the last three months on the ringgit stablecoin, real, and SGD. Yeah, so that's pretty much it.
Okay. So it is 2026, and we heard Vitalik talk about security and AI, or superintelligence in general. And this is not just a market trend — it's actually a potential inflection point and a technology capable of reshaping how economic activity is created, coordinated, and scaled. And that transformation has also arrived at a time whereby, to some people, trust may be eroding. There may be problems with centralized systems and institutions, and there's always going to be problems because where there's money, there's always going to be problems — not apart from what was mentioned about the difference between monkey and money. It's okay in between, I think we learned today.
Now, as AI systems become more capable of earning, spending, investing, and even reinvesting, they will need financial infrastructures that can actually operate across borders and remain reliable amid those broader shifts. So for trade to remain global, many argue that businesses across the East and the West need something that is more decentralized, jurisdictionally neutral, a financial system that they can build on. So our theme today of discussion is a programmable economy at the speed of AI, and what role can Ethereum play in that economy, and what opportunities does it create for the companies building within that ecosystem. So I'm going to have my first question to Vitalik. What changes when millions, if not billions, of AI agents can start to earn money, hire one another, invest, reinvest their earnings? What does that look like? What could that drive financial systems onto Ethereum, for example?
Yeah, I think that level of increase in volume and basically everything — it's both a source of big opportunities and it's a source of, obviously, big challenges as well. On the opportunity side, one thing I find again and again is that there are many markets that could exist and that even make logical sense to exist, but which just end up either being extremely inefficient or not existing. And often it's just very hard for people to wrap their heads around participating in them effectively.
I think part of what we'll see with AI is actually a lot of financial market situations, especially even on-chain, coming much closer to neoclassical economic predictions — the complete markets vision of what they expect should happen. And basically the reason why is because the price of thinking about what you're supposed to do is dropping to zero. And if AI makes mistakes, that's sometimes fine, because all meaningful frontier arbitrage and statistical arbitrage anyway. So that, I think, will enable many more types of markets, many more opportunities to hedge against different things, to reduce risk in different ways, to manage your exposure in different ways.
At the same time, in terms of the challenge, one big one is that, of course, volume is going way, way up. And another one is all of these security challenges that I just talked about. Those types of security challenges — someone literally trying to steal money by exploiting a bug — is sort of the most direct and the most blatant. But there's also a more general thing in society, where there are a lot of mechanisms in places that people are not even looking, that are only findable by using them the wrong way, and that friction is going away. And that's something we'll just have to — we'll discover a thousand little instances of that, and we'll just have to find the thousand ways to adapt to each one of them.
So it will start to evolve, and there will be lots of, like you said, bugs or incidents that will crop up, and then we'll discover the resolution to some of these issues along the way. Does that give us a lot of comfort in how the future looks like? We'll figure it out along the way. Well, it seems that that has to be the one. I'm not expecting to actually have all the answers to all the problems in the world. And so maybe turning to Liash, right? If an AI has already traded and reinvested money thousands of times, or even tens of thousands, hundreds of thousands of times, what happens, do you think, when the bank or court disputes the original payment? How does Ethereum change the way we can establish trust and keep that economy going?
Well, first, I'm disappointed that probably could not predict the future of AI society. I mean, my public market portfolio is going to be up like 16% this year. So great, solid prediction. Yeah. So again, I think when you have some assumptions that are baked into a system, it can be hard when those assumptions break. And the way we've set up the global financial system, there are a lot of assumptions. We try to keep things safe, and it's much easier to add rules than to remove them, especially when removing them is a global coordination problem. And in fact, we still have T5 — it takes five days to do an international transfer still in 2026.
And AI, I feel we should expect to become an increasingly dominant part of the economy. So you have non-human actors, and their involvement — more intense involvement — is probably the most competitive version of that activity. So I've had experience recently developing with AI, and my KPI is how do I get myself out of the loop as much as possible so I can saturate my human attention, which looks like giving the AIs more rope, basically. So I think we should expect that also to happen in our economies, and the result of that would be an inevitable increase in the amount of complexity that we see, because before the complexity was limited by human cognition, and that's no longer going to be true. It's now going to be limited maybe by human supervision.
So how is that compatible with the way we're currently doing things? It's probably not. And then if you can figure out a way to make it compatible, to host the same activity but in a way that is, for example, providing finality guarantees. So when there's a big scam, in a regular financial world there is the potential to reverse the damage by appealing to the court system, but that does have costs. It can be expensive, everything is uncertain for a while, and then it slows things down. But we should expect in the future to have an economy that's much more complicated. Everything is happening much faster. So even if you technically would have the ability to appeal to human institutions, it's just going to be like — in the time that you appeal, the world could change so much you don't know, like, you know, the recovery basically becomes impossible.
And then I feel that it makes a lot of sense to just have a technological substrate that can be politically neutral and host all this economic activity and provide not only finality but also direct access to AIs that are very good at programming things. So now instead of setting up institutions, which takes a long time, they can set up, you know, 10 companies before the hour is over, right? And anyway, that's — I think that's going to change quite a bit. Okay.
Okay. Turning to Douglas here, right? You've done a lot of stuff. As I said, you're a serial entrepreneur. You build cities, you build dating applications, all kinds, right? Very, very innovative. So how could stablecoins and AI let a retailer or conglomerate build a financial business as valuable as its original business, right? I mean, are there compelling case studies that you know of that we can learn from?
Yeah. So there's one very interesting one in Latin America. It's called Li. They're listed on NASDAQ. So back before they launched their loyalty points, right, they were doing about 1.2 billion in revenues. Then after, like, shortly later, about four or five years later, they started to launch their loyalty token on Ethereum. Then that went kind of okay for two years. Their cashback program was reliant on the loyalty token. And about two years later, they decided to retire the loyalty token in favor of a stablecoin. Right. So then at that time, at that point, they've really grown to about 10-ish billion in revenues. And after they launched their stablecoin, just this quarter alone in 2026, they're doing an average of like 10 to 11 billion every quarter, right? So it's significant. By the end of this year, they're probably going to do 40, 50 billion, right? And their cash holding went up from $2 billion in less than three years to doing three billion, right? So that's like a very significant company, and a financial business now represents about 44% of their revenues.
So from an online marketplace in Latin America, that's, you know, selling electronics, groceries and stuff, they've transformed themselves, I think, pretty well to become a fintech company and very sizable, right? Then there's also, like, you know, other companies. Like, for example, there's one in Malaysia that I found. It's called UAB, right? So they went from a travel business, like, providing host services, right? And then they expanded to become — to allow travelers to pay like a local. So these days they're called the crypto cards, where you can use, like, USD stablecoins and spend worldwide, right? So what they're doing is that they're allowing USD stablecoins across multiple intermediaries, licensed intermediaries, to enable the spending of — do it now. So you can just scan and you can pay using your USD stables against the local Malaysia rates table. So those are two very good examples.
Okay, great. So turning to Vitalik again. Why does an economy of AI agents trading across, you know, be it from the East to the West or the West to the East, need a financial system that perhaps no single country can actually dominate or control, or have, you know, absolute control? And what gives Ethereum the strongest claim to that role? You know, like, for example, why couldn't a database that is run by, you know, be it an American startup or a Chinese startup, do the same job?
Yeah, I mean, I think the answer for agents is in many ways similar to the answer for humans, right? Which is, finance is often one of those things where the kind of losses and gains are very asymmetric, right? And, you know, a lot of people who kind of do finance the wrong way spend a lot of time thinking about the difference between, like, 4.59% APY and 4.89% APY, but not enough time thinking about making sure they don't have minus 100% APY. And, you know, like, fundamentally, like, decentralization is a strategy for trustlessness, and trustlessness means not having counterparty risk, right? And counterparty risk is one of the more important risks out there. And, like, especially as I think the, you know, the world becomes more unstable, and this kind of illusion that people might have had 15 years ago that sort of finance, you know, like, basically sort of just is almost blockchain already has kind of kept on, you know, like, falling away in more and more obvious ways.
And if you want, like, just a reliable assurance that you'll be able to both keep your assets and be able to continue operating, then blockchains do provide that, right? And the unique thing that decentralized networks, and particularly Ethereum, have as a technology is that they are genuinely independent of any one single choke point that either could subvert them or that is necessary for their ongoing operation. Right. And anytime when there is a choke point, it's very easy for choke points to appear invisible for a long time, and then suddenly when those choke points get kind of taken advantage of, then everyone notices, right? And I think a lot of that was even, like, the story of the early internet, right?
And, you know, like, blockchains are one of those things that, like, on the level of the actual network, it is not dependent on any one single company or any one single country or even any one single operating system to keep running in service of the protocol. It's not dependent on any one particular team that keeps building it. There isn't a, you know, a structure that has, like, a king. There isn't a king. There isn't a high council. I mean, I think I hold the record for being the person who has proposed the highest number of Ethereum improvement proposals that got rejected. So that's — so there's a lot of — this keeps — right. So it's like the process of maintaining it is something that's also kind of very, very distributed.
And I mean, even at every level, right? At the level of, you know, like, at the level of kind of countries and regions of the world, at the level of kind of companies, even at the level of technology stacks — like, these are all things that we think about very explicitly. And it's actually quite hard to, like, make and maintain even a blockchain that gets to that kind of level, right? Like, there are a lot of projects that kind of describe themselves as being sufficiently decentralized, but, you know, in reality, it's — they're, you know, there definitely are choke points. They just haven't been activated yet. And knowing that difference is really important.
Okay. So I'm going to turn to Las again, right? And again, a bit of a caveat, without getting you into soliciting investment here. As AI drives more of the economy on-chain, what do investors still underestimate about the Ethereum ecosystem as an investment opportunity? You know, where could the biggest return come from?
Well, I'll divide that into two pieces. One, there has been, like, some social media sentiment that I feel grossly misrepresents the historical returns in the Ethereum ecosystem. It's been one of the best investments of all time. ETH itself did, like, 11,000x. So you invested $1,000 in the ICO, it's worth 11 million now. So — and that's — yeah, if you look beyond that, if you just do a thought experiment: what if instead of giving grants to some of the early projects in the space, we had made investments? We did the math, and it would have been a 100x multiple with multiple projects. Of course, not all the investments, not all the grants turned into successful commercial companies, but the returns from the — you know, you can only lose 1x, right? But when there's multiple, there's no limit. And, you know, we have 500x. So yeah, that's what happened in the past, right? But somehow people got so rich so fast they became super impatient. It's like, what, only 11,000x? Where's my other — I want to go to 100,000x. I'm disappointed.
So I think we said that I think a lot of what happens on social media is a kind of, like, psychological warfare between competing tribes and ecosystems. So we should look at the actual facts and have some critical thinking about it. And then when you look at the future, I think we're still at a very early stage. If you look — I mean, stablecoins are currently 2% of bank deposits for, like, USD, right? So 17 trillion in USD, and stablecoins are just a better form of money. There's no counterparty risk if you structure it properly. And it's, you know, so — or a very short chain of custody which is transparent, and then super fast, almost immediate settlement times. And this is programmable money, right? And you can ask your AI to program whatever you want. So that's just a better form of money that will drive a lot more innovation. So it's a good ecosystem to be in.
And then you look at, okay, how do we — there's an opportunity to bridge basically all of the existing financial assets, bring them on-chain so they benefit from this inherent property of programmability. And if you look into the future, given how ossified the traditional financial system is, how difficult it is to change rules, especially at a global level, it's a very easy prediction to make that everything on-chain will out-innovate everything that doesn't get to benefit from these properties, especially when you look at things that actually need to be jurisdictionally neutral.
So I think that's another big thing that people — they see Ethereum and they compare Ethereum with database clusters, like, wait, why isn't it going as fast as something that I could host on one rack in a data center or on AWS? But that's not a useful comparison. A useful comparison is: wait, can you — does the best version of this product actually need this property of neutrality? Is it bringing together, for example, you know, participants from the East and West that don't want to be exposed to only one jurisdiction, like the whims of the next political administration, whatever that is? And there are a lot of opportunities where that's the best version of that market, whether it's data markets, even things that are pedestrian like shipping.
So, and then I think, you know, AI is sucking all of the oxygen out of the room, but that is, of course, another area where we're very excited, in the sense that the best version of AI would be an AI that you don't have to trust. Just like the best version of a financial system is one where, you know, someone can't just, like, arbitrarily seize your assets, especially in a jurisdiction you don't even live in. So we see AI becoming more Web3 native. We see — but I think that there's a whole sovereign AI movement that we're very excited about, and this is very, very early. So we expect to see in the future, you know, very significant advances in terms of can we actually host not only AIs using blockchains as a substrate, but also AIs that give — basically you get guaranteed access to a model, it can't be taken away from you because it's too powerful, and you don't have to worry about the company just taking your data away and using it for training to commoditize your business.
And you can also monetize your own proprietary data as a business. So, you know, if you contribute to the training of a model, then that's something that you should benefit from. And how do we coordinate all the proprietary data sources out there that could be used to make specialized models and even general-purpose models much more useful? That's a coordination problem on a global scale because not everything is happening in one country. And the best substrate for that would be a neutral substrate, in a similar way to the way the internet is neutral, right? So how do we make the internet a native financial layer for finance, also for AI? And it's very, very early. So yeah, we're excited about what the future holds.
Okay, cool. I'm turning to Douglas here again, right? And, you know, I think you made some predictions before around the number of stablecoins there should be. You know, right now there's a couple of them that are pretty dominated. So, you know, as AI brings more global commerce on-chain, you know, why will the world still need more stablecoins and currencies beyond the US dollar?
Yeah. So I think if we look at stablecoins before the Genius Act was passed last year, in total fiat-backed stablecoins, you have less than 30. But fast forward to July this year, that number ballooned from 30 to more than 700, right? With about 500 of them being US dollar stables and the remaining being the rest of the world. So, and you have about 70 countries that have said that they'll be announcing about launching their own stablecoins in various formats. So from your central bank, your CBDCs, to your tokenized deposits, your stablecoins, right? And then if you look at, like, the world's money, the total global money supply for M2, right? You're looking at 102 trillion, and about 22-ish trillion is US dollar. The rest of that is non-USD.
So from a non-USD perspective, every country does not want to be dollarized, right? So if USD stablecoin were to — I mean, were to be dominant in a sense, then it will mean that a lot of these other sovereign nations who have their own central banks, who have their own currencies, would now have to be dollarized. But that's not possible. Like, when Singapore with a Malaysian team, with a Korean team and Japanese team, right? So they need to pay bills, they need to buy food, they need to pay their home loans and everything else, right? And domestic transfers. And a lot of countries, like rupiah in Indonesia, Malaysia in ringgit, they're all paid in local currencies, right? So we have also seen countries that have moved away. So, like Vietnam — I remember 10 years ago when I was in Vietnam, you can spend USD pretty freely. But now, if you want to spend USD freely in Vietnam, it's not that easy, you know? So they only want Vietnam, right? So a lot of countries will resist spending dollars, which means non-USD stables will rise.
And why they're not rising yet is because if we look at after the Genius Act, right? So typically most of the tech comes from the US, right? So the rest of the world will start to absorb it one to two years or three years later. So in this case, after the US has, you know, sort of told the whole world, okay, great, you know, so now in a big-four audit, right, it can be in your balance sheet. So Fortune 500, you know, they're all issuing their stables. Okay, you're in pretty safe hands. But by the time that's all done, you'll be at what, 2025, 2026, right? So once the rest of the world look at what the US have done, what the big four has audited, and what all the balance sheet looks like, okay, great, so now we adopt it to the rest of the world. So then you will see a very huge uptake of non-USD stables being launched, issued.
And a lot of these countries, about 22 of them right now, are actually licensed. So they actually have licensing frameworks that anyone can issue a stable. And this is very much like back in the 1800s when you have the free banking era, where you have more than 8,000 US dollar notes from different corporate institutions, right, worldwide. And that also happens in Canada, in Macau, in China, you know, everywhere in the world, right? So it's like history repeating itself. So now everyone will be issuing their stables because it's far more efficient. There'll be more companies like Mac Libé. They'll be issuing their stablecoin program with cashback, you know. So in Thailand, we're looking at a pretty big group that intends to launch their own stable as well. They, you know, they own a lot of small chains. I think about 30,000 outlets and so forth, right? So non-USD will have a very huge uptake next year.
Okay. I'm just going to ask a last question to all our panelists here. So maybe starting with the rest. It's the same question. So, you know, it's a bit of a what keeps them awake. So if we really achieve a programmable economy operating at the speed of AI, right, like the theme of this panel, what worries you guys more? Is it that the technology won't become powerful enough, or that it becomes powerful much faster than our ability to make it safe? Maybe with the rest, you want to share some thoughts on that. So if I'm worried that it will not become powerful fast enough, or that it will become too powerful, too powerful, then faster than our ability to keep it safe.
Oh. So I mean, I think any sensible person is worried about things that they're relying on becoming disrupted, as opposed to — oh, like, I think things — change is happening so fast now that it's hard to be upset that it's not happening fast enough. I feel — I mean, I loved sci-fi as a kid, and I feel constantly future-shocked. In the last — and it feels like a year now, it's like it's a decade before. And if it keeps going, I mean, how fast can things get? So I'm definitely in the, like, worrying that things will move, like, faster than we can cope with it, and then just change the underlying assumptions of our society in ways that just break everything.
Yeah. What about you, Douglas?
Um, AI is quite scary for me as a layman because, you know, it'll get more and more powerful. He'll start hacking stuff, you know, and then later there'll be robotics and he'll start planting AI into robotic brains. Then there might be surrogates. Then maybe there'll be a next pandemic and we'll all be at home, and then we'll just plant our brain to an AI and he'll just move around like the Bruce Willis show you around my age when I watched Surrogates in 2009, right? So yeah, you know, kind of thing, right? So it's pretty scary for me, and I think — I don't know, is there anything that...
I realized one thing that I missed in my answer. Well, maybe the talk will have more to say about that, but things happening fast doesn't have to be one frontier, right? I am worried that some kinds of things will not happen fast enough, especially defensive technologies.
Yeah. And so I think a lot of this is basically, you know, a race between this kind of raw capabilities of, you know, like, AI models and eventually a couple of other technologies, and together with the fact that they're kind of in a world that has evolved and, like, where the infrastructure is still only really equipped to defend against the previous version of the thing. And the fact that it's, like, faster to upgrade the frontier than it is to upgrade the world's infrastructure basically means that if, like, that balance keeps proceeding very — keeps going very quickly and it keeps becoming more imbalanced, then effectively, you know, we will be getting sort of, you know, like, the attackers of 2035 against the defenders of — I mean, sometimes 2015 and sometimes even 1980, right? And that is, like, that kind — like, that kind of lag is a big problem, right? Especially if all of these technologies are moving so quickly, right?
And I think one of the things that — I mean, this has, like, two concerns, right? I think one of them is in terms of kind of attack-defense balance, and then the other one is just kind of raw inequality of power, in the sense that there are a few particular groups of people that have the ability to hack, attack, interfere with, economically exploit, do all kinds of things to almost whatever they want, and kind of basically everyone else is only getting the same capabilities kind at their permission. Or, you know, like, sometimes only thanks to the benefit of, you know, what gets called distillation attacks. And so the — like, those are, like, both kind of significant issues that really compound on each other, right?
And so I think one of the things that we really need to work on is to kind of think explicitly about the defense side of the equation, right? Which in cybersecurity is, like, I think probably the most urgent frontier of that right now, and, like, rapidly upgrading all of the infrastructure so that it at least maintains, like, the most basic safety properties that we expect, and also basically even the, quote, reasonable privacy properties that we expect against this in a much more newer and more — against basically certain types of very intelligent thought having their cost almost drop to zero. But then, you know, there also will be other frontiers, and, you know, biology and the physical world and all of these things, and we will need kind of similar-scale efforts and similar programs there too.
Okay, cool. And so basically that's a very, very important reminder. So if you're in a business, startups, fintech, whichever tech you're in, I think what's important is, as F have mentioned a few times here, information security, cyber security is fundamentally important. Then the second pillar obviously is data privacy. The data side of things is super important. And when we talk about AI, there's absolutely another vertical called AI governance, which is already up and coming with lots of people or regulators getting into that space. So I think with all that, hopefully we're able to actually be alive to witness a programmable economy operating at a speed of AI not very far down the future. So with that, thank you very much for our panelists here, Vitalik, Las, and Douglas. Please give them a warm round of applause.
Okay, what we're going to do now, do we want to have a break?
Oh, we just go for Q&A.
I'm happy to go for Q&A.
Okay, cool.
Great stamina here. So he's going to be talking straight for three and a half hours.
So, okay, we're going to do a Q&A now. And then, in between the Q&A, if you guys want to grab some coffee, drinks, a little tea, please do so at the back. So if you want to... Okay, maybe you can stay. So if you got any questions, please raise your hand. And the people with the mic will help to bring the mic around.
Okay. Sorry about that, gentleman over there.
So hi, I have a village in Thailand and also involved in the Zuzalu calls. So that's my perspective. And I'm interested how does all this affect communities like people who are outside of this possible bubble that we are now in — farmers or dancers or artists, you know. And the other one is, are you seeing any stablecoins that are not pegged to a government currency, that their purchasing power is kept stable, but it's not pegged to the US dollar?
Yeah.
Sorry. Who's...
Let's start from Vitalik and go to the one as Vitalik.
Yeah. How do these things affect communities? I think once a community becomes organized enough, that community has resources that it collectively manages in some way. This can include physical resources. It can include digital resources. It can include information. It can include assets. It can include software. It includes all kinds of things. And I think all of those things have exactly the same security concerns as finance in many cases. And I think also the same type of potential in terms of cryptography making it possible to make whatever relationship of authentication and authority who wants to be implementable in a programmable way, but then needing tools like formal verification to make sure that it's actually secure. So that... and then in the physical world, you know, there's an increase like we're only going to continue to see this growth of IoT, and that especially matters if a community has physical spaces. And IoT, the way that it's been done up until now, is like basically a dystopian hell in terms of its security and privacy properties. And I think many people who rightfully just completely reject it and proudly have dumb homes for that exact reason. But I think at some point, you know, like that will have to and is going to flip, and we need to figure out how to make devices that are actually secure. And that's actually the other big reason to also care about the silicon side of things, right? Is that we want to actually be able to verify not just things about the digital world, but also things where the ultimate objects that we care about kind of seeing and modifying are in the physical world. And like we need to verify that, you know, the bridges, which are basically the sensors and motors that we have, actually are kind of making that connection in the correct way.
Okay. I would add to that that it used to be that you had to get permission to leave your country in some parts of the world, and then you would have had to forfeit all of your assets. That happened a lot in the 20th century to a lot of people. And it's already been the case that we have now a growing class of digital nomads. They're mobile, you know, they can take their assets with them even if they show up like naked on a shore. With the right security setup, that's not a problem necessarily. It's a very new situation to be in. I think if you add AI into the mix, AI reduces the need for specialization because it aggregates human knowledge in a way where it could be reapplied very cheaply anywhere, especially if you combine that with body-like robotic systems or like farming, or you can have a much more even quality of life throughout the world. So it matters less like if you ended up winning the birth lottery. Maybe you could just hang out in nicer spaces like your village.
So if you look at the history of money, you know, we start from bartering, right? So you're starting a new civilization, right? Then there'll be some form of bartering. And then if you progress that, you know, there'll be some currency of value, so like tea or silk in China. And then you have like, you know, all the other items and stuff that they trade with one another, right, in currency value. So I guess what would happen is that when a civilization starts to have its own currency that they recognize, that they start using as payment, whether it's in kind or whether it's for products or services or value from each other, you will progress to a stage where you will want distribution because then you will want probably being able to handle remote work locally. You'll want a transient bridge. So like for example, USD is a very good bridge today for multiple currencies, but they're not the destination. So in your case, the destination would be the civilization's currency, in this case your village currency, right? So yeah, you can post-issue anything that the community believes in, and people can either buy into that because you will start developing stuff that people can either invest in or partake or even contribute to. And I think it's an amazing effort, especially if you can in a decentralized manner too.
I remember you asked one more question. I think actually it was about how do you stabilize it? I can issue equity, but that's not the... So, you know, I think that's the holy grail of having like a native, native money that we can use online. I mean, if anything gets big enough, eventually it will become stable just by the property of it being big.
Still waiting on Bitcoin.
Yeah. Yeah. But that's a very long ride. So I think we're going to have like VR currencies in some form or another, even if they're, you know, tokenized, but that's here to stay for a while. But then as they get more adopted, you get like bigger and bigger networks. So the transition between that and maybe other forms of money that could be interesting, like maybe you want money that's pegged to purchasing power and doesn't just deflate away, but then you would need a lot of people to sign up for that. I think that's the challenge. How do you bootstrap network effects?
Okay, everyone here, we could start our own currency.
Who here wants to bootstrap a network effect?
Okay, we have a second question over here.
Okay, so maybe before you ask the question, state your name, the organization you represent, and who you are asking the questions to.
Sure. Thank you, panelists. Jason here from Hashkit, exporter of Ethereum. So I have one comment and one question, if I may. Should I tell it? Comment: I agree with what you said just now, what you shared. Just one thing I didn't agree with was SI, because if I may quote our favorite president: 'Welcome to the new world of super intelligence, SI.' Let's see how that goes. It sounds much better. It is much better, and it's way more accurate. Let's see if I have any power. Maybe I do, maybe I don't.
I mean, I'm from the 51st state, so we'll have to see what happens.
Thank you so much. So my question to you, Vitalik, is: you mentioned a lot about the safety issue on AI and the way forward. I know you are an advanced supporter of the whole idea of agentic agents and also your zk proof of personhood agenda. My question is, how do you draw the line between giving access to your AI agent, your proof of personhood details, and is there even a way whereby it could not be misused or abused? And what is the way forward in terms of what it looks like?
Yeah, it's a good question. I think proof of personhood is like one of those things where you have to be careful about kind of what it means and what it doesn't mean, right? Because like the maximum that that type of proof of personhood can prove is that a human authorized an action at least indirectly right now. And like if you see any big application dependent on proof of personhood succeeds, then there will be people who just give like their LLM full access to it, and you know you will see bots, right? So I see like that doesn't mean that it's useless, right? Because like as an anti-sybil measure, as a way of like preventing someone from making millions of fake accounts, it's very valuable. But it does mean that like that's not the same as a proof that like this thing was written by a person and was not written by a model, right? It's not like it's not proof of human effort in any particular way, right? And you just have to kind of be cognizant of that limitation. I mean, I expect, you know, like in situations where we do value having the guarantee that human effort went into something, like I imagine it will just happen in formats where you'll be able to like verify that in a more high-bandwidth way. And just in-person interaction is one of those big examples. And like probably, yeah, you know, there will be lots of in-person things happening for that exact reason.
Which is why, sorry to interrupt, which is why we have KYC, KYB, KYT, KYA address, and now we have KYA for agents.
Sorry. Yeah, I'm... I made a company that makes fun of people when they're wrong on the internet. Have you ever made fun of me?
Oh, we'll get there. I think yes, I have been wrong. Of course, everybody's been wrong. I think that there seems to be a lot of technological changes that are almost kind of separate that are somehow happening at the same time. And it seems like users can tolerate more complexity and developers can build faster. And I'm wondering, for programmable money, does this and the changes to Ethereum that are slowly coming online, does it unlock new categories of use cases that you guys are particularly excited about? And how do you think those would change society and the relationship to money?
Yeah. So first of all, I like your shirt. I think, you know, many people around the world are worried that we're heading into a dark future with all of these trends. So I'm glad you're going to make us go into a duck future instead. So I think in terms of like new capabilities that will get unlocked to kind of reimagine money, I mean, one of them that like I'm interested in is just like basically having more complete markets in terms of being able to hedge things, right? And like this is actually my own pet radical theory of the future of money and how money might go away, which is basically that the reason why you would hold fiat at all instead of, you know, like holding stocks or holding ETH or holding whatever else is that like you want like predictability, right? Like stability of value. But stability of value for everyone is always with respect to their own personal purchase patterns, right? Like the underlying good that you probably want is like not even, you know, like a particular type of dollar. It's closest to the guarantee that you will have a particular, like you will be able to continue your present-day purchases for some particular number of days or months, right? And that's the thing that fiat is like kind of good at providing. But the problem is it's like still sort of too, like it's too universal, right? Because everyone's spending patterns are different, and like that's even true across countries, right? And that's, you know, like part of the reason why things like the Eurozone enter into problems once they get bigger, right? And if you then ask, well, what's the opposite extreme of that? Then like you realize that, you know, you versus an AI company versus someone who has a totally different lifestyle versus, you know, like someone with a family or totally different kind of family versus someone who lives in a different place will have like very different expectations in terms of spending. And so everyone could potentially like basically just hold shares of indices of the categories of things that they expect to spend in the future, and like this would be a different way of getting stable value. So I think for this to work, like you need a really extreme level of kind of complete markets that like basically might only be possible with AI. But I'm like I'm hopeful and excited to see if that can happen.
