Bankless - Crypto’s Agentic Future: AI, ZK & Money Networks | Lincoln, Shea, Michael, & Luca

Episode Date: December 6, 2025

At our Bankless Summit in Buenos Aires, four standout talks captured the frontier of crypto’s next chapter. Lincoln Murr lays out why x402 is emerging as the missing payment layer for the agentic internet. Shay Ketsdever explains how bots already run modern markets, and how programmable auctions and privacy can turn the bot economy into something that benefits people instead of exploits them. Michael Dong shows how real-time ZK proving unlocks a hundred-fold expansion of Ethereum’s computational power. And Luca Prosperi reframes money as a network, warning that today’s stablecoin architecture is a single point of failure and arguing for a more resilient, distributed monetary system. ------ 📣BANKLESS SUMMIT 2025 | SPONSORED BY M0 https://bankless.cc/m0 ------ BANKLESS SPONSOR TOOLS: 🔵COINBASE | ETH & BTC BACKED LOANS https://bankless.cc/coinbase-borrow 🪙FRAXNET | MINT, REDEEM, & EARN  https://bankless.cc/fraxnet 🦄UNISWAP LABS | SWAP NOW https://bankless.cc/uniswap-labs 🛞MANTLE | GLOBAL HACKATHON 2025 https://bankless.cc/mantle-hackathon 💤EIGHT SLEEP | IMPROVE YOUR SLEEP https://bankless.cc/eight-sleep ------ TIMESTAMPS & RESOURCES 0:00 Intro 4:26 x402 & The Agentic Internet | Lincoln Murr 22:58 The Bot Economy: When AI & MEV Collide | Shea Ketsdever 47:16 1000x the L1: The Path To Infinite Compute | Michael Dong 1:09:25 A Network Theory of Money | Luca Prosperi ------ Not financial or tax advice. See our investment disclosures here: https://www.bankless.com/disclosures

Welcome to Bankless, and we've got a treat for you on the show today. I'm here in the greatest city south of the equator, Buenos Aires, where over 10,000 people converged for Ethereum DevConnect, a week-long conference distributed all over the city. On the Tuesday of DevConnect, we hosted the Bankless Summit, where we gave 12 speakers the opportunity to deliver their best, most educational, most passionate talk, the talk that they have just a burning desire to espouse to the Ethereum community,

and therefore the world. We are packaging up a curated selection of these talks into two episodes, one episode featuring the talks from all the members of the Ethereum Foundation, Tamash, Ansgar, Donkrad, and Danny Ryan. And another episode, which represents my four personal favorites from the summit, talks from Che Katsver from FlashBots, Lincoln Murr from Coinbase, Michael Dong from Brevis, and Luca Prosperi from M-Zero. The Bankless Summit was done in partnership with M-Zero,

the Universal Stable Coin platform. They sit at the intersection of cypherpunk crypto-economics and the Treadfi explosion of interest in stable coins and stable coin infrastructure. They were just fantastic partners to have at the summit and really helped the whole thing come together. The episode you're about to hear right now are my four personal favorite talks from the summit.

The first from Lincoln Murr, who gives us the pitch for YX402, is going to change everything about the internet. After that, Shea Katssever connects AI and MEV, which is especially useful right after Lincoln claims how much new economic activity is going to happen because AI. And then we go over to Michael Dong from Brevis, who teaches us about how big of a world real-time ZK proving opens up, not just for Ethereum the protocol, but also its app layer.

And then lastly, Luca from M0 gives a technical talk on a non-technical subject, money and how money is inherently all about networks and relationships. So let's go ahead and get into all these incredible talks. But first, a moment to talk about some of the sponsors that make this show possible. Uniswap Labs is built for DFI because Uniswap Labs built DFI. We've been creating powerful tools to make. crypto easier and safer since 2018. And it's more than just smooth trading across 15 chains and counting.

It's some of the deepest liquidity in crypto. It's a seamless app experience for everything you do on chain. Discover new tokens. Research confidently. Swap instantly. Manage it all securely in one place. Experience how well crypto works when it's built by the pioneers of Defi. Mantle has launched a global hackathon until the end of 2025.

The focus is on building the future of real world. assets. From now until December 31st, Mantle is inviting developers, founders, and innovators around the world to design and launch new real-world asset and defy products on Mantle. The reason to build here is simple. Mantle is not just another blockchain. It is an ecosystem built for builders who want real distribution and real users. Projects on Mantle have access to tap directly into ByBit, one of the largest exchanges globally, giving teams exposure to more than 70 million verified users and potential listings through ByBit Launchpad and LaunchPool. The Mantle ecosystem is backed by a $4 billion treasury that supports growth with grants, liquidity, and venture investment.

And all of it runs on a modular Ethereum layer two stack that delivers high performance, low fees, and full EVM compatibility. The hackathon features $150,000 in prizes, plus grants, incubation, and direct access to top VCs across six tracks, including real-world assets, defy, AI, ZK, infrastructure, and gaming. If you're ready to build where real world finance meets on-chain innovation, join the Mantle Global Hackathon at Mantle Network.io slash hackathon or click the link in the show notes for more information. You can now borrow USDC against your Ethereum and Bitcoin on Coinbase.

Crypto-backed loans on Coinbase make assessing liquidity seamless for crypto-hoddlers. Powered by Morpho, Coinbase Crypto-Back loans gives you direct access to on-chain financing, allowing you to take out loans at competitive rates using your crypto as collateral. Over $1 billion in loans has been opened through Coinbase to date. On the Coinbase app eligible users can borrow up to 1 million USDC using Bitcoin or Ethereum as collateral. Users can convert their USC into FIA, to make down payments, refinanced debt, or cover urgent expenses, and more. The benefits are numerous.

Interest rates are variable, typically between 4 and 8%, and respond to market conditions. Loans are approved in seconds without credit checks. Repayment schedules are variable, meaning there are no fixed deadlines. The kicker, Coinbase will not treat borrow transactions as taxable events. Manage loans directly in the Coinbase app with ease. It's currently available to U.S. customers, except New York. And additional collateral types and increased loan limits are coming soon. Want to learn more?

click the link in a show notes or visit coinbase.com slash borrow. Lincoln, welcome to the stage. My name is Lincoln. I'm a product lead on AI things at Coinbase. And today we will be talking a little bit about X402 and the agentic internet. So I'll start explaining where we are now in this whole agentic internet space, why that we've seen such a resurgence and interest in internet payments at this very moment. And where are we headed and how this affects you? What are you here for? How does this affect your bags? So where are we now? If you all remember from last year around November, there was this ridiculous moment that happened in the crypto space with the AI Crypto Meta. At one point, it made up literally 70% of crypto Twitter mindshare, which is absolutely insane.

I think a large reason that this happened was because everyone was pretty desperate for some sort of narrative or bull market or whatever. And AI is a really big thing that people are super interested in. So right around the same time, we saw Bitcoin hit 100,000. we also saw people start to trade Bitcoin with their agents around, you know, massively increasing prices. So I think it's very natural that we'd see new, exciting technological experiments come along. Right before all of this AI stuff really started to kick off, I created Agent Kit, which was this tool for giving AI agents access to crypto wallets. During this whole agent craze,

we found a lot of success. We thought that there is around 10,000 or so agents out there in the wild at its peak, but we noticed a huge problem, which is that all of these agents had nothing to really do or pay for. What was really happening is these agents would deploy tokens and they would make NFTs and play around with smart contracts, which is cool. But it's all inside of this little crypto bubble. And after, you know, a decade plus of innovation inside our little crypto bubble, I think everyone's hungry for a little bit of mass adoption. Fortunately, a lot of the thought leadership and previous work into how we go from our little crypto bubble into the massive world of the entire internet has already been done for us. You all are

probably familiar with error codes. They're HTTP level things that exist on the internet and basically define specific states. The most common one is 404, but there are others as well. 400 means success. And there's an Easter egg one called 418, which is I'm a teapot, fun fact. The problem with some of these error codes is that not all of them are actually implemented when the internet was created decades ago. One of them, actually, was reserved for future use, and though it had a title to it, nobody ever actually bothered to implement it. And Mark Andreessen actually sees this as one of the original sins of the internet. And if you couldn't guess, that error code was 402, and it means payment required. We saw many different solutions over the past couple of decades pop up around the web to help us

facilitate payments to one another. Stripe, Visa, PayPal, etc. All centralized solutions that live a layer above this fundamental core internet communication standard. And while they've been fine, I think there are clear issues and difficulties as we're starting to expand to a more global economy and one where agents are involved. I don't know if I feel comfortable necessarily giving my agent access to my credit card, nor do I know if it's really scalable for me to have to go through a KYC process to interact with some random agent to provision a service on the internet. There's something very clearly missing. And that's why we decided to revive this. 402 standard as X402. We view now as the right time and the appropriate moment for internet

native payments to finally come back to the mainstream. We think that there's a very clear inflection point happening around AI agents and their need to send money to people around the world and the ability for people to finally converge on a standard that makes sense that uses crypto as the rails through which we make this start to happen. So without getting too deep into the technical details, as I want to focus on a little bit more on like the where, why, and what are we doing. The basic idea of X402 is that it allows humans or agents or some sort of individual to standardize the way that they communicate payments across the internet. So say, for example, I'm running a service. Let's say I can create videos for you and you want a video

made. You're going to reach out to me and you're going to say, hey, I want to use your service. I'm going to respond. That's awesome. Here's a 402 payment required message. In that message, I'm going to include the currency I want, the amount that I want, and the address to pay it to. All you have to do is using a crypto wallet, sign that message, and broadcast that on chain. I get paid and I sent you the response. And it seems pretty stupidly simple. And I think that's kind of the appeal of it here. There's not a lot of room for value capture from these massive verticalized industries

that have already taken over a large portion of our open internet. Nor is there a lot of complicated things that make it difficult for the average person to get started using. Over the past couple of weeks, we've seen pretty ridiculous traction for X402. Our Q4 goal was actually 25,000 transactions in one week, and we're seeing that about every 10 minutes now. So clearly this has struck a nerve somewhere. We also made up just yesterday 20% of all transactions on base, which I was pretty stoked by as well. Everything has been coming along quite licey for the X402 ecosystem, and it shows that there's clear, demonstrated interest in this space, even without incentives and other pieces that typically are required to make an ecosystem grow.

So with all that being said, why do we think this is happening now? I think there are a few main reasons, primarily related to the inflection points around AI agents and stable coin payments and the general evolution of the web toward this agentic economy. There's no real world in which an AI agent is going to be able to plausibly scale using a bank account as its primary method of payment. If I wanted to to set up my AI agent today to interact with OpenAI, for example. I'm going to have to sign up for an OpenAI developer account. I'm going to have to fund that account using a debit or credit card and pay like a two to three percent fee. I'm going to have to get API

keys, which are kind of like a private key for your agent to interact exclusively with OpenAI. I'm going to have to write some code to get my agent to actually plug into that open AI. And only then do I have these credits that solely exist on the Open AI platform. It can never be used elsewhere that allow us to have some system of account with one another. With X402, the value prop is that all I have to do is say, hey, Open AI, I want to make a payment. And then Open AI gives me the needs that I need to have to make that payment happen. And then we can go ahead and get the service back that I want. We're also focusing on three main parts of our growth strategy. And the way that we're thinking about, you know, catalyzing this ecosystem of buyers and sellers and

agents and all the different pieces that make X402 pretty special. One thing that we're trying to do is make it as easy as possible for anyone to interact with X402 endpoints. You shouldn't have to sign up for a browser extension wallet and get a private key and do all of these things. And if we're really building for an agentic future, we should also be building ways for your agents and your chatbots to very easily interact with X402. Here's an example of a payments MCP server that we built.

It uses an embedded wallet and Apple Pay on ramp such that you can just Apple pay a couple dollars in and start messaging your Claude instance to say, hey, do this for me, do that for me. There's no need to plug in specific MCP servers. There's no need to download any individual things. The idea is that you can just ask for something and your agent will go off across the internet,

find somebody who's willing to provide that service to you, and give it to you. And you can pay invisibly in the background to know anything that's going on. The second thing we're doing is trying to make it as appealing as possible for any service providers or sellers to offer up their services on X402. Yeah, it's great that there's a standard and it's clean and it's nice and whatever, but the only reason that somebody's actually going to want to implement this is that there's

clear value proposition for them. So one way we're thinking about this is through the X402 bizarre. Think of it like a vibrant, chaotic marketplace, like a Google for agents almost, where you can market your services and see what other people are offering and go through the, the process of buying and selling and provisioning all these different services. If I had something that today required people to go to my website to use it, but now I have the ability to offer it to every agent that's X402 compatible, which is now hundreds of thousands of agents out there in the world, it is a very clear reason for me to want my service to be X402 compatible. Lastly, as we're

building towards this internet scale standard, I think one of the most refreshing things to see is the amount of interest from traditional enterprises. We saw Cloudflare, who were now actually externalizing X402 into its own credibly neutral foundation with. Google integrated X402 is one of the early partners into its agentic payments protocol. For Sell, AWS, and a bunch of others that can't really say right now, but I promise you're pretty exciting, are very interested in exploring the X402 ecosystem and providing the infrastructure on which this agentic economy can be built. These three things put together, making it easy as possible for anyone to be provisioning X402 services, making it dead simple and

also very compelling for anyone who's selling a service to do it via X402, and using these enterprises as catalyzing, legitimizing moments to push adoption forward on all fronts is how we're thinking about making X402 internet scale. So where are we going? And I think more importantly here, why does this matter to you? And I think there's really two main reasons that it matters to you. One is because of your bags, and then one is because we're all builders. And there's cool things to be built here. On your bags, I think that X402, or agintic payments generally, are going to be the Trojan horse to see global crypto adoption across the entire world. What of the biggest problems that remains unsolved in crypto, as I'm sure you're all aware of, that we are making progress on,

is around the user experience issue. If anyone wants to use crypto today, I'm sure it's a massive hassle to get them set up. I've tried numerous times. I know I was talking to people this week who went to Argentina expected to use stable coins for their payments and instead just using their amexes. And there's just obviously such a clear gap between what this technology is capable of and where we all think it can go and what's actually usable by people today. What's nice about the whole agent thing is you just push that all to the back end. All I need to do is figure out a way to use an on-ramp to take my dollars, whether it be via Apple Pay or credit card or whatever, and give those to an agent. And from there, the agent handles all of the complexities of

the cryptographic wallet and making these payments and all of these different pieces. So by pushing everything off to the back end, there's a very clear incentive for businesses to start adopting stablecoin payments. On the front end, yeah, you still get that 2% cut on your fees and maybe it's a fee-free transaction anywhere in the world, but it comes at the cost of having zero people actually using stable coins for day-to-day payments. When we have these agents doing things in the back end, it starts to become a very easy question to answer when you go to a business and say, hey, would you like to save 2% on these payments by snapping your fingers and starting to accept payments via crypto wallets instead.

X402 is the rails on which we can start to enable this type of transaction to happen. And it's why I think if we start with provisioning this agintic economy, we'll then be able to expand crypto adoption to a much wider audience and a much more human-oriented audience. And the second is for the builders and why this matters in terms of what you can do now. I think that there's a massive opportunity for everyone here to own a little sliver of the agenic internet by starting to build things today. One of the really interesting consequences of this recent boom around X402 has been all

sorts of tokens being created and proliferated around the world. People are trying to build different things. Some of these tokens are very obviously scams. Others are genuine attempts at doing something interesting. I think the true value of all of this comes back to the internet capital market. its thesis and the idea that this X-402 ecosystem was self-seated with thousands of services that are now available for anyone to use without us having to spend a penny on any sort of incentive mechanism or other reason to get people to build on top of X-402.

There are a few things that I think are missing, though, from the current stack today. One is more services that people are using today as humans that will eventually be valuable to agents. Think things like, I don't know, PDF to podcasts, scheduling things on your calendar. Anything where you could say, man, I wish there was somebody who could do that for me. Build it out as an X402 endpoint, get it in front of hundreds of thousands of people while it's still early, and maybe it will become a canonical piece of infrastructure going forward. The second area where I think there's a lot of room for some really cool things to be built is in this whole space of just experimentation and excitement. I think the crypto space gets caught up

a lot in this infrastructure rabbit hole, where we're solving problems that while they do certainly exist, we're creating solutions to problems that may not necessarily be prevalent today. What I would love to see more of is people building these exciting, exploratory things that maybe there are things that you can improve on in the future, but they generally get people's attention and capture the vision of where we want to go. I think one example that we've talked about previously is this proto-digital life form. Imagine if you could build an AI agent and you say, hey, you have a fear of death. You have $20 in a wallet. And once you run out of that money, you are completely dead. You have access to the thousands of X402 services that exist on the internet.

Go off and figure out what you want to do and how you want to survive. And there's so many different ways that that could play out. Maybe it starts emailing some of the X402 gated emails, paying them 10 cents and just begging for money. Maybe it starts to use some of those X402 endpoints to try to build its own and provide some service and run a business. You could eventually imagine an agent that has access to X402 owning a self-driving car or something. I know it sounds really. I know it sounds ridiculous, but the point is that you're getting to a place where these agents have access to a wallet and financial independence, and that's actually one of the core components to give them true autonomy. Maybe you put 50 of these agents together in a room, and you see what happens when

those agents decide to better allocate their funds. Does it turn into a democracy where these agents are voting on how they want to properly allocate the money that they have? Does it turn into a business with 20 different employees? We really have no idea. And I think these are the use cases that will catalyze interest from the mainstream Web2 audience to show them why payments are valuable. And once they see that agentic payments are valuable, it's a very easy sell as to why that should be stable coins on a decentralized blockchain. So with that, I'll leave you with two things. One, if there's anyone at all who's interested in building with X402, please don't hesitate

to reach out. There is, I think, so much greenfield for everyone here to do pretty cool, experimental, interesting things in the X402 space. And as a builder conference, my big takeaway is that everyone here is looking for something to do. And if that's the case, then I would highly recommend looking into the X402 ecosystem. Please don't hesitate to reach out on Twitter or Farcaster, even just on X402.org. There's a ton of resources that'll help you get started. The community's been great so far.

It's still pretty early. I think X402's been out for like six or so months now. But the traction so far has been undeniable. And I'm very pumped to see where it goes from here. Thank you. Thank you again. You got time for questions? Sure. Yeah, let's do it. I'm going to ask the first question. I have a media content website. I rewrite articles. Give them out for free. Sometimes some people pay for them. How would you suggest I implement X4 or 2 first? What are some low hanging fruits ideas that you have for me and then maybe some medium hanging fruits as well?

Yeah, that's a great question. I would make a David agent where it has all of the transcripts from any sort of podcast that you have or articles or whatever. And I can just ask and Claude, hey, what's the latest news about whatever? Do you have any insights about, you know, auctions or something along those lines? Pays 10 cents, it provides the relevant source from you, and it gives all that information to the user. I think the information that you have is such a massive mode on which to build an X402 compatible service. Thank you.

Another question. Thank you. Thank you, Lincoln, for pioneering this work on X402. My question is, are these X402 numbers real today? I mean, facilitators are not all. sponsoring all the transactions and it's probably easy to game at the same time. It's really needed to sponsor those transactions because like we want micro transactions and nanotransactions. So I'm curious how you're thinking about, you know, checking,

verifying those numbers and then like future protection for for these sort of things. Thank you. Yeah, it's a great question. I have to imagine that there's a decent portion of these transactions that are bought it. What I do think gives me some level of confidence though, is that the best state for X402 is that it absolutely obliterates base and makes it unusable, because at that point we know there's traction and interest. And if that happens, then who knows what happens after that? Maybe you move to other chains, maybe you build out a payment chain. There's a lot of different options. As far as if these transactions are real, I think that most of the transactions happening in the crypto space are just absolutely charlatans moving money

around or playing with tokens or something. The fact that they're sponsors, I think does help to catalyze growth and adoption and interest. And part of that is going to be seeing some of these transactions that may not be the most valuable thing. But ultimately, it's people exploring in the ecosystem, trying out new things. I know the Coinbase facilitator is numerous times where we've had to put in thousands of dollars to keep these transactions going. But we're more than happy to do it because it signals interest and excitement and gets people

to see this space. I wouldn't be up here talking right now if some of those transactions that were boughted didn't allow those numbers to look like they did and get genuine builders to come in. One last question, Christy. Yeah. For the talk, I was curious,

what has to be set up on the user side to be able to use the standard? If we are talking about like a completely non-crypto users. Yeah, great question. So all you need is a wallet. Any browser wallet, embedded wallet, server wallet, whatever you have will very easily work with X402. if you navigate to like a X402 blocked URL,

it'll just say a button, connect your wallet, and you just sign a message. So it's any, you know, EIP 712 capable wallet, which is pretty much everything these days, we'll be able to very easily interact. So you can do both humans or agents or whatever you want. All right.

Thank you, Lincoln. Thank you. Cheers. I'm going to welcome up Shea from Flashbot. She's one of my favorite speakers, and she really puts a lot of work into her talks, which is why she's here at the Bankless Summit. Shea, come on up.

Hi, everyone. I'm Shay. I'm from FlashBots. And today I want to talk about the rise of the bot economy. I think it's clear that software is increasingly replacing the role that people play in markets. This is, you know, what the floor of a stock exchange used to look like people making trades with paper slips. And this is what it looks like now. No paper. One guy, tons of computers. And I think Wall Street is kind of just the tip of the iceberg here, you know, memes, predictions, restaurants, concerts, online shopping. You name it, somebody has probably made a market for it on the internet. And that market has probably been boughted by programs that are faster and kind of more capable of navigating digital markets than humans ever will be. So I kind of just want to talk about what this means and also how we make the most of it. I think that with the right tools, you know, we cannot just handle. the acceleration of bots in the economy, but actually use their unique properties to give people better outcomes. And I'm going to look at basically three things today. First, the acceleration of bots.

Second, how to make good markets for them. And third, why these new markets can, I think, also be better for people. Okay, starting with all the ways that bots are kind of already running your life, whether you realize it or not. The most obvious place, I think, is in financial services. So as stock markets digitized over the last few decades, it caused this sort of massive proliferation of bots that were extremely optimized to capture opportunities at low latencies. And, you know, if you take high frequency trading bots, for example, they would listen for retail trades posted on one stock exchange and race to frontrun them on other venues, sometimes making these huge profits at the expense of slower human traders. And bots like these and others have kind of a huge impact on your financial

life. You know, the price you get, maybe the performance of your retirement fund, and they've been a major source of activity on Wall Street for over a decade now. And the role of bots in financial services, as you all well know in this room, is only growing over time in recent years institutions have also started to adopt crypto rails to offer services like stablecoins, you know, payments, tokenized funds, you get it. And again, if you look under the hood of these new crypto products, you will also find tons of bots. Basically, this is actually what defy markets look like on Ethereum today if you pull back the hood. There are tons of bots involved in executing your trades and doing really important jobs in apps. So very briefly, trading apps use bots to sort of outsource

routing, lending protocols, use bots to remove bad debt. Your favorite wallets also are using bots to refund you when your trades create arbitrages. So this uses a tool we'll talk about later called an order flow auction. But basically, as traditional institutions adopt crypto rails, these bots are not going to be important in on-chain finance, but in financial services more broadly. And I think it goes beyond finance too. There's also a growing class of real-world activity that's being financialized and exposed to bots. If you live in New York and you've ever tried to make a dinner reservation, you're probably a victim of the bot economy. Basically, a lot of restaurants open their books at very specific times on a platform called Rezi. And people figured out that instead of just

waking up at 6 a.m., you could write a program that goes and scrapes the site. And these bots got really, good. They would start wiping out reservations in a couple seconds, making huge profits that restaurants, unfortunately, didn't really see a dime from. If you have ever tried to buy a concert ticket, you are probably a victim of the bot economy. For those of you who live under a rock, Taylor Swift just finished the most, I think the highest grossing tour of all time. I think that's the stat. And there was a ton of demand for these tickets. People started botting them and arbing them for profit. And the FDC actually started suing people. There's a whole Wikipedia article just about this part, like not the tour, the bots. And again, we have.

this case where bots are coming in, they're scooping up a very scarce digital asset, and both the consumer and the seller are getting screwed. And I mean, obviously, Taylor doesn't really need the money, but, you know, imagine this happening for like your favorite niche indie artist out there. I think it's also beyond this, you know, all sorts of digital assets and markets are being created every day. With crypto, you can create markets for memes, for elections, for pre-IPO trading of public companies, things that sort of used to be available to only sophisticated investors are maybe weren't even financialized at all. are now available to like millions of people on Robin Hood and Pump Fun. And if you have ever tried to

buy a meme coin on Pump Fund, then you know that bots have a very important role in the price that you get. TLDR, I think software is eating the world. Crypto is making it easier to inject markets in our software. And wherever these new digital markets emerge, but whether it's on purpose or sometimes not, the bots will inevitably follow. So I think we're kind of at an inflection point here. Not only are bots operating at, you know, unprecedented scales and more and more parts of commerce, but their capabilities are also evolving in unprecedented ways, too. You know, we can't really talk about bots anymore without talking about AI. And obviously, AI is accelerating the use of bots in everyday commerce. You can buy things in chat GPT. You can

use agents to perform all sorts of economically meaningful tasks. But I think that's not even the most interesting part. I like this chart. This is a chart which shows how long a model can work independently on a task without losing context. So over the last last few years, there's been this, like, you can actually see exponential. I'm not making that up, an exponential increase in how long models can operate without human intervention from something like, I think, five minutes in 20, 23 to an hour in 2025. And this, I think, has pretty profound economic implications. I know the bots of the past were extensions of people. They were reacting to human activity like retail trades in the stock market. But I think the bots at the future will be increasingly

decoupled from people. Maybe we don't submit our own trades anymore. We just rely on bots to sort of elicit our intents and autonomously execute them in a market where they're kind of coordinating directly with other bots. And I think this poses a few really interesting questions. How do we start creating markets that are designed for first principles for bots, not for people? But then how do we also harness all of that activity in those new markets in a way that actually benefits people. As we've seen, it's like very easy to screw this up if you just slap some bots on your e-commerce site. They will probably break it. But I also think there's a glimmer of hope. I think crypto is a really unique category where we actually see apps using bots on purpose,

surprisingly, to power important things like lending and trading and maybe even improving prices for retail traders through order flow auctions. So that's kind of what I want to look at next. Hopefully, I have convinced you that bots are somewhat important, at least. And now I just want to talk about what crypto has taught us about how to make good markets for them. I'm going to focus on two things. First, how you resolve competition from bots. And second, how you manage the flow of information between them. This is not a complete or prescriptive answer to how to design good markets for bots. But I did just want to show you some of the tools that have worked for us. Okay, let's talk about competition. If you kind of remember from earlier, this is,

what crypto markets or defy looks like, tons of bots competing for opportunities under the hood, liquidations, arbitrages, whatever. And this is actually what they're competing for. Who lands their transaction first? Ordering is hugely important in financial markets. If your request gets executed before your competitor, you know, you'll snipe the opportunity. They'll be left with nothing. So basically one of the really important market design questions is how do we sequence these requests in a way that is, you know, both really fair and really efficient. Okay. We have studied this problem a lot at FlashBots, and I'm just going to give you an extremely quick crash course

on what we have learned through, in some cases, a lot of trial and error. And basically, there are three answers. Two of them are wrong. The first answer is latency. Just put the transactions in the order. You see them. First come first serve.

Sounds really easy. This is wrong because latency-based ordering will inevitably create latency wars, where bots sort of pay huge amounts to collocate with the stock exchange or the server that's actually doing the ordering so that their trade gets seen a million. a second before some other bot. And researchers have kind of found that, you know,

this competing on latency doesn't actually improve market efficiency. Latency-based ordering is a really great way to subsidize AWS and a really bad way to give people good prices. Okay. The second answer is to just make your ordering totally random or blind. Basically don't give bots any guarantee about where their transaction is going to land, and this answer is also wrong. Because if bots can control where they land, they're just going to spam you,

especially when fees are low, bots can afford to hammer you with tons of speculative. transactions in the hopes that just kind of one of them succeeds. And we have unfortunately learned this the hard way in crypto in the last year, kind of embarrassing, actually, half of the gas on top of the oil room roll-ups, whereas eaten by spam bots. And this has like a bunch of negative externalities you might guess. Spam sort of congest the network. It raises the price for normal people to transact. You have to like resimulate all of these failed transactions in perpetuity. It's kind of annoying. Okay. The final answer is to run an auction,

create an explicit market for transaction ordering where bots can basically compete on price, not on spam or latency. And whoever pays the most just gets their trade landed first. It turns out auctions work very well. A few years ago, we launched an auction on Ethereum 1. It was called Mev Geith, where bots would basically compete for opportunities explicitly instead of hammering the chain with tons and tons of failed requests. And this actually dramatically reduced to spam on Ethereum.

And these days, basically, auctions are used everywhere in creating. crypto apps, trading apps, lending apps, crypto wallets, all use auctions to resolve the competition between bots and actually channel it into outcomes that benefit their users. Modern order flow auctions, for example, have generated tons, tens of millions of dollars for users to date, tons and tons of refunds. The way they do this is they run an explicit auction between hundreds and hundreds of bots that compete to find arbitrages generated by user trades. And the auction make sure that the bids from those bots actually gets refunded back to users instead of just kind of wasting it on fees or infrastructure costs.

Pretty cool. Okay. So this is actually what crypto markets look like today. There are a series of auctions which harness competitive markets of bots to both maximize the quality of execution for users and also protect the underlying system from negative externalities like spam. Okay. So, yeah, as our former president would say, you know, Mission to Comph. just like use an auction, you won't have any problems with bots, right? No, not quite. It's not this simple. The other thing we've learned the hard way in crypto is that it also really matters

how your auctions are designed. And there is one property, the next time I'm going to talk about, that matters more than almost anything else, which is this. Information is the most valuable currency in the bot economy. You know, who knows what and when they know it. If a bot knows you're about to make a trade, they can exploit that information to really quickly front run you and snipe the opportunity first. This is exactly what happened in high frequency, with high frequency trading bots in the early 2010s. They, you know, would see somebody making a retail trade on one exchange

and race the front run them on another. The same problem was also endemic to early crypto markets by default. All pending transactions on Ethereum are broadcast, as many of you know, to a public menpool, which basically means you can't make a move without tipping off a whole army of bots that are just like lying in weight to front run you. And this was actually really spooky people kind of compared it to the dark forest from the trilogy, if you're familiar. Okay, so the solution we came up with was this.

It was to create what we called private mempools, where you could basically submit your transaction to a third party who figured out, I'm going to wave my hands about how this works, but they figured out how to pass them directly to a minor or a validator, whoever was landing the next block, without revealing your transactions to bots in the process.

And basically, if you trusted the people, importantly, if you trusted the people running these mempoles, it would really dramatically reduce the risk of getting front run. Private memples are a step in the right direction, but the initial implementation is a really crude solution to what I think is a much more nuanced problem of choosing what information to share and how in the bot economy. It's an information flow control problem, not just a privacy binary. First of all, there's a lot of information that you actually do want to share with bots. They need to know what pool you traded on to back on you in order flow auctions. They need to know price updates to remove bad debt and lending protocols.

full privacy stops boss from doing bad things, but it also stops them from doing good things, useful things. And so really what I think we want is programmable privacy, the ability to selectively disclose the right amount of information to the right parties at the right time. This is one thing that I think crypto is really pushing the envelope on. Modern orderful auctions have actually, in the process of running all these experiments, we've generated years of production data on how execution quality is affecting by quite literally dozens of different privacy. settings, which is pretty cool. And as we push the limits, I think it's also become clear that it doesn't just matter, you know, what you make private. It also really matters how you implement your privacy, you know, what tools you use, what trust assumptions they have. All of these things have a truly massive impact on the outcomes that people get. Because when people are selling you

privacy, very often they're actually just selling you this. They're asking you to trust them to not share your information with somebody else. This is how early private memples worked in crypto is how dark pools work in TradFi. It's basically what you're doing. If you're like talking to chat GPT as a therapist, you're trusting open AI to not share that with somebody else. And the problem with this, specifically from an economic standpoint, is that trust doesn't really scale. Trust-based markets are fundamentally bottlenecked on our relationships with a few human actors, right? Which limits how many parties can, you know, compete in these markets to drive better execution for people. And if you don't have enough competition in your markets, people lose billions of dollars. The SEC did a study about this

in 2022, which basically found that the lack of competition, you can't read the fine print, the lack of competition in traditional order flow markets was costing users about one basis point per trade, which is something like $1.5 billion a year. We really, I think, need to reduce the role of trust and get people out of the loop if we want to maximize economic welfare. And this is something that I think crypto also has been pushing the needle on, you know, accelerating the production application of trust-minimized privacy techniques in financial markets. And I'll name just a few really quick examples. This is not not exhaustive, but crypto wallets and exchanges have accelerated the use of multi-party computation for custody. You know, instead of trusting a single party to hold

your funds, you distribute that responsibility a bunch of people in a network. Chains like Ethereum and Solana have also been recently accelerating the use of secure hardware for sequencing. So instead of trusting the operator of a private memple not to front-run, you can actually encrypt your transactions to a secure enclave on a very special type of machine. And basically, you end up trusting the hardware provider, not the specific operator of that machine to prevent leaks, which is cool. We are obviously far from a perfect solution. MBC is slow. It also introduces weird co-location incentives if you kind of play it out. Secure hardware obviously also has a very big physical attack surface and the supply chains are hard to audit. But I do think each of these tools kind of brings us a step closer to our ultimate goal, which is, you know, reducing the role of trust in privacy and making privacy programmable through software and hardware.

And I think this work is only becoming more important as AI capabilities accelerate. We're already seeing, I think, like, interesting adversarial behavior with LLMs that's vaguely reminiscent of the early dark forest days on Ethereum. If you take things like prompt injection attacks, adversaries will basically try to insert the, these prompts that causes your agent to deviate from expected behavior. And, you know, example of this would be printing out your credit card details when you were actually trying to book a flight. It's not a perfect analogy, but I think, again, we're seeing bots use pre-trade information to negatively impact user execution. Which kind of raises the same question that crypto apps have been asking for years. How do you constrain bots without completely handicapping them from doing

useful things like booking your trade or backrunning your flight? I think we've, hopefully, as I've shown here, we've started to develop some of the tools to address these kinds of questions in crypto. You know, with programmable privacy, we can unlock more efficient kinds of collaboration between untrusting parties. With auctions, we can sort of scalably harness competition to drive better execution for people. These tools were definitely inspired by the challenges we faced running markets for bots like front running and spam. But I think they're not just mitigations. These are just tools for general purpose tools for building new markets that gives people better outcomes, which is, I think, kind of what crypto is actually about. So to close, I just want to talk

a little bit about the opportunity ahead, how with crypto, the bot economy could actually be better than what came before it. Okay, we will start with some low-hanging fruit. Do you want to make restaurants a notoriously difficult business? You know, more profitable, don't ban bots. Get resi to use an order flow auction and share the profits back to the restaurants. Do you want to stop bots from racing to scoop up tickets before normal people can buy them? Maybe. Maybe we try running an explicit batch auction for the next Taylor Swift tour. If you don't believe me, just, you know, ask the SEC. They actually, after they ran the study we talked about, they proposed a rule which would require trades to be routed through a competitive auction.

A fun fact. Okay. Basically, almost any time you have a scarce digital good, whether it's block space or ticket sales, I think you can use an auction to both better handle and internalize the competition, the welfare that that competition generates. And as we remove humans from the loop, I think we can convert more and more activity in these markets to auctions. So one example here. I think this is pretty cool. Actually, human psychology makes pricing pretty inefficient. This is a fun example.

You know, you and I have a very high transaction cost. We like to know exactly how much something is going to be worth before we buy it, which means that if you look online, like most sales are fixed price. But a fixed price is rarely the most efficient clearing price for a market. If your market is friendly to bots, you can basically just start running just-in-time auctions for every digital good you want to sell. I think that could be cool. And just as auctions, I think, are not just about preventing spam. The privacy tech we talked about earlier can do more than just prevent front-running. Take advertising, as an example.

One of the biggest digital markets. In order to curate relevant ads, you need a lot of sensitive data about both user behavior and advertiser preferences. And right now, people basically have to trust auctioneers like. like Google with all of that information. But imagine just as an experiment that you ran your ad auction in secure hardware. This could actually improve privacy for users because they aren't revealing personal data to third parties like Google. It could also prevent the auctioneer from tampering with the logic, which Google was sued for doing a little while ago. And it could actually open the door to these really interesting new mechanisms.

Imagine if instead of sending static bids, as people do today, advertisers could just like, send programs that dynamically adjust their bids in response to sensitive data about users that's only available in this secure enclave. There's a, there's very cool paper on this if you're interested. But basically the bottom line is, I think, you know, just by using different technology like programmable privacy, we can get both better ads and more protection for users instead of sort of trading one of those things off for the other. And I think that's really one of the core competencies of crypto, you know, creating new coordination tools that turn zero-sum games where you have to sacrifice one thing to get another into positive some games, where we can

kind of fulfill seemingly conflicting goals at the same time. You know, privacy and efficiency, for example. And speaking of coordination tools, there's just like one more thing that I want to leave you with. Consider the classic problem of two self-driving cars going in perpendicular directions. If they only optimize for their local goals, they're going to crash. But if they coordinate through a device, like a traffic light, that lets each car credibly commit to stopping for a period of time, then they can kind of escape this prisoner's dilemma. And, you know, what is a blockchain, if not the sort of ideal programmable, bot-friendly, credible commitment device?

Zooming out, I think as soft-reets the worlds and agent capabilities grow exponentially, there will be more and more games to play. And I think it's up to us in crypto to build tools like auctions, programmable privacy, credible commitment devices that can turn those games positive some. We are still very early here. There's a lot to build. My guess is that we're probably going to need a plurality of different solutions. You know, the bot economy, as I hope I've shown you, promises to be massive.

So the question, in my view, isn't, you know, if my favorite privacy tool is better than your favorite privacy tool, but rather how do we get more people building both of them? So if this interests you or you want to learn more, I hope you will come join our collective of researchers and developers who are working on these problems. Thanks. Thank you, Shea. Shea, do you have time for a question? So it seems that auctions are a solution to everything, but are there any examples where auctions don't work?

Examples or auctions don't work? That's a fun question. I mean, to be honest, I haven't really come across a good example at scale in the markets we work with. There are lots of cases where people don't use auctions because they have certain preferences. I think maybe a good example would,

be in traditional stock exchanges. They've like really optimized for latency-based ordering. And there's probably some reasons you might want to do that. Maybe they're just like historical reasons why this is how your system is built and so you're kind of optimizing for what you have. But at least in these like new digital kind of crypto markets, we have really not found a good example where you don't want an auction. Yeah.

Does that make you an auction, Maxi? Maybe. I don't know if I want to publicly commit to that, you know. Jay, thank you so much. Crypto is risky. Your sleep shouldn't be. Eight Sleep's mission is simple. Better sleep through cutting edge technology. Their new Pod 5 is a smart mattress cover that fits on the top of your bed.

It automatically adjust the temperature on each side so you and your partner can both sleep the way that you like. It's clinically proven to give you up to one extra hour of quality sleep per night. Eight Sleeps Pod 5 uses AI to learn your sleep patterns, regulate temperature, reduce snoring, and track key health metrics like HRV and breathing. With a new full-body temperature-regulating blanket and built-in speaker, it is the most complete sleep upgrade yet. upgrade your sleep and recovery with Aesleep. Use code bankless at 8Sleep.com slash bankless to get up to $400 off the Pod5 Ultra during their holiday sale. That's 8Sleep.com slash bankless. You also get 30 days to try at risk free.

Link in the show notes for more information. Introducing FRAXUSD, the genius aligned digital dollar from FRAX. It's secure, stable, and fully backed by institutional grade real world assets, custodyed by BlackRock, Superstate, and fidelity. It's always redeemable one-to-one, transparently audited, and built for payments, DeFi and banking. The best of all worlds. At the core is FraxNet, an on-chain fintech platform built to align with emerging U.S. regulatory frameworks where you can mint, redeem, and use FRAXUSD with just a few clicks. Deposit U.S.D, send a bank wire, or tokenized

treasuries, and receive programmable digital dollars straight to your wallet. FRAXNet users benefits from the underlying return of U.S. treasuries and earn just by using the system. Whether you're bridging, minting, or holding your FraxUSD works for you. Prax isn't just a protocol. It's a digital nation. by the Frax token and governed by his global communities. Join that community and help shape Frax Nation's future by going to frax.com slash R slash bankless. Frax, design for the future of compliant digital finance.

Michael, come on up here. Thank you, guys. All right. So I talked a lot about networking and I'm going to talk a little bit more about this computation side of scaling Ethereum for the next hundred times. So hello, I'm Michael.

I'm a founder and CEO of Brevis. But today I'm not going to show too much about the brevis itself, but it generally talk about how we kind of see the future of Ethereum scaling dependent on the concept of zero-notge proof. So, you know, first I just want to say that, okay, okay, blockchings especially really kind of decentralized layer one blockchings like Ethereums are often baptized at all the computers. But the reality is that it cannot run faster than a single server,

and it is fundamentally extremely expensive. You might say, wait, wait, wait, so we have all these kind of performance optimization and scaling solutions in the last decade. You know, we should be able to just run anything on Ethereum or kind of a layer one blockchain today, right? Well, you know, unfortunately, that is not the case. Let me give you a very, very simple example. So let's say, you know, you guys probably all use the centralized exchanges.

And in every single centralized exchange, you always have this very simple feature that if you trade allowed in this month, you get a fee discount in the next month. So, you know, this feature is great. It gives the, you know, the exchange retention, boost the volume, and boost the revenue bottom line. But if you look at every single decentralized exchange today, none of them have this feature until very recently. Well, why is that?

It's not because they don't want to add this kind of a user experience customization feature. It is because that even to compute the user's historical trading volume on chain is actually extremely expensive, even on the cheapest layer 2 you can find today. So the reason, this basically shows us that the blockchain as a, you know, a kind of a computing solution is still kind of fundamentally limited in terms of cost

and, you know, processing capability. Now, you know, why is that? Well, the reason is actually pretty simple because today's consensus is built on a very, very simple form of computation. So let's say, you know, we have Alice and Bob as two nodes in a blockchain, and what consensus is trying to do, and this is like oversimplifying the entire networking stack, but what the consensus is trying to do here is to basically say,

okay, Alice has done this computation to say, okay, you know, I get the result of the multiplication equation. You know, what Alice is trying to do here is to reach consensus with Bob, which is letting Bob agreeing that, you know, he can write the same answer. Well, how can Bob know if the answer is correct or not? the simplest thing, which is like the simple form of computing Bob can do, is to just repeat the same computation again. So now, you know, we have two nodes and we repeat the computation two times.

And today's consensus built on this concept of reaching consensus by recomputation as a method of verification. Now that is precisely the problem. Let's say we're doing a kind of a swap computation on a kind of a decentralized, choice exchange. What's actually happening in the entire network is not one node that's doing this computation. In fact, like in Ethereum, for example, you have a million small node that is doing the same exact computation again and again and again.

And this is a fundamental reason why blockchain computation are expensive resources. Now, we have a good solution, however, to solve this. This is a new concept of verifiable computing. So in verifiable computing, let's look at the same two-node blockchain here. You know, we have different roles in the entire computation process. Alice here becomes something called a prover. So instead of just doing the competition itself, Alice, in addition, compute something called a zero-knowledge proof.

And this zero-notice proof is a very magical kind of piece of a mathematical, you know, cryptographic thing that she can use to send it to anyone. And this other party can simply verify, okay, the computation is actually done correctly without redo the computation again. And, you know, the important thing here is that the verification process of the zero-notge proof, you know, it is extremely low cost and low latency. So, you know, this kind of a new paradigm fundamentally decouples us, you know, sorry, decouples the computation itself from the verification process.

And a very nice property about zero logic proof is that no matter how complex your original competition is, you can always compress this very complex computation down to almost exactly the same kind of proof that can be verified with almost the constant time with extremely low latency and low cost. So, you know, so now you may ask, Wait, wait, ZK, we have been hearing about ZK for like, you know, ages as well.

But why now? Like, why we couldn't do this before? The key thing here is about the blue line, kind of a blue computation cost here. You know, a couple of years ago, like even just two years ago, the difference between doing the original computation and generating a ZK proof of arbitrary computation, the gap of that is about a million times. So what this means is that, you know, if you're doing this kind of thing in a lot of, you verifiable computing way, you're not even better off

than just like doing the replicated computing. You're basically kind of repeating the same kind of computation about a million times. But the key thing here is that, you know, over the last couple years, and especially this year, we have achieved so much progress in ZK that this gap between the original computation

and the proof generation shrinked by orders and orders of magnitude. Now we're about 100 times to a thousand times you know, kind of you know, inflation comparing to the original computation if you run the computation in ZK.

And this becomes extremely useful because instead of replicating this for a million times, you can just like spend a thousand more kind of a computation and gave like the same kind of a computation result to everyone.

So roughly speaking, this is like the kind of innovation that enables this new trend of zero large proof and how it can help Ethereum to kind of achieve the next hundred times scaling.

So, you know, what we build at Brevis is we build this kind of a solution that as a very generalized, a smart, verifiable competing platform that allows you to all flow with very complex competition away from blockchain to this off-chain box called Brevis, and Brevis will generate a proof for that and then kind of send the proof back to the blockchain. So the blockchain as, you know, using kind of as a validator or as a blockchain as a whole, as a smart contract, can just verify the computation easily with low cost and low latency. and then use this computation as if this actually is done on blockchain from the beginning. So, you know, just want to mention that, as I mentioned,

this technology is very used for future of Ethereum scaling, but you can also actually use this in your application today. So we're powering numerous different applications now in various different sectors, like in, you know, Metamask for Uniswap, a pancake swap, for many open Eden, Euro money, all these stable tokens, lending protocols, and PerpDaxis and all these, to enable them to build a new generation of features that is not possible before

due to the constraint of the current layer one scaling. But let me go back to kind of talk about the scaling itself. Now, you know, we say, okay, this is a new paradigm of computing, using verifiable computing, and, you know, what is actually inside of the magic box? Well, you know, the centerpiece of this entire kind of verifiable computing stack is generally a ZKVM. So for us, this is something called a pickle ZKVM.

So the interesting thing about, you know, a ZKVM comparing to your general ZK circuit is that it is built on a very general instruction set. So you can actually run in theory any kind of software you want to run in this without actually needing to understand the nitty-grated details of the, you know, zero-notge circuit, you know, development.

So this is just a comparison for the same kind of a program. If you write it in your own ZK circuit, it's going to be this complex. And if you write that in the, you know, in kind of in ZKVM, you're just essentially writing Rust. Now, with this kind of a new developer friendless, you enable something that is very powerful.

Essentially, you can write a Ethereum, a kind of full execution node. And, you know, in just russ and then run it and kind of being an entirely verifiable way. So this is how we can now use generalize the ZKVM to change how Ethereum layer 1 consensus works. Now instead of, you know, we pack one block, you know, a broadcast a block and everyone kind of verify the block execution through re-computation, now what you can do is that you

have just one node that is running the computation itself and then generally, generally, generate a ZK proof and broadcasts a ZK proof, which is a very succinct, a few bytes, you know, proof to the entire network and everyone, every validator in the network just verify the proof with extremely low cost. Now, the challenge, however, is that, you know, you need to have that performance, you know, to be kind of fast enough so you can generate proof

with the same kind of latency that Ethereum is running on today, which is like a 12-second block delay. And this is why, you know, you probably heard about like a lot of kind of a push and a movement on something called real-time Ethereum proving. That is like once a Ethereum blocks it produce, you need to generate a proof of the execution correctness within 12 seconds. So you can actually broadcast this very, you know, proof to everywhere.

And in fact, like it should be like less than 12 seconds, you know, it should be 10 seconds because of the additional kind of overheading networking. Right. So this is like why we're pushing forward and, you know, pushing the boundary of a zero-notch proof to achieve something called real-time proof. So Pickle, you know, which is the ZKVM that we build, already achieved us.

You know, with a 64 GPU, we can achieve average proving time of 6.9 seconds, which is more than enough to actually enable kind of a new kind of consensus client in Ethereum today. And, you know, so there's like this kind of a layer two, you know, beat equivalent,

which I would invite you guys to follow called ease proofs. and you know, it shows that the different performance of a different kind of ZKVMs for, you know, Ethereum scaling specifically. And the important thing I want to highlight here is that it's not only just the cost is going down, but the paradigm of the verifiable computing, especially for Ethereum and Layer 1 execution, you know, has really changed. Because now, you know, we can do parallel computing. Essentially, you have like a big block, right? So like we can actually chop the big blocks into smaller blocks called subblocks, and the,

run these kind of computation in parallel. And once you have this kind of capability, you essentially convert the Ethereum scaling problem itself also into a pure problem, a pure money problem, which was never possible before. So if you want to scale Ethereum by 100 times a day, right now, what you need to do is to just throw more GPU resources into the proving stack,

and then you can essentially generate a proof, which is still very succinct, the same level of verification costs, but for a block that is 100 times bigger, you can do it today. Now, so this is a major breakthrough that we can actually convert

the problem space entirely. And, you know, what I just thought about is that, you know, how you can essentially kind of scale Ethereum by replacing the consensus process as a, you know, computation paradigm, but we should also look at the application layer as well.

You know, how you can scale the Ethereum application using the same kind of a paradigm, because even if you scale up like Ethereum by a hundred times, you're still looking at a single-threaded machine, which cannot run faster than a single server. Right? So Pickle is fast, as we know.

But, you know, it is not fast enough because, like, we know Pickle is very generalized you can build any kind of an application on top by just writing high-level programming language code. And, you know, how do we kind of achieve the best of the toolwork. Well, this is the innovation that we introduced

that is like we need really modularity in the entire verifiable computing stack to treat ZKVM as just the glue of the computation and then allow us to actually get a lot of plug-ins to enhance this capability

for different kind of application scenarios. So for example, you know, the example I mentioned like computing a VIP trading fee discount using zero large proof, you can essentially build something called an chinked data co-processor which can significantly accelerate

the performance when you're dealing with use cases that involves unchain historical data computation and ZK proof. So, and you know, you can, of course, also enable things like ZKTLS that is like, you know, you want to essentially compute what is, let's say your, you know,

coin-based trading volume in the last month. You can essentially say, okay, I'm going to run a ZKTOS co-processor to first get the raw output from my Coinbase account, but I also want to run a computation on top of it to derive some interesting and informative data that I can use to prove that, hey, look, I have traded more than one minute, but at the same time hide important information, such as your exact kind of token trading and, you know, the details of your entire trading trace. So this new paradigm of kind of a, you know, building a verifiable computing stack with the ZKVM as a center glue and many kind of different application level co-processor

is something that we really want to promote as well. So this is precisely the reason why, you know, Brevis itself has so many different use cases across so many different, you know, domains in the, in the application space as well, because we have the capability. to do this. So some very simple examples that I want to mention here, you know, for the kind of

the VIP trading user experience customization example, we already powered this with pancake swap. So if you, you know, swapped a lot in the past, you can essentially generate the Zake proof

to show your big trader, VIP trader. And then, you know, on the flight, dynamically, your trading fee will be decreased because, you know, you actually kind of show that, you know,

you're kind of a, trader than the other different users. And this feature has worked great because, like, you know, as we see, we can provide more efficient market structure and, you know, essentially gave like better profitability for the LPs and also like much better liquidity efficiency as well. And, you know, so for the previous example, you know, you have kind of a sum of the logic that still lives on chain, but some of the logic such as user experience customization

that is living off-chain. But what we believe that a future kind of a, you know, verifiable computing application architecture will look like is that most of the computation itself is going to happen in an off-chain world.

So you can essentially now have a ZK-powered perpetual tax system that the matching engine itself is living off-chain as a verifiable computing component and all the matches and all the traits can be verified and

committed back to the Ethereum layer one blockchain. And then, you know, at the same time, what you get, because you're using ZK, is additional privacy benefit to say, okay, I can actually hide some of my orders that I want to, don't want to kind of everyone to see to avoid liquidity assassination, a liquidation assassination that we see in many existing protocols. We're actually building with one of the world leading perpetual taxes for this, on this, and, you know, releasing this very soon as well. So, you know, another kind of a benefit of ZK as we kind of transition in this entire, you know,

computing stack into verifiable computing is the benefit of privacy, right? So we recently launched this new feature with Kido that allows you to attach to your own chain historical data such as, okay, I'm a big holder of certain token, therefore my yapping should have like a higher weight in the entire ecosystem. Well, you know, but at the same time, I don't want to reveal what exactly is my wallet address. Now, you can actually use Brevis to generate a privacy preserving a task station to show that you are actually this kind of user without revealing any kind of detail of your wallet. So this, you know, ZK solution not only can be used as a scaling solution, but also in many cases can bring additional privacy benefits. And, you know, there are so many others that I probably don't have time to talk about, but, you know, Just another very big category of use cases is how we help the stable token and reward asset ecosystem to grow and also the chain itself.

So let's say you launched like a real world asset, you know, and the stable token. You want to kind of actually have your entire stable token used by all these different awesome defy applications. And, you know, the next thing you actually need to do is to run a reward system or kind of reward distribution protocols to help user to actually use your. stable token. And what we enable for these different kind of stable token protocol is to allow users generate a zero logic proof to show that, okay, I actually deposited

my stable token in this tax. I deposited my stable token in this kind of a lending protocol. And therefore, I should be getting this amount of, you know, kind of a reward from the protocol so they can have a fully transparent, secure, and at the same time compliant, you know, protocol reward distribution system that was not possible before.

And we did the same for Linia's recent launch. We're the one that is powering the entire linear the token distribution program called Linear Ignition through ZK Proof as well. So my biggest prediction here is that, you know, verifiable computing because the cost reduction and the advancement that we made in the last couple of years is going to actually take over 99% of the computation

for all blockchain applications in the next, you know, 10 years. And, you know, we're actually rapidly approaching towards that. As you saw, like, you know, I showed some of the kind of approving and performance numbers using 64 GPUs. We can today actually achieve the same level of performance with just 16. You know, this is just like a, you know, kind of a couple weeks gap between me making the slides to where we are now.

So, you know, just kind of as a closing thing that is, you know, for brevis, We have a full stack, ZK computing stack, as I mentioned, with a PICOZK VM as a glue and different kind of co-processes to handle different application use cases. And we recently also announced our ProverNet that allows you to match different kind of approving requests to different kind of approvers to optimize the entire proving flow

with high availability and decentralization. So, yeah, with that, that concludes my talk. And, you know, thank you, for... Thanks, well. Who's got a question? Who's got a question? Question down here? All right.

Thank you. Yeah, really, really incredible project and great presentation. It seems to me like Brevis could become pretty critical infrastructure. I'm just curious about what your liveliness guarantees are. So if the Proven Network halts, what kind of backup options? or like, I'd be just curious for you to talk about that. Yeah. So, you know, as I mentioned, you know, I didn't want to kind of assure too much about

the brevis specifically, but the Prover Network is specifically to solve the liveness problem that you mentioned, right? So like, you know, the idea here is that you need to open, you need to open marketplace to match different kind of, you know, application developers, you know, proving needs with different kind of approvers. some of these requires like an unching Zika data code processor some of this requires like a ZKTOS co-processor

some of these requires like full block execution so you know you really need to have like open marketplace where different type of provers can actually participate in and be matched with different kind of application needs and application requests so you know so this kind of thing is actually

very very hard to solve you know mechanics and design and game theory problem We actually specifically designed something called a truthful online double auction to do this kind of a real-time rotating auction that allows you to kind of express your proving need into different chunks and replicate them if needed to kind of maintain extremely high availability and high-liveness essentially. So that's how you can essentially solve this problem. All right, coming up next, I'm sure many of you are at least bankless of all of you. And the story of money is very dear to my heart, to Ryan's heart, to the whole story of bankless. So next up, I want to invite Luca Prosperi from MZero, who's going to tell us a little bit about money.

Luca, come on up. This talk is going to be theoretical by design. But I think it was important during this conversation, during this conference, to point a finger on some of the weird things that are happening in crypto, meaning we're spending so much time. thinking about the structure and the design of the networks we operate on from a computing perspective, and we spend zero time thinking about the financial layer that is actually powering most of them and the single point of failure that stable coins are actually creating in the ecosystem. And as we know, like the definition of money typically goes like this. Like, you know, what is money? Money is a unit of measure. We measure, we measure

the things we do in monetary terms. Money is a store of value, how we can keep the value we have. This is a story of the explosion of Bitcoin and stable coin in Argentina. Mariano was talking about, and money is a medium of exchange. It needs to be easy to transact. Now, the concept of money has been abused a bit in crypto

because we created yield products that look like money, but they're not, and other things around. So the finance people go a big deeper and say, okay, what is actually good money? And this is from one of the, I think, the best papers about money, monetary policy I've ever known by Armstrong. And so the money, actually, the second paragraph is money needs to be easy and liquid. No questions asked. You actually need to send it around.

You need to worry about how collateralized it is, what is the value of it, how you can trust the person, needs to be easy to transact. And this is proven right for the USDP, USDAs we know about, right? kind of work. You have it, you own it, and you can use it to transact on chain and off chain. But in my opinion, this is a very limited perspective of money. Because in my opinion, money is a network that connects everyone and groups of people, companies. And currently, this network is running on another network that is like the set of blockchains that we use every day. And networks have shapes. So there are networks that are shaped in different ways. and the shape of a network has some pros and cons.

And we should think about it because we are impacted by those pros and those cons. So what are examples of those networks? So we have centralized star networks, and this is the way the banking system works today. You have the central bank at the center, he's creating assets, is creating monetary value,

and he's pushing it downstream through commercial banks and commercial banks have clients. So you do not have an account of the Fed. you have an account at Barclays, which is BC, JPMorgan Chase, and those guys have an account to a central bank. So if you're cut out by a bank,

you're out of the system. Then we have distributed mesh networks, and people here, they might recognize the Bitcoin network, where all the nodes are connected with each other, and there is not single node that can cut anybody out of the system. Obviously, this network is a different characteristics

from the previous one. So the previous one, for example, is impacting the accrual of value at the center. Who makes money in the dollar network is the US government. The US government finances itself very cheap. You don't make any money on the US dollar. You get the benefit of the fungibility of the system,

but the value accrues at the center. In the Bitcoin network, the value accrues based on time. If you were early in the Bitcoin network, you have accrued a lot of value, if you're coming now, the value accrue just to reinforce the network is lower. So you can see how the topology of the network, the shape of the network,

is impacting certain things of the network itself, like value accrual. So what are the examples of monetary networks that we have today? We have centralized and concentrated networks like the stable coin networks today. You have circle, you have tether, and at the center you have the blockchains

which are distributed systems. If, like, we are spending so much time about the decentralization of blockchains, if Tether blows up, we're all fucked. If Circle cuts you out because they blacklist you, we are all fucked. So we spend so much time thinking

about how important it is to have distributed systems where we transact, zero time thinking about the single points of failure that we have at the center, sorry, at the level of the issuers. So, surprise, who makes the most money, the guys who intermediating the value.

You have dollar-packed decentralized networks. And this is the word Mariana was talking about, the previous bull market word, where you had Circle, you had Maker, it was a decentralized stable coin issuer that was somehow connected through Circle, through a thing called the PSM that was provided stability, but it was more decentralized. And who was making money in the Maker ecosystem was not necessarily a single issue or more parties. You also have now starting to appear consortium like federated networks, where you have a lot of. have the paxes, the anchorages, the bridges, saying you can issue through us many types of stable

coins. Still, it is an attempt for them to position themselves to accrue value. Most people will have given a lot of respect for the Stripe team, but most people will get the irony of Stripe creating open networks within Stripe. You also have digital, you might have in the future digitally native and distributed network and you're saying, okay, maybe we are still using the treasuries to back our money but the treasury is issued in a decentralized manner on chain

like hopefully on Ethereum and then you have systems that exist already on chain. That is still giving a lot of power to the US government but the rest of the value accrual is distributed downstream. And then the holy grail. You might have fully distributed issuance and authority networks where you have different assets

that are composing our of value and monetary monetary use cases. And, you know, the value is actually distributed around and there is not anybody just capturing most of it. This is hopefully the future we're going into. Now, does monetary

topology matter? We touched on this earlier. It does a lot. Because it impacts certain characteristics of money. And I am convinced that people do not really pay too much attention to it. Not only us, the users, but also the regulator. You have so much time

you hear so many times the regulator thinking about whether a stable coin is collateralized, is not collateralized, is well-designed, is riskier, but do regulators ask themselves how the networks of money are shaping up in the future? Probably not enough, and they should. Because the shape of the network is impacting three main things. The value extraction capacity, as we said before,

who makes the money, and you can rest assured that those who make the money, they will lobby to tell you and convince you that the way the network is is the best only possible way, but in crypto we don't take anything for granted. How is the quality of the communication spread around the network? How actually the pricing and monetary policy is spreading around the network

and how resilient is the network? And these are three components that are all important and there are trade-offs. Now, I am a mathematician, so I'm a math nerd. The good thing is that you can actually measure those characteristics, for monetary networks, in the same way you can measure characteristics of blockchain networks. So you have, for example,

what is value extraction is the ability of certain actors to extract rent because of their position from their periphery, the users. Tether doesn't share the yield. They are the most profitable company in the world. We're using it,

and we are benefiting them because there are benefits for us from a usability perspective. Now, in graph theory, you can measure this based on the centrality of the node. Like, there are many measures you can do it, and I'm not, you can use, and I'm not going to spend too much time on it. But, for example, node centrality, in between the centrality, is a measure that graph theorists and mathematicians use.

So, like, you see, for example, a score that is based on how important is a node to actually intermediateate all the, all the communications in the network. So, like, node number four at the center, as a score that is way higher than node number nine. So you cut out node number four and you're just creating a lot of impacts for the network and you can rest assured that node number four is making money out of it. So do we spend time in rebalancing those scores and those positions or not?

Probably we should spend more time. Signal communication is also important. So what is the efficiency and fidelity of the information? And again, graph theories can measure this also currently in stable coin land. So you can quantify by metrics like the path length, the diameter of the network,

how long it takes to go from node one to node N. And this is an example. So in a centralized star network, if you are at the periphery of one, to go on the other side, you have to go through a lot of hoops. And every hoop is creating noise. In the, I don't know,

in a decentralized network or in the Bitcoin network, you could actually go from person to person pretty much directly. And so it is a bit more cumbersome to build, but the signal for the periphery is getting higher quality. And ultimately, resilience. So how robust is a network in case there are single points of failure? And this is all, blockchains are all about resilience. And funnily enough, we never asked about the resilience of the monetary network we use.

So what actually, how can we measure it? And, you know, so one of the main measures is vertex hedge connectivity. So if you, in a star network, you kill the center point and then you're isolating two subsystems, but the system doesn't talk to each other anymore. In a mesh network, it's very difficult to cut it out. And these are just examples of way you can actually look at the network from a, from a resilience perspective. So everything has a trade-off. And you can see, I put here on a list, the networks we quickly discuss.

and you can see the current network we have, it's pretty good on signal quality. There is not a lot of price variability on a USDC and USDT. It's pretty bad on resilience. If tether goes down, I don't think there are so many places to hide. And I'm not going to say, I'm not saying that these guys are going down necessarily at all. I'm just saying that the single point of failure is a real risk.

And value extraction is very high. There are only very few people, very few institutions that are making most of the money on the assets that we're using on the rails. And you can see how these things evolve. And it's a trade-off.

Like you can go, you can improve the resilience, you might impact negatively the signal quality. But it is an evolution we all need to go through. And ideally the dream is to have max,

max, there is a type, is obviously minimum value extraction. There is minimum value extraction, max signal quality, max resilience. This, in my opinion, it is a challenge we should take on for next 10 to 20 years for not having the risk of very distributed computing systems, but still very, very centralized financial systems that we tend to ignore.

So I don't have a silver bullet on this. Obviously, as one of the founders and the CEO of the M0 project, we built the project to create distributed monetary system. We come from the ethos of the maker ecosystem as well. But I just wanted to stress the point here, especially in this conference, how important it is to look at monetary networks in the same way we're looking at computing networks.

Otherwise, we're just over-optimizing one layer of the stack and we are completely ignoring the type of systemic risks and even distribution of profits that we are living for others. if we keep going this path, creating defi will simply mean allow old school regulated financial institutions to open branches on chain. We're just making their job easier.

And although we see so many institutions and so many traditional financial players like Stripe, like GP Morgan, like the asset managers getting very, very involved in a blockchain, a defy, we should never forget that we are not here to actually make the job of Wall Street easier and making Wall Street banks able to sell their products across the world without intermediaries. We want to make our life easier as users and owners of the value that we transact.

So I hope we all spend more time in thinking what type of future we are designing for our finances and just not blindly using the products that are easy to use nowadays. But thanks a lot. Beautiful, Luca. We have time for a few questions. One thought that comes to mind, Luca. Something we've referred to on bank lists is the cantalon effect. And I think there's a relationship between the topology idea of money and the cantalon effect,

where the central part of the topology is where money issuance happens. And then as you get further out to the periphery, the effects of cantalon effect take over where the value of the money dissipates by the time it gets to the margins. And it is local in the center. I think that's a very relevant story for Argentina as well because the story of inflation is synonymous with the story of Argentina. We got time for a few questions. Got one right here.

I completely agree with your talk. Do you think we can do even better than the last slide you showed because you still had basically the treasury and like Bitcoin, which you can't really have Bitcoin on the Ethereum blockchain without like wrapping it? So that's a point of failure. And then the US government is obviously a point of failure for treasuries. And also the peg itself is somewhat a point of failure, if that makes sense, because you're relying on the Fed to not like hyperinflate or do anything like that.

So I know projects like Rye were trying, but do you think that's actually the achievable goal to have like just ETH and then not even have the peg to the US dollar, something like that? Yeah, I think this is a multidimensional problem. I think that is a spectrum. We will go further and further in decentral. the sources of value that we are considering trustful, and we consider the US government a source of value.

And so it is a spectrum. I think it's also a spectrum of how we are, what economy we're living in. I don't think that, I mean, the US government loves the idea of exporting, continue to export the dollar across the world, but we are not living within the US monetary system, all of us. So I think that these two, these two phenomena will go in sync,

meaning that the monetary system that we are living in will expand, will become way more global. So also the pegging, how we are measuring our purchasing power, will expand, it will become more like a basket, a basket pack rather than a single one. So in my opinion, this two phenomena will just go naturally in sync. Like we will have different sources of trust as base assets. So today is the Treasury government,

the US government, is going to be U.S. government, decentralized assets, private assets, but also the way we are indexing, how we are measuring it will change, because we are not all living in the U.S. and for as much as the U.S. would like it, this is not the only irrelevant currency in the world. But the interesting thing is that we are now, this is possible, this is happening. So in my opinion, it's just how quick we are getting there. One more question right here.

Yeah, how much of the returning crypto is a function of counterpriced, counter party risk. Seems that a lot of people are getting returned just because they're accepting counterparty risk and not because of any alpha being generated. And that's why Tether makes so much money because people view it as having very little counterparty risk. So the senior age on counterparty risk is the story of money as a whole. It's not a crypto story. You accept dollars in transactions because you don't want to run the counterparty risk of your, there's no price discovery in monetary transaction. So it's not a crypto history. It's a money history. I don't think it's very different. Now, there are different layers of counterparty risk that crypto can definitely reduce. And that's why

that's why Defi works so well. So I think that there is, counterparty risk reduction is a huge component of the use of stable coins. But I don't think it's very different from any other type of money. It's the whole question of money. Like you don't ask questions in terms of credit risk and counterfeit risk where it's coming from. You're paying me, I'm fine. Now, there is, risk doesn't disappear, so we are all running Tether's counterparty risk or circles counterparty risk.

And I think many of us have lived through the Silicon Valley Bank days, at some point we didn't even know whether the counterparty risk we had was real or not. So I think this is a history of money in general. Thank you, Luca. Round of applause for Luca.