On Ep. 2 of Agentic Commerce, Cuy Sheffield, Head of Crypto @ Visa, and Bam Azizi, CEO & Founder @ Mesh are joined by Louis Amira, Co-Founder & CEO @ Circuit & Chisel to discuss the evolution of agentic commerce protocols, explosive growth of the x402 protocol and more!
On Ep. 2 of Agentic Commerce, Cuy Sheffield, Head of Crypto @ Visa, and Bam Azizi, CEO & Founder @ Mesh are joined by Louis Amira, Co-Founder & CEO @ Circuit & Chisel to discuss the evolution of agentic commerce protocols, explosive growth of the x402 protocol and more!
Timestamps:
Tokenized is sponsored by Visa
A world leader in digital payments, Visa is bridging the gap between traditional financial institutions and innovative blockchain networks, helping players in the payments ecosystem navigate the ever-evolving world of tokenized fiat currencies with confidence and ease. Learn more at visa.com/crypto.
Tokenized is also presented by Mesh
As the first global crypto payments network, Mesh connects over 300 wallets, exchanges and payments platforms, and enables anyone to pay and get paid instantly, anywhere, in any asset. Mesh makes digital transactions seamless, secure and universal, fuelling the next era of agentic commerce. Learn more at meshpay.com
***
We’d also like to remind you that the views or opinions of our contributors today are their own and do not necessarily reflect those of the companies they are representing. Nothing we say should be taken as tax, financial, investment or legal advice, do your own research!
Music by Henry McLean
Cuy Sheffield 00:00
Kai, welcome to tokenized. The show focused on stable coins and the institutional adoption of real world assets. My name is Kai Sheffield, head of crypto at visa, and today we are back for another episode of our agentic commerce series. Back again. Stepping into Simon's seat today is my co host, the CEO and founder of mesh. Bam. Azizi, welcome bam. Great to have you
Bam Azizi 00:33
back. Thank you, Kai for CO hosting this I'm excited for this
Cuy Sheffield 00:37
episode. Awesome. And joining us today is Louis Amira, co founder and CEO of circuit and chisel. Awesome to have you here, Louis. How are you doing
Speaker 1 00:47
today? I am great, excited to join you guys. My agent didn't join the chat, but that's okay. We'll solve that next time.
Cuy Sheffield 00:53
All right, so two quick bits before we get into the content, I need to remind you that the views or opinions of our contributors today are their own. Do not necessarily reflect those at the companies they represent. Nothing we say should be taken as tax, financial investment or legal advice. Do your own research and a reminder, this podcast series is made possible by mesh and visa. All right, let's jump in. Louis, why don't you start by giving us an overview of what you're building. What's the backstory? I think you've got a really interesting history at the intersection of crypto and AI like, how did circuit and result come about?
Speaker 1 01:25
Yeah, well, thank you guys for having me. We'll try not to make story time too long here, if you rewind the clock way back, I had I worked at Google for a while, sort of read the Bitcoin white paper, did a bunch of interesting things, but didn't have the full commitment to go make crypto and fintech. My full career, until many years later, did some interesting things at Google that might end up becoming relevant. Saw DeepMind up close when they were acquired, watched some of the AlphaGo things happen here and there. Watched a bunch of very smart college and Google friends go to stripe, and was like, Man, those Collison guys really know what's going on over there, ended up working at chain link, which was fascinating, which we can come back to, and then joined stripe the week that Elon Musk bought Twitter now x and Stripe launched a crypto payouts partnership with Twitter, which was fascinating to land in the middle of about a month later, the Taira Luna stable coin blew up, and I got to write the internal recap on what that was. And then six months later, FTX blew up, and I was like, what I was at Google. What am I doing here at this FinTech company and working in this crypto space? Why did I make this my career? And fast forward a little bit, and to both Patrick and John Collison will gabrick and a bunch of others credit. They were pretty committed to crypto, actually going all the way back to their investment in stellar long before I had figured out which way was up. This is 2013 I want to say, accepting bitcoin here and there. They decided to put a crypto team together because they said there's something happening here. There's something that will be very transformational in terms of what we're trying to do. We should be experimenting with it. We should be at the forefront. That's kind of like the ethos of stripe is, like, we'll listen to people, but we need to go see for ourselves. And that led to a bunch of interesting conversations, sort of, where we all met along the path 2023 there were a lot fewer tourists wandering around the crypto space. We had some meaningful conversations. And I think towards the end of 2023 and into early 2024 companies like bridge and a handful of others, like the mesh folks, you were building things that actually allowed people to do real things. You were separating, as John Collison likes to say. And we talk this audience is particularly conversant in money 2020 so I think his famous metaphor is like walking through the casino to get to the real business at the back. Like that was 2023 and 2024 fast forward a bit, 2024 bunch of things happened that were very positive. We launched pay with crypto, we made the bridge acquisition, and then sort of full steam ahead, and many others in the space, and I would say Congress overall, Trump embracing it like there were a bunch of things that were very positive alongside that, to make a much longer story, very slightly longer, I had been working on the AI side of things, running stripes AI partnerships for the first bit of its formation, and then exploring what we should be doing. I had been pushing internally saying, Hey, if you thought that we were right two years ago with stable coins like I can't express how big of an opportunity AI is and Stripe was using it for purposes of making the docs easier to navigate and a handful of other things. End of 2024 I was having some conversations with the leadership team on the partnership side, some of the product folks looking at like, how do we start projects internally? There's now an experimental bets team at stripe that didn't exist at that point. The whole premise was, there's an intersection between crypto and AI, we actually would argue this was before agents were like a commonly used term. Agents are going to go out and do a lot more. We need to get out ahead of that. And there was so much low hanging fruit for stripe to pick up around the AI space, similar to what was happening on the stable coin side. My co founder, David Noel, Romas, DNR, hopefully none of the med. Cool people care too much about that nickname, but DNR and I were sort of like, we're not all that interested in grabbing a lot of the low hanging fruit. We want to go a few years further out on the curve. And that led us to a lot of what we're seeing a lot more of today with all these protocols and some of the things we're starting to build, actually building the agents, doing the interesting things. And if you rewind back to crypto in 2022 and late 22 and early 2023 okay, there are a bunch of tourists, but who's actually doing interesting things, who's moving volume, who's building agents, who's trying to run into these corners? That's a lot of what we've been running into the last handful of months, since we left, and we can get more into it, but that's the sort of arc of the company we left January 2025, so earlier this year, bunch of great investors, bunch of great partners, Stripe ended up investing, which is a very cool, surreal experience to go back and then say, Yes, complicated things, because investors were saying, We're the stripe for agents. And I was like, I don't know if this is a good or a bad thing for that. It's not actually what we're trying to do. But anyways, long explanation, but that's what we set out to do. And I would say that the agentic space, while much more hyped, the actual building and utility and all that sort of stuff is still probably back in the 2023, crypto and stable Coin World, like we know there's something there. We don't know what it is. There's a lot of people that are enthusiastic, but nothing that we all use on a daily basis. So that had factual headlines are written about so long intro, but that's the last decade in a nutshell,
Cuy Sheffield 06:16
when you started the company, what was the state of agentic commerce, and particularly agent to agent commerce. And then, if you fast forward today, you've got a new protocol being announced every week. You have x 402, now making headlines. You got AP two agentic commerce protocol. We announced the trusted agent protocol. And so help ground the audience with what existed six months ago. Was there anything and like, what's happening now? And how do we help make sense of the different protocols, and then where your company and your products, where you fit within that landscape? Are you building another protocol? Are you building products around those protocols? Like, help us unpack that?
Speaker 1 06:59
Yes, many questions that can all hopefully be boiled down to a simple one. So roughly a year ago, anthropic launched the model, context protocol, and that was the first thing that you could shorten that to say, like agents now have tools. It used to be. Developers had to go to an API. They had to get it exactly right if you miss a semicolon or your formatting is wrong, or bad news everywhere. Mcps are much more forgiving. The MCP framework, I guess, or protocol, is much more forgiving. You can use natural language, you can do a bunch of different things, to package things up in tools and go from there. The original insight was, wow, these tools are going to be really interesting for agents to be able to discover and then pay for. Agents don't have the ability to go out and express their identity. They don't have the ability to pay for discovery is one problem, but they couldn't go out and access any of these tools. So end of last year, as we were forming this thesis and leaving, there are now agents. There are now tools. Really interesting. How are we going to all go figure out what these tools are? How are we going to recognize them? How are we going to pay for them? To us, it was obvious that it was going to be stable coins, because you need very low transaction fees. If you're going to be paying two cents for a thing, or five cents for a thing, or 10 cents for a thing, the visa folks and everybody else could get into a lot more depth on batching and all sorts of other things that make this possible moving forward. But to my small brain, it was all right. Well, stable coins, like smaller transactions, this will make sense. So dozens of people had similar ideas. They were all sort of aggregated at specific companies. Coinbase, ended up launching something maybe two months after we left stripe to start working on a lot of this stuff. That was the x 402 protocol. We can get into specifically what the x 402 protocol is. But for folks who've ever seen the 404 little dinosaur in Chrome like this page doesn't exist. Or the if you really geek out, you can go into the two hundreds, the four hundreds, that all these other things. There's these little sort of like things that servers and clients will end up writing at the top of the paper to make sure that things are working properly. Gross oversimplification, but the broader point is, like, it's a way for everybody to be literally on the same page and expressing what needs to happen. We were like, oh, that's an interesting start. Nobody cared for probably four months. And Coinbase talks about this, I think they launched in March or May, or something like that. And literally, nobody cared, kind of the same as MCP when it first launched, for that first month or two, we were sitting there going, Okay, we were doing a thing, kind of like x 402, maybe we don't need to work on X 402, we had some, I would say. And Coinbase is an investor in us, so I should be maybe more careful than I am in saying this like we had some concerns. We had some incentive misalignment concerns. We were like Coinbase doing this in theory, if it takes off, others will want to expand and make this more neutral and all that. If they don't, this is a great way for Coinbase to start charging money and make sure that it's USDC on base and all the value accrues to them, which I don't think is a cynical way of breaking it down. But going back to the 90s, that's exactly how the modern browser ended up working. There was a group in Minnesota they were going to charge and then Tim Berners Lee and CERN, hey, HTTP is the free way to do it. And luckily for all of us, HTTP ended up working out ad businesses got kicked off. But that's a whole other tangent. So anyways, getting back to, like, March to April and May, our observation was okay, well, Coinbase is going to do a thing. We'll just say, based on who we've known for the last bunch of years, like, it was pretty obvious others were going to do their version of this. And we're like, if there end up being three or four or six or 15 different protocols, like, we kind of need them all to work well together. I don't think Coinbase launching something is going to stop stripe from working with open AI to launch ACP. I don't think any of them are going to scare visa out of launching the trusted agent protocol. I don't think that's going to stop MasterCard or AmEx, or anybody else from stopping launching their own thing. And the worst world would be, we have 15 of these different little kingdoms, and they can't really, like you can't travel freely between them. So our approach, which ties back into my background in the partnerships world and my entire career, as well as DNR, my co founder, was sort of like, is there a layer above or below all of this that can enable everybody else? And everybody should have the best the go on whatever path they want. If something is strictly better, it's faster, it's safer, it's more comprehensive, whatever. Like, great, everything should default in that direction, but we need a neutral way where all of this is low cost, low friction, whatever for that to work. And people looked at us like, you know, that's not a business. It's like an open source project. You're like, giving things away for free. And we're like, yes, it's not your job to ask us how we make money off of this. It's our job to deliver a product that actually works and goes from there. So we can get into each of the individual protocols. But the parent company, circuit and chisel that we created atxp was, I hesitate to call it another protocol. I mean it is by definition. But the aim is to be compatible with absolutely everything, so that there's an easier front door to walk through to get access to all of the other protocols down the line. So we hope they all fit in. We have a hunch that if Solana launched a competing one to Coinbase, or if Adian launched a competing one to stripe, they might not play nicely together in the short term. And we think it's our job to go try and make them easier to navigate in a drop rate makes
Cuy Sheffield 11:57
sense. Bam. How do you think about all this from your seat at mesh. And particularly, it's interesting, if someone talks about, okay for agentic payments, you could use Fiat rails, or you could use stable coin rails, but then stable coin rails, it's not monolithic. It's a which stable coin, which blockchain. And so you could then have a bunch of fragmentation around different standards that are focused and optimized for one of the other like, how do you look at the crypto ecosystem trying to fit into this use case?
Bam Azizi 12:26
Yeah, so fragmentation is a dear topic to me, so I keep talking about fragmentation. I'm glad that Luis is doing what Luis is doing, because we don't want another fragmentation on the protocol side, because we already have enough wallets and exchanges and different networks, different tokens, different stable issuers. And I wish that we were living in a war, that everything was HTTP, that's it, right? But that's not the case. And I think that's one thing. The second part is the user experience, right? Fragmentation directly changes how you build for users and how the user experiences the technology. And I think if you really want the next billion users to join this momentum, both in crypto and also in the AI agent, you need to make it as dumb as possible, as easy as possible, so everyone can be able, like the barrier should be lowered as much as possible, so anyone can use it. So I call it the grandma test. So if you have a piece of technology, if your grandma cannot use it, then throw it to the trash can, because you're not going to get the adoption from the majority of the users. So I'm really curious about that neutrality that you mentioned Luis, like, if Coinbase is doing or stripe is doing, is it something that PayPal is going to use that, or alien is going to use that, or how you can build that community around it? Because we are, like, it's super painful on the crypto side of it. And I will tell you, after you mentioned this, about, like, how we are solving it on the crypto and stable coin side. But I want to hear your opinion about, like, how this neutrality will be formed around Coinbase X 402,
Speaker 1 13:55
so Coinbase started X 402, you need a couple Coinbase specific things, the Coinbase, developer platform, when you get into facilitators and all of that, which they've been pretty vocal that they are going to move out into a foundation. They're moving it in with Cloudflare. They're talking with a bunch of others. I think they announced something specific with Cloudflare. And we know the Cloudflare folks quite well, like, if you look at our funding announcement, the CEO of Cloudflare was speaking about us. So we we know, I would say both sides of the equation, and really respect and hope that that's the direction things go sooner rather than later, because it would be nice to have more folks around the table to say, You know what, we all agree we shouldn't have 15 different protocols. Maybe X forward to in its current iteration is not necessarily like the perfect one. There's a v2 that they're working on Hey by v4 whether you're tether or stripe with tempo or bridge or privy, thinking about other wallets like none of this should actually matter to the people that are going out and trying to create agents, let alone all of the FinTech infrastructure that they should not care about, like people that. Sell rubber ducks. Go to Shopify, to sell rubber ducks, they don't go to find out, like, well, who's the best payment processor? Is that a stripe thing? Like, which cards should I accept and do I take crypto? Like, they don't actually care. And our the reason that we left was there's a bunch of things that need to be packaged up and put sort of in a bit of a black box, or an easier to access thing hidden to some extent from the users that actually want to go build agents or vibe code things on lovable so like, luckily open AI and lovable and cursor, and basically everybody out there is trying to lower the bar so that it's easier and easier to build these agents. We think that it should be easier and easier to not care about all of the things behind the scenes in terms of what the different protocols are. So I would say our grand vision is we need to be careful, because we have a bunch of investors that are all the different blockchains, we more or less created this foundation that they're aiming for intentionally on our side, because we want to say we are not going to favor a specific blockchain. We're not going to favor a specific wallet or stable coin or any of that sort of stuff like we want them all to flow freely to the extent that visa is tokenizing cards, and that's the cheapest, best way to do it, and that's more competitive than using USDC on pace, like the end developer and consumer down the line and eventually another agent that's getting paid by an agent, like they absolutely don't care how the money shows up. Can it get here quickly? Can it get here with no fee? Can it land in the thing that I want? Like, great. I don't actually need to care about most of that stuff anyways. That's our sort of opinionated take, like the technology is now there to enable each of these things to happen. And I think Zach Abrams from Bridget said in the podcast, like anybody that has money at rest should be thinking about stable coins, and they're probably going to be opinionated about who gets the interest. Okay, great. The interest should eventually get shared back with whoever is holding those dollars at any point in time, and we can speed things up along the way. And that's kind of, again, the ethos like, there's a lot more friction and resistance in the system than there should be, and we're trying to remove it. So hopefully that's a somewhat non answer to your question, because it's not entirely clear to us how we work with Google, with their very early AP two, which, like, kind of talks about X, 402, but also mentions that now in the next sentence, and like, talks about picks and UPI and we're sort of like, All right, well, I guess they're kind of heading in the same direction. But anyways, we're working with everybody to try and understand their protocols, to understand them and
Cuy Sheffield 17:14
fit them better together. As much as many of these protocols are designed to be somewhat payment method agnostic or blockchain agnostic, like in principle, and it seems like it's going to take time for like. What that looks like? Isn't it fair to say that some of the protocols seem more focused on specific use cases than others. When I think about a lot of the work that we've done at visa, it's been very much on agentic commerce, in the context of a consumer delegating access to an agent to purchase something at a merchant, and that's where we think that there's a ton that we can do with Agent friendly cards and tokenized credentials and like all these things. Then there's a whole nother set of use cases that are like B to B, agent to agent, transactions of a developer, like buying a resource like very different than a consumer booking travel. And so do you see the protocols optimizing for or being better designed for certain use cases than others? And then, as you look at what you're building in AX TP, what are the specific use cases that you're looking to solve for, particularly in the early days that you think will be the most relevant. And like, what are the capabilities that you think are be important
Speaker 1 18:27
for those use case? Yeah, even the essence of that question is, like, I think to take a step back, and probably all the listeners are thinking the same thing, I don't actually interact with any agents on a daily basis, whether it's at work, whether it's at home. We could define agents, and there's a bunch of competing definitions, but most people think about their agent as chat, GBT, maybe if you're a power user, you're talking about Claude and anthropic, and you have Gemini and all this other stuff. But like, your agents aren't actually doing anything. They don't exist really. There's a bunch of developers over the last few months that are creating them, and we're trying to figure out what they're used for. A lot of what people call agents are really just workflows that are a little bit smarter, because AI is like, there's GPT five plugged in behind them. What we are doing is we're actually trying to build agents on our own. So we have a few engineers. We rapidly prototype a bunch of things. We use them for things internally. We have some pilots going with public companies that I can't really get too far into. But the whole point is like, what are your problems? Can we actually solve them? Whether they're an agent or a web app or any number of different form factors? Doesn't actually matter. What we view as the future is that at some point, you're probably going to start to think about them more like another human you work with, or collaborator, or whatever else it might be. And to answer your specific question, I think you were asking something along the lines of, well, why is Google doing ACP? And their demo was buying a refrigerator from Lowe's. And we've got other B to B ones that are like, invoice factoring and like, how do those work? And the honest answer is, like, we go to these hackathons, we try and. To help people in building things. We talk with individual companies, and we're like, Well, what have you done? And they're like, Wow, we haven't done anything. We've read the blog posts. Okay, well, then we'll go to, I won't incriminate any particular big companies, but we're like, Well, what are you using them for internally? And it's like, oh, well, we like, book our travel faster. And sort of like, okay, can we collectively look at your P and L together and figure out where you spend a bunch of money? We could do interesting things like, look at how many subscriptions you have to Bloomberg and like, try and cut half of them or something like that. But it's much more interesting to say, what are some things in the bottom of your product roadmap that you can't get to can we use Devon or Codex or cloud code whatever, to vibe codes into that stuff into existence and see how that goes. So my suspicion is that many of these protocols originated from each of the companies to solve slightly different things that are specific to them. And stripe and open AI is a good one to pick on, because I angels and investors at both and we like both of them, I gave both sides a bit of grief after stripe and open I announced it. I was like, you realize that was really just instant checkout? Like, you deleted a couple screens and, like, more or less, that was a human driving an LLM to find some products and then remove the checkout. Like it's gonna be much better. I'm like, I don't doubt whatever is going on back there. I just don't think that meets the bar for agent tic commerce. Like it wasn't an agent sending me a text message with some products saying, Hey, is it good react with the green emoji and iMessage and that's the checkout. Like, that's the world we expect to come and Stripe knows this open. Ai knows this. Like, there's some steps, there's some fruit to grab that's sort of low on the way there. But these tools are so new and not many people have actually built agents yet that I don't know that these protocols are actually close to their final form. They're just close to wherever that particular entity had a right to win, if that makes sense,
Cuy Sheffield 21:48
yeah, but I remember when when you came by our office a few months ago, was there's this idea that resonated a lot to me. Of you have a bunch of people building protocols for agents and infrastructure for agents, but one of the problems is the agents don't exist yet, like they're just they're not real. It's just very hard.
Speaker 1 22:07
You have 20 stores selling pickaxes and shovels around gold mines, and there's nobody walking into the stores, and we're like, wait a
Cuy Sheffield 22:12
second, it's very hard to design a protocol for an agent if you've never built an agent. And then as agents emerge, there's also the question of, okay, whoever builds the agent, how do you go and convince them to use your protocol versus someone else? And so if I understand correctly, like the strategy of the company has been you have a protocol, but then you're not waiting and just going out trying to convince other agents to use your protocol. You want to actually build the agents yourself to demonstrate how the protocol is going to work and to bootstrap the protocol. And so could I think about like the company is, like the protocol is open source, and then there's like a dev shop to like, build agents that add value to it, and like how you see those worlds coming together. And like you think that's you actually need agents to exist yet to even figure out what the protocols should do with it. Yeah, 100%
Speaker 1 23:07
we started off very early in our journey, just rapidly prototyping using lovable DNR. My co founder was using cursor. He was building them the right way. I was vibe coding them on replit and lovable and every other thing under the sun, because I'm slightly less technical, but also more indicative of the mass market. And I went, Man, these things are getting a lot better, but they still can't. I can't go generate an image or a video on Google's video model without creating a Google account, getting an API key, all of that and like, that shouldn't happen. I want to make a music video for my son's birthday party. Like, I should just be able to ask my replit or vercel or whatever agent to go do this and, like, it couldn't for a bunch of reasons. We ended up building our protocol, x4 two got launched. We ended up getting funding because people were like, This is an amazing protocol. You guys are onto it. And we're like, I don't really know that the protocol is the thing that necessarily makes a bunch of money in the grand scheme. Okay, sure, if you want us to go pursue that. We pass forward a few more months. It plays out that everybody has a protocol. And we're like, great, this is the outcome we wanted. Can we go back to what we were saying, which is, we want to be the layer that connects all of them, and then let's take advantage of being early. We're going to go out and build agents. We're going to run some pilots. We're going to probably make some money running the pilots. And then, by the way, if we find agents that are actually worthwhile, here's a take for everybody. I am pretty confident that by July 1 next year, some individual agent, probably wrapped in a company goes and raises money at $100 million valuation, there will be funds by the end of next year that are set up to invest in agents, because they are creating so much value and able to capture them. So anyways, I think you summarized it quite well in saying, like there's a protocol we are investing in the protocol. We want there to be at least one neutral one neutral one that interacts with everything and plays nicely with everybody else. And we currently think we're the only one that's actually all the way there. It leverages X 402, it plugs in is compatible with everything else. But in the meantime, before any developers show up at the party like we should just be building them ourselves, which is what we've been doing all. Longer spending even more time
Cuy Sheffield 25:01
on bam. Are you buying that prediction? Agents are gonna be raising money directly next year. Like, how do you look at all this?
Bam Azizi 25:07
Not next year, but maybe next five years. But maybe Andreessen Horowitz has a different plan, but I don't know. Maybe Luis has insider take. But anyways, I want to talk a little bit about where you draw the line in terms of, like, when you want to basically monetize a protocol when it needs to be neutral and free. So we know that opening in nonprofit doesn't help. Like, look at open AI, right? It was nonprofit now it's for profit, with all all the discussion around it. But like, how can Luis Amira or Amira create something that basically is widespread, adopted by the industry, by people who are using agents or building agents, but at the same time, you can build it for profit and make money out of it, like where you draw the line.
Speaker 1 25:53
So I think there's a bunch of services, and this is the brilliance of Coinbase. They have the facilitators. I explained it to somebody else yesterday, is grossly oversimplifying. Like that is escrow. You should have escrow on a bunch of these transactions, especially as the dollar amounts go up. Okay, I'm going to give you the 1000 bucks for the whatever thing, but, like, I kind of need to be sure that you're going to bring that thing back. I don't want to just send $1,000 for a flight that turns out like that wasn't actually the, you know, Delta giving me the flight or the seat back, or, I think people understand that circumstance. So that's going to cost some money. Okay, how did chargebacks end up working if you're not happy with the quality, who's the right network, if you're a new merchant or a new agent, like, how do you go get any of that trust? Should you be able to deal with high dollar amount transactions or high volume transactions? This is where I think the visas and everybody else of the world. I don't want to put too many words in Kai's mouth, like they're a technology company more so than a financial oriented company, like they provide value added services through technology. And I think there's absolutely that same right to win here. If you're asking if that's something that we're trying to chase, I would say that atxp, like might have a right to play in that world. I don't think we're trying to go out and figure out what the best wallet setup is. Or should we be doing hosting of individual agents? Should we be making the passport equivalent? Should we be building escrow like we want to have a very easy place for others to plug into and build, and there should just be far more transactions out there, like we wanted to be agent builders, not protocol builders, I think, is the way of putting it. So we are not out here trying to make a $10 billion protocol. I think we're trying to go out and make it to where agents are easier to build. We have some ideas about the platform we're using and creating in service of building our own agents. We think that's really valuable. And we can get into some of the pieces of that, some of it are still a bit confidential, but if there are more agents that are more freely able to transact, we suspect we will be able to monetize that a bunch of different ways, in terms of the agents we create. But we think that the cost of a lot of that sort of stuff, to just send something between two known entities should be very low, discovery, reputation, all of that starts to cost more money, and that's where I think a bunch of others are going to innovate and play
Cuy Sheffield 28:06
before we move on. Let's hear from the sponsors that make the series possible.
Sy Taylor 28:14
This episode, if it's not obvious, is brought to you by our friends at visa, a global leader in payments. Visa's tokenized assets platform, vtap uses smart contracts and cryptography to help banks bring fiat currencies on chain. Vtap allows financial institutions to issue Fiat backed tokens, improving financial efficiency and enabling programmable finance. You can check out the links in this episode's description to express your interest in vtap. This episode is brought to you by mesh, the first global crypto payments network. It connects over 300 wallets, exchanges and payments platforms. Mesh enables anyone to pay and get paid instantly, anywhere in any asset. Mesh makes digital transactions seamless, secure and universal, fueling the next era of agentic commerce. Learn more at mesh. Pay.com forward slash AI
Cuy Sheffield 29:19
X 402, it's been hot lately. It seems like the past few weeks. I remember I was talking about it a few weeks ago, it was just, like, very theoretical, like, oh, like, you know, it's the story of the original sin of the internet. It's an interesting story. But then you look at the data, and all the data is public because it's on chain, and you're like, there's been $500 the last week. Nobody's like, actually using this thing. Then all of a sudden, the usage explodes. I think there have been multiple days that have been 500k plus in volume a day. Shout out to Merit Systems X 402, scan. Amazing resource to follow. What's happening in real time? What happened? Where did the excitement and enthusiasm? Enthusiasm come from what's driving the growth of x 402, now. And like, how do you separate some of the noise from signal? And like, real things that are being built that give us a view into what some of the use cases will be in agentic flows.
Speaker 1 30:14
I don't know what the initial spark was. I have some suspicions. I would say it works. It is easier to get a facilitator up and running. Like there is a lot of Coinbase developer platform work behind the scenes that org that has been put back together, reconstructed under Brian, Eric and Alec and a bunch of others that are amazing over there building it. Like they have a thing that works well, and they know it works well. They have a bunch of very excitable crypto folks that if there's a chance to make money, we'll go figure out how to make money off of it. A lot of it was meme coin driven stuff. Some of the early demos and things they talked about at hackathons were very cool. You go ping an API and you get back ping tokens. I would just say, like broader crypto Twitter very quickly took over, because it was very reminiscent of like the ICO boom and some of the we'll call them alt coins, because I don't remember what this podcast is allowed to say. Little while ago, a lot of enthusiasm behind it. It caused a bunch of conversations. We went and built more on X 402, ourselves. So we're actually, I don't know when this is going to be released, but at some point the last two weeks, we were at the top of the 24 hour, seven day, all of those different charts, because we actually set something up to where developers could end up getting tokens on our platform at a slight discount. So like, if you're going to end up making images and videos and LM credits, we were like, hey, you know what? We want to show people how this works. We want to use XOR two. We want to support the volume. We very quickly were at the top of a leaderboard because people were like, Oh, wait, this is kind of an arbitrage. It's like, well, yeah, that's the whole point of this. So we, to some extent, contributed to a different version of the meme coin trend. We like to think that we were doing it by giving people discounts, and sorry, investors like, we used a little bit of your money to give away, like, dollars for 97 cents or whatever ended up being. But it played out exactly. I don't want to put words in Brian Armstrong's mouth, because I think he does a great job of speaking on his own, but I have a lot of respect for their team's ability to say, if we go get 100 times or 1000 times the traffic and eyeballs on this thing, 10s of eyeballs, might actually end up going and building the real thing, the use cases that actually matter, we might be able to get the merchants talking about it. And we've been in some rooms where it's like, there are very legitimate, I don't wanna give away any companies, but like, very legitimate folks thinking about this now, not because they've seen any volume. They know that a bunch of its meme coins, but they're like, it seems like it now works. So we should take it seriously, especially if everything coalesces around x 402, and our approach is like, yes, we are landing on something that works. Let's not forget the second half of that, which is make sure that it supports all of the blockchains, all the tokens and Coinbase is not they said that they intend to do that. I do believe that they are going to do that. I think the ecosystem is going to more or less require it as the foundation gets set up and they need to play nice with everybody. There's a ton of initial volume. I don't want people to go out and lose money. I think we said at the start, this was not financial advice. My advice would be to not go necessarily buy some of these meme coins, because I don't think X, 402, is a viable thing. So maybe that is financial advice. But the broader point is, there's a lot of eyeballs, a lot of excitement it works, which is amazing. Now let's like, take it from working to working for everybody, and that's what we're trying to, like, gently nudge and contribute and help in all of that way, hopefully in saying that it does work and
Bam Azizi 33:25
it's amazing. Bam, you, you've been in crypto for a while, like we've seen this story before. Is this another like Solana? In many ways, meme coins helped to bootstrap the chain, led to it hardening, led to the growth of phantom now it's going into more stable coin institutional use cases, like, when you see something like a flurry of speculation, like, how do you look at it as like a crypto builder? Is there an interesting arc where that could turn into utility? Yeah, I think these little sparks, I would call it, could potentially ignite a bigger movement, right? I think both me and Luis believe that the future is Agent tech. There is way, way more things to do than just market arbitrage or making some money here and there, but you really need some financial incentive to bring more eyeballs and more attention to the industry. And I think what they did is absolutely smart, and that hopefully ignites a bigger movement. I personally believe that we will have 5 billion users or human connected to the internet as of today, but we will have 500 billion agents and humans, they interact like twice a day. Agents will interact like maybe 200 times a day with each other. And what e commerce did to commerce. Agent e commerce will do like 1,000x bigger. It's going to be 1,000x bigger catalyst for the whole economy. And I'm personally excited to see where crypto comes in to that picture of helping agent e commerce to become a bigger thing. X 402, is one of the tools. Is not all of it. There are more that we can do on that end. But I'm very curious to understand, or maybe the audience, to understand, why agents need to do micro transactions, right? Because that's, I think, where crypto comes in. So if you have to pay, like a half a cent, how you want to pay it, like you want to pay with your Visa card or you want to wire it, it really doesn't work with the traditional type of financial services, you have to use crypto in a stable coin, especially because it's cross border, like one agent from Philippine wants to talk with another agent in Colombia. They can't really use their local currencies, or even US dollar, because it takes time. These are, like real time type of transactions. But why agents need micro transactions. Maybe Lewis can share some insight on that. Well,
Speaker 1 35:45
some of the first ones we built were, I think many people that are maybe AI Ninjas will have gotten the minute chat GPT launched deep research. They were like, Okay, I'm meeting with this person tomorrow. Tell me everything about them. That's almost the same as going and like googling yourself once upon a time, like, what does Google know about me? Google know about me? Go ask chatgpt to go come back for the deep research report and then grade it. One of the first things we built, this is many months ago at this point, was, can we go get a better version of that produced? So if chatgpt gave you a c plus, if you guys were doing the background research on me, what's his background? What did he do? Maybe it gives you a c plus. Can you go make an agent that is an A minus version of that? Okay, well, how would you do it? There are some paywalled articles, there are some YouTube videos, there's some podcasts that have been on. There's a bunch of stuff that like either you need to pay for an individual article. This is what Cloudflare has been working on and with paper Well, paper crawls a different thing, but the ability to be able to pay for an individual article is there. I think what they would like to see in the world, and I think the content folks like eventually will agree that that is a really interesting world where you have 500 billion agents, and they all go spend 10 cents reading an article. It's a much bigger business than all the subscribers of everybody else. If you're trying to go learn about me, you probably need to go read some articles that might be pay walled. You might have an agent go out and listen to five hours of podcasts or watch five hours of videos. You could have chat, G, P, T, do it. They're basically paying the five cents in the background for all of the servers and everything else they're using. If they could hand that task off to another agent that's the best podcast distiller and YouTube video watch or whatever. You might end up in a world where somebody creates that agent, it comes out of your chat. GPT bounds would be fascinating. Great. 10 cents. Go off watch all the content. Yep, sounds good. As part of this deep research report, it's unclear to me whether open AI creates a wallet off to the side of your existing subscription, or they kick that out to you and say, Well, you have a wallet, and you go use that and pay the 10 cents and like approve it. We suspect it's more the latter, but there's no reason it couldn't be the former. And to us, it doesn't actually matter. There's some developer out there that built an amazing agent that allows you to go consume other content, plug that into the deep research side of it, and that person probably human, but maybe agent that created the deep research agent got paid 10 cents. It only cost them seven cents, and they made three cents on that transaction or that deep research report. If there are a billion deep research reports being run every week, like that is a massive business, and that's where I come back to the point about agents getting funded with 100 million dollar valuations. Like we have built early versions of that, but we can't go read Bloomberg or anything else, because Bloomberg doesn't have an option for us to pay two cents to read a Bloomberg article. So there's still a bunch of that adoption, tying it all back together with X 402 if not just X 402 if stripe and ACP, or anybody else is out there getting more merchants and tools and APIs payable in this pay per use world. That's what we're really advocating for. And we have been advocating for for the last year, not really because we're going and trying to make a bunch of money off of it. We think everybody will benefit, but because we need that to build the agents that we want to see in the world, the only people that take us seriously in building agents are people that are payments people. They're like, You don't know anything about anything else. Just tell us how to do the payment side. And we're like, Okay, fine. Like it works. To this destination. Like, we'll tell you about the payment side, but we're really just trying to get into that destination. So that's one of many, many different use cases I mentioned, deep research, but the same thing applies for if you're trying to get a workout plan together, like you could get the free one, or somebody out there might make a better version of an agent that is able to read more of the latest medical research, or can access my fitness pal and do nutrition tracking, all those little things. You're not going to pay $500 for once. You're probably going to pay per workout or per deep research report, anything like that. And that's where the micro payments make a lot more sense. It's a lot more composable.
Bam Azizi 39:35
Do you see micro payments being done with tradfi or more on the stable coin side,
Speaker 1 39:42
I would imagine that more Trad fi folks accept that the difference should be in name, not really in practice. Like, I think more of the TRad five folks, if there are Trad five folks that are not using crypto rails, like, somebody should go talk to them, find out, like, what rock they've been living under, it should be listening to the show that's. Yeah, unfortunately, we have a the echo chamber of all the people that are in the know. There are people out there that have no idea that any of this exists. But this is, candidly what I talked with folks like Kai about, I'm like, how do you view this vision, like this version of the world? And I did the same thing. I tried two and three years ago. It was like, we have this crazy set of ideas of what might be coming. And I think people on their side were very correct in saying, like, none of this has taken off. Why would we invest in that now, similar to the agent side, I think if you're a merchant and you hear this x 402, pitch, or you hear us come in, say you should make your API or store discoverable from atxp, and they're like, how many agents are going to show up to buy anything? And the answer is like, well, we're working on some. We hope others make some. They're very right to sit on the sidelines, but I suspect in short order, and this is why I have a lot of respect for what Coinbase and others are doing. This meta it is in the crypto Twitter timeline when there are better agents. And why I hope one of these 100 million dollar agents happen is because the second somebody gets funded with $100 million agents, more VCs go out there and fund more of them. It's like that first big iPhone app, like you got a bunch more iPhone developers. When being an iPhone developer, it turns out it could actually be a lifestyle, a business and a lifestyle. So it'll be interesting to hear how this ages over the next 12 months and 24 months, but it's a lot of we are very close to having all the pieces. I think all the pieces do work now. It's just kind of people appreciating that they do work. Appreciating that they do work and getting the right use cases
Cuy Sheffield 41:25
rolled out. Yeah, my starting point on this, in your question, Pam, is, I think all these things end up coexisting and like, that's how things usually work, where there will be transactions that are end to end in Fiat, that could be a visa tokenized credential, that is a consumer spending at a merchant or an agent that accepts a card, there will be transactions that are stable coin end to end, where the consumer, the buyer, has a wallet and they're paying to a merchant who directly accepts stable coins. And then they're going to be these interesting hybrids, where, what if you have a consumer who has a wallet on one side and they're trying to pay a micro payment to a merchant who doesn't even know what a stable coin is, or what. If you have a merchant that's accepting a stable coin, but the consumer doesn't want to have to buy a wallet like you need to have all of these things just work. And I think that there'll be many different flows there. The other thing I'm interested in, the way that I've thought about x4, or two specifically, is, while the vision is that this is an agentic protocol, and at some point, the examples you walk through of your agent's gonna go out and do a research report and it's gonna have to pay for other tools. Is there an argument that it becomes just another way that developers buy software? Because isn't one of the other value propositions killing the API key? And I thought that marketing from Coinbase is really interesting. Of traditionally, if you are coding up a project in your cursor environment and you need some paid service, whatever that paid service is, you usually have to leave that environment. You have to go sign up, create an account, put a payment method on file. You have to grab an API key. You have to then bring that back into cursor. And so there's this dance kind of going back and forth, where it is interesting to think about this idea of like a headless checkout that you're inside cursor the whole time. You could then discover, get access to some paid software, and it's your wallet that is the API key. You never have to leave the environment. And so even if it is, like a developer clicking buttons, it's not a agent autonomously going out and like, sorting and executing a transaction. Do you see a market where just B to B software could be purchased and consumed? Oh, absolutely, where the wallet is an API key and, like, is that going to come before some of the agentic stuff. Are you building for that? Like, how do you see that use case? It's not even necessarily like an
Speaker 1 43:45
agent Well, it's actually so it's not just limited to B to B. There's a consumer version of this. And I'll make two points here. Hopefully I don't forget when we come back to the vercel V zero, lovable vibe code version of this. But the first version to answer your question, the original name that we had for atxp. This was before x 402, this was, I think, in our pitch docs and all that sort of stuff, dating all the way back to January when we first left. We called it pay NCP, and I talked about MCP and all that sort of stuff. At the beginning it was, I'm a developer. I don't really care about APIs. I don't really care about API keys or not. Developers care about APIs. I am a vibe coding equivalent person. I just want the image. I just want the video. I just want the whatever else. Companies were setting themselves up to say, hey, we have an MCP server. It's gonna be able to do more and more over time. And we were like, Great, I'm gonna show up in an MCP server, and I'm going to pay the MCP server. That could happen in cursor, eventually, that could happen in a bunch of different places, and it almost abstracts away the fact there is an artifact that needs to be delivered. Maybe it's a PDF, maybe it's a video. Maybe who knows what it might be, somebody needs to pay so our assumption is that, basically, every human out there, and eventually agent, is going to have a crypto wallet, whether they know it or not. It's a different point, but they're going to be equipped with a crypto wallet. They will then be. Able to pay for things. And when merchants can assume that the human and the agent shows up with a wallet, they're going to probably think more about accepting crypto by default, like they want the lowest fee option. So if you assume both of those things are true, we had set up saying these MCP servers are going to see a lot more traffic, we should make them payable export to hell of a lot better branding than pay MCP. So kudos to them. Like, yeah, take the protocol that was meant for payments and, like, slap an X on it. And now everybody it should have been 0x 402, but that's a different point. So we jettisoned pay MCP when we were sort of like, okay, well, that's actually the right way to do exactly what you just described. Don't need API keys all that. We should just rely on X 402, for a lot of that sort of stuff. We built our own implementation. But for all of that discovery, and what I think you're describing is pay per use, consumption, or token based billing, is how a lot of people are now talking about it in stripe and other lands. That's very helpful for developers. You don't need to go create accounts. Okay? I'm gonna use this database once, so I have to go create an account through the $20 a month. This is dumb. Can't I just pay for the nine cents that they need from whatever your service is this one time. So yes, on the developer side, coming back to the second point, on the consumer side, or prosumer, how we think about it in terms of vibe coding? To take a very specific example, if you've ever gone to replit or lovable or one of these places, and you've been vibe coding for five minutes, and it's like you ran out of tokens, you need to pay $20 to upgrade. I understand why there's an upgrade path. I understand why they need to, like, throw the roadblock in front of you and get you to convert. But if I said, Listen, if your tokens cost three cents, I'm willing to pay nine cents. I don't care. Mark it up. Make money off of me for every token. Let me just keep going until I've convinced myself that I am now in a position where I actually want to pay you 20 bucks a month or whatever. That works for lovable that works for Netflix, that works for every other entity out there. Netflix is maybe a bad example, because you actually want something where you might only ever interact with the service once, assume I have a wallet. Assume I'm good for it. We'll pay as we go. I'll give you the 10 cents, 50 cents, whatever. Please don't make me give you $20 I think it's actually much bigger than the B to B side of things, and that's where that demand is already there. And that's the initial use case when you assume that every site is payable with one of these protocols. That's when you can build much more powerful agents, because you don't have to interrupt the human nearly as often to go approve some transaction, pull out their card, any of this other stuff. So it's all intertwined 12 months from now, nobody, or maybe to Dan's point, a little over eager. 60 months from now, we'll look back at this and be like nobody should have ever cared about these parts of it. But we need to overcome these gaps to get to the point where you're actually able to do some of the more interesting puzzle piece connecting that we're talking about.
Cuy Sheffield 47:36
It's fascinating. We could talk about this for hours. Unfortunately, we're out of time. Thank you so much for listening. Bam, where can people find out more about you?
Bam Azizi 47:44
Yeah, you can find me at BAM as easy mesh on X or go to meshpay.com or our Twitter handle is mesh pay. Love to talk to people like Luis Amira. More about agenda, commerce, crypto stable coin. Love to hear from you, awesome, Louis. What about you?
Speaker 1 48:03
My Twitter is easy, or x is easy. It's just at Louis Amira or at Louis Amira. I don't discriminate, and then the Twitter account is at atxp, underscore AI. We're trying to do the handshake to get at atxp with the Twitter folks. So find us there and come build with us. And we've got plenty of credits for folks that are actually building agents. There's not that many. We're trying to get to all of them, so please
Cuy Sheffield 48:26
call us nice, and you can find me at Kai Sheffield on x and LinkedIn and visa comm slash crypto. If you haven't already, please subscribe to tokenize on Apple, Spotify, wherever you get your podcasts. If you enjoyed this and want more, leave us a review. Please. It helps others find the show. Thank you, bam. Thank you, Louis. Thanks everyone for listening. Bye for now.