Why Your AI Agent Needs a Card and a Stablecoin Wallet

Cuy Sheffield  0:10  

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. Welcome to another episode of our agentic commerce series back again, stepping in as my co host today, the CEO and founder of mesh Connect. Bam. Azizi, welcome, bam. How are you doing today? Thank you,

Bam Azizi  0:31  

Kai, excited for this specific episode with Nicole, super pumped to have this conversation. It's a

Cuy Sheffield  0:36  

great time for it. So joining us today is Nikhil chandok, Chief Product Officer, and CTO at Circle. Welcome Nikhil, how are you doing good? How are you? Thank you. Fantastic. So two quick bits before we get to the content, I need to remind you that the views or opinions of our contributors today are their own and do not necessarily reflect those 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. Nikhil, tell us more about yourself and your background. You've had an incredible career at Google and YouTube and now at Circle today. Kind of how did you end up at Circle? What was some of the background? Yeah. I mean,

Speaker 1  1:18  

thanks first. Thank you for having me. I'm very excited to be here. Long time listener, first time attendee, so very excited. As far as my background goes, I think it's my fifth year at Circle. I got here in late 21 early 22 and came to work on stable coins because I thought stable coins were the most obvious at scale, consumer product that would come out of, like, the crypto ecosystem, right? Like, so Bitcoin was already there when I first came in. Ethereum was already there. But, like, in my mind, everybody in the world wants us dollars, and now there was a way to get access. Like people everywhere in the world could essentially hold dollars in their wallets, which was not possible before. So thematically, my entire career has been on the internet. Like I was early at Google, I was early at YouTube. I worked at meta for a while, and I've only worked on consumer products my whole life. So for me, stable coins were the most obvious consumer product that came out of this ecosystem, how it happened, or, like, in all the the technical wizardry around it was interesting. But I think like the outcome being like, hey, a billion people could now hold dollars, where, like, only about three 30 million people in the world can hold dollars. And I think like that at scale opportunity, or like, 5 billion people can hold dollars. I think that opportunity is still in front of us. And I think, like, that's what's exciting about this. And then I got to know everything that is complicated about, like, financial technology and, like, why some of it is not the best, and what does settlement mean. So I didn't know any of that when I first came in. It was just obvious to me that this was important. This is how things should work. It's kind of when I first went to YouTube, it was just obvious to me that like this is how all videos will work. And even if it was just cat videos who cared, there was no concept of like this that we're doing right now. And before this, before circle, I was building Ray van sunglasses at meta. I used to run their augmented reality product team. And in my mind, like, you know, that was also a great consumer product, because all glasses in the world would have to be smart. So I have this like idea that I come up with these like conviction every few years that, like, hey, something needs to exist in the world. And then I follow my curiosity. I find the best people in the world to go do it. Jeremy definitely is one of the best in the world and has built an incredible company and is very successful at doing something that is so important to the world. So that's how I got here. It's been an exciting journey so far.

Cuy Sheffield  3:50  

So then, as you think about consumer products, where is AI energetic for you? Of like, having seen early video with YouTube and social with matter, like, where do you think we are in the cycle of AI from a consumer standpoint, and specifically agentic commerce, how do you

Speaker 1  4:08  

think about that? I'd say two things. One is all timelines are compressed, right? Like, everything that used to take five years takes a year, everything that took a year takes a month. So one, I think, like speculating on like where we are is, like, really difficult. It's incredibly hard to say like this is exactly where we are. And like, you know, I'm going to sort of analogize with like previous technology trends and say like this is where we are. I will say that it feels early. It's still sort of pre lonely girl 15. Like, so lonely girl 15 is this big milestone in YouTube's history.

Cuy Sheffield  4:44  

Lonely go, what was, I don't think I'm familiar with it. What is it? So in

Speaker 1  4:48  

like, the late 2000 eights, I could be getting my dates wrong. So, like, late 2008 2009 there was, like, this creator who came out with this, like, web series, and nobody knew who they were. And, like, this was the first. Concept of, like, a web series that comes out, and it was just intriguing. It would come out, like every week, and people were interested in it. But like, if you can believe it, there was a time when there were no channels, there were no video channels on the on the internet. And so we took those ideas that we were inspired by at YouTube and made channels back in, like, 2010 and so it kind of feels pre channels, like, in the sense we haven't figured out, like, we have these like sparks of interest, you know, like, hey, agents are really good at, like, paying other agents using on chain infrastructure. They sort of fumble about with like, sign up pages. They fumble about with like, card fields and like, there are really cool marketplaces where agents can start paying other agents for a set of services. But nothing seems to work at scale, because we don't know what that's going to look like. So that's number one second is nobody could have predicted that there would be, I don't know, millions and millions of useful channels in 2010 so our goal back in 2010 was, can we have 100,000 channels that are really good, that pay the creators more than $10,000 a year. And so it kind of feels like that, can we have like, 100,000 like, Agent tech endpoints that are, like, really, really, really useful that other agents can use? How should we, like, define metrics on that, what parameters should exist? So it's early, but then it took like, 15 years to create, like, I don't know, 5 million useful channels. I think this is probably two years. You have, like, at least 100,000 endpoints that are really, really useful. I think that could just be the end of this year, who knows, or could be in three months, who knows, from a sophistication perspective, like, I think that's a good analogy. I would use to say, like, Hey, this is how mature we are. It's early, but not that early.

Bam Azizi  6:45  

What does that mean for consumers? Like, how agent e commerce can change, like, a day to day people's life?

Speaker 1  6:51  

I mean, this is the big question for the broad internet overall, right? Like, how do interfaces evolve? Like, what is a computer like my computer right now is me opening up a computer, signing in, getting on the browser, doing a thing. What is the computer of the future? Does it require me to, like, go to an interface and interact with that interface, or am I just going to interact with the whole world using this agentic interface that I have? I think that question is very open, and I think about it as like, hey, what's the right architecture for a personal computer? And I think, like, from there, like, I think the question becomes like, Okay, what is the right payments architecture for this new personal computer that people will have? I think for now, I think it's clearly obvious that for developers like agentic, like commerce is like, real they can, like, start paying for services using agentic infrastructure. For consumers, it'll happen. The question is like, what kind of transactions are you happy to like, allow an agent to pursue on your behalf? When do you want to be in the loop and versus when do you not want to be in the loop? Is it a transaction size threshold? Are you primarily going to use your agents for research and then still be involved when you finally need to press the button, and are there new models to emerge? Because, like, now agents can, like, transact really, really cheaply on really, really small values. So like, as a consumer, should I be expecting new kinds of services? One of the things that sort of, I think about all the time is like the internet, in its current generation, evolved at a time, and ads were the main way you could monetize content. You could monetize websites using ADS. So circuit 2008 when, like Google puts accents inside of Android, iPhone comes out in 2007 ads is the dominant way of essentially creating, creating and monetizing websites in 2026 you can have using USDC. You can pay another endpoint, like a millionth of $1 for no gas. And so if that's the case, like, what kind of services can you build, like, you know, and like, what kind of endpoints now start to exist that, like, offer services, will consumers experience a change this year? Maybe, I'm certainly looking forward to being able to give my agent, like, like a personal finance agent, like, the ability to, like, do things with good guardrails. I haven't seen the product in the market yet.

Cuy Sheffield  9:05  

Yeah, it's super interesting to me that even six months ago, I had never done any agentic commerce, and I didn't know anyone who'd done any agentic commerce like, it was like, purely a concept. It was like, Yes, we have AI. We're going to have agents soon, and then the agents are going to want to spend. But like, we don't know exactly what's going to look like now I do agentic commerce. And as you said, there are developers who can do agentic commerce through this new ecosystem of mcps and CLIs and, like, we'll talk kind of more about the infrastructure later. And so it's like ejected commerce is here for this, like, very early adopting more developer, like crowd. But I think the thing that gets me really excited is I think about it less as it's here for developers, and then at some point it shows up for consumers. And more of I would have never self identified. A developer like ever, because I'm not a developer, and so it's like vibe coders. Now everyone is becoming a developer. And so if agentic commerce starts to work for developers, and then before we know it, there are 2 billion developers, because anyone that can speak English or any language can now write code. It feels like that is a path where agentic commerce could go mainstream. So I guess, how do you think about vibe coding with your lens of kind of seeing early consumer internet? Like, you know, the experience of content creators on YouTube. Like, is code just becoming another form of content? And are you do you drink the Kool Aid that everyone is going to write their own personal software and like, what does vibe coding look like over the next few years?

Speaker 1  10:44  

I think, like, the cost of building software goes down to zero. I know it's a big claim. I had this experience over the weekend where a friend of mine was first on Friday texted me, hey, like, check out this new service. And it was great. Like, it's a great UI to interact with your agents and all of that. I was like, okay, like, I'm flying today. I'll land tomorrow, and then I'll check it out. And then I sort of landed on Saturday. Didn't really have the time, so on Sunday afternoon, I get to it, and I started using it. And he's like, Oh, can I call you? I was like, Yeah, she'll call me. And then he shows me his version of the same thing. And I was like, how do you do this, and he's like, Well, I just byte coded it. And so the total cost, it's not zero, it's the cost of, like, building software is basically a one day worth of tokens for this guy. And the cost of tokens also has to go down, right? Like, it's not going to go up. And so when cost of, like, doing something approaches zero, like, you know, data transmission on the internet is a great example of it. Like, the marginal cost of, like sending another packet is almost zero or zero, for that matter, right? The amount of activity goes up. You know, it's very hard to sort of imagine 2026, levels of, you know, data transmission on the Internet back in like, 2005 or even 2002 when, like, the first sort of big boom of, like, laying down cables on the internet came about, right? You could not imagine that. People could not I remember having conversations in 2002 with people being like, why do you need to watch this video on the internet? You can watch it on your TV. And I was like, no, no, no, there's gonna be new video. And they were like, No, you're crazy. Like, you know? And like, it happened. It happened, like, five years later. So, I mean, I'm not trying to dodge the question that I just don't know if, Will everyone, like, write white coded apps. Anybody who wants to write a byte code app will probably write their own application. Because, like, and so then the question is, like, what is the interface and like, When are interfaces pre packaged? Do kids get, like, pre packaged interfaces? Who pre packages them? Like, are the parents responsible for pre packaging those interfaces? Like, how does a workplace work? Because, like, this is a challenge that circle has that like that we actively work on, which is like, what is the AI interface for the average employee with, like, all of circles data and like circles content? So it's unclear, but I do think one sort of consequence will be, like, we will have an internet that is primarily built for agents and a lot of endpoints that actually have to exchange value without humans having to be in the loop, essentially. And so that much is true. It doesn't necessarily require us to take a position on whether everybody will White code their own application or not. There will be billions of agents, and they will be like interacting with each other, and they will be trying to transact real value, and ideally at really, really small scale, which allows for all kinds of new, like, economic models to emerge, right? Like that is cool, but, like, my hunch is, like, everybody will write vibe, coded ads. But I'm not sure it's a hunch essentially, like, because, like, what you're doing is my experience as well, right? Is also my 12 year old's experience, like, she writes her own application, right? And it's not very all she does is, like, tells Claude, like, hey, I want to build a website that sells stuffies. And it gets done. And then she's like, shows it to her friend. She's like, Hey, do you want to buy my stuffy? And it's cute and it's fun, and so, like, it's so obvious. Like, how does her experience evolve from here? It's not clear at all to me. What does that look like here five years from now? Because the computer itself is being reinvented. So like, what are the interfaces of that computer? Yeah, but the services to, like, inform that computer are being built right now. Are the things that we're talking about.

Cuy Sheffield  14:12  

It's interesting. This parallel that comes to mind, you say, is of like in the early days of bringing video onto the internet, it seems like the natural starting point for people will be like, Okay, how do I watch a TV show or a movie on the internet? Of like these existing legacy media companies building websites to show their videos on the internet, but it seemed like the biggest impact was this brand new generation of independent content creators which never would have been able to put content on TV, that were able to put content on the internet. Now you have like, Mr. Beast, and so in the agentic world, when you say endpoints, it resonates with me that a lot of people focus on, okay, how are Walmart and Amazon? How are like, existing large e commerce merchants going to sell to agents and like, yes, that's gonna happen to some. Fix that. But there's this whole other world of like, who are going to be these new one person companies that stand up a merchant endpoint that can then sell to an agent that never would have existed because the barrier was too high to start a company offering some niche service. And so I love the way you talk about endpoints as, like most of our listeners probably like, it's not that familiar of a term, but it's like, how many endpoints will exist, and then how do those endpoints monetize and get plugged in? And so I guess, how do you think about endpoints in the context of circle talks about this economic operating system like, what is the economic operating system like? Where do endpoints fit into that. And like, the broader vision that circles building towards, right?

Speaker 1  15:45  

Yeah, thanks. And like, I'll just say one thing, like, I remember sitting in review in Google in 2006 and like, Eric Schmidt, who was the then CEO of Google, was like, you know, we're just bought YouTube in October, and so deal closed in December. And he's like, you know, what we should do is make sure I can watch Lost on YouTube. And lost is this old show. And, like, and so and my head, I was, like, very young. I was mid 20s, and I'm thinking to myself, I was like, Why do you want to watch it on YouTube, which is kind of, it's not great. Like, you know, the quality is terrible. And like, you can watch it on your TV. Who cares? And like, just buy a DVR. And I think like that still applies, like in the in the sense that and Mr. Beast as a concept could not be imagined by the best of us in like, 2006 2010 even 2015 I don't know how old Mr. Beast is. I think is started around 2014 is my guess. But like, it's very hard to imagine the ecosystem like 10 years out, right? I think for us it's going to be even harder as far as circle goes, and like the economic OS goes, look for us, we are the largest regulated stable coin in the world. So first to have any kind of economic activity, you need a good dollar, and you need other currencies as well. And we have a Euro, and now we have a yield bidding asset called us YC, right? So we need, like, this core financial infrastructure which is trustworthy and reliable. Then you need like, primitives, and then you need, like, essentially, a performing blockchain on which all of those things can exist. Then you need a bunch of developer services on that performant blockchain that can take advantage of both, like the regulated dollar, and can take advantage of the core services that a performing blockchain can present. And so that's art. For us. We are working on moving towards like a main net launch soon. And so for us, like as far as like agentic endpoints go, it's everything from making sure that we have robust and reliable settlement infrastructure, making sure that we have good identity systems so like your agent knows who it is interacting with, making sure that we have wallets that the agents can spin up to essentially interact both with, like the on chain world and the off chain world, making sure that agents can transact in like a privacy preserving manner on this blockchain, and then having like the right reputation system so the agents know who to interact with versus who not to interact with. And this happened with like the web as well, because when the web first came out, anybody could publish a web page. So now what happened? Well, you need page rank, and so we will have something at some point which will be like, Okay, well, this is how you determine reputation for an agent. Are they a good economic actor? Are they able to perform the economic activity that they are advertising? Should you exchange value with them? And then a whole set of like payment primitives that we're building, an example being nano payments, which allow you to do like, you know, millionth of $1 transaction and millionth of $1 transaction with any endpoint on arc and in other places. The other thing we're doing with Arc is we're trying to build an ecosystem that is rose the pie for everybody. So we want to make sure that like arc essentially utilizes our multi chain infrastructure that we have built, like USDC now moves over. I think I could be wrong about this, like more than 25 chains right now, and we have core bridging infrastructure that allows us to move that USDC across all these chains. So like arc is going to sort of like ride on that. So issuing on arc, you essentially inherit all of this multi chain infrastructure that USDC has, all of this performance cross chain bridging everything that comes around, like, you know, being able to utilize, like USDC and other tokens, by the way, on arc as well, like to being able to do nano payments, for that matter. So it is a pretty broad, comprehensive point of view on like hey in the future, you can imagine, to your point Kai, like the one person company that has 10 employees that are all agentic and that are creating real economic output, and they're interacting with hundreds or 10s of 1000s of like other one person companies that are out there that are trying to create some economic value. And what is economic value? It is basically solving a problem that others will pay you to solve. That is what economic value is. Now there are a lot of problems in the world, essentially, and a lot of problems are not solved because it's uneconomical to solve those problems. More of those problems will be solved now because, like agents are able to do things for individuals, which. Is your experience as well as white coding, right? Like, because you're now solving a problem that could not be solved before. It wasn't even it was not possible for you to go hire a developer to byte code your application. And so you're like, you know, I'm just gonna wait. I'll give a quick example, like my bike coded application, and I think parents will appreciate this. It's just tracks the emails that come from the school and, like, essentially updates my calendar and my calendar and my wife's calendar and creates a bunch of action items for me on a daily basis, because I have three children. And, like, there's a lot of communication. There was nobody was going to build that for me. And like, till AI came about, and I spent like, $10

Bam Azizi  20:35  

worth of tokens doing it. Yeah, I'm super interested to dig deeper on the Nano payment that millions of dollars, super exciting to talk about. And then especially now, we have Kai as a co host here, so you can't really do that on Visa or MasterCard. So that's the traditional rails. And you're talking about the new unlock that you can pay for a service that could be millions of cents. Tell us more about that. What are the examples like, why consumers should care about this type of functionalities in on lock? Yeah.

Speaker 1  21:04  

I mean, I think like it could be things like, I mean, developer services are easily in that bucket. Like developers have, like, small value payments to do all the time for consumers. I think like it'll show up as packaged around other services. Like, whether it's translation services, whether it is like, you know, you want to go create a song, you want to go and get, like, a 15 second remix for something a video, like, since we were talking about video, I want to add music to my video. Like, who should create it? How much should it cost? Do I need to go sign up for a whole service to, like, actually do that? Or can I just pay, like, some service that says, gonna go spend, like, I don't know, a cent or 10th of a cent, like trying to create that 15 second song for my video, or maybe

Bam Azizi  21:44  

your agent wants to pay another agent for something that needs your approval, right?

Speaker 1  21:49  

Yeah, and an agent wants to pay another agent, but like for machine payments, especially like, where I think the machine payment is essentially fronting, like consumer service, you will have a ton of these interactions. Originally, like this was taught about this concept of, like, look, people are going to pay for content on the internet. Like and like micro payments were talked about for a long time as like, hey, people pay for content on the internet. That word hasn't come about in the way that we had thought it will come about. Like content in the internet. On the internet is primarily, like, ad supported, or there are subscription tables, essentially, like, preventing you from accessing it. How will that evolve into a world in which like these endpoints are providing specific services, if I must know a certain piece of data that only a few people in the world can know about, like, how should my agent go out and, like, discover their data and then consume the data? It probably will require like these, like, really, really small payments to do that, yeah, yeah.

Cuy Sheffield  22:43  

I think micro payments are fascinating. We're super excited about them, yeah, visa and spending a bunch of time on, like, how can we enable them and and I think one of the big questions that that I have is, how often is it that one consumer developer is going to make one fraction of a cent payment one time. And like, if it's just that, it's very hard, if it's a repeated you're going to be transacting over time and with many different services, then I think there are really interesting techniques around like, that's kind of how transit and fuel work, where, like, can you off up to a certain amount and then be able to have people pay and then settle at a later period. And so I think we're going to see a bunch of really cool innovation in different designs and approaches that can enable micro payments and and I think it's, it's a new type of business model, and I always thought that the reason that it didn't take off from a consumer standpoint, it wasn't just infrastructure. It was more like consumer behavior and the mental transaction cost. When you're used to, like, ad supported, it's hard to be like, do I want to pay one cent to watch this video? Like, it's just, it's something that's new with agents. It's kind of like you have brand new consumer behaviors that are emerging, of how people interact with their agents, what those agents do. And so anytime you have these, like, new platforms with new types of behavior, there's an opportunity to have different types of monetization built into them, versus trying to take something I'm used to doing today and have me do it differently. Like, that's like, very hard to get

Speaker 1  24:12  

someone to change. One of the things I'm excited about, and this is, like, this is not something that we're working on at all, but, like, it's an open question. It's like, how will the devices of the future be built, right, like the originally, like, if the devices of the past have been built around this idea of an ad supported internet, like the devices of the future could be built around this idea of micro payments. And if that happens, then like, you should not have, like, these problems that you have around, like, you know, these devices stealing our attention, or, like, you know, people trying to, like, break us with constantly, constantly feeding us content. All of that actually is rooted in how these things are monetized, and those things, and those things are monetized the way they are monetized, because the payment systems then evolve to allow it to build a device I can't go tell my phone, hey, here's $20 go spend it. Every month and bring me all the content I need. And, like, check with me anytime that needs to go up. That's not a thing that is available to me, but it could be available to me in like, five years time, because there would be, like, all these endpoints that would actually take, like, you know, the 10th of a cent of a payment and my phone, who is primarily going to be my agent, who's my main agent, is going to be able to interact with this world and construct an experience for me without me having to, sort of like, pay the mental tax that Tai you were talking about in the past when I was sitting on a browser and I'm hitting like a pay wall and, like, the question is, like, Hey, do you want to spend 10 cents on this? Like, what is this? Stop asking me these questions. The other thing is, like, I can now express a natural language to my agent to do things and like, they can have judgment. Software didn't have judgment before. Software was a bunch of rules. Now you have alignment and like issues and like, you know, you have to you gave software, like, the ability to, like, decide like a human and so you can expect software to, like, you know, make these payment decisions

Cuy Sheffield  25:55  

as well. Yeah, it's a great point. So before we move on, let's hear from the sponsors that make this series possible.

Sy Taylor  26:03  

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 bat 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  27:09  

all right, coming back in, let's talk about protocols. We've got this alphabet soup. We've got eight, ACP x4 two, MPP, UCP, trying to think, if I missed any we have. Tai P, you've been around protocols for a while, from early internet to now blockchains. Like, how do you see these protocol wars playing out? Are we going to have many? Are they going to interoperate together? Like, what's the circle view on how you look at the emerging agentic commerce protocols?

Speaker 1  27:40  

I mean, I kind of think about it like the protocols being like, you know, operating systems on mobile, when there were, like six or seven before, like the iOS operating system and like before there was Android. So when I used to first work on, like YouTube on mobile, we would issue on all these different operating systems. Symbian danger had a phone like, you know, and we would like issue, we made an app for danger. I did the iOS, the first iOS app, not knowing that like iOS would be so dominant from here. So from circles perspective, we're going to work with all of them. We don't want to necessarily like, exclude the benefit of a protocol to like our core asset, which is USDC, right? Like, so we're going to go work with them. That being said, like we also joined like the exporter to Foundation, and because we also want a place where we can sort of direct the protocol like we will also work with the other protocols as much as like that we are able to essentially like so we can advance the standard for USDC, all of these protocols as much as they need a stable point we'll find in USDC and in circle, the best partner that they can get right, like, because we are technically savvy, we are ahead of the curve than most when it comes to like, having an asset, having the assets to be multi chain, sort of understanding natively, what it means to run like an on chain ecosystem. So I don't know which of these protocols is going to win? I know that all of the protocols are going to have to be comprehensive of all payment types, like they have to be comprehensive of stable coins, they have to be comprehensive of cards. They have to be comprehensive of the country specific, like the local flavor of the way payments are done in those countries. And so I think everybody will try to move towards that. I don't think there will be like, a stable coin only Protocol, or like a card only protocol. I mean, Tai, I'm sure you guys have a point of view on this, but like, I don't think there's going to be just or, like an issuer specific protocol. I think it's going to be truly like, hey, there are all these payment types, and there are all of these endpoints, and when an agent sort of reaches them and wants to initiate a session, the endpoints needs to reveal, like, all the types of payment methods that they will take, and ideally, the protocol is comprehensive of all the payment methods that because we want value exchanged with the lowest friction possible, right? Like, so that's how we're thinking about it, like we are. But one player in this right? We care about stable coins deeply, but we also care about like, other. Other payment methods, sort of like coexisting with stable coins, because otherwise the merchant experience is broken and the consumer

Cuy Sheffield  30:07  

experience is broken. Yeah, we've talked about this a ton, and anytime people try and create this kind of false dichotomy of it's cards versus stable coins on ejected commerce, it's just like the market's going to be huge and both of them are going to play a big role. And sometimes you have a single transaction that uses both. And so it's like, I think there's so much opportunity within that, and particularly in my opinion, kind of my personal view on this is I think that the best possible experience, and I've said this on the show many times, I'm repeating myself, like the best possible experience is to give your agent a card, even though I'm a crypto guy, even though I have wallets, even though I have USCC, it's like there is a certain level of protection that you get giving a card where you have controls at the agent layer, and you have controls at the issuer layer, and I could spend on credit, and I can get rewards, and I get, like, there's just so many things that I could do if something goes wrong. I've got kind of more rights there. But on the merchant side, when I think about these brand new, emerging merchant endpoints, they can't wait three days to get settled. If you're running an image generation company and you're generating images as an endpoint, you got GPU bills like, you can't be like, waiting for a wire transfer to come in and say, All right, now I could go pay for that GPU that's burning in the data center from all these requests. And so I think there's this major role for stable coins like USDC to be a part of that real time settlement, where when I go and generate an image with my card, USDC should immediately go to that merchant and then it shouldn't just stop at the merchant endpoint, they should be able to use that USCC to then pay their supplier, pay their cloud provider, who should then be able to use that USCC to pay their electricity bill. And so being able to have this velocity of money increase on the back end, I feel like is like a critical prerequisite for agentic commerce to work, even if the front end of using a card looks the same that it does today. And so, Dan, how do you think about that intersection before, of how cards and stable coins kind of fit together

Bam Azizi  32:11  

within agentic? Yeah, before answering that, I think my prediction on different protocols is like we will see some sort of a consolidation, even though it's not good for my business, personally. So I want to see more fragmentation, because we can hop there. The more fragmented the market is, the better mesh can perform. But eventually, that's what's going to happen, similar to, right? We don't have Symbian anymore, Niko, right? So everything is consolidated around iOS and Android and web. We'll see some sort of a consolidation on the protocol level, but in terms of like, I agree with Kai on the merchant side, but I disagree with him on the consumer side. I think the reason is Visa and MasterCard, they're like 6070, year old. Company. Blockchain is 15 year old, like a stable coin is six, seven year old. I think at some point you will get the same type of rewards. Get the speed you get, the faster settlement. You get, like, nano transactions. There is a lot you can do in the consumer side. And I think everything will come in chain, and Visa and MasterCard can be a key players here to embrace it versus fighting it, which you guys are doing it. Kudos to you, Kai, that's, that's my opinion about, like, how this things will shake up there.

Speaker 1  33:19  

There's no reason why you can't replicate it on chain. Like, the question is, what is the reason to replicate it on chain? Yeah, so I think the way I think about it is, like, what are the network incentives to do the things that, like, Kai is describing, and like, When do those incentives shift so that you know, you were like, Okay, I must do this on chain, versus, like, doing it like off chain. And I don't know yet what those incentives are going to be, but like replicating those services, yes, if you're today, you're starting out as a consumer, you're better off giving like, you know, a card to like, you know, your agent, and then go and like, if that agent may instantiate a US, a wallet on chain with like USDC in it, because they need to interact with an endpoint. And I'm completely comfortable with that like flow today, because it is, it is solving a problem. And like the merchant problem that you described, yes, like one of the things Kai is like, when you're running these services on the internet, you are running them on the internet, you don't want local currency to because I need to pay my GPU bill in dollars, so I want to settle with Visa in dollars. And I may be in like, random country somewhere in the world. And so, how can you give me USDC? How can you give me $1 stable coin so I can then go pay my GPU bills in like, dollars, because more things are getting getting priced in dollars as the economy shifts towards the internet. So I think there's that as well. Like, which is not just the three days. It is also like, Hey, can I get dollars? Which is like, the original premise of USDC. I want to hold dollars because I want to create crypto that's at least eight years ago, right?

Cuy Sheffield  34:44  

So, yeah, it's really interesting. I was actually going the other direction of, I think that there's this big question around non USD, stable coins and local currency, stable coins and like, what's the future? What are the use cases? Like, where is this going to go? I think the facts are right now, 99% of the supply is in dollars and, like, that's what the demand has been. And I think people say, Okay, well, what do I need a non USD stable coin for? And particularly if I'm just, like, transacting domestically, like, what? What do I use it for? And I think there's an argument to say, well, agentic commerce is a domestic use case as well. It could be cross border or domestic, and there could be value in having local currency, stable coins, that are used within those transactions. And I think a lot of governments and regulators, when you talk to them, if they believe agentic commerce is going to be big. They don't want agentic commerce to be what dollarizes Their economy. And so like they want their consumers to be able to participate in agentic commerce, to get the speed, to get the programmability, but to not have to move the entire economy over to dollars. And so can you talk a little bit more about with Arc? To me, one of the most unique, differentiated things that you're doing is you've got a bunch of partner, non USD, local stable coin issuers that are coming into the arc ecosystem. And then you're looking at, how do you do on chain? FX, how do you create a environment where you could have a transaction that initiates domestically in a australian dollar stable coin that then lands in a US dollar stable coin. And like, being able to plug those local economies in. Like, how do you see that, that playing out?

Speaker 1  36:29  

It's a great point, and I agree, by the way. Like, I don't think circle has a point of view that USDC should and like arc, for that matter, should replace like, national payments infrastructure, right? I think for every country, it's really important for them to sort of like, control and manage that and like, it's a matter of security. It's a matter of being sovereign as a nation. So So I agree on arc. Specifically, we have about six partners who have signed up with, like, local stable coins. One of the sort of consequences of genius is that other countries are now adopting, like, stable coin frameworks. So when it's an X country, like, you know, local stable coin, it's much harder. Like, for example, if it's Country A, and like, you don't really have a stable coin framework in country A, you create a country a stable coin from country B. That's more complicated, right? Because that's when the regulator doesn't like it, and they're like, look, we don't want, like, an X country stable coin, sort of like taking on flows inside our country. As that matures, there are going to be more and more local stable coins. So, like, we have a partnering program. We go and vet these stable coins. We try to make sure that they are good actors, that, like, they have good reserve management methods, and when we are convinced about that, we bring them onto arc, we issue them on arc. We also promise them being able to take them across all these other chains. So we want their stable coin to go wherever the liquidity is needed. And so our value proposition is that we do that, and then we go and work with like liquidity providers and market makers to make sure that there's good liquidity in the pairs where it matters, like, and most of the time the most important pair is the dollar pair, which you have a local currency, stable coin and like, against USDC, because if you want to swap between any two currencies, your typical path is going through the dollar. That's how everything gets priced. And so how do we go and make sure that those marketplaces exist? So we have a protocol, which we call stable FX, which is an RFQ protocol that runs on R which allows for spot delivery of like local stable coins. So if you need currency in a local market for whatever reason, because you want to settle it, you can actually go make an RFQ request inside of stable FX, you know, can get quotes and then you can act on it. How big does this scale? Kind of depends to your point. Kai, how fast do these countries take on local stable coins? The one thing I will say is a lot of these national payment systems are not programmatic, right? Like they're just settlement infrastructure. And for agentic commerce to exist, you need like, the ability to like program that infrastructure. So one of the things that these national infrastructure have to do is figure out, like, how they're going to make an infrastructure programmatic, or are they going to migrate onto some sort of blockchain based infrastructure, which allows them to have programmatic flows that can work well with agentic flows? Short of that, the Dora based agentic flow is always going to be much more fully featured than like anything else that is there out there, that's that's the risk, and I think all of us have to get more sophisticated about like how to manage that risk.

Bam Azizi  39:24  

That's interesting. I have one question for you, Nikhil, what's the role of wallet in the agentic commerce? Are agents are going to have their own wallet, or they will share the human's wallet? I saw tether today, they launched their human wallet as well. So what circles roadmap on that, and how you connect it to agentic commerce?

Speaker 1  39:46  

Our roadmap is we've been building like developer control wallets for some time soon, we will have a CLI we will have the ability for an agent to essentially spin up a wallet, essentially take all the actions on a wallet using an agentic interface and. Um, the human wallet, I think, is, I imagine these, like, developer control wallets to be temporary. They tend to be like, you know, they tend to have like, strong policy on them, like, there's good governance on them, and that governance is, like, programmatically expressed and it's controlled. That's where I think, like, most of the value is going to be, I'm not going to give an agent, like, full access to my bank account. Like, that is what I am. I would find very, very difficult to do, right? Like, and I don't think that technology is there where I'm like, okay, cool. Like, you know, I'm gonna connect via my agent onto plaid and, like, connect it to my Chase, and be like, do what you need to do. I think, like, that day is like, far away. So I think about these, maybe you're all

Bam Azizi  40:39  

old, maybe the New Generation Trust agents more than humans.

Speaker 1  40:42  

Maybe I'm too old. Maybe I'm told, maybe you guys are doing it, and, like, you know, I'm not, and I'm happy to learn. And it's the first age based joke in this conversation, which is good, like, you know, I gave you a lot of, like, lot of water really early on, at least these temporary wallets that have, like, you know, for sure, when I say temporary, I don't mean like, you know, temporary in the sense that they come and go in a few seconds. They could last days. They could last months. But fundamentally, these are, like policy constraint wallets that have like, some kind of spending ability and have some goal associated with it, expressed in like, you know, natural language that the agent can then act on, that makes a resource, in my mind, that, like, the agent has access to to accomplish a task, and they have, like, some goals and, like, some constraints around it. That's how I'm thinking about it. We have the we are building the infrastructure for it. And we also have discussed, like, how much should we, like, you know, hypothesize on, like, what is going to be the right use case. We don't know in the level of detail that we're used to knowing. And so we're just shipping the thing and, like, we're seeing what activity is happening, and we plan on learning

Cuy Sheffield  41:42  

from that as well. Yeah, it's fascinating to see this new class of wallets. Of used to be wallets from mobile apps, and now they're CLIs and MCPS. It's like another form factor of a wallet that has a different design pattern, that every wallet that exists as mobile apps probably gonna end up as a CLI as well. But I guess last question like, how are you using AI and agents around building products inside of circle? How have things changed over the past six months in terms of your internal kind of product development cycles? And then any advice for other product and technology leaders as you've kind of learned and

Speaker 1  42:16  

experimented with AI? I can't give advice like because, like, I think I am still learning I can share my experience. Like, which is still like, late last year, it was like, Yeah, will this work? Will this not work? And then come December, when the new models dropped, the world changed, and it changed in a way that was not comprehensible. And it's still not comprehensible to most. So like, I guess if I have advice, I have, like, you know, go be a practitioner. If you haven't been a practitioner in a while. I'm technical. I have a master's in CS, like, so it's very easy for me to go back to that world and be like, All right. Like, what has changed and what was revealed to me was, like, just incredible. So what we've been doing at Circle is sort of like getting everybody tooled up. Our developers have like best practices on like, how to use agentic methods to, like, write code, how to do PRs, how to make sure that like our environments are set up so that like agents can essentially maintain context, like, locally, globally. How does that all come about? So all of those practices we just formed up in the last, I want to say, month, and now they were in the process of rolling them out to the whole organization. I am expecting most people to be natively enabled by the summer where, like, I guess the harness works, which is like, hey, we have local context. We have global context. This is how you build, this is how you push. This is how you do a code review, I have a view which is, like, the way I express it is, I say, like, look, we used to build the widget. Now we're building the factory that builds the widget, and then we're going to build it be building the monitoring systems that monitor the factory that build the widget. And so we're not there on the third thing yet. And so we're in the process of building the factory that builds the widget and trying to convince people, trying to convince my engineers, that they should stop just building the widget, like, try to build a factory that builds the widget and like, that is not easy for most people. And like, the monitoring system that, like, monitors the factory that builds the widget, I think is much more of a it's a vision for next year. I think, I don't think it's going to happen. Like, realistically, this year, we have those monitoring systems. They're very, very rudimentary, like, they're not at any level of sophistication that I would want to build systems on. So that's where we are. We're also trying to tool up our XFN. We're like, hey, like, look, I run power can engineering some like, hey, what about the other control functions? How can they get tooled up? Like, how? And then we recently read, like, Jack's company, AGI, like, blog post and like that has given us a lot of fodder to think about. We are much smaller than, like, you know, visa is, so we can experiment in a sandbox, and we will try some experiments, and we'll try to learn from that. But we're not there where, like, I think Jack's vision is the company's the AGI and the humans are the endpoint.

Cuy Sheffield  44:56  

6000 reports to Jack, right? Like everybody on the company reports. To Jack. So I expect next time we talk to kill the the entire company will be reporting to

Speaker 1  45:05  

Jeremy just flat out. I think it's just, I don't have strong counter arguments yet, because we're still sort of exploring, like, what was written that I thought it was really insightful. It changed the way I thought about what is the company, but I haven't fully digested it. And like, that is one of the challenges right now, which is, like, there's just so much to digest, like, so fast. So as a leader, you have to be very specific about, like, how to be concrete, where you need to be concrete, and then where to be not concrete, and allow for, like, a lot of experimentation. And so striking that balance constantly is, like, really hard. I was out on vacation for four days last week. And day one of my vacation, Claude mythos launched. And then day two, I'm sure, something else launched. And by day four, I was like, Oh, my God, I'm behind again, right? I think, I'm sure everybody's struggling with it, but like, I think, like, it's definitely a challenge for us, because we think of ourselves as, like, an AI forward organization, but we have, like, I think we had like, 90% plus DAU on, like, AI usage inside the companies. I want to, I want to guess that that's the case.

Cuy Sheffield  46:08  

It's a fascinating time. We could go so much longer, but this all the time we have. Thank you so much for listening. Nikhil, where can people find more about you?

Speaker 1  46:17  

On Twitter, like, I don't know. Like, I'm not very active, but like I am, my last name on Twitter, which is chandok and C, H, A, N, D, H, O,

Cuy Sheffield  46:24  

K, bam. What about you?

Bam Azizi  46:26  

You can follow mesh pay on LinkedIn and X or Twitter, and you can find me bam, as easy mesh on both

Cuy Sheffield  46:32  

platforms, and you can find me at Kai Sheffield on x in visa comm slash crypto. If you haven't already, please subscribe to tokenize on Apple, Spotify or wherever you listen to your podcast. If you enjoy this. You want more, please leave us a review. It really helps us. Thank you, everyone for listening. Hope you enjoyed the show. Bye for now.