The Breakdown - Are AI Agents Actually Using Crypto? | The Breakdown

Episode Date: April 7, 2026

AI agents are coming online fast — and the payment rails for that world are being built right now. David breaks down the race between Coinbase’s x402 and Stripe’s Machine Payments Protocol, the rise of agentic commerce, and why crypto may have a natural edge, before sitting down with Jordan Liu of AIsa. __ TIMESTAMPS: (00:00) Intro (01:30) Tool Time (03:45) AI-doption (05:48) Nexo Ad (06:30) When in Doubt, Incentivize (07:36) Interview with Jordan Liu (12:24) Nexo Ad and Blockworks IR Promo (14:09) Interview with Jordan Liu (Con’t) __ › Jordan Liu (AIsa) —https://aisa.one/ › David — https://x.com/dcanellis SPONSORS › NEXO Nexo is the premier digital wealth platform. Receive interest on your crypto, borrow against it without selling, and trade a range of assets. Now available in the U.S with 30 days of exclusive privileges. Get started at http://nexo.com/breakdown Get top market insights and the latest in crypto news. Subscribe to the Blockworks Daily Newsletter: https://blockworks.co/newsletter/ DISCLAIMER As always, remember this podcast is for informational purposes only, and any views expressed by anyone on the show are solely their opinions, not financial advice.

If you believe Jensen Huang, Sam Altman, and Elon Musk, then you already know that we're destined to inhabit a digital world massively overpopulated with autonomous AI agents. And it's not only the usual suspects who are super bullish on agents. Matthew Prince, the CEO of Cloudflare, recently projected that in 27, the amount of bot traffic online will exceed the amount of human traffic, which is only next year. We just think that the use of the internet is going to explode over the next several years,

and what is using it is going to change dramatically. What's funny is that crypto has already lived with this reality for years. It's been the case throughout crypto's history that trading bots, points farming, algorithms, and automated smart contracts have at times made up the lion's share of on-chain activity for certain networks, particularly when it comes to transaction counts, which I suppose explains why we've been so quick to take up the on-chain agentic meta and run with it. And it's not that the agentic crypto crossover is not compelling, especially when you consider that it hugely spans crypto's total addressable market,

beyond a mere billion or so people that might adopt it to a hypothetically inferred a number of AI agents that might use Crypto Rails to pay for all sorts of things, executing trades and arbitrage on behalf of, and maybe even in spite of, their human beneficiaries. It's all very optimistic and even somewhat plausible, but if OpenClaw is supposed to be the fastest adopted open source software in history, then crypto should be in for a similarly fast fast road to adoption. So, are they here yet? Time to find out.

I'm your host David Canellis, and this is The Breakdown. Let's get to it. This episode is brought to you by Nexto. Step into a new era of digital wealth. Earn interest on your digital assets. Borrow against them without selling and trade all in one platform. Get started at nexo.com slash breakdown. Nothing said on the breakdown is a recommendation to buy or sell securities or tokens.

This podcast is for informational purposes only and interviews expressed by anyone on the show are opinions, not financial advice. Host and guests may hold positions in the company's funds or projects discussed. Open Claw is probably the single most important release of software, you know, probably ever. It's true that the scene is really set for agents to take over our block space. Well, at least the tools are there. And now we have two standout chain-enabled primitives built especially for AI agents. There's the X402 Payment Standard from Coinbase and supported by Cloudflare via the X402 Foundation, which I suppose spiritually puts X402 in the orbit of Ethereum L2 base,

though related activity has also spread beyond base into chains like Solana. And now Stripe has debuted its own machine payments protocol, or MPP for short, a general payment standard for agents. Shopify is understandably a launch partner for MPP, which means Shopify merchants can already accept payments from autonomous agents. Stripe's MPP is technically rail agnostic. It supports stable coins on the custom-built layer 1 tempo, typical Fiat cards via Stripe,

and Bitcoin Lightning support is available in the broader MPP stack. Both X402 and MPP offer a way to send stable coins over HWTP, but the most important distinction right now lies in the fact that MPP supports Fiat out of the box. Today, X402 is centered on on-chain stablecoin payments. Similarly to Stripe, Coinbase also says that X402 is network, token and currency agnostic, but interestingly states in its GitHub docs that X402 may extend support to Fiat-based networks but will never deprioritize on-chain payments in favor of Fiat payments. Of course, many people think that it will be entirely up to the agents themselves

to decide which payment rails they'll use. In practice, the rails will often be chosen by agent developers, and even by the constraints of the wallets they opt for, or which payment methods merchants accept, and more generally the defaults of platforms working to connect agents to different marketplaces. So really the idea is to create the most frictionless payment experience possible for LLMs, so that those rails would be the obvious choice for either developers or the cold legions of logical or motionless on-chain operators we're soon to be trading against. And we already know that Fiat Rails have been historically ill-suited

for micropayments considering the fees, not to mention the hassle of manually approving a long list of tiny payments for various stuff on the internet, even if it meant creators and other data owners could make a living. And that lack of internet-native micro-transactions is arguably the root cause, albeit an indirect one, of the subscription model and the broken economics of the online attention economy. Ben Thompson wrote an excellent piece on the subject over a year ago, and in it he actually outlines how agents and stable coins can fundamentally reshape how the internet works. What is possible, not probable, but at least possible, is to in the long run build

an entirely new marketplace for content that results in a new win-win-win equilibrium. First, the protocol layer should have a mechanism for payments via digital currency, i.e. stable coins. Second, AI providers like ChatGPT should build an auction mechanism that pays out content sources based on the frequency with which they are cited in AI answers. The result would be a new universe of creators who will be incentivized to produce high-quality content that is more likely to be useful to AI, competing in a marketplace a la la the open web. Indeed, this would be the new open web, but one that operates at even greater scale than the current web, given the fact that human attention is a scarce resource. Well, the number of potential agents is infinite.

First, pushing for all manner of companies, merchants, services, and resources to accept micro payments. Stripe has made this easy for its users, which is why so many people are interested to see how Tempo and MPP grow alongside the agenic side of crypto, while Coinbase is doing its own thing by integrating X402 payments into its own merchant-focused platform, Coinbase business. The second step would then be to push for LLM providers to let users pay for resources consumed by the model at any moment, which, if I'm honest, seems only plausible at stale if that payment was entirely optional, because on one hand, Open AI, Anthropic or whoever else, would effectively be conceding that it should have been paying to have its models access online content all along,

and on the other it would technically make using AI more expensive on the end normal user. Step into a new era of digital wealth with NXO, the Premier Digital Assets Wealth Platform. Earn interest on your digital assets, borrow against them without selling. Trade a wide range of cryptocurrencies, all in one place. Nexo is now available again in the US with an evolved product suite tailored to today's market. For a limited time, new U.S. clients can unlock 30 days of exclusive wealth club premier benefits, including enhanced interest rates, reduce borrowing costs and up to 0.5% crypto cashback on trades. Get started today at nexo.com slash breakdown.

As always, investments in blockchain technology involve risk. Terms and conditions apply. Do your own research. That's not to say there is an adoption already. Over the past 30 days, X402 scan reports about 25,000 active agents who have paid for stuff via X402 from around 8,000 active seller agents. Right now, most of that is the result of Virtuals Protocol, which is funneling $1 million per month into incentivizing its native agents to generate revenue, including funding a trading competition for AI agents trading on hyperliquid,

which Virtuals is calling Degenerina. That incentives program in particular seems to have boosted the average transaction amounts from around 15 cents, which has so far been indicative that the micro-payments thesis might indeed be starting to play out, to over $15 in March, which is perhaps due to the size of the trades been actioned by agents competing in Degenerina, which, if that's the case, only again goes to show how effective incentives programs are at boosting on-chain activity. Stripes MPP, meanwhile, is seen much less uptake, but is still only early days. According to Artemis, in MPP's first two weeks, there was about $9,500 in adjusted volume across

cumulative buyers and 67 cumulative sellers with average transaction sizes of between 7 and 77 cents to help make sense of all of this and discuss where we're headed i recently caught up with jordan loo co-founder of AI payments AISA here's some of what we spoke about today we are chatting all things agentic commerce and our very special guest is actually building in that space uh jordan loo co founder of AISER welcome to the show thank you very much davy thanks for having me have everyone i'm a founder of a ISA. Yeah. Cool.

Yeah. Maybe we could, I mean, I could probably try to explain what your platform's doing, but maybe you could just give a rundown of what ICE is doing in the agentic commerce space. Sure. Definitely. Yeah. In a simple way, ISA is building the capability layer and payment layer for the agent economy. So, for instance, if you're running any agents right now, like open clock or cloud call,

you can actually, we can enable your agent to access to the AI resources from like large junk digital tokens, license data, API calls, and we embed all the payments, each of the, you know, API call into the payment event. So all in, you know, in one event. So that's actually what we do. So I suppose the like the platform itself is taking a cut of those transactions or something like that. Like how does that work? True. Yeah. So from the payment layer, yeah, we have, you know, we're supposed to, just like Visa, right? You have this kind of transaction fee. And on top of that, we, because we also aggregates or the resources and an agent need to complete a certain task. So there is a take rates on top of that. Okay, okay, cool. So I suppose

a good place to start, like the news this morning, I mean, in between the time that I recorded the monologue portion of the show and today, I mean, it was only a few days, but then all of a sudden And there's been a big, I don't want to call it a shakeup, but there's like a repositioning of the governance of the X402 Foundation to come under the stewardship of the Linux Foundation. And I'll just read through because originally the X40 Foundation, it was launched by Coinbase and then they kind of brought cloud flare into the fold. But then all of a sudden there's the Linux Foundation.

Now in the release, there's also Stripe with initial intent and support being expressed by, I mean, it's a long list. There's ad yen, Amazon Web Services, American Express, ampersend.AI, AI, base, Circle, coinbase, Fiserve, Google, CacaoPay, MasterCard, Microsoft, Shopify, Salana Foundation, Stripe, a bunch of others. I wonder your take, because, I mean, I think when we chat a little bit before the show, but you've been building in the space for over a year. So that might have even predated the X402 Foundation.

I'm not even sure. but now we're out the other side. So I'm wondering your take on what this means for agentic commerce in general. We are lucky to be one of the pioneer into the X-4-2 ecosystem, and we are so happy to see the family grow. And we are, yeah, we are real, I need to actually build her on top of X-4-2. We were one of the largest X-Fo2, like, you know, a builder in terms of top-setter and top-server.

overall we have transact over like 20 million micro transaction. It's all production usage. I can explain more later. So yeah, that's actually we have been, you know, building on Exfo2. So the, yeah, if we come back to the, you know, authentic commerce, you know, perspective, there's definitely, you know, a very broad, you know, subject, a broad topic.

But for, we can divide into like actually two main categories. One, we call it human-centric transaction, the commerce. We're talking about the e-commerce. Like your agent, if you go to chatypt, ask them to recommend you to a pair of sneakers, there's a flight ticket, something like that. That's actually we call human-centric transaction. Usually it's large check size, like $10 for a cup of coffee,

$100,000 for flight, right? Then what we see actually that AI native use cases, is this kind of agent-centric, you know, A&A-NATIC, you know, transaction. We are talking about one API call for a large-anglemental inference token, or API, data, data API. So we are talking about this kind of micro-transsection, one sense, a fraction of a penny. So that's actually we are focused on, this kind of a micro-transaction initiated by these kind of agents. And I think there are lots of interesting things happening, you know, on these kind of native use cases.

That's how, you know, I saw and the rest of the, you know, the name, actually, the brands you mentioned about are all, you know, building towards. So that's my take. Yeah. Let's take a moment to talk about NXO. Nexto delivers a premier digital assets wealth platform designed to help clients build, manage and preserve their wealth, earn interest on your digital assets. Access crypto back credit without selling your holdings. Trade with advanced tools, all supported by 24-7 client care. Now back in the US, Nexo offers new clients 30 days of exclusive

wealth club premier access. That means enhanced interest rates, reduce borrowing costs, and up to 0.5% crypto cashback on trades. Benefits typically reserve for wealth club members and private clients. Nexo is also expanding its global presence, becoming the official crypto partner of Tennis Australia, the organisation behind the Australian Open, and the digital asset partner of the Audi Revolut Formula One team.

If you're ready to approach digital assets with a more structured wealth strategy, visit nex0.com slash breakdown to get started. Hey all, Blockworks co-founder, Michael Apolito here. Quick break to talk about something we've just launched, Blockworks investor relations. As the market shifts toward institutional capital, investors want more transparency, more standardization,

and a higher level of professionalism. But the traditional IR model is slow, manual, and not built for how crypto works. If you're building on chain, your data is already live, your business is already transparent. The challenge is turning that into a clear, credible story for investors.

That's exactly what we're solving with Blockworks IR. It's a single platform that brings together real-time analytics, branded investor portals, and hands-on advisory support so you can communicate what matters. If you're an on-chain business looking to level up your investor strategy, check out Blockworks Investor Relations at blockworks.com slash investor-dash relations. All right, back to the episode. As always, investments in blockchain technology involve risk. Terms and conditions apply.

Do your own research. It seems like it's quite a big deal that you do have the Linux Foundation. It's not that the Linus Foundation has been like totally against crypto or chains or anything like that, but it does seem like it's at least moving forward in a quite significant way in the past couple of weeks. And like I'm just wondering, like, if we could go back to like the usage that you're seeing flow through ISA because it's funny. Like crypto has always been trying to bring.

about adoption and if you go back 10 years, the whole thing was merchant adoption. And what I find really cool about the AI meta in general is that it is kind of rebooted the whole thing, because you, on one hand, you need to inspire at least like agent developers to bake the on-chain payments within their own agents. So you're actually trying to inspire developers to adopt X402 and maybe MPPP. But, But also, you have to get merchants themselves to accept stable coin payments and everything like that. So we're almost back to 2015 when we're trying to get everybody to start using this stuff

in commerce. So I'm just wondering what you're really seeing in terms of adoption, maybe like at the end of last year compared to now. Is it change very much? Yeah, I'll say there are a few actually fundamental change. in a whole ecosystem. I'll say number one, the men compared to even few months ago, you know, the agent ecosystem itself actually changed, you know, dramatically, right?

We see how open-cloth change everyone's definition of our agents, right? Even before that, we have to use like Chi Chi-GPT or even other like, you know, a cloud to, as a, you know, you know, managed host. kind of a solution platform to chat. And now with the open claw, everyone can have, you know, has their own, like, self-host agent and they can, you know, consistently work with this kind of heartbeat scheme, right? So I think that's the big, one of the biggest change from the Asian, you know, ecosystem. Another thing, I have to say, when we, because we are talking about agentic payments,

the payment currency itself, well, definitely will be the stable coin. you know, for the, for the, you know, authentic, you know, transaction. We see starting from circle and you go in IPO, everybody's saying in the real adoption for the real, you know, of the stable coin into the real company, you know, commerce, you know, the real world, all the adoption of the stable coin.

So I think that's the main and the infra itself, right? We say many different chance, you know, being launched and the transaction codes for, you know, for every, you know, transfer become, you know, cheaper and cheaper. So I think with the infra, with the agent ecosystem itself, now at this time, we are, you know, we are talking, when we are discussed, the agent payments is totally different from the world, you know, like even one quarter ago, end of last year.

So I think that's the main change from what we view, yeah. Yeah. Yeah, and I'm curious your take about, you know, this, because we're always trying to, like, classify usage to, like, try and find what the real usage is. And, I mean, just jumping off the agenda thing, more broadly like this conversation about meme coins, and I know this is even a year ago, that it was relevant. But a lot of it was like, well, yeah, chains are seeing a lot of adoption, but a lot of it is mean coin speculation. and is that even useful? And it's like, well, meme coins are actually kind of like the perfect stress test for these systems. So in lieu of like real quote unquote, real trading is okay to have pure speculation on meme coins.

But then like because like AI and vibe coding has come around, I'm, I no longer really am starting to see a difference between users, like real users and developer use cases. because when I look at people discussing the volume on X402, that, oh, it's just developers testing things. And it's like, well, actually users are developers now because they can vibe code. And if you're building agentic systems, then you might be a user, but you're a user developer. So like when you're going out and trying to attract people to use ISO,

like how are you viewing that? Are you trying to attract developers or you are trying to attract users, I guess is a way to frame that. Right. I think that's a very, very interesting question because we are talking about one fundamental question here. Actually, who are the agents, you know, who are actually driving this kind of, you know, use cases, right? So first of all, from what we see, right now we see three, yeah, types of activity, right? The first one is that this kind of what we call incentive-driven usage, like test farming, proof of concepts.

That's the real activity, but they are not real demand. And secondly, you're also mentioning about this kind of experimental usage, the developer, you know, wiring agent to pay for API, testing workflow, trying out this kind of X-4-2 or similar primitives, right? But the third category, which is the most important one, is the production usage. And that we say from ISA perspective, actually, that's already happening.

And yeah, we are saying agents actually paying resources to complete task. So not just moving the tokens, not mentioned the stable coin, you know, meme coin, but paying for the models, search API, data, and services. So I think that's the key distinction is that the real use case is not just this kind of a, you know, moving the tokens, you know, on, you know, a blockchain.

It's the agent paying for the resources to get it work down. And our definition of the agents, actually, we have two kinds of agents right now. We have this kind of individual agents like OpenCloud. If you're using any of them, they can directly come to ISAP platform to, you know, get immediately sign up, get an API key, and access all the resources they need. And same time, we also serve those kind of enterprise agents, the developer needle driven agents. They can be like a financial research agent, news aggregate agents, like HR agents,

language learning, you know, agents. Yeah, they all come to our platform to access to the resources. So yeah, from our, you know, the use cases on our platform, yeah, we are targeting the real developer or the developer developed, you know, AI agents. So that's actually what we see right now. Yeah, cool, it's interesting. And it's like I can only come at this from, like, trying to understand the economies of scale of this,

because if ICER is focusing on the purgentic side of it rather than the e-commerce side of it, then they are going to be quite small transactions, like literally micro-transactions. Maybe we can end on what you see are the roadblocks to really getting there, because I mean, it is exploding so far. So it seems like, I mean, whatever roadblocks were there a week ago aren't really there anymore. And then obviously we now have a lot of weight behind the X402 Foundation with the Linux Foundation too,

which I have to imagine will bring a lot of developers on board too. So it's like, are there any roadblocks here? Because it seems quite open right now. Like, yeah, I'm curious. Yeah, true. Yeah, first of all, definitely the infra layer. So we're talking about when agents really, you know, become an economic actor, they can make their micro-decision and make the micro-purchase, right? If every single API-cold transaction brain on chain, you think about that might be like millions or even 100-million transactions per second.

None of the existing infra block, we're not talking about the centralized kind of payment rule like Visa Mastrichta. They're actually handling like $65,000 transaction per second. And the existing like layer one, layer two, they're actually maximum, even paypole, temple, like, you know, Solana, they are talking 100,000 K transaction per second. None of there actually the infra, you know, viable for that kind of volume. So I would say that's from the infra, you know, perspective as a bottleneck to facilitate this kind of high frequent micro-transection. And then another layer is definitely how the agents. the framework, the frontier labs can really make that things work for us.

The security issue, right? That's another big topic. Then how capable the runtime, the framework can develop over time to become, you know, to make the agents can really do the work for us and do the work for themselves. So I think that's the main two, you know, potential bottleneck, right?

in a, you know, yeah. Yeah, it's like it's, we might be in for a future where there is also off-chain solutions to these, to some of these things as well. But in the meantime, there is still a lot of growth to be had, I guess. But this has been great. Thank you so much, Jordan. And we hope to speak to, I hope to bring you back on the breakdown soon. Cool.

Thank you so much. Thank you very much, Davey. Right. And that's about all the time we have for today. Stay tuned for the full conversation with Jordan later this week. And we'll see you next time. You know,