x402: The Key to Internet Money, Micropayments & The AI Agent Economy | Sam Ragsdale

David:
[0:03] Bankless Nation, I'm here with Sam Ragsdale from Merit Systems. Sam, welcome to Bankless.

Sam:
[0:08] Excited to be here.

David:
[0:10] X402 is the topic of DeJure on crypto Twitter. People loosely know it as a payment protocol for AI agents, but I think we need to go into a lot more detail here on the episode. So maybe I will just start with the question, what is X402?

Sam:
[0:26] Yeah, so starting from the top, X402 is a protocol, more general than agents, for gating access to API resources, in other words, off-chain servers via crypto payments. I think it helps to go back to the beginning. HTTP is the transport protocol for network packets on the internet. And it was designed in concert with the browser by Tim Berners-Lee, Mark Andreessen, and a bunch of other people that I'm not going to name drop right now in sort of the early 90s while the browser was being created.

Sam:
[0:55] The status codes 400 and 500 are sort of the error status codes. The 200 ones are the success status codes. 402 is very low. It's very early on because it was very important to them that they reserve that one for payment required. It was very obvious to the people creating the internet that the most obvious way that a server would give you resources would be that you would pay directly for that resource. Particularly in the 90s, when the status code was created, servers were like three orders of magnitude more expensive than they are now. And so the question to the people creating the internet, the first browser, was why would you put a server online? And then why would you respond to either a human or another server asking that server for things? And the answer to them at the time was that you would pay for it. Unfortunately, microtransactions did not exist at the time. This was pre-PayPal. Payments on the internet didn't really exist either. And so there wasn't actually the technology to be able to do micropayments on the internet. So it has remained unused since then.

David:
[1:56] So it's just been a placeholder. It's just been this 402 thing that just sat on the shelf and be like, hey, once we figure out money, we'll take it down on the shelf and build this thing. But otherwise, it's just been this empty and empty vehicle that we will figure out payments in later.

Sam:
[2:12] Yeah, exactly. And some people do implement on the server side, like a 402 saying, you know, you haven't paid for this thing when your API key doesn't have any credits in it. But there was never a way to respond to that. It would just be an error that you couldn't deal with and you would go through some admin console or something like that.

David:
[2:29] I see where this goes. So I feel like now with the excitement around X402, we now have a way to respond to that.

Sam:
[2:37] Yes, exactly. So how it works is that a client, whether that's a browser, a human on a browser, or a crawler, or an agent, or some sort of other entity you could come up with, pings the server at a specified URL. That server responds with a 402 status code. In an X402 land, it provides some additional details on how it can be called and how much that thing costs, and that it will accept X402 payments on some chain in some currency. And then the client signs a crypto transfer, usually USDC on base, but Solana is coming, people are doing other coins as well. And it attaches that signed crypto transfer to its request in the headers as an X payment header, sends another request to the same server, the server takes that crypto transaction, it settles it, and then it returns the resources back.

David:
[3:27] Introducing FraxUSD, the genius-aligned digital dollar from Frax. It's secure, stable, and fully backed by institutional-grade real-world assets, custodied by BlackRock, SuperState, and Fidelity. It's always redeemable one-to-one, transparently audited, and built for payments, DeFi, and banking. The best of all worlds. At the core is FraxNet, an on-chain fintech platform built to align with emerging U.S. Regulatory frameworks where you can mint, redeem, and use FraxUSD with just a few clicks. deposit USDC, send a bank wire or tokenized treasuries, and receive programmable digital dollars straight to your wallet. FraxNet users benefits from the underlying return of U.S. treasuries and earn just by using the system. Whether you're bridging, minting, or holding, your Frax USD works for you. Frax isn't just a protocol. It's a digital nation powered by the Frax token and governed by its global communities. Join that community and help shape Frax Nation's future by going to frax.com slash r slash bankless. Frax, designed for the future of compliant digital finance. Imagine a world

David:
[4:24] where traditional finance meets the power of blockchain seamlessly. That's what Mantle is pioneering, with blockchain for banking, a revolutionary new category at the intersection of TradFi and Web3. At the heart is UR, the world's first money app built fully on-chain. It gives you a Swiss iBand account, blending fiat currencies like the euro, the Swiss franc, the United States dollar, or the renminbi with crypto, all in one place. Enjoy real-world usability and blockchain's trust and programmability. Transactions post directly to the blockchain, compatible with TradFi Rails and packed with integrated DeFi futures.

David:
[4:54] UR transforms Mantle Network into the ultimate platform for on-chain financial services, unifying payments, trading, and assets like the MI4, the MEth protocol, and Functions FBTC, backed by developer grants, ecosystem incentives,

David:
[5:07] And top distribution through the UR app, reward stations, and Bybit launch pool. For MNT holders, every economic activity in UR drives value back to you, embodying the entire stack and future growth of this super app ecosystem. Follow Mantle on X at at mantle underscore official for the latest updates on blockchain for banking. That's x.com slash mantle underscore official. Crypto is risky. Your sleep shouldn't be. Eight Sleep's mission is simple, better sleep through cutting edge technology. Their new Pod5 is a smart mattress cover that fits on the top of your bed. It automatically adjusts the temperature on each side so you and your partner can both sleep the way that you like. It's clinically proven to give you up to one extra hour of quality sleep per night. Eight Sleep's Pod5 uses AI to learn your sleep patterns, regulate temperature, reduce snoring, and track key health metrics like HRV and breathing. With a new full body temperature regulating blanket and built-in speaker, it is the most complete sleep upgrade yet. Upgrade your sleep and recovery with A-Sleep. Use code bankless at A-Sleep.com slash bankless to get up to $350 off the Pod 5 Ultra. That's A-Sleep.com slash bankless. You also get 30 days to try it risk-free. Link in the show notes for more information. We couldn't do this back when they were building out this in the first place because we like just didn't know what would go there, right? You can't put a credit card there, or could you? What were some of the first theorized ideas of how this would have worked anyways?

Sam:
[6:27] Yeah, so I think theoretically you could put a credit card there if the resources were like $100 or something. But microtransactions were not really a concept. I don't actually know the fee rates in the 90s. That was before I was around. But they were very expensive. I think sort of even now, like Stripe will charge you a 30% minimum or sorry, a 30 cent minimum on each transaction. So microtransaction, one cent transactions for credit cards doesn't really make sense when there wasn't the technology in place to do a cent or a tenth of a cent transactions. This is also even before PayPal existed. And so even with credit cards, there was not a great way to transact with credit cards on the internet, even if you had an $100 transaction.

David:
[7:09] Yeah, I guess the premise of, because of the economics of credit cards, you could only pay for something large. And if you're going to pay for something large, just write it into the straight web page. Like if it's so important that it's $10, $20, $30, $100 plus dollars, then they can take out their credit card and they can manually type it in the transaction. They're probably going to want to do that. anyways.

Sam:
[7:29] Yes, exactly.

David:
[7:31] So was it always theorized to be micropayments?

Sam:
[7:33] I'm not sure. I think my guess, we'd have to go ask Tim Berners-Lee and Mark Andreessen, which maybe we can get them on the next episode. But I think it would be quite hard to price that at the time when you're creating the browser. People didn't have, the internet as a concept didn't really exist. Blogs didn't really exist. E-commerce didn't exist. Crawlers didn't really exist because there wasn't enough interesting content to crawl. There wasn't an internet industry. There wasn't streaming video, any of this stuff. And so I think it would be very hard to, for those people theorizing the internet to price what these resources could be or what they could cost. All they knew was that the cost of running servers was so egregiously expensive that we needed to come up with some business model. The business model that fell out of it, which was not planned by people creating

Sam:
[8:23] the internet and the browser was ads. And we can get into that in a second and see how that transformed the 21st century.

David:
[8:31] Yeah, ads filled the niche of micropayments because we didn't yet have micropayments. And I guess you could have like, just to kind of continue down this thought process, you could have also have put in bank account numbers too and be like, oh yeah, I want to access this webpage, yoink, 30 cents for my bank account, here's my bank account number. But maybe it's one problem with that was like, A, banks can't handle that amount of traffic, that amount of volume. It's like almost a DDoS attack. I don't know, maybe technically it would work, but it would be a lot. And that's not really what bank competency is, is not like high volume of internet traffic. So that's one problem.

Sam:
[9:07] Wires are $5 to $20, you know, like they're even worse than credit cards.

David:
[9:11] Yeah, right. True. Yeah, good point. Yeah, yeah, yeah. But then the other problem with that is like, well, then you're just exposing your bank account number. Yeah. And that's kind of how checks are. So we actually do kind of expose our bank account numbers quite frequently, but putting it and posting it on the internet and having it part of the internet protocol seems just like far more dubious in that context. And so I guess that brings me to the question of just like, so when I do an X402, to whatever degree that means, is this encrypted? Is this not encrypted? What is the level of... Surveillance that goes with this or exposure to the rest of the internet that this has?

Sam:
[9:48] Yeah. So it's a stable coin transfer on either an EVM chain, most are supporting it now, or Solana, which means that it's not shielded in any capacity. It is settled on the EVM, which means that you can look at it and you can look it up. Of course, unlike providing your credit, the naive way of doing credit card transfers or bank transfers, if you provided that raw, somebody else would be able to draw down against your credit card or your bank account. Of course, with crypto, you have a one-time signature that's provided that can never be used again.

David:
[10:20] And that's the key secret sauce.

Sam:
[10:22] Yeah, exactly. And then I think the other thing that's interesting to call out in contrast to credit cards and bank transfers is, and this is a common thing in crypto, but the reason that stable coins, in my opinion, are a fundamentally better financial technology than the existing TradFi system and everything that comes with it is that they don't have chargebacks. There's no concept of chargeback risk or a dispute or anything like that. Every single transfer in the traditional financial, traditional banking system can be clawed back at some point. Credit cards are the most obvious. Any e-commerce merchant is familiar with this. You are responsible for your own chargeback risk. That's somebody that comes and buys shoes, you can ship them the shoes, and then they can cancel that credit card transaction. And there's a massive fintech industry built around this problem. Same thing exists at the banking level, wires, ACH, RTP, all of this stuff can be clawed back, it's a little bit harder to do. And then you can even keep going up the stack like international wires, you know, wires at the level of the Fed and the Treasury, these things can be clawed back through massive of international disputes. And all of that leaks its way down into the devx of people trying to accept payments and build payments processing infrastructure. With stablecoins, you don't have that. It can't be charged back. Once it's settled and confirmed on the chain, depending on which chain it is.

Sam:
[11:46] You're done. It's one payment, it goes through and you're good to go. You've accepted that payment and you have it forever.

David:
[11:53] So when Tim Berners-Lee and Mark Andreessen built the browser and built JavaScript and all that kind of stuff, they didn't know what would fill the X402 container. They just said, like, okay, we have the notion of money, don't really know what goes there. And now we invented crypto, crypto money on the internet, no chargebacks, internet native, nothing to do with a legal meatspace system. And now we have a payment system that is resonant with what is required of it for it to work inside of an internet. And like they didn't even know that that was coming.

David:
[12:26] Maybe they didn't even need to know, but it doesn't really matter. Now we actually have it. And so what, like Brian Armstrong and Coinbase decide to take this thing down off the shelf and build a standard around it? How did that step work?

Sam:
[12:38] Yeah, so I think it's really just recently that blockchain's stables actually have become sufficiently cheap that microtransactions are plausible in volume. That's a good point. And so X402 is agnostic. It works on Solana, which Solana is significantly cheaper than most EVM chains. And then the primary usage, about 99% of it so far, is USDC on base. Base is an L2, an OP stack L2. That technology is sort of very recently around. So had you tried to do this in, I don't know, 2016 with Bitcoin or 2017 with ETH, it wouldn't have been that plausible. But very recently, sort of last cycle, it became plausible that you could send one cent transactions and that the fixed costs there. because of gas, we're not egregious. It is important to note that X402 is the crypto standard for this, for using USDC or something else. Cloudflare also has a version of this that's a 402 standard, which is pay per crawl. And so far, the Cloudflare version of this operates in fiat on basically database credits on the Cloudflare side with the net token and other things they might lean into using crypto as well. Interesting.

David:
[13:51] What are the rules behind who can just make a standard? Can I just make a 402 standard?

Sam:
[13:57] Yes, you could. You would need some traction. You would need lots of people using it. The main thing, I guess there are internet standards bodies like the IETF, the Internet Task Force, and others that create real internet standards. But in some sense like you know every eip erc20 etc goes through the the ethereum foundation and they have credibility because lots of people build on ethereum and trust the process so it's a group of people that agree to a common standard and then the benefit of the standard is that kind of like erc20 if you create a common set of abstractions then everybody that is using that set of abstractions gets to, interoperate with all of the others. In the RC20's case, you create a network effect, one that's very familiar to X402 as well. If all of the buyers of X402 resources can find all of the sellers of X402 resources and you can interact, you get an exponential effect on the value of the overall standard.

David:
[14:55] Of the standard. Yeah, the standard turns into a network. And so like no one's forcing anyone. It's just being proposed. I'm sure a bunch of developers could put up their heads together. And it's like, let's think about designing this smartly in a good way that's scalable and accessible to the most participants on the internet.

David:
[15:12] So this network effect can grow and then we can get money on the internet. Can we like deconstruct an X402 transaction? I think you kind of did it, but I want to go through it a little bit more methodically. Like what are the steps? What's the information? There's got to be some public keys somewhere. There's a signing message. There's a currency that I accept. Can you just like deconstruct how one is formed.

Sam:
[15:35] Yeah. So step one, you need to know the URL of your resource. And the URL is where it lives on the internet. It's its address. And that largely today comes out of the bizarre or comes from X402 scan, which is what we shipped about three weeks ago, which we can get into in a bit. But once you have this URL, you can ping that resource and you can send a request to it with basically an empty body. And you say, I would like access to your resources. You don't include any payment information in that because you don't know what it costs yet. And then the resource responds with a 402 status code, and then the x402 spec of what it requires as input. And so, for example, if it's an LLM generation, GPT-5 generation over x402, It will respond with like, please type in the prompt, the temperature of the prompt, and like max output tokens or something like that. You will then fill out that schema. You will sign a transaction. It's probably a USDC transfer on base.

David:
[16:37] So USDC contract address, base chain ID number, all that stuff kind of goes in there?

Sam:
[16:44] Yeah, exactly. It's the gasless transfer on base. There's an EIP that I'm not going to be able to cite the number of, but it's a gasless transfer of USDC. You don't actually send the transaction. It's not actually a full transaction. You submit that to the server in the header. The server takes that. It settles with what's known as a facilitator, which is basically just a server that will pay for the gas on your behalf, which makes it really much simpler to run one of these servers because you don't need to put an RPC URL and all that stuff in. And then it will submit that. It'll settle it. It'll know it has the money in its account. and then it can send back the original resource requested. In our case, it would be the LLM generation.

David:
[17:23] Right, okay. And you call this like a single signature, a one-time signature. It sounds like they're writing a check. I'm just writing a check. I'm just like, hey, 100 USDC, here is a signed transaction and here it is. And then the receiver can execute that signed transaction and upon execution, the specified amount of USDC gets transferred from address A to address B.

Sam:
[17:47] Yeah, exactly. And it has a nonce on it, which can only be used once. So the idea there is if I sent you one of these transactions, I wouldn't want you to be able to submit it seven times and debit my account.

David:
[17:59] Cool. And then what that does is that just kind of just unlocks, it just like smooths over some frictions. It's like, oh, I can start to do things a little bit faster because I don't have to pull my credit card. Like loading the website is synonymous with making a payment. What if, does like a MetaMask pop-up happen? Or like, what's the user end experience like?

Sam:
[18:18] Yeah, so this is, so we started, we launched X402 Scan about three weeks ago now. Before we launched it, the ecosystem, the Coinbase launched X402 in May, roughly. And then before three weeks ago, X402 is about 25 transactions overall. Did about 10K of volume and there was about 200 resources on it. Since we launched X402 Scan, so in the past three weeks, we're up to around 3 million transactions, about $3 million of volume transacted, and about 10,000 resources. So like on those metrics, somewhere between 10,000 and 200,000% growth on pretty much everything. And when we looked at the ecosystem, we saw three things missing or making it harder for it to grow. This comes back to answering your question, I promise.

Sam:
[19:06] One was discoverability. At these resources, there was immense friction between supply and demand because it was hard to find those URLs and then hard to know what services they provided, what they cost, and then how you could call them. And then two was that the schemas were, it was quite hard to understand the schemas and because there was no constant place to call them, there was no MetaMask equivalent. A lot of them were actually broken and they couldn't be called. Like I think we were some of the only people to actually call all 200 resources that existed before X402 scan. And about 75% of them just straight up didn't work because there was no common standard for how they should be called. There was no Meadowmask. There was no Window.Ethereum. There was no Phantom, Rainbow, etc. equivalent. And then three was that these things couldn't really be composed. And I'll get back to that one.

Sam:
[19:53] Discover ability helped quite a bit. It made it much clearer which resources could be called, how many people have called them in the past, whether they're reputable as a result of that. And then what we built in directly was the ability to call all of these things. So people could just, you know, hit the one click button to test it out. And I think that has made it much easier because when a developer is building one of these servers, they think about, OK, is it going to work under the X402 scan standard? And therefore, can users actually use it? And then when users go and look at the server, they can see it's been hit 500 times by 10 different users. And they can believe that if they send their $1 across the wire, they're actually going to get resources back.

David:
[20:34] So you guys are writing the standard, the calling standard.

Sam:
[20:36] So it's technically in the spec, the spec that Coinbase provided, and they did a great job with it. The thing that was missing was that there wasn't a standard place where people could call these resources and try them out. But as a result, people were often mis-implementing the spec and not realizing it. It's really what we've done is provide observability tools and testing tools so that there's one interface that you want to conform around, and people build servers that work and are reliable.

David:
[21:05] Isn't this still a little bit more manual than it would otherwise be theorized to be in the future? Like, isn't a lot of this supposed to just kind of be more in the background?

Sam:
[21:15] Yes, meaning that today people are calling resources individually rather than sort of agentically.

David:
[21:22] Yes, I think so. Yeah.

Sam:
[21:24] Yeah. So this is sort of the last piece that we are really excited about and are working on and are pushing towards over the next two weeks, which is composability. The dream of X402 is really that you don't have, so the standard way that you build web apps is you have your API layer, which is your server, and then you have a client for that server. And so you have like your front end and you have your back end. And these things talk to each other very closely. And you really don't build your API for any other web apps to be able to call it. You wouldn't have Facebook calling out to the Twitter API and calling out to the YouTube API. Rather, the Facebook web app calls the Facebook server. And X402 thinks about this differently. Like the dream is a composable future where you have five different APIs that serve different purposes. Maybe you have your generative image one, you have your newsfeed one, you have your custom shirt printer one, and then you get your agent, you load 10 bucks into it, and you say, I'd like to create a relevant news posting

Sam:
[22:23] shirt and send it to my friend for his birthday. And you compose all these APIs together. It's kind of like the money Legos of DeFi, that idea, except for APIs off chain.

Sam:
[22:33] The issue that we saw when we started working on the APIs is that a lot of people were building them in the skeuomorphic way, which is like the single client, single web app that are bundled together and have strong opinions. And they weren't very easy to compose. They weren't designed with interfaces to. And so X402 Scan now has a composer tab. We called it composer rather than agents because we really want to get across this idea that people should be building API resources that can be composed together. And now anybody can build an agentic style flow pull any of the 10,000 X402 resources off the shelf, put it within their agent and compose between five or 10 or 200 different resources.

David:
[23:12] I think we skipped over the idea of like, does a MetaMask transaction window like pop up? And like part of that is like, well, how do I even load up money in the first place? So like, how do I get started? What are the first steps I need to do to make this work?

Sam:
[23:26] Yeah, so you do need a client. X402 scan should work as your client for now, but a bunch of other people are building them. Hopefully, they'll build them into their website as well.

David:
[23:35] A client as in a place where I can go and press the connect wallet button?

Sam:
[23:38] Yeah, exactly. Similar concept to Uniswap, V2, V3, V4 are smart contracts that live on ETH, but you still need a website that has an opinion of how to call those things. Same thing exists for X402. You have X402 servers. You still need a website that has an opinion on how to call them. X402 Scan will do this either through embedded wallets or connected wallets. So we have embedded Coinbase wallets built in. You can sign in with your email. You can load it up with some ETH, and then you can call any X402 resource, or you can connect your external wallet directly. And then if you hit any of the X402 fetch endpoints, we will have you sign a transaction in line, and we'll submit it for you.

David:
[24:16] So a window does pop up, and I do have to manually sign a transaction.

Sam:
[24:20] Yes, exactly.

David:
[24:21] But I could also tell my agent to do that too. And the agent has the keys.

Sam:
[24:25] Yeah, if you load money into the agent, which would be a signature or a transaction on chain, then your agent can sign on your behalf in the background.

David:
[24:33] Okay, cool. Cool, cool, cool, cool, cool. And then I can give my agent more or less autonomy as I see fit.

Sam:
[24:39] Yes, exactly. And the idea is sort of like the agent models, largely Claude 4.5 Sauna and GPT-5, have gotten really good at only doing exactly what you tell them to do, particularly when you put in context that this thing costs five cents and so please don't call it unless you're sure. And so they're very good at knowing when to tool call and when not to.

David:
[25:01] I would imagine that would be a primary concern.

Sam:
[25:04] Yeah, but you load up sort of like $5 at a time.

David:
[25:08] You try it out. You don't load in your cold storage wallet.

Sam:
[25:11] Yeah, exactly. I think the most expensive one that we've ever done a few times is the shirt.sh x402 endpoint, which is one of our ones that we've built as a side project, but it is the Printify API. You can load up an image and then it will print a shirt and it'll send it to your house. And so you can load up an agent that hits, you know, GPT image one generates a picture of like a guy driving a sports car in the mountains, and then sends it off to the shirt.sh API, prints a shirt, then we send it to our friends as a joke. It's about 20 bucks per time. And so it's, you're giving the agent a lot of autonomy to spend there, but we haven't seen it screw up yet.

David:
[25:49] And then I guess really the big, like bummer about that whole process is probably it takes a couple of days to ship to you because That's the slowest part.

Sam:
[25:57] Yes, totally. We definitely need atomic shipping, but I don't think we're getting that anytime soon.

Sam:
[26:04] Oh my God.

David:
[26:05] Amazon drone it just straight into my doorstep. Okay. What are the resources that people are spending? So like you kind of like ran through some numbers and you're like, yeah, there's like a, we're up 10 bajillion percentage of transactions and like resources and, and what are people doing?

Sam:
[26:21] Yeah. So of the, the, the increase in volume, the 200,000% increase of volume. I've never thought I'd say numbers like that. In the past three weeks, about 99.5% of it by volume is meme coins. People are launching, you know, it's crypto, you might expect. People are launching X402 meme coins and meme NFTs. It's not a particularly good way to launch a meme coin because you rely on a trusted server in the middle. I think that's, and so if you go on x402scan.com and sort by volume over the past seven days, you'll see, you know, 50 or 100 of these um but i think the.

David:
[26:58] Most like first and obvious thing that i would do is like yeah i want to play with this a new protocol let me like trial this out what's the resource that i'm going to do fuck it i'll make a token and that's the token and then like i'll let people call the buy token button and like haha that's funny and then people also think that's funny and then some degenerate gamblers get in and it's like oh i'm i'm actually gonna buy this and that's kind of how the first wave of all of these things go i

Sam:
[27:22] I think the first the first few of them actually did rip and it was actually like a pump fund style mechanism where you send the money, which builds up the liquidity pool over time. And then they put that into a Univy two pool or Univy four with limited liquidity.

David:
[27:36] Yeah, that's just a great way to stress test the system. I'm a fan.

Sam:
[27:42] It definitely stress tested our systems as well. We were not expecting 200,000 unique visitors over the weekend. So there was some significant database migration.

David:
[27:50] 200,000? How many are actually unique, do you think?

Sam:
[27:53] That's based on Vercel and PostHog Analytics, which are generally pretty good. Crypto users have, let's say, more active VPN usage than anybody else. So I think it's probably actually between 100 and 200,000 unique users. Unique wallets is.

David:
[28:09] Much higher than that. I'm actually surprised by that. This seems very high. Wow, that's crazy.

Sam:
[28:12] Okay. So, okay.

David:
[28:13] Meme coins, taking out meme coins, what's the most common resource that people are buying?

Sam:
[28:17] So I think this is sort of the important thing to note is that by volume, either transaction count or dollars, the majority of the $3 million transacted, like 99.5% is meme coins.

Sam:
[28:30] But if you look at resources, so we've gone through the 10,000 resources and the resources in the past three weeks have gone from roughly 200 to 10,000. And of the resources, only about 50% of the net new ones are meme coin related, which means that there's like, you know, about 4,000, 4,500 resources that are newly added in the past three weeks that people sort of just heard about it and started adding stuff that have nothing to do with meme coins. And I think that's sort of like the shift you see a lot of times during, I had a tweet about this last week, during bubbles where rapid price increases make people really excited. It gets a lot of attention. Users come, real developers come in to fill and meet that demand. And then, you know, even as the price fluctuates, the developers fluctuate on a slower cadence.

Sam:
[29:19] They stay around, they build real stuff, and that gets the ecosystem going and kickstarts. And I think we've very much seen that. And so the stuff people are building, I think there's the raw AI stuff is kind of the most obvious. I think the reason for this is that maybe the most interesting API that has ever existed is like GPT-5 over API. It's like a very, very rich API resource that you'd be very excited to spend, you know, tens of thousands of dollars on. Then there's the Gen AI, so GPT-Image 1, Nano Banana from Google, and then the video stuff. So Sora and Sora 2 and VO3 from Google.

Sam:
[29:54] People have, I think there's five to 10 endpoints for each of those. Then there's a category I'd call enriched AI resources, which are one of those services, usually an LLM, but with some additional context. And so like AIXBT has their special...

Sam:
[30:10] Endpoint, which is a fine-tuned LLM that also adds in all of the crypto Twitter data, as well as some token prices and stuff like that.

Sam:
[30:19] And then there's been a bunch of people building like stock market bots that can pull in all of the prices. They can pull in indicators like RSI and MCD and stuff like that. And then they can run an LLM over that as well. A bunch of other data resources. People have put up polymarket APIs that are a bit richer than the raw poly market ones that charge you a cent, weather APIs, horoscope APIs, stuff like that. I think some of the most interesting ones actually are the walled garden data APIs. So some of the hardest to get data today is the data from Facebook, X, and LinkedIn. And there's a lot of Web2 startups that are scraping this stuff at a violation. I won't name names, but are scraping this stuff at a violation of TOS and then reselling it. And then there are some CRM style tools that buy this data and then create a CRM experience where you can take a thousand rows that you want to do outbound on, enrich with their LinkedIn history, their X history, and then write a much more compelling outbound email. This is the type of thing that actually makes a ton of sense over X402, ignoring the toss violation, because you are very willing to pay like a cent per enrichment. And that data is actually extremely hard to get. The Twitter API costs something on the order of $50,000 a year and scaling depending on what you want to do with it. And really what but what you want is like one person's social graph. It's very, very hard to get.

Sam:
[31:40] And then I think I'm excited about the custom goods category as well, the shirt.sh style things where people compose a few of these APIs together and then get a physical good out of it. And so there's headless checkout from Amazon and Shopify as well, where you can buy pretty much anything that's available on the internet over X402.

David:
[31:58] That was very useful to really get my imagination going. And kind of what I'm thinking right now is that not having... This power, this internet superpower of microtransactions and like automated commerce and resources, being able to call resources, resources as like a very broad generalized term has stunted the growth of just like internet native commerce in an incredible fashion. Like if we had had native stable coin or native micropayments a decade ago, the internet would look so different than how it looks today. And like that was what I just like kind of felt hearing you is like, oh, you could do anything with this. You could, the possibilities here are truly limitless.

Sam:
[32:45] Yeah, I think all of, most of the ones we've seen so far are pretty skeuomorphic in terms of their like, take all of the existing API stuff we have and then wrap it and then see what happens. And take something that used to cost, I guess an interesting thing that people are doing is take an API that costs $1,000 a month and basically fractionalize it, right? $1,000 a month for infinite usage for a single enterprise. One enterprise buys that, rewraps it, and sells usage-based pricing on top of it. And I think that one is pretty interesting, but it's still fairly skeuomorphic. It's like, take the way that enterprises used to buy one-off services and then turn them into X402. And what we're hoping is that this giant spike in demand and then the supply that has followed will start to build out, like, what is the new shape of resource

Sam:
[33:32] that makes sense in this paradigm? Like how can you build an API that's particularly interesting at one cent per call that's something other than an LLM or something like that?

David:
[33:42] Right, because if we can really cut down the cost of an individual transaction, then the permutations, combatorial permutations of what can be built can explode. And that just means a lot more commerce can happen. It's like, what can you do if you can just like, call ChatGPT once and Amazon once and like three other things a few times and that's wrapped into one package. Like it really is just like up to the imagination of the developer to be able to do cool things.

Sam:
[34:12] Yeah, exactly. And I think the other thing is also that if you were to call an API before, you were sort of bundling payment and identity as two different things. When you wanted to sign up for an API, you would typically sign in, you'd convince them that you're a human, You'd sign an enterprise agreement. You'd put in a credit card and you'd agree to pay whatever bill that they send you at the end of the month. And then you would draw down against that bill over time. They're implicitly extending credit to you. You're taking on debt against them over the course of the month. And then they trust that you'll pay that at the end of the month. In agent land and in the land where you have an agent on your behalf or on behalf of your enterprise that is using lots of resources on the internet for something like outbound CRM work or something like that, it's actually very cumbersome to set up an identity for that agent and then pay for their work at the end of the month. You'd prefer a system actually where you can go to 300 of these enrichment APIs. You can spend a cent on each of them and then the agent can determine if those results are good and then decide to load up the truck on the 299th enrichment API without having to sign a legal agreement with all 300 of those API providers.

David:
[35:21] Yeah. And importantly, it gives agents, AI or human directed, I guess they're all human directed, first class access to the internet, to resources on the internet.

Sam:
[35:31] Yeah, exactly. Because all you need in the case of a payment that cannot be charged back is that the payment has been settled. If that's settled, then you're happy to provide your resources.

David:
[35:43] Yeah, you don't need a contract. You've received payment, you provide resources. Why do you provide resources even after you've already received the money? Well, because if you are a good resource provider, you will get more money again in the future for more people.

Sam:
[35:58] Yeah, exactly. And then particularly in the case where there is a proxy layer like X402 Scan, we run the request, and so we know whether or not they provided the service that you paid for, and we know whether you paid for it. And so we can show sort of this uptime percentage thing, which is like, how often did this server actually return the resource that it requested? And it actually was very common before X402 scan existed, like 75% of these 200 servers that were around beforehand where they just totally would not conform to schema or provide responses. And so the hope is that over time, Us and other providers can build out reputation scores here where you'll know in advance how many other people have called this thing, how reliable those people are, and then how reliable the server was to each of those responses.

David:
[36:48] Ethereum's Layer 2 universe is exploding with choices. But if you're looking for the best place to park and move your tokens, make your next stop, Unichain. First, liquidity. Unichain hosts the most liquid Uniswap V4 deployment on any Layer 2, giving you deeper pools for flagship pairs like ETH USDC. More liquidity means better prices, less slippage, and smoother swaps, exactly what traders crave. The numbers back it up. Unichain leads all Layer 2s in total value locked for Uniswap V4. And it's not just deep, it's fast and fully transparent, purpose-built to be the home base for DeFi and cross-chain liquidity. When it comes to costs, Unichain is a no-brainer. Transaction fees come in about 95% cheaper than Ethereum mainnet, slashing the price of creating or accessing liquidity. Want to stay in the loop on Unichain? Visit unichain.org or follow at Unichain on X for all the updates. BitDigital, ticker BTBT, is a publicly traded ETH treasury company that combines the two biggest metas of our time, Ethereum and AI Compute. BitDigital believes that ETH will power finance and AI Compute will power everything. BitDigital gives you direct exposure to both. BitDigital holds more than 150,000 ETH with institutional-grade staking and validator operations. On top of that, the company owns roughly 73% of White Fiber, an AI infrastructure business that runs high-performance GPU data centers that adds a meaningful exposure to the growth of AI Compute with over 27 million shares. This is an ETH treasury backed by real operations, designed to capture staking yield today while positioning for the future of intelligent computing tomorrow. The ticker is BTBT. This ad is not financial advice.

David:
[38:16] Do your own research, learn more about BitDigital and try their MNAV calculator at bit-digital.com. That's bit-digital.com. Bankless is being compensated by BitDigital for this ad. You can find out more information by clicking the link in the show notes. Sam, this is incredibly cool.

David:
[38:31] I'm pretty stoked about this. Mainly just because there's so much potential here. There's a lot of GDP potential. There's a lot of wealth creation potential. There's a lot of startup opportunities. I think the only thing that I know that is truly tangible to me as an example to make this as real as possible to the listeners is...

Sam:
[38:49] Ads.

David:
[38:50] I don't know if you've noticed, but the internet has gotten so fugly lately. Like I log into like Forbes or Wall Street Journal or trying to read an article on the internet, even Coindesk. It's just so cumbersome. There's like 17 pop-ups. There's a bunch of ads. They make me sign in. I have to pay a subscription to anything to read anything. And it's just such a human friendly and so unfriendly to humans. And I'm just like, man, if I could just pay them five cents to just never bother me again and just serve me the goddamn article that I've wanted to read so I can do my job as a podcaster. Five cents every time I load up that website for sure I would pay that. I'd pay them even more. And I would assume this ought to be a very low-hanging piece of fruit for some forward thinking, forward-looking media website which I own one, maybe they should just do that and stop asking people to sign in and pay the money.

Sam:
[39:53] Yeah, I think that there's a great irony here that I've really enjoyed seeing develop, honestly. So my first job out of college, I worked as a software engineer for Google Ads, the great ad monopoly of the past 20 years. And when the HTTP 402 status code didn't work out and they didn't figure out microtransactions, shortly thereafter, the internet, largely Google, figured out and monopolized ads on the internet. And in the process, they subsidized the free and open internet, right? Google is the biggest supporter of the free and open internet that has ever existed. They've done really, really great work there. The reason that the free and open internet exists is because ads created and incentivized...

David:
[40:32] Ads funded the growth of the internet, yeah.

Sam:
[40:35] Yes, exactly. You can monetize through distracting the human explorer, right? They come with either a search goal, an information retrieval goal, or a high intent thing to execute. And in the process, you can put up either like a blue link somewhere in their feed and convince them to get distracted and go elsewhere. Or you can put up a box next to your sports scores and put in some other information about something that they might buy.

David:
[41:03] Like a time to go pen. Hot singles in your area.

Sam:
[41:06] Yeah, exactly. And so you could distract the human explorer, and that was worth a lot. And it was worth even more depending on who that explorer is, how much you knew about them, and how much they were worth to the advertisers for the instant ad auction. And ironically, this created the 5 trillion token data set of the internet, which is human-generated text, that was able to train the LLMs in the first place. Of course, there was some other stuff that happened. The GPUs got fast and the algorithms got good and stuff like that. The other thing that happened is that we had a corpus of 5 trillion tokens that

Sam:
[41:38] we could train these LLMs on to make them as good as they are. And in the process, we've killed the advertising business model. I mean, it hasn't happened completely and entirely yet, but it'll happen over the next 5 or 10 years. Whether it's the search and retrieval information that will come from the context the LLM has in pre-training, or it will come from the searches that it does and then does not get distracted along the way and then exposes just the response that you need. Or if it's through the agent that you use to execute a high intent purchase, like, I don't know, a handbag or a flight or a vacation. That explorer, that LOM-based explorer that was built on the back of the free and open internet has basically killed the business model for the free and open internet. And so in the process, we're going to need something else to incentivize people to put up information like sports scores or the weather or how fast a cheetah runs, etc.

David:
[42:31] Do you think this will change the topology of the internet? Because if I have a browser that is like, the ChatGPT, OpenAI just reintroduced the ChatGPT browsers, so web browser with ChatGPT in it. And if we theorize that we're going to have a proliferation of the growth of X402, and then therefore agents that can go out onto the internet and buy resources and retrieve them for me, and then bring them back to me in some particular way. Do I really have to go surf the internet anymore, or can I bring the internet to me? And what does that even look like?

Sam:
[43:04] Yeah, I don't know. It definitely will change the topology of the internet. I think the very early indications, I think it'll change the topology of the open internet, for sure, which is sort of the edge leaf nodes that were monetized through ads. And then the direct distribution stuff, which is like, I want to buy something on Shopify, I want to buy something on unitedairlines.com, et cetera, definitely get changed. The beginning of this, you're seeing through the OpenAI Stripe collaboration on inline checkout in chat GPT. This will remain sort of like the upper end of the things that you can purchase. So it'll be sort of like the Ebays and the Etsys and the Shopify's of the world. The stuff that has very high trust is the vast majority of volume that happens on the internet that OpenAI is comfortable putting within their search experience or within their chat experience because they believe that the merchant is high quality and will provide the services, et cetera, and is high reputation and they have a good enterprise agreement with, et cetera. And the question is like, what happens with everybody else, right? What happens with you're a new business, you just got started, you don't have a lot of volume, you don't have a lot of reputation, and OpenAI shouldn't actually spend time on you. That's like the correct decision out of them. And X402 feels like the solution there. You can list anything permissionlessly, you can promise that you'll provide the resources and then people can buy it if they want to.

Sam:
[44:29] And so there's this new, bizarre search layer that happens where people are crawling for stuff that they would like to purchase, they would like to execute on. And the way that search has always worked historically has been based on ads. And so I think that thing has a completely different shape than we've ever seen before. But I can't speculate specifically on what it is.

Sam:
[44:51] We don't know what.

David:
[44:52] We just know that the internet's going to look very different in a short order.

Sam:
[44:55] Yeah, for sure. I think that one of the ways that I've been thinking about it is like, what does it look like if every website you go to, it has the existing website layout, but it also has like a bunch of X402 things, like where you can just pay the website a dollar to get something out of it. And so I've been exploring this on my website, samragsdale.com. There is a $10 endpoint to send me an email that you won't be CC'd on. There's a $50 endpoint to send me a priority email, which you will be cc'd on and I'll respond. And then there's a $300 endpoint to book directly on my calendar. I mean, I think that's sort of like the interesting shape of it, which is like you land on all these websites. Like, what does it mean to land on, you know, some cement company's website and then immediately have like a booking with them 15 minutes later? Like, how much would that cost? And could you expose stuff like that?

David:
[45:46] Wow, that's cool. That is cool. I don't check my email at all. And I could for money. I could check my email finally.

Sam:
[45:56] Put it up on X4 too.

David:
[45:58] Wow. That's great.

Sam:
[46:00] What else excites you?

David:
[46:02] Like what? Okay, so say I want to get into Merit a little bit in a second, but like say you are, say you're a builder listening to this podcast and they're like, wow, that is the coolest thing ever. I want to start building. Give them some ideas. What should they go build? What should they go do? What should they tinker with? What ideas do you have?

Sam:
[46:18] Yes. So I think X402 is a green field right now. All of the obvious stuff should exist on X402. People are willing to pay up for it. They just want a single place. They want a single interface for accessing all these tools. And so you can go through basically everything that people put on MCP about six months ago when that was all the rage after Anthropic dropped that. And you could put it on X402 and put a cent on it. And people almost certainly will be willing to pay a cent per call for something that costs you a lot less than that to run just for the convenience of being able to call it over X402 because all of their other good stuff is over X402. So I think that's a good starting point. all of the enrichment APIs, all of the context APIs, all the data sets, all that stuff you put on X402. And then I want people to explore more of this personal X402 thing, which is like what I've done with my website, which is what are the things that you would take a transaction for that can automatically book against you, contact you, etc. And then the third leg is if you go to x402scan.com slash composer, it's our new composer mode, which is like chat GPT, except fully over X402.

Sam:
[47:26] Making resources that compose really well with the other ones. And so a few of the ones we really like are the search endpoints, the news endpoints, and then the crawling endpoints. And you can compose them really well together. And if people create many more of those that are really easy to use in chat, I think that you will get a lot of users, at the very least through X402 Scan and maybe whatever comes next.

David:
[47:48] And what's your perch in all of this? What is Merit Systems doing here? What do you guys do?

Sam:
[47:53] So in July, we released the terminal. Terminal is a bank account for every user and repo on GitHub. And the dream here is repos that fund themselves. It's similar to the DAO idea from 2017, but without the liquid token voting stuff. Without the liquid token voting stuff. And so the idea is that these repos collect money into them, and then they can pay out to their contributors weighted by their impact or their merit, hence merit systems.

Sam:
[48:22] And the goal is that these repos can pay out to people to make the repo better, then more people use the repo, it becomes more valuable, and then they can continue to pay out. And so you basically have an internet-native company that does not have a Delaware C-Corp, it doesn't have lots of paperwork, it does not have lots of fixed costs, but it does become more productive over time.

Sam:
[48:41] The obvious thing to do after we launched that was come up with ways to monetize directly out of these things. It's not very interesting to have an internet-native company if it doesn't have any money to distribute to its contributors based on Merit. So we launched Echo since then, which is a user-pays AI SDK. It allows you to build AI apps, have your users pay directly for them, collect a markup on that, and put it directly into your repo. So the main thing that people are building right now across all of GitHub is AI apps. If you're building one of those with Echo, you can monetize it trivially into a Merit repo and make it a multiplayer paid experience. While we were building that, we saw the X402 hype that it was a bit small at the time. And so we asked ourselves how much we could grow it. And if it gets to sort of on the order of $100 million a year, then it becomes very interesting. And it is kind of like a counterbalancing force to the more centralized equivalents of like the Stripe OpenAI style arrangement where people can build things at the edges. And they don't have to ask one of the big interfaces like Claude.com, Claude.ai, or ChatGPT to integrate. And rather, they can permissionlessly list things like they could do with smart contracts. And then you can build API software on the internet that then can monetize itself directly into a Merit instance or otherwise.

David:
[49:58] Sam, there was a lot of imagination that I think that's left the crypto industry in the last two years as like, A, regulation really wasn't helpful. And like, B, meme coins, perhaps downstream of the lack of like good quality regulation and just the oppression from like the Gary Gensler era, just like sucked a lot of the imagination out of like crypto. And I'm very hopeful, cautiously optimistic that stuff like this can reinvigorate a lot of imagination into what you can do here, especially when you talk to me about GitHub's that build themselves. Like we've like totally lost the original idea of the word DAO, but a DAO was always supposed to be code at the center and humans at the margins. And now in 2023 and beyond, a DAO is just like, no, it's like a human organization on the internet and there's humans at the center and it's just like a digital, digital company. But no, no, no, that was never the sci-fi thing that brought me into crypto was like, oh, Uber, but everything is autonomous and everything end to end is just code and it's just service providers on the outsides. And that's what I see happening here. That's where I see some of the groundwork being laid.

Sam:
[51:08] Yeah, exactly. So liquid token voting was not a big part of the original DAO paper, but that is the mechanism that we have figured out to make it work. Merit asked sort of like a different question, which is like, if the majority of the productivity that happens within one of these organizations is people contributing to code and making code better, then let's just pay out weighted by their merit, the amount of work that they've done. And you actually can measure that quite well. And so that's what we've focused on. And then things like X402, if there's a significant demand side there, people willing to pay in stablecoins, then it becomes very easy for somebody building an API resource on GitHub to monetize itself and then pay out to those people. And so our sole goal in X402 is Blue Ocean. We need to grow the space as much as possible, get it to hundreds of millions of dollars in volume, build a massive supply side there, people that already

Sam:
[51:59] have USDC or whatever else that people want to monetize their resources in. And once those people have that and they're used to transacting and paying a few cents microtransactions for server resources that they find valuable, then it's fairly trivial for the developers to collect their cut along the way with no middlemen in the process.

David:
[52:18] Sam, I feel like we are maybe a little bit more than halfway down the X402 rabbit hole, but I would imagine it goes a lot deeper. Where can we point listeners to if they want to go learn more about X402? Like what threads should they pull on? What blogs should they go read? Anything you can point them to.

Sam:
[52:36] Yeah, so I would start with, gentle plug here, I would start with x402scan.com. It was built for this purpose. We looked at all of Twitter and we looked at the other websites, we looked at the GitHub repos and you couldn't really get a, like you could get an idea of what it was intended to be, but you couldn't really get a feel for it. X402 Scan is built for you to get a feel for it. You couldn't click around, you can try different servers out. If you load it up with 10 bucks or so, you can actually fetch the resources directly. Composer mode, we actually sponsor, I think about 25 bucks of anybody using it. So go and try out Composer and you can try sort of the agentic version of this. There is an ecosystem tab on there, which is very similar to the x402.org slash ecosystem tab, which you can look at. The Coinbase guys put out great work. So it's Eric Reppel over there and Kevin. So you can look at their Twitters. I would highly recommend the Stratcher T article. This is Ben Thompson from May this year called The Original Sin of the Internet. If you want a lot more depth on the ad model, the ad based business model of the Internet and, you know, Mark Andreessen and Tim Berners-Lee's thoughts at the time. It's a really good article. And then I think the Coinbase slash X402 repo as well is the most up-to-date source of the spec itself.

David:
[53:55] Sam, this has been great. I'm pretty excited about this. And I really, really hope that this actually blossoms into something pretty huge. So I guess, yeah, thanks for coming on and just educating me. There's a lot more here than I thought. And I'm very glad for it.

Sam:
[54:07] Yes, I'm hoping that we can maintain the 200,000% growth rate, but I fear that it might go down a little bit from there.

David:
[54:14] If you can do it just one more week, then I think you'll be in a pretty good place.

Sam:
[54:19] Yeah. So if anybody out there is looking at building X402 resources, please reach out to us or get building them. I think there's an immense amount of opportunity there and you're very early in this ecosystem.

David:
[54:31] Yeah. It feels good to like be early and know that we are very, very early here. Anything that can go up 200,000% that's an indication of being early. Sam, thanks for coming on Bankless.

Sam:
[54:40] Absolutely. Thank you for having me.

David:
[54:42] Bankless Nation, you guys know the deal. crypto is risky x402 brand new frontier you're probably going to hear some more episodes about this in the coming future uh but nonetheless we are headed west there's frontiers not for everyone but we are glad you're with us on the bankless journey thanks a lot