On Ep. 3 of Agentic Commerce, Simon Taylor, Head of Market Development @ Tempo, and Bam Azizi, CEO & Founder @ Mesh are joined by Erik Reppel, Author of x402 and Head of Eng @ Coinbase Developer Platform to discuss the history and original purpose of HTTP 402 status code, the convergence of blockchain, AI, native internet payments and more!
On Ep. 3 of Agentic Commerce, Simon Taylor, Head of Market Development @ Tempo, and Bam Azizi, CEO & Founder @ Mesh are joined by Erik Reppel, Author of x402 and Head of Eng @ Coinbase Developer Platform to discuss the history and original purpose of HTTP 402 status code, the convergence of blockchain, AI, native internet payments and more!
Timestamps:
Tokenized is sponsored by Visa
A world leader in digital payments, Visa is bridging the gap between traditional financial institutions and innovative blockchain networks, helping players in the payments ecosystem navigate the ever-evolving world of tokenized fiat currencies with confidence and ease. Learn more at visa.com/crypto.
Tokenized is also presented by Mesh
As the first global crypto payments network, Mesh connects over 300 wallets, exchanges and payments platforms, and enables anyone to pay and get paid instantly, anywhere, in any asset. Mesh makes digital transactions seamless, secure and universal, fuelling the next era of agentic commerce. Learn more at meshpay.com
***
We’d also like to remind you that the views or opinions of our contributors today are their own and do not necessarily reflect those of the companies they are representing. Nothing we say should be taken as tax, financial, investment or legal advice, do your own research!
Music by Henry McLean
Sy Taylor 00:00
Simon, welcome to tokenized. The show focused on stable coins and the institutional adoption of tokenized real world assets. My name is Simon Taylor. I'm your host for today, author of FinTech brain food and head of market dev over at tempo, enjoying me once again, stepping into Kai's shoes is my co host for this agentic commerce series, the CEO and founder of mesh. Bam. Azizy, how you doing? Bam, good. Thanks for having me again. I always appreciate it. I'm like that you wore a shirt today, and with good reason as well, because Joining us today is Eric reppel, who's the head of engineering of the Coinbase developer platform, and, of course, the creator of x 402 So Eric, thank you so much for being on the show. How are you feeling?
Speaker 1 00:53
Got in a little late last night, didn't get quite enough sleep, didn't have the wherewithal to wear a shirt today. So apologies for the t shirt, but as an engineer, maybe I get away with not wearing a button up shirt all the time.
Sy Taylor 01:04
No, no, yeah. I wonder what BAM has been up to, because I'm a t shirt guy myself. But before we dive into the show with Eric, I want to remind everybody that views and opinions of contributors today are their own and might not always reflect those of the companies they represent. Please don't take anything we say as tax, legal or financial advice, and always do your own research, and of course, reminding you that this show is brought to you by our friends at visa and our good friends at mesh. So Eric, I want you to take us back to the very Genesis pre X 402, the internet itself had something called HTTP 402 Can you explain to people a little bit of the history of what that was and what it was intended to be, and then sort of the time gap between that happening and X 402
Speaker 1 01:51
for sure. So 402 is literally just a number. When you talk about the internet, everyone says the internet. What they mean is HTTP based systems where clients and servers communicate with each other, and part of the HTTP standard is what's called a status code. The status code, it's a summary of what went wrong that the server can send a client. And so you might have heard of status code 404 which is not found. If you go to a page that doesn't exist, you'll get back a 404 generally, the 400 codes are when a client has made an error in asking for something from a server. 200 codes are when everything's good. So you'll see a lot of like HTTP. Two hundreds that's kind of the happy path. It just means okay. And HTTP status code 402 I have a party trick. I can recite the entirety of the HTTP, 1.1 specs, definition of the 402 status code.
Speaker 2 02:44
Wow, that's that. You got to do it. I'll do
Speaker 1 02:48
it for you right now. It's HTTP 402, payment required, reserved for future use. And so back when the internet was being built into what it is now, they knew that we would need to do payments on the internet, and they wanted to make payments native to the internet, but the technology just didn't exist at the time. Mark andreesson actually spent a bunch of time trying to make payments native to the Internet back in the Netscape days, but we just didn't have the right tech. They knew it was going to need to happen at some point, and so they reserved 402, to represent Hey, a payment is needed to access content or to access this web page, but they didn't have the technology to actually make it do anything, and so this is a problem that goes back to the 80s. This is actually Coinbase, his third attempt to do something to activate the 402 status code. Brian Armstrong tried something in 2015 Balaji Srinivasan tried something in 2016 if you Google Coinbase, 402 wired and then there's probably some more recent results there, but you can find a Wired article on Brian's exploits with 402 back from 2015 that commemorates that attempt. Fascinating.
Sy Taylor 03:54
Did you ever look at 402 before you got into stable coins? Yeah.
Bam Azizi 03:58
So I'm also deep down in engineering, maybe I'm not as good as Eric, but I memorize all the 402 100 codes because you have to kind of know it when you're building an app or anything related that was the whole idea of web two was read and write, and web three is read and write and pay, right? So you should be able to combine the crypto or blockchain technology, plus the web technology. Both of them are like global phenomenon, like, if you have an access to the internet by nature, you have an access to the blockchain as well. So combining these two to offer a native thing that addresses one of the basic needs of a human being, that would be very, very interesting. Well, like you want to browse, but you don't want to going to get a stop at the intent, right? So you want to be able to pay but what makes it even more interesting is there is another trend that's happening, which is AI, right? So we were talking about a lot of human behavior, but soon we need to think about agentic behavior too, right? So agents don't have the limitations of. Humans, right? We need to go and browse and click here, click there, and then type in our credit card information. That's not how agents work. And they want to kind of get something native, API based, HTTP based, so they can move as fast as they can. So that's why I'm very, very interested and intrigued by what Brian and Tim is doing. Do you know
Sy Taylor 05:20
it's interesting that we sort of evolved so many other business models as a result of not having this micro payments infrastructure. We start typing in our card numbers, or we get into Subscriptions so that we don't have to think about the small individual payments. But we're seeing the limitations of subscriptions already with llms and some of the larger usage based billing that's now coming along, and the fee structure isn't necessarily working. So I'd love to hear the Genesis story, Eric, of how you actually got x4 or two off the ground, how it works. Just unpack it for me. Take me on the journey. Because, as you said, this is the third go around. Was this just like, hey, I think I got it. Was there some light bulb moment? Like, how did that happen?
Speaker 1 06:01
So I've been aware of this problem for a very long time. I was at Coinbase from 2017 to 2021 and then was leading engineering at a company called Zora for a number of years. After that, then I took some time off from Zora, and I was thinking about what I wanted to do next. And it kind of occurred to me that with the advancements that blockchains have had, this is kind of the moment to try to reactivate and make an internet native payment standard. In parallel to that, Coinbase had reached out and asked if I wanted to rejoin to lead engineering for CDP, and Brian Armstrong was on my interview panel, and he asked me this question, what do you think we should be doing at CDP? And I answered, well, I think it's time to try to make an open standard for Internet payments. I think the advancements we've had in base and Solana and modern l twos and performing l ones make it so you actually have such low gas fees that you can do sent transactions. And you now, I think have this incumbent force of AI that is a reason to reevaluate. I think anytime you have like a step function change in user experience or a shift in paradigm, it needs to have a really strong reason why to change the way things are done. And I think to bam's point, to your point. Simon AI, really is this complete alteration and how the internet works, and it introduces this new economic actor that has all these second order effects on how businesses are run. One, I think, which is the most obvious, is that ai disintermediates content publishers from their users in a way that is kind of unintuitive, in that more and more people are browsing the web through their chat interface, their chat GPT or cloud or equivalent, instead of going directly to websites, and that kind of disrupts the existing ad model, because now a human isn't seeing an ad. Roughly, you can describe the economic internet as a human views some content on a website and is shown an ad, and some seller is paying that website to show that ad, and then, probabilistically, that user converts and buys some piece of content or some good or service from the publisher of that ad, but if the user is no longer shown the ad, well, the merchant doesn't have a reason to pay the website for ad placement, and the website isn't going to make any money. This is so disruptive, we need to think of new ways to have business models work on the internet. You also saw this with training of all these llms, functionally, they scraped the internet, took all the content from the websites and then built these models that are generating huge value, but the content publishers and the web properties whose data was used to train these models, in the majority case, aren't being compensated. And you see some cases where they are now retroactively making deals like there. I believe there was a big deal announced between anthropic and the New York Times so that Claude can more formally train on New York Times data. But that requires, like major institutions to do a lot of BD, whereas if you can just pay 50 cents for the content, that's kind of like micro business development.
Sy Taylor 09:00
And what about the Long Tai as well? What about all of the other pages on the internet? Talk to me about how a transaction is actually constructed, like I'm a developer. I want to make a transaction. I have some resource. It's gated. What do I need in order for an x4 or two transaction to exist? And talk me through a payments fund.
Speaker 1 09:20
So as someone who's publishing content, all you need is a place for funds to settle so that could be a self custodial wallet that you manage yourself, that could be your Coinbase account, that could be any place that crypto funds can be received to. And X 402, as a standard, makes a lot of the decisions on how to receive payment and a lot of the decisions on the user journey on both sides for you, and also has this optional component called a facilitator, that means that you, as a merchant actually don't need to interact with the blockchain at all. And I think this has been like a huge reason why we've started to see adoption from companies that don't traditionally do things. In crypto is because one, they get stable coins, they understand, oh, I can think of this as $1.02 it's so easy to adopt as a method of payment. Typically, I see companies start to be able to accept X 402 payments in under an hour. It really is that easy to start accepting payments with X 402 and so a lot of companies think, well, yeah, if it's an hour of time, and this is potentially a new revenue line item, why shouldn't I give it
Sy Taylor 10:24
a shot? That makes complete sense. Am I understanding a facilitator, right? Like almost, it's almost like a PSP. So historically, I had a payment service provider who would help me access various card networks and accept payments of those sorts, and then handle all of the settlement on the back end between the banks and that sort of stuff. The facilitator is sort of like a PSP. Is that?
Speaker 1 10:46
Right? It's analogous, but it's different in a few critical ways. If you think of a PSP as an abstraction over the underlying networks that move money, then yes, the facilitator and a PSP both abstract away the complexities of dealing with the visa network or the MasterCard network or the CH network away from you as a developer, and gives you a consistent surface to build against. The critical difference is that facilitators aren't in flow of funds, and they're not at any point a custodian of funds on your behalf, and so they don't need money transmitter licenses, and so anyone can operate a facilitator, because functionally, all you're doing is abstracting the complexity of interacting with blockchains and providing gas sponsorship.
Sy Taylor 11:30
Bam, curious in your thoughts here and the new types of payments flows that you're seeing and the need for for people to be able to make that work,
Bam Azizi 11:38
maybe it's the first time that agents will come first kiss sooner than humans, because they really need this right away. And as you mentioned, it takes time to see the adoption. I think it would be very, very important, from my perspective, two things to happen to x, 421, is making it as neutral as possible, like making it as open as possible, so they don't feel like this is just a coin based thing. Maybe, Eric, you can add some insights there in terms of, like, what's the direction this is like another coin based app product, or it's just no it's just more often, like what Google is doing with Agent protocol. It's just want to make it as neutral as possible. The second thing is, how we can bring it up to the current standard, or table stake of payment products like we need to have chargeback, we need to have even KYC, AML. We need to have, like, recurring payments, right? All smaller features that make this payment even more complete, because the idea is great, but the bells and whistles and the detail is important too.
Speaker 1 12:41
For sure, totally agree on neutrality. I think the web works because you have neutral technologies that are uncontroversial. No one is like, I shouldn't use HTTP because Google contributes a ton to the HTTP protocol, right? And so that's kind of the analogy I'd make with X 402 That being said, we're moving X 402, into a foundation, because we want to create the most level playing field possible, and so we're hoping to have that done by the end of the year. We think it's really important to make sure that foundation is set up correctly, so that it has, like, both maximum neutrality and maximum ability to advance the standard, and doesn't get bogged down in process as we make that transition, so 100% agree on neutrality. What from the start X 402 has been the goal has been any network, any token, you see remnants of it being rather EVM centric, because I happen to know the EVM best as an engineer, but we've worked closely with the Solana foundation and Cardano and avalanche and a bunch of other blockchains on adding support for those blockchains. And now I believe there's, like at least 10 different blockchains that you can do X 402 transactions on some of them, EVM, some of them not EVM. And then it's also not just the USDC thing. Any token that supports EIP 309 works on EVM blockchains. And then we have a proposal coming out soon that will unlock any token that supports ERC 20 to be used within the x 402 standard. And again, that's largely a design decision goal. I think it's extremely important to have a gasless story for stable coin payments and EIP 309 is, in my opinion, the best EIP for doing gasless transfers of tokens. It's just that it's not super broadly adopted. And so now we need to go after the majority Tai L of tokens that don't support EIP 309 in terms of these small features, I think it comes down to feature by feature case. And this is why we've added this concept of extensions in the v2 of x 402 which should be coming out probably around the same time that this podcast release, so that you can add components that you may not need in every transaction, but are really desirable in some transactions. I don't think every single transaction that X 402 processes needs to support chargebacks, but some of them, you definitely want chargebacks or higher. Forms of consumer protections, but if you're paying a cent for something, you're probably not going to charge back a cent, and it's adding chargebacks, and adding these features dramatically increases the complexity, because now who's the arbitrator of the chargeback? I think it's a better setup to make refunds really easy for the merchants, where if someone goes to the merchant and says, Hey, your service didn't do it. It said it would. They can refund. And the market risk and the reputational risk is kind of the thing that keeps companies accountable. We probably do need chargebacks for larger purchases, but I think you address that through extensions and do it on kind of a case by case basis, in terms of KYC, AML, X 402, and x4 two. Facilitators that are compliant with the like spec of the facilitator perform k, y, t, which is the AML and KYC equivalent for stable coins. And so anything that goes through the CDP facilitator does perform Kyt on these transactions to make sure that sanctioned actors aren't using x, 402, and then the other thing I'd call out is most stable coins, definitely USDC, bake in sanction screening into the transaction list, and so any sanction actors have their assets frozen or at a contract level are unable to transact
Bam Azizi 16:15
with USDC from our audience. What is k y t? Stands for k y t is,
Speaker 1 16:19
know your transaction. If you're an operator in crypto, you need to provide screening such that like a sanctioned actor isn't using your product to move funds on chain, which,
Sy Taylor 16:30
yeah, comes out of the far off recommendation 16 for vasps and casps. And in the self custodial space, it's quite difficult to know your customer, because it's a self custodial like you don't have to it's cash. But if I'm going to move 1000s of dollars a day, I definitely don't want it to be sent out to a sanctioned address, and I don't want to be the creator of the software that enabled that, or the creator of the stable coin itself that enabled that. So it's layering in protections without necessarily doxing somebody's privacy. It's an interesting balance to try and strike there, for sure.
Speaker 1 17:01
I think that point you made, Simon is actually the exact one that I I'd make a lot, is that we did commerce for 1000s of years without a lot of these concepts that we think of in kind of the PSP world, chargebacks, KYC, etc, because we use cash. If I hand you $5 that's kind of the end of our transaction. I don't really have a way to do a chargeback. I just have to come to you and say, Hey, give me my $5 back. And you know, if it's like $10,000 I might sue you to get my money back. But a lot of these stable coins really are analogous for cash. And I think the mental model for folks is cash versus kind of the credit based PSP systems that we're more accustomed to on the internet.
Sy Taylor 17:40
You know, it's just a great way to think about it, because when you do the big problem with cash is carrying it everywhere with you. It was really inconvenient. And so by introducing this credit system, originally the credit card, then we were able to make acceptance work everywhere without carrying the cash around. And that became a convenience. It also meant that I could bake in a bunch of other insurance products around it, like chargebacks, and those became really popular, and they built a lot of trust into that network. But when you build trust into the network, you also create cost, as you say. But it'd be interesting to get into the use cases that might be more suitable, because if somebody's making a big dollar purchase, then they might like those consumer protections. If it's for agents. They might like something else. So before we get into use cases, we're just going to take a quick break here while we hear from our sponsors. 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 back 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. All right. Thank you to our sponsors, Eric, I'm interested in your happy path. Core use case, are you seeing AI agents buying resources or people getting access to pay walls with micro payments like, give me a spread of the use cases you're seeing in early days and which ones you're targeting. Yeah.
Speaker 1 19:53
Interestingly, I think API usage has been even more than content access, if you think about what using. Chatgpt or cloud is like, they're kind of useful because they know stuff, but they can't do much on your behalf unless you integrate a ton of MCP servers or other connected services to them. And some of them are good, like you want to connect them to your calendar or something, so it can schedule for you, but it's a finite set of things that it can do, and you kind of have to add each one, one at a time. And so what x4 two unlocks is, if your chat interface has a wallet, anything that it knows about that can accept X 402 payments, it can now do. And so what I've seen a lot of is using x 402 to enrich the existing AI usage that people have an example of that is like, maybe I'm doing research on the popularity or the trends of the next 402, I can do a web search and have my chat agent summarize like, what's available in a web search, but there's news articles or APIs that have like social media influence data that are not on the public Internet. You can't just search for and get access to that data, because that data is more valuable that my LLM can now pay for, and that gives me a more robust, rich understanding and a better artifact out of the agent. So I've seen a lot of that. And Coinbase is a product called payments MCP that gives your LLM access to a dedicated wallet, and you can fund it with stable coins and put in like, spending policies and so on. And so I think access to data via APIs is probably the thing that is happening the most right now. There's a bunch of like, crypto related use cases as well where people are using x 402 to then, like, distribute an NFT or to distribute a token, and you're basically purchasing a digital good using x 402 that's definitely happening as well, but I think the highest value actions that I'm seeing are being able to pay for one time API access.
Bam Azizi 21:47
Do you see any trend in terms of like getting access to GPUs for AI is like getting access to resources and paying with x4 or two, or potentially paying with x or two?
Speaker 1 21:59
Yeah, so there's been a bunch of folks who have integrated various LLM models, or image generation models. So a common one is maybe you want access to Sora or a Sora equivalent model, because you want to generate a video, but you don't want to now have a $20 a month subscription, you can pay like $1 to generate the video, and that's kind of a hard thing to fund with a credit card because of the fees, but if you're charging $1 with X for OT, you receive $1 and so now that's actually an economically viable action. That's actually a demo I give all the time is, yeah, generating a video via Claude. Claude doesn't have video generation capabilities, and paying for one time generation of a Video Direct like raw GPU access. I've seen people fund credits for GPUs, but you kind of need to include, like, a code artifact along with the access to the GPU. And so it's like a little bit more difficult to do. So I haven't quite seen that advanced of a use case yet.
Bam Azizi 22:56
Another big problem with credit cards is there is a lot of fraud there, right? So if you really want to build something that is high adoption, high conversion, low fraud, it's very hard to build on top of credit cards. So what if I generate that 2000 images that is not like $1 in words, $2,000 now and then you go and dispute debt, and I don't get my money, even if I want to get it. I probably takes me 30 days to get it back, but export it to or blockchain in general, that problem is kind of resolved. And then I'm getting asked this question all the time that what happens to charge back? By nature, Blockchain doesn't offer chargeback, and it could be a curse and blessing. You should think of it as like some sort of a wire transfer. So the same way that there is no expectation that you can dispute the wire transfer unless the merchant really wants to pay you back, then they can. But the difference is wires, first of all, it's very expensive. Second is local type of transfer. It's not global, and it's not fast, so I think it's a curse and blessing, but at the same time, like you're seeing a lot of fraud on the Visa card and MasterCard and credit card in general. That maybe makes it harder to trust someone else out there as a stranger on the internet, right? So crypto kind of solved that problem for us. And then if there is a merchant that, as Eric mentioned, wants to add that functionality or feature, they should be able to do that, versus in credit card, it's on by default for everyone, which is very limiting.
Speaker 1 24:25
I like to say that stable coins support refunds, but not charge packs.
Sy Taylor 24:28
Yeah, same. Eric, where does escrow fit into all of this? And is there a role for that sort of capability, where you might be doing some complex set of transactions, or you might have sent off for some work to be done by some resource, and you want to validate it before you complete a payment, and you need some third party escrow. Does? Does X 402 enable that?
Speaker 1 24:50
As of today, it doesn't. One of the, I think, important technical decisions of x 402 is, there's this concept called a scheme, and you can think of the scheme as the. Logical way money moves. And so today, the scheme that we support is called exact and it's the hand you $1 equivalent, like, I'm just handing you a $1 bill, and that's it. That's the logical payment that's being made. So you absolutely could have an escrow scheme in X 402, and in fact, one is being developed in order to support nano payments, because even a lot of APIs are priced at $1 per million requests for things that aren't LLM generation, and even crypto can't handle that small of a transaction size. And so there's a scheme being developed that would allow for you to escrow $1 and then do all these requests, and then for the merchant to pull out of escrow as they have the proof points for you and as you consume the API credits that you're functionally purchasing. In theory, any logical way of moving money, whether it's escrowing or kind of streamed payouts, or one I get asked about a lot is like dynamic payment amounts, like with an LLM generation, you don't necessarily know how much is going to be charged? Because you don't know how many tokens are going to be generated until the end of the generation. And so any logical way that money can be transmitted, in theory can represent in X, 402, via this scheme mechanism, it just requires a consensus on how that scheme should operate, like, what the mechanism of that scheme is going to be,
Sy Taylor 26:21
yeah, the more complex, the harder it is to get people to buy in, and then you need roles and a rule book, and there's a whole thing that goes with that.
Speaker 1 26:29
Anytime you add third party trust as well, it becomes like a whole can of worms, right? And so I with X 402, I thought it best to start as simple as possible and build up the complexity, versus trying to solve every problem all at once, and start very lightweight and expand, versus trying to say, like, here's the closed form solution to every problem, because I don't think any person in the world can solve all of the problems that we have in payments in one shot.
Sy Taylor 26:54
No, the closer you get to payments, the more you realize it's edge cases all the way down. I'm interested in the partnership with CloudFlare, I know there's some work that they've been doing with yourselves around HTTP, four or two and x4 or two. How did that collaboration come about? And can you tell me a little bit about how it's gone?
Speaker 1 27:12
Yeah, so CloudFlare, it was probably six months ago now, announced this concept of pay per crawl paper crawl is a third option. So for those unfamiliar, Cloudflare is a CDN, which is basically a host of websites and a host of domains, and one of the services that they provide is the ability to block bots. So if you've ever seen one of those, like, Are you human click the checkbox? Tests, that's probably Cloudflare. And up to this point in history, there's been block or don't block because there wasn't really an economic reason to allow bots to access your content. You just want to humans to access your content and then consume the content, consume the ads. But what Cloud flow did with paper crawl is add a third option, which is now you can have a bot or an agent pay to access your content, and so we were connected post them announcing paper crawl, because obviously there's a lot of overlap with X 402 and that concept of paying for content access. And we found that we thought about internet native payments in a very similar way. And so Cloudflare is going to be moving their pay per crawl infrastructure to leverage the x 402 standard. And the value there is if we all align on one standard as an industry, you get interoperability. The analogy I like to make is you don't have to download a browser for each website that you visit. You download one browser and it works on any website. And that's kind of what we're aiming to do with X 402 as well as if you as a merchant accept X 402 or you as a client can pay using x 402 it should work
Sy Taylor 28:40
everywhere, and it's a great concept as well. And CloudFlare, I think, sees something like 20% of the internet's traffic. So that paper crawl could end up being quite substantial. And it really is coming from that agentic space of like needing to access so many different types of resources. It's creating a bunch of new How will they economically access those challenges? Where do you think the growth volume is going to come from in the next sort of six months? So there any early trends you're seeing in your usage patterns? You mentioned APIs being a big one. Is that likely for the foreseeable? Is there anything that you're excited by?
Speaker 1 29:15
I think pay per crawl is going to be really big. I think that once it becomes like a common standard, I suspect that we'll see mass adoption on this concept of paper crawling for most CDNs and most hosts in terms of consumer use. I think the agentic Commerce has two buckets, one of which is I'd call conversational commerce, where you're kind of talking to chatgpt, but you're ultimately in the loop for a last mile purchase, and then the other being autonomous commerce, and sometimes you may mix these in workflows. An example I'll give is maybe you want chatgpt to book you a vacation. Let's say I'm usually in Vancouver. Let's say I want to go to somewhere that's sunny and warm, and I want to go on a beach, and I want it to. Be within like, a six hour flight. I don't really care on the country. I just want a nice beach to go to. And I can ask chat GBT, hey, do some research and figure out where it's sunny and book me some travel for two weeks from now. And chat GBT might pay for access to, like, some localized weather data to understand, like, what the weather is going to be like in a couple of weeks. It may engage a service to plan the itinerary and research local things to do and touristy things to visit. And then it may pay a small amount to make that itinerary for me, where it's doing autonomous commerce, to pay for access to those pieces of information. And then at the last mile, it may bring it that itinerary to me and the hotel and flight that it wants to book, and then I'm in the loop for that last larger purchase. And so I think that the most common use case we're going to see over the next handful of years is just AI becoming substantially better, because it increases the number of capabilities that it has and how helpful it can be, because it's paying to basically be better. And you give agents a budget to accomplish a task on your behalf. Bam.
Sy Taylor 31:05
We've seen so many standards. Now in the agentic commerce space, there's ACP, there's Google's ATP, there's a to a now there's x4 or two in the entering the chat. How do these things come together? And how do we interlink all of those? Because I'm reminded of the classic. There are 14 standards. Let's make a standard that rules them all now there are 15 standards, XKCD thing, how are we going to overcome that? Like, what's the way around it?
Bam Azizi 31:29
Yeah, I think we need to accept that. We will have fragmentation for sure. And this is space. The great example of that. There are many example of that in the history of tech. The biggest one is chargers that, like every mobile phone had a different charger, and then, so, like, stop it. Let's let everyone use USB. And then now USB C, which,
Sy Taylor 31:49
funnily enough, is one of the best things the European Union has done. Like you can, you can thank the EU for that. Credit to you guys. You might not like GDPR and cookie consent, but that was the EU. They can reasonably claim that one,
Bam Azizi 32:01
yeah, they pushed very hard. That's one thing. The other one I remember is, if you remember, like, before Bluetooth, some devices had infrared, there was like a red eye protocol, and then we had Bluetooth. We had some other protocol that nobody knows, but I know, because I was, like, working in robotics called ZigBee. But Bluetooth got the mass adoption. Now everyone is using for many different reasons. This is one of the areas that we are going to see some consolidation. Believe it or not, HTTP was not the first protocol that the internet was running on. There were many different protocols built over the course of years, and then HTTP become this consolidated version of the internet. So I think we will see some consolidation. I think the way that Coinbase and Eric's team are doing it is going to be an add on to whatever protocol that we have on the agentic side. We don't have to pick either like Google's protocol or Coinbase protocol. They can well work together. And I don't think any other protocol I've seen that is doing anything around payment that can serve both humans and agents. So that's kind of where it differentiates from my perspective. Overall, it all boils down to the adoption, like which company or which protocol can get the most distribution, and then we take it from there.
Speaker 1 33:13
I think people miss something when they think about AP two and ACP and X 402 X 402 has actually been around since March, and we actually worked with the Google team. And x 402 is two is kind of a first class payment mechanism within AP two, because AP two and x 402 and ACP all actually have very different goals and very different audiences that they serve. X 402 is kind of internet native transfer of value. It doesn't try to make a bunch of assumptions about use case, or about identity, or about these other things. It's really meant to be layered on with other standards. AP two is very much used within the agent to agent payment space. It's really an extension of Google's A to A framework. And so using AP two, at least right now, is very much intended to be done within the context of the A to a framework that Google has. And then what the underlying payment mechanism is, is kind of a separate task. That's how AP two actually works with X 402 where, if these agents want to exchange value in stable coins, they can communicate how the value is going to be exchanged with AP two, and then perform the transfer value with X 402 and then ACP is really meant for kind of this conversational commerce, e commerce analogous, you're going to get shipped something use case, and it's very much meant for the kind of existing e commerce workflows that we already experienced. And if you go read the spec, it has things built into it, like a shipping address. It's like, if you're buying an API credit, do you need a shipping address? Probably not. And so you actually can just use x 402, as a transport mechanism within ACP as well. I've actually talked to the stripe team about this, and at some point, when we have capacity, I think we're probably going to just make that work. And what you saw in the early web is you it's not just HTTP. You have HTTP and HTML and CSS and JavaScript,
Sy Taylor 34:54
FTP and SMTP and
Speaker 1 34:57
yeah, but on the web, like you very rarely using. HTTP without using HTML or JSON or CSS, like these are, I think, complementary ecosystems of standards. I'd throw MCP in there as well that are kind of the new generation of standards that we need for the agentic web. And so I see all of these as very collaborative and complementary, rather than being competitive.
Sy Taylor 35:17
It's a great point, and I think there's a giant table to be made there of standard and use case, and then sort of the matrix of how those interoperate, because it's a point that I think a lot of people get lost on. So thank you for clarifying that. One last question before we let you go, which is, how did X 402, become quite the meme? It has, because my x feed is full of people explaining why agentic commerce is taking off, and how x4 or two is solving it. Like, what did you guys do?
Speaker 1 35:45
Honestly, I have no idea. I wish I had a secret to Tai y Simon, but it's sometimes Twitter just does the thing and it gets the right niche in the algorithm and things blow up. I was slightly annoyed because we had this really nice, like, 30% week over week compounding growth chart. And I was like, wow, this is growing like a YC startup. This is like, great growth numbers. And then it kind of ruined six months of very nice compounding growth by just hockey sticking up and kind of sustaining now, good problems to have, but no, I have no idea. I think it's an idea whose time has come. I think it's one of those brain worms that you can't stop thinking about once you kind of realize what the implications are.
Sy Taylor 36:24
No, it's spot on, because I remember going through the same thing with MCP about 1218, months ago, where I couldn't move for somebody explaining what MCP was to me. It's like, No, I just want a coffee. Like, I don't want you to explain MCP to me. And same sort of happening now, but that's a sign that you've really caught the Zeitgeist and done something. There was an idea whose time has come. So Eric, I want to thank you for building it, and I want to thank you for continuing to build it in an open standard way. If people want to find out more about you or the x 402 standard, why do they go to do that?
Speaker 1 36:56
Best way to get in contact with me would be on x. My handle is just at programmer, the English word programmer.
Sy Taylor 37:03
Nice handle. Thank you. Thank you. That's a flex I gotta give you that says flakes. Well done.
Speaker 1 37:08
I mean, these days I my time is split between management and coding, and so unfortunately, I don't have head of engineering, but I'll settle for programmer. And then if you want to learn more about X 402, x4 Oh two.org, as the best place to go, or github.com/coinbase/x, 402, for the repo where you can read more details about the spec and the
Sy Taylor 37:27
standard fantastic. And bam, if they want to find out more about you. And mesh, of course,
Bam Azizi 37:32
yeah, my Twitter is bam, as easy mesh. And you can check out mesh pay on Twitter. Mesh pay.com, or LinkedIn. Love to hear from the audience.
Sy Taylor 37:41
Yeah, heck yeah. If you're doing anything at genetic commerce, please do check out our good friends over at mesh, and you'll find me at sy Taylor on Twitter, FinTech, brain food.com, or tempo, dot x, y, z. And if you haven't already, I'm going to ask you to do the thing for conversations like this. If you want to hear more of it, you've got to hit the subscribe button. You've got to hit Like buttons. You've got to leave us reviews. You got to tell your friends to find this show, because conversations like this one with Eric are truly why we do this. So thank you Eric, thank you everybody, and we'll catch you next time.