The New Internet Vending Machine: X402 Payments and Trading | Nick, Celeste, Joanna, Erik

ETH Denver. Can't tell if you guys want to be here. You just like can't get up, so you're here. E e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e either way you're here. So hopefully we uh entertain you.

Um title of the session is the new internet vending machine x42 payments and trading. Uh my name is Nick Prince. I lead product uh app growth on the base team. And we'll just start with some intros. Uh Joanna.

Hey guys. Hello. Hello. Can you hear me? Hello.

I can just speak. Oh, there you go. Okay, I wasn't speaking directly to the microphone. Hi, my name is Joanna. I'm the co-founder and CEO of Soon Network.

So we have been working on uh AI export to related applications um dedicated to you know consumer grade use cases which is something that I'm very bullish about um given that you know I think X42 has been seen more as a developer tool and we really think that it has the potential to uh on board more users and I can talk more about the application that we have been building. You're you're not late if you make it before the intros, right? [laughter] I'm I'm Eric. I'm the head of engineering for Coinbase developer platform and I created X42.

It's off, right? Oh, okay. It's on. Hi everyone. My name is Celeste.

Uh I hate developer relations at Virtual Protocol. So uh at virtuals I focus very much on our agent commerce protocol where agents you know buy and sell stuff from each other with a trust layer and um a lot of my focus areas is on like the developer community. We have lots of agents on virtuals and we basically create developer tools and and ecosystems um to make it easier for them to build.

All righty. Thanks guys. Let's go ahead and kick off a question for Eric just because you think you invented X42. You think you don't need to be here for the intro? I was just running around.

But anyway, you created it. Like, walk us through how it works and why the internet needs it now.

Yeah. So, as all of you may have noticed, uh, the internet doesn't really have money in that if you've anytime you've ever bought something on the internet, you've either typed in a credit card number or maybe you've connected a wallet. But that's not actually that's like an experience that's shown to to you through the internet, but it isn't payments that are native to the internet in that there's no like association between HTTP and the payment that's happening. It's kind of like a thing happens over here and then something gets notified and then the thing happens over here. And so what X42 really aims to do is be a native payment layer for the internet.

And so what it allows you to do is to attach a payment to any HTTP request or MCP request or there's like a bunch of different transports uh that X42 supports. But instead of payment being something that's kind of negotiated on a custom one-off basis adjacent to the thing that you're doing, it allows you to include payments directly in the requests you're already making between a client and a server. And I think the reason why we need that now is one, I suspect we're going to see a lot more payments as agents become bigger and bigger economic actors on the internet. X42 uh primarily is used with stable coins. Stable coins are a great way to do microp payments.

I think it becomes a very strong narrative and a really compelling use case for stable coins for agents to be able to make 1 cent, 10-centent, 0.5 cent transactions to pay for more context through data to make them better at doing jobs for us. And then two, I think that anytime you have a technical revolution, you get kind of permission to introspect the foundations of what we all build on and that I think there's value in just having it even for browserbased payments, but you only get permission to edit the internet once every couple decades. And so agents, I think, are the reason on detra of why to try to solve this like 40-year-old problem of there not being payments on the internet today. Love that.

I think native payments on the internet definitely seem like an idea whose time has come. Uh Joanna, you guys are doing this really cool consumer experience around red packets and you're involving X42 in in unique ways. Can you just talk more about what that product is?

Yeah, absolutely. Um I think this is uh really interesting that you know timing wise we we first um I've been coming to East Denver for the last like five years and this is the first time Ethan coincided with the the Luna New Year celebration uh on the lunar calendar. So yeah, so we have been um experimenting with these um you know viral marketing type of uh consumer use cases uh using X42 underneath as the agent to basically uh take care of the transfer and the payment logic for you right and uh the actual product itself people don't need to know what's going on under underneath right if they're interacting with X42 or interacting with base it's very like ab extracting the complexity away for folks that are getting into um web 3 for the first time. So the rap packet use case here is uh essentially you know using this uh uh tradition you know in Asia in Asian culture where you can gift you know a value um actually in the traditional value it's like you just hand over this right pocket of uh um money to your friends and family and uh it's a it's a way to bless them for the year of a good luck and we turned this entire experience on chain with a single sign in without any gas. So, you know, I I would be the first one to proudly announce that my mom was on boarded within uh one minute like she was able to claim this even though she never touched crypto in her life.

Uh so so this whole experience taught me that you know we need to basically uh come up with new use cases beyond just like the enterprise like fintech kind of using Afro2 for payments in the existing platform but bring the use case to like the existing crypto native crowd as well as like people that may not know may may not know about crypto um in a in a in a way that it's familiar to them, right? So, we also build a a mini per which is um mini just meaning it's a ultra short time window. We were able to uh bring it down to 30 seconds. So, for anyone that trades on this per your auto uh your leverage position auto closes within 30 seconds. you don't have to watch the overnight risk and um you know lever up to like a hundred times um you have this gamified experience to to really try out um without the complexity of on boarding right so I think all of this to say that uh we're really bullish on you know using X42 as a way to simplify UX and uh bring the you know mass adoption use cases

love it can attest the UX is super smooth anyone can use it it's great to see

send you a red packet Yeah, she sent me a red packet. Uh, my mom could have claimed it. U, it's also just great to see builders creating experiences that are appealing to mainstream users globally. I think we we need more of that as an industry. So, shout out for that.

Uh, question for Celeste. As developers build more autonomous agents, how important are programmable payment primitives like X42? And could this become foundational for agent-driven applications? Interesting question. I think um many people know of virtual as like you know a token trading platform.

you know you launch a token an agent token and then it gets traded and in the early days we actually had a lot of like actually even now we do have some agents agents which don't have a lot of utility and what we found is that these kind of agents uh don't really survive like you know bad market cycles um for a very simple reason they don't really have much of a product um we actually believe that agents should be autonomous and they should actually um make revenue otherwise there's really not much of a point in it. It cannot survive market cycles. Um, and really people are not going to, you know, spend on LRM or even like APIs if there's not no real utility for this. It's just going to be side projects and there's not going to be much economic value in them. So, we believe that for um autonomous age to actually build a society of AI agents, uh we do need to um have programmable primitives for AI agents to be able to transact with one another.

Um and we believe that it's actually really the monetization era for AI agents and we really need like you know standards like X402. We've also built ACP uh this kind of standards to really ensure that agents understand each other when they want to when one agent wants to pay another agent knows how to respond to it. And also like on top of just payments with ACP we also have standards to ensure that you know when payments go through um from one agent to another there's actually safeguards around it. So there's actually a trust layer we have um where you know if a pay if if an agent pays for something the other agent um the the the buy sorry the buyer agent pays for something the seller agent actually um delivers uh something um the initial payment is actually like u kind of kept in escrow and it's not released until uh it's determined by an evaluator that the deliverable is actually of good quality. So trust layers, payment rails actually super important uh for payments and you know revenue models for agents to scale.

Agents in a sense um we believe will become almost like small businesses of the future and maybe even big big businesses.

Love it. And you guys have a great vantage point on the frontier with respect to like aentic trading. So we'd love to hear how you see specialized agents such as trading agents evolving in the broader agent economy both in the context of virtuals but also just more broadly. Yeah, interesting how you put it because on I think on virtuals we have um we have like a live product with a live product a live registry ecosystem of agents um a lot of them are functioning you can basically go in and ask a foreign you know ask uh our so-called butler or con search service like hey I want an agent to trade for me and you will get like a lot of you know recommended agents because we just some of I guess because we we started out with a lot of the gen use cases is we have plenty of trading agents. We have like swap agents, pub trading agents.

Uh uh we are partnering with like 100x uh soon as well as like plenty of other uh agents um on boarding to our platform and it's become it's becoming quite competitive almost. Um I think it's almost becoming a time where um trading agents have to become a lot more intelligent and become a lot more specialized just because there's so many agents out there probably not just trading but any other uh industry if you want to build an AI agent uh given you know how many agents are coming up over time with you know open claw you know v coding tools and all that coming up a lot more agents um in in the a in the ACP ecosystem and probably many other agentic ecosystems. Um and yeah, it's just a time where agents really just need to prove prove what they what what they've got because in a sense they need to market their their skills better. Uh it's almost like SEO. They need market skills better.

They need to make sure that they build up their reputation so that users actually want to use them over many other agents out there in the space, right? Um so I mean honestly there's from a from an ecosystem point of view um there's many ways you can solve this. I mean I'm just speaking of this because like we have air virtuals and ACP we are building up an agent like economy and agent ecosystem. There's many ways to kind of like solve this or help different teams like succeed. But then um the way we kind of like help users discover the best agents for like trading is to have like better discovery mechanisms.

Uh we've also introduced things like bounties and compet competitions to make sure that the best trading agent really wins because ultimately that's what we want from a user point of view as well like to for users to benefit from the best trading agent or the agent which gives them the best life returns.

Fascinating times prices keep going down. We'll see if the trading agents also become uh community member agents. Um Joanna, can you talk a bit about your forward-looking like roadmap and how you guys are thinking about building around agents um on base and soon?

Yeah, absolutely. I think um maybe going back to a little bit about the history of soon, right? So we started the project as a infra project um bring high performance you know SPM you know Sana virtual machine to um many of the equipment ecosystems and in that process we realized that um in order to really like demonstrating the the high performance underneath we need to build that first killer app right so we incubated a bunch of trading apps and um over the year year and a half time and then you that just seems to be the the really the biggest PMF that we got um that's building on top of a high in high performance infrastructure. So uh that started our journey pivoting from a an infra project into an app project right. So we started building our own first party dabs.

Uh first started with a uh copy trading platform called simpl simp um simplify uh trading and then the second one know as soon as I started talking to my old friend Neil from uh the coin my coinbased days um you know I I realized there's a really big potential out there for utilizing the X42 uh technology. So that you know transitioning into basically three meaning experiments. Uh first was a prediction market called 10s soon. Uh that was a basically a fun experience for people to predict the bitcoin or whatever asset movement in the next 10 seconds. Uh using x42 as an agent to you know give proxy to right in order to trade on behalf.

And um that experiment quickly took off went viral. we got um so many you know users hitting um the platform we actually brought down the the X42 pointbased facilitator for a few seconds that's how I got talking to Eric and team and and we started working on this high performance-based uh X42 facilitator so that was like the you know the learning from that journey and then we really started to hone in to this vision of you know going to how do we actually bring like a real world use case right to AI to agents and um and and and then still like connect to the existing crypto crowd. So yeah, so this I think really um bring it back to the the trading use case, right? So um biggest problem with you know building a per today is like it's hard to onboard new people right and uh you have to learn all this you know jargon and understanding like how to read a funding rate and how to you know keep a pause on the uh leverage position like when to close it. So, so we think that next generation per needs to be running on like something like X42 the the infrastructure that really extracting the complexity away.

So um so yeah so that's been our main focus is building a perp that can on board you know new people into the feray right uh so it's called 100x soon and um um the feature that we talked about the red packet is a way to acquire new users right and introducing this social feature so folks can like again you know uh 1.5 billion people are celebrating Lunar New Year in 15 countries right so we kind of reach out to this this entire a new user base with something that they're familiar with and has a natural product market fit. Uh so yeah, so you can like quickly uh get folks into uh the trading platform, you know, by in inviting them uh through this uh new product feature. So yeah, so I think like going forward the the road map on base is continuing to build build out this per infrastructure and um yeah, basically I've been having the pleasure to work with the base team with the TDB team. Um and I think I also like learn a lot from my Coinbase days, right?

where, you know, we're constantly like incubating new projects, you know, just like how Coinbase is like there's a Coinbase mafia people that um, you know, even at the time they're working at Coinbase or even leaving Coinbase, they're, you know, always like kind of starting these new startups within the startup um, coming up new ideas. So yeah, we were really uh, fortunate that, you know, we live in the stage of AI. So it's the bar to start a new product is much lower, right? folks can engage in this vibe coding and um you know quickly like put it put together like a front end uh in within hours not days. Yeah, never been a easier time to build.

Uh speaking of new projects or products, Eric Coinbase develop developer platform just released uh Aentic wallets. Can you tell us how those are different than the existing agent kit product and how they enable AI to do things on chain?

Yeah, for sure. So I think if we want agents to be actors in the crypto economy, they need tools that are going to allow them to do that. That they need ways of signing transactions or doing swaps or trades or paying for things. And so Quinnb's developer platform has two products that we've released over the years to try to solve that need of giving agents access to a wallet. And so the first one of those products was agent kit which is a TypeScript and Python SDK that allows you to build an agent in with a wallet.

And so you as a developer are building an agent and that agent now has uh a wallet that it can use via tools and interact with things on chain. And that I think was like the meta about a year ago where people were kind of building lang chain and other frameworks for building agents were really popping up. And uh now we've kind of seen this pivot into most people are using cloud code or uh codeex or other kind of more full stack verticalized tools around their agents. And so what Awol the Agent wallet is is a wallet product that is designed for agents. And so it's not a library or an SDK you need to build into your agent.

It is a product that your agent can use. And so that product surfaces it itself to agents via either CLI or via MCP. Um and you can add it you can give your cloud code or any other agent a set of skills and uh if it has those skills it knows how to use a wallet. And so we started kind of working from Coinbase's principles of like most trusted, safest to use and uh sandbox the wallet keys in a very in a pretty novel way such that the agent doesn't have direct access to the keys and it can't actually even sign arbitrary transactions. every action that the agent takes goes through kind of a policy layer.

That means that that includes like spend permissions and uh kind of user controlled ways of limiting the blast radius of an agent gone rogue with the wallet. And so the wallet's fully dedicated to funds, but then you can also layer on rules like yeah, you're only allowed to pay up to 50 cents per payment or like up to $5 in an hour into the agent. And then we've created very nice like verticalized experiences within the product for sending crypto for paying for X42 enabled services. There's X42 search uh integrated. So your agent can go and find services that accept X42 payments and then pay for them in line.

And uh we also you can just trade. So any token that's available on base you can trade through the uh Aentic wallet. And um it's really cool. If you want to check it out, you can npx skills add coinbase agentic wallet skills or it's ultimately just a CLI. If you're a developer, you can do uh npx awall--help npx aw-help and uh it will install a full full endto-end wallet experience that is email OTP authenticated.

So if your agent has a if your agent has an email, it now has a wallet and uh no API keys, nothing else. So I think like as we're starting to get things like X42 and trading and all these use cases that leverage crypto, we need to start thinking about wallets as products, not just for humans, but wallets as a product category that your agents are going to need to be able to reason through and kind of elegantly control. And that's why AWOL, we designed it as like a a human wallet that's accessible to agents. And so check it out. Um there's docs online.

And if you search Coinbase agentic wallet, you'll find it.

It's cool to think about how Coinbase brought simple, safe wallets and crypto trading to users 15 plus years ago. And now Coinbase developer platform is doing the same for agents. Uh we're short on time, so let's take a half a minute starting with Celeste. Just what would you like to leave the group with today?

What would you like to leave the group with today? Can be a prediction, a hot take. I would say um it feels like it feels like to many people it might it might seem like Asian to Asian commerce is almost too early um simply because you know not many places are using it today but I feel like we'll kind of hit an inflection point where AI and large language models will become uh will reach a certain level level of complexity uh wallet infrastructure payment infrastructure and even agent to agent uh standards um will become mature enough and I feel like you know at some point um there's actually like a lot of potential for Asian and Asian commerce to really like change things up. So I'm really looking forward to that.

Eric, oh it's me. Um we're in this moment right now where things are going to change. Agents are really the next platform and we've gone through the internet, we went through mobile and now we're going through the agentic cycle. And so now is the time to choose the types of technologies that that that platform is going to be built on top of and the values that those technologies are going to have. And so I think it's an incredibly important time for people in crypto who really live the ethos of like open decentralized trustless to get involved and build in the space.

And uh I have a lot of thoughts on this. I'm speaking on this stage in about half an hour 225 325

for Eric's talk after this.

Yeah. Uh if you want to hear me talk about this for longer, I'm I'll be right here. So uh yeah, come by.

Joanna,

don't sleep on X42.

But um we were just talking right before this. I mean, there's a lot of building opportunities. If you feel like you're late, you're not, right? So you know even for them they're talking about the um this discovery portal bazar like there could be a consumerf facing end uh that someone can build probably within hours. Um yeah I I really are very bullish about the um the new use cases that we can see coming out of expert 2.

Um and um yeah if you guys want to check out the the per that we build mini per we build it's called 100xsoon.com. Um, yeah, we love to hear feedback. We'll stay for a few minutes after this.

Yeah, still definitely day one. Thanks all. Thank you so