Tempo Mainnet: The Race to Agentic Commerce
David:
[0:02] Bankless Station, I'm here with Giorgios Constantopoulos. He's an engineer at
David:
[0:05] Tempo and also joined with him is Brendan Ryan, another engineer at Tempo. Brendan, Giorgios, welcome to Bankless.
Georgios:
[0:12] Good to be with you guys. Big day. Thank you.
David:
[0:15] Yes. So congrats on the Tempo launch. I want to know what just like the launch actually looks like, you know, day one or just in the near term, the short term. What are some of the first movers that are coming online? And then also just like kind of the first categories of activities that's happening on Tempo. What is just like the launch kind of look like?
Georgios:
[0:33] We've been building Tempo since August with a lot of wonderful partners. We're working with Tribe on this. The goal is to make stable coins and web scale payments to work finally using a lot of crypto blockchain technology that we've been building over the last few years. Today's launch is focused on AI agents using the machine payments protocol to pay for things on the web autonomously. And we're continuing to push on our work on the enterprise work streams around post-border payments, remittances, and things which really are truly what attracted
Georgios:
[1:05] us to the crypto world in the first place, 24-7 borderless finance and payments. So there's more to come on that in the next few weeks, and we'll continue sharing on that. But today's launch was all about the agentic payments.
David:
[1:18] It's all about agentic payments. Well, I mean, there's mainnet, right? But I do notice that there is just like a very large emphasis on this MPP thing, which I think we'll go into. Tempo is known in the crypto industry as like you guys do stablecoins, you guys talk about remittances, tokenized deposits. If somebody wants a stablecoin, they might go to Stripe and Tempo and get a stablecoin minted on Tempo. But it really seems like that wasn't really the focus of today's mainnet launch, really. The focus, the emphasis is on this machine payments protocol, MPP, which I think kind of gets us into the topic of agentic commerce. That's what I'm reading into. That's what you just said. Why the focus on agentic commerce as like a primary, just like part of the actual mainnet launch?
Georgios:
[2:05] The mainnet as it launched today supports all the use cases that you mentioned. So we already have some flows live on it which are for a particular normal payments. For example, Bridge has already, Bridge, the Skype company has already gotten some funds on Tempo and we're working on further expanding support for that. So this continues to be a lot of the focus. At the same time,
Georgios:
[2:30] The AI payments world seems to be happening so all of our people on the team, we use Cloud, AMP, Codex all day and we're already seeing that even to us as developers, it's kind of too much to go and log into a service, auth, add card, get API key, put that API key back. It's just too much when we're just supercharged with these new tools. So this really came out of our own need in a way for okay guys it looks like this ai agents want to do more but they're not able to they get bottlenecked on the human or just to give you another example let's say i have my deep research agent and i'm browsing around it can many times just improve its quality of response if you had access to some piece of content that was paywalled for example new york times article or anything else so we just thought hey what if we didn't have to do anything around that and what if we just gave the agent a wallet and we just let it rip. And we really went with that thesis that, okay, we have the enterprise stable coins things. These are very important. We should continue doing them, but we cannot ignore this wave that's coming, this tailwind about which could affect materially the focus for everyone in the crypto payments industry. And as we decided to really put a lot of energy, make the launch, be focused on this, and then continue with the rest of our work because it was
Georgios:
[3:54] just too important to not make a move on this.
Ryan:
[3:57] We've already mentioned something called MPP. Can you describe what that is? And how is that different from other agentic payment standards that we've seen, something like X402?
Brendan:
[4:06] MPP or machine payments protocol is a open and payment method agnostic protocol for machine-to-machine payments. And I think the best way to think about it is it is like the payment form for agents. So today, I think all of us are very familiar with coming on a page. You see kind of a standard payment form. It's all the same layout, but you can plug in hundreds of payment methods into that. If it's cards, Klarna, even paying with crypto, all of that plugs into that. And humans are used to that UX. It's very efficient. But if you expand that, and I think this is why we are so interested at Tempo about machine-to-machine penguins, is we just see it at a huge level. Velocity, and a huge compounding number, month over month growth. So we think as that velocity increases, you just need more efficient interfaces.
Brendan:
[5:06] So MPP is, we view as a payment method interface, which agents can interact with really, really efficiently. We've done a bunch of benchmarks to make sure this is true, and allows them to transmit payments over multiple payment methods, seamlessly in HTTP requests. So we just see this today, a lot of API services, people ordering sandwiches today, you can use it in MCP servers, but you can also use it even for if you just wanted to host a video or host some content, kind of this classical micro payments for content use case that people have been talking about since the 90s, but really hasn't been feasible just because of the interface it's exposed in. And people thought about it at the time. The status code was created in the core HTTP spec, but never really formalized. And we think MPP is the formalization of it. And we have designed it in such a way that is entirely neutral. Payment method, currency agnostic, works with web standards. And we actually submitted it to the IETF this morning in order to be the true
Brendan:
[6:20] spec for 402. and we think it's the best chance for a totally neutral approach.
Ryan:
[6:25] Wait, be the spec for 402. I guess maybe the question is, so we've definitely heard this story before and we're excited about it.
Ryan:
[6:34] But we've heard it more in the context of another emerging standard,
Ryan:
[6:37] Which is X402, Coinbase's champion that Cloudflare has been behind it. I also thought somewhere in the world, like Stripe was involved as well. Just to be clear, is this like a competing standard to that or is it some sort of superset? Like how is it similar versus different? How much is it competitive with X402 versus collaborative? What does it look like in contrast to X402?
Georgios:
[7:02] With regards to the Cloudflare and Stripe component, Cloudflare and Stripe are both companies that are on the record that they want to be neutral and want to be supporting all of the systems. So today, Stripe has support for both MPP and 4x402. Cloudflare, the same. Cloudflare also this morning added an MPP proxy GitHub repo where you can use to make any service MPP enabled. So from their perspective, all these platforms, they will just adopt everything and they will allow their users to choose what they want to choose based on other reasons that the protocol to differentiate on.
Georgios:
[7:39] Now, how do they actually differentiate on? I think it's three reasons. One is performance. Two is developer ergonomics. Three is platform support. And let me unpack them.
Georgios:
[7:52] With regards to developer ergonomics, David, Ryan, we've been in many podcasts like this together. Our team is a team that shipped the Foundry project, which is the thing that powers all of the 100 billion IndieFi that has existed over time. We know how to build, how to test developer tools for backend developers. Our team also has built Wagme and VM, which are the most used front-end frameworks right now for any Web3 crypto app. It's used probably by 90% of every crypto website right now. So we really know developers. And I think that really manifests in our API design, in our library design, how ergonomic and how intuitive things are when they're built.
Georgios:
[8:35] And when we looked at the ergonomics of X4.2 and other things on the market, we just weren't satisfied. So we thought it's just easier for us to go back to first principles and think, okay, hey, what's the simplest developer-friendly ergonomic API I could use? And the way that I saw it is let's go back to the basics. What is the most basic thing that we're doing here? We're doing auth. That's literally what we're doing. We're just saying instead of authorizing with an API key, let's authorize with a payment. So we went back and we looked at all the literature and our basic auth, bearer auth, all of that. And that was what informed the original design of the first iteration we did for MVP, which is actually in a funny way what made it so easy to make it payments method agnostic. Versus if you look at X402 or whatever other approach, X402 is a bit tied to the facilitator, which is a very specific implementation detail almost, but should never be surfaced all the way up to the protocol. So we had a different approach and we said, okay, this will be too much effort to make it work with that. Why don't we do it with our own and see how that could work?
Ryan:
[9:41] Now, in the history of the web,
Georgios:
[9:43] It always has been better to have more than one solution because then these
Georgios:
[9:47] two solutions or many more solutions, then they end up iterating to what's optimal for the consumer. So, yeah, I think there are like two different approaches that you can use to use agentic payments. And I think it's going to be a beautiful, you know, step by step on evolution of how do people make it work.
Ryan:
[10:04] But to be clear, it's a it's a competing standard then to X402. You think it's better, but it's competing.
Georgios:
[10:10] I think they could compete. I think there's a world where they converge. I think either could happen. You could, well, technically because MPP is more general, you can express X402 in MPP. And we had a draft that we hope to publish soon about this. I don't think you can express MPP in X402 terms.
Ryan:
[10:28] Okay. So MPP could be a superset and you say it's more broad. When you say it's more broad, I noticed like Visa integrating it, for example. So it's not just Stablecoin smart contract types of payments. It's also some of the traditional type payments. Is that what you mean by more abroad? Like, you know, it works with a Visa card?
Georgios:
[10:47] And Ryan, to be clear, I want to still man the X402 side, which there's a blog about X402 V2, which specifies that, okay, it will be more payments method agnostic and so on. That's fantastic. But what we have right now is live and it works with, and it works already with a bunch of like other methods. The methods that we support right now are on Temple, which supports both one-time charge payments and sessions, which we should talk about in a second. We support Stripe, where the interesting thing on Stripe, and Brendan used to be at Stripe, so he can say more about how that works, that it works with anything that Stripe supports. So we could be doing Klarna over MPP to do payments. It supports Visa cards, so Visa wrote an extension to MPP. And we also have Bitcoin Lightning support, which I think is just remarkable that we spoke to the Spark team. Just gave them the spec and same day or the next day they had proof of concept repository extension to spec just one shot. Of course, their agents did a lot of work. But I think that was just remarkable and it just showed how like easy this thing is to extend.
Ryan:
[11:55] So may the best standard win. I guess that's where we are right now. Can NPP be exported to other chains besides Tempo, other EVM chains?
Georgios:
[12:03] Absolutely. And again, we just didn't have time to do, you know, integration games are just so hard because somebody will always say, hey, you didn't do my integration. So of course, yes, it can. And we want to do it. I actually think I have a pull request draft up about that, or maybe it's a thread in our agent. But yeah, it works because it's just call API, get response back, respond back with a signed transaction. This works everywhere. The thing that works nicer on Tempo is that you can pay fees in any stable coin, for example, without having to do more work just because the chain supports it natively. But yeah, of course it works anywhere because what is an agentic payment? Just a signed payload that says, hey, transfer five bucks to Ryan. Absolutely, it works anywhere. And even the sessions things that we're going to talk about in a second also works anywhere. So because it's just a smart contract that can be deployed anywhere and we've known how to do these things for many years. But yeah, it can be deployed on any chain. It can be deployed on EVM, SVM, whatever you want. we started with the tempo transactions because that's what we're working on. We're not going to do everyone's integration work up front, but we do want to expand it to more things. So whoever wants to work with us feel free.
David:
[13:18] Giorgio, so I get a sense of urgency with some of the, just the way that you're talking. I mean, you're in San Francisco. It's the epicenter, the ground zero of what kind of seems like the AI arms race, at least on the United States side of things, the domestic side of things. And I think maybe when we look back, the transition from 2025 to 2026 will be remembered as the time in which society went from chatting with their AI as a chatbot to chatting with their AI as an agent. Like, you know, the agentic word is like a huge buzzword these days. And we're just kind of imbuing our agents with capabilities as we develop them. Like, you know, agents got memory recently and like the MCP protocol was issued and that was like communications and like with the whole agentic commerce side of things, both X402 and MPP, the standards that you guys are issuing. This is like the tool that we were adding to the agentic tool belt is like, you guys can pay for things now. This is like, you guys now have money and you guys have a native way of paying for stuff on the internet. And sell things, not just paying. And also receive money too. Yeah, like send and receive.
David:
[14:23] Can you just kind of illuminate, illustrate, expand on just like the agentic commerce side of this story, which is the thing that I think the AI arms race is really focused on right now is like trying to unlock that part of the agent tech tree.
David:
[14:40] When we do unlock that, which is seemingly being unlocked, What comes out of that? Like how big of a deal is like agentic commerce broadly in your mind?
Brendan:
[14:50] I think it's huge.
Georgios:
[14:51] Why? So think about it, like I use my agent to do things more than I click around in Chrome, Brave, whatever. That in itself means that I want to make bookings, I want to buy things, I want to do all sorts of web crawling, auth into various services. Like people in our Slack, they want to use, for example, 11 Labs. But to them, even the friction of going to log into the service and get billing set up, it's kind of painful. They just don't have, you know, the attention span or the patience to like try out new things. Why? Because we're just in a world where like... We move fast.
David:
[15:31] The world moves fast these days.
Georgios:
[15:33] Yeah. And so people just have low patience to do things. And so while they're trained now to be doing things in their agents, what this means is that they want their agents to be equipped with money and such that their agents can then go and do payments on the web for that. So for sure, I think this makes sense from the user side. Now from the seller side and i think david orion you guys interacted with like this ben thompson blog that i shared the other day strategy yeah where the the insight there for the seller side is that basically as a seller your website has to become an api asap not just a ui why because you want it to be served the agent in the best way possible such as they go and buy things from you Why? Because the user isn't going to go and click around. It's going to be a machine making payments to another machine, hence the name.
Georgios:
[16:27] And so in these cases, like Ben makes a very interesting case about ads, which I don't know if that will play out or not, because there's many ways this can play out. But he says, hey, if like way less people are browsing the web with their eyes, then way less people are going to be converting from ads. And maybe that means that this forces all of the sellers or of the people running websites to actually switch to paid APIs. Why? Because, well, they need to monetize somehow. This is already the case, by the way, on Cloudflare, where Cloudflare is adding like methods on their endpoints to like fend off the bots. I think there's something more beautiful here, which could be, hey, just embrace the bots, just browsing and like dosing the web and just say, hey, to get through this, just pay me. And so I think there's just like such a strong tailwind to agentic payments that just wasn't there, frankly, last year. Where you just give five bucks to your agent. And we showed that in the demo that we did with Brendan in our launch video where we gave our agent some bucks. We called it search, generate image, upload, send email. I've never myself used any of these services alone, but because it could just crawl the web of services, it could then go and do a multi-step paid workflow across everything, almost like a map reduce or a waterfall, however you want to visualize it. And then that's beautiful. You just give your machine money and it can do amazing things that you never thought were possible.
Ryan:
[17:52] For people who aren't familiar maybe with that Stratechery article, I'm wondering if you guys could summarize that. I know, David, you sent it to me earlier this week and you said this is one of your favorite articles in Georgios. You just tweeted it out.
Ryan:
[18:03] I just skimmed it and it's called The Agenic Web and Original Sin. We'll include it in the link in the show notes. In my retelling of it, and you guys tell me if I left something off, it's basically back to Marc Andreessen's idea that the internet, as we created it, and he was there in the early days of Netscape and Mosaic, and the internet had this original sin, which is we didn't have a payment standard, a 402 type of standard. It was left blank, and we didn't work with credit card companies effectively anyway. And because of this original sin, the entire model for the internet was an ad-based model that we still have today, that we still maybe suffer through today. I think Ben was talking about that and saying, well, you know, it had to be that way because with the human web that like the most logical business model was always going to be ad-based anyway. So it wasn't just because you forgot to add a standard. It was just always destined to be this way. But he's saying that the agentic web has, I think what he's saying is it has the potential to completely re-architect the business model of the internet. And so if the human- His argument is that
David:
[19:12] The sin is now. It wasn't Zen, but if we don't do the payment things now, that's the sin. The sin is modern.
Ryan:
[19:19] I think the agentic web will drive a different business model in that AIs scanning different websites, they're not going to click in on banner ads. They don't care. They're not going to be persuaded or be distracted. Okay. But they will want good content and we need good content. We need an incentive mechanism for good content. And so the natural business model that falls out from that is AI agents paying for data providers, paying for content, creating an incentivization mechanism. So the new web actually is kind of this AI agentic web where AI agents are paying directly for content and it's no longer ad funded. And maybe the human web, you know, just kind of is legacy. That's in our past. Is that basically what he's saying? What did I leave out?
Georgios:
[20:08] I think you... That was an impressive scheme, Ryan. Because that's kind of exactly what he's saying. I think there's a bit of juries out on ads, and I'm by no means an ads expert. There's juries out on how ads will evolve on this because the whole thesis is that ad revenue doesn't convert in the agentic web. Why? Because agents got no eyes, so they don't care about what number you're on a list or whether it's surfaced or not because you just tell your agent, And hey, focus on like the actual content, not on the ads, kind of like a native ad blog kind of thing. I think jury is out. Why? Because ads may evolve into prompt injection style things where, hey, you have the website and the ad is, you know, text on the website, prompt injection. So it could go anyway.
Ryan:
[20:58] I mean, also the jury is out too in that some of the AI companies seem to be looking at advertising as a way to monetize and as a way to grow. You know, most of the open AI is very much leaning in this direction, whereas maybe Anthropic is saying, hey, no ads, that's our special feature. I don't know if you guys agree with this, but let me throw out a statement that it would be a much healthier future for the agentic web and for the next platform. Internet and all of AI to be payments funded rather than ad funded. And here's why I say that, because I get very worried about a super intelligent force I'm telling all of my life's problems to, knows me better than I know myself, and has a super ability to glaze me and persuade me to do things. I worry about that super intelligent agent who's my best friend being able to upsell me things all of the time and basically manipulate me into whatever it wants to do, including buying all sorts of things maybe I don't need. I get very worried about that model. And so it's more comforting to me to look at a model where it's actually paying for things. I know what's going on in the background. So I would contend that world is a better world and something that we actually want. I don't know if you guys agree with that,
Georgios:
[22:23] But that's where I arrive. episode where they have like a Neuralink style chip on a person's brain in a freemium model and then suddenly it's a crazy it really reframes your thinking on a lot of things the idea is that you have a chip on your head and in the freemium model every now and then you say something that's an ad to your if you pay you don't have an ad so that's great so I think that's a fair statement what you're saying I think people want to feel safety when in the when telling people their secrets, right? And it's true that people use the agent or the chat GPT or whatever interface as a more trusted confidati that, you know, the Google search bar or maybe, you know, the Google search bar, of course, itself is evolving. So, yeah, I think that's a fair statement. But again, I think it's too early to tell how things will evolve. And the best thing we can do right now is observe, understand, and try to have a notion of preparedness. But as David said, and absolutely, we're in San Francisco, things are crazy, things are moving really fast. It's hard to make predictions about things. Even a year out from now, it's really hard to make predictions. And I think the people that can adapt the fastest are the ones that are going to be the most prepared.
David:
[23:40] Yeah. This next part also came from the Stratechery article, but George, you said it's hard to make predictions, but it's easy to spot the trends. As you just said, people really trust their ChatGPT interface or their Cloud interface, the things that come out of that, the text that I read, I'm like, generally trusting of.
David:
[24:01] And I either I realize that, oh, I typed the wrong prompt, or I can actually realize that like, this is actually a hallucination. But like, it's intuitive to me either way. So like, I trust, I trust these websites much more than I do when I go and I see like a Facebook ad, for example. Now, there's, there's two other technologies that are very related to this. There's MCP, the model context protocol produced by Enthropic. And this protocol is basically just how agents talk to services. It allows an agent to basically read a website. And then there's also NL Web. This one came from Microsoft and...
David:
[24:38] Excuse me, this is the one where agents can like read a website. MCP is communication. And so like NL web for Microsoft is like the agent's eyes. MCP is the communication between agents. So this is all agent to agent stuff or like a native internet to an agent and vice versa. And now with MMP or X402, now we also have pay. Now we just had Ilya from Near on the show. He's also one of the authors of the Transformer paper. And he said one of his early predictions is that, you know, in the future, you're just never going to go to a website ever again. There is no more internet, actually. Like, the websites are gone. You're just going to talk to your AI in the same way that, who's the name of Iron Man? What's the Iron Man guy?
Ryan:
[25:23] Tony Stark.
David:
[25:23] Tony Stark. The same way Tony Stark talks to Jarvis. And there's no more internet anymore. And I can see... Well, that is the internet.
Ryan:
[25:31] Basically. That is the internet.
David:
[25:32] Yeah, the internet's in the background. And I can see...
Georgios:
[25:36] Right? And you could easily imagine building a browser right now that actually doesn't care about the JavaScript DOM. I think, Brennan, I forget if I told you about this crazy idea. Like we could build like a browser that just doesn't care about rendering the web in the proper way. You know, people say web browsers are really hard to build. Why? Because you can support all of the JavaScript, basically all of the whole web, you know, gigantic test suite. But what if you just didn't and just said, hey, give me raw HTML back so you don't even need the browser the interface is no longer the browser you just don't interpret even the raw html and you just tell the agent hey parse this into like a interface that i would like and then you get you know maybe you're tony stark or maybe you're tom cruise in minority report doing doing things yeah
Brendan:
[26:22] We view and we really intentionally designed MPP as a composable standard that works well with a lot of other things. I think you've seen this today where you can plug MPP. Yes, it works in the standard HTTP request flow, but it also works in MCP. It works over JSON RPC. And so you can do all these various transports. But we think what the things that people are really excited about next and what we do often developers is like, okay, how do I translate identity? How do I translate things like reputation? How do you track that across? And we don't intend MPPM intentionally designed it where we're not going to try and jam all these things together. We want to compose with, just as we compose with multiple different payment methods by design, we're going to compose with multiple components of what are other things that people and machines need to do and are useful. So we see a bunch of identity proliferation. There's a lot of different standards for discovery, etc. And we want to work with as many of those as possible. And that's why we designed the protocol to be simple as possible and as neutral as possible, because it's a massive tailwind towards journalists.
Georgios:
[27:45] Bernard is too humble to share this, but he has also written a great discovery proposal for MPP co-authored with the Merit Systems team, which are good friends of ours, and we remember it closely. And the idea is that every MPP service can define its schema via, again, very well-owned, literally, there's a dot well-known path that is a well-established web standard thing to be discovered by services. We're not building a search engine ourselves, but we're just building the ways for people to plug into their own search engines.
Ryan:
[28:15] The MPP libraries, they support MCP,
Georgios:
[28:18] Or you can just call them over standard REST. So again, it's not prescriptive
Georgios:
[28:23] about these things. These are just layers on top. And there's other things. For example, there's UCP, I forget the acronym, by Shopify. There's A2A by Google. There's AP2. There's a bunch of things, and they all do different things, but none of them really nails the pay angle. So we've made it so that you can do the payments based on the things that we talked about earlier, and then you can compose it however you want with whatever is on top.
David:
[28:52] What I hear there is that you guys don't have an opinion about the direction of the internet. If the Ben Thompson Ilya outcome of just like, you know, there's just your AI and it renders the internet for you to visually appeal to how you like it, maybe that's great. Maybe natural forces point us that way. Maybe the MPP or agentic payments is like a very important puzzle piece to get us there. But you guys are unopinionated about where it.
Ryan:
[29:17] Goes after this.
Brendan:
[29:18] I think it's hard to draw a specific like long-term bet. I think like we'll look at the structural trends is there is just going to be more things you see on GitHub, more code being generated than ever before. There are more services going live on Stripe than ever before. There are more people just building things and building things that produce valuable work and we think that those things should accrue value because they're providing value so that's really the purpose you just think there's going to be more things they're going to proliferate, they're going to get built faster and
Brendan:
[29:56] That very much is why we're excited about Tempo and especially machine payments on Tempo is because it draws to the natural conclusion of like, okay, what is the fastest way to get started today as a developer, spin up a service, start monetizing it and provide value and get discovered. We think that is MPP and specifically MPP on tempo today because you don't need to touch a single API key. This is really the dream of stable coins of crypto is I just been up a website, integrate MPP, host it and start getting payments in stable coins immediately and not have to go through large set of flows. You will need to do that for offering deeper integrations, which we're really excited about. So we just want to see more things in the world, different ways to monetize. And we've been super excited to see like new things to develop.
Georgios:
[30:50] So the interesting thing about MPP then is that it's great from the least sci-fi to the most sci-fi scenario. The least sci-fi being just a paid API. The most sci-fi being, you know, open, close, run the internet, and they all have wallets. Actually, there's like millions of open closing, like so many of them that they all just like nonstop pay each other. And there's a lot of like the in-between stuff that is exciting, which is all of the, you know, the agents crawling the web and paying for services. And I think we're not in the least sci-fi. I think we're definitely in like a bit sci-fi world based again on the AI tailwinds that we're experiencing. And I think the very sci-fi world is actually not out of reach.
Ryan:
[31:32] Yeah, I mean, to give an example of this, so in the announcement post for Tempo, right, there's like one of the first things you can do is go create a wallet. And so I kind of expected to go see, okay, there's a wallet, maybe Tempo has rolled at its own wallet, or maybe there's a link to like whatever, MetaMask or Phantom Wallet or everything I'm used to. That's not what happens here. What happens here instead is I'm greeted with a page that says, supercharge your agent. It's almost like about spinning up a wallet for my agent instead. And there's a button I could click called try with your agent. I'm not sure what that does. I haven't tried that out, but if it's going to connect into my cloud or my open AI or what, but the way to use this is kind of like it's a chat type interface. Find me a hotel and flight to a conference in New York December 28th to 29th. No red eyes, keep it under $700, right? That's the example that you're supposed to send to your agent and I guess it gets a wallet, it spins it up and it does this work and it pays for things. So even your onboarding flow is like agentic, right?
Georgios:
[32:34] Absolutely. So Ryan, the thing that, so we've sent this demo flow to a lot of people and they all get this wow moment. So yesterday we actually told one of our colleagues, hey, install the tempo skill. So every agent, agents now have a thing called skills, right? Where the skills are just prompts for how to use the thing. So we literally told one of our colleagues to say, hey, tell your OpenClo on Telegram, tell it hey install the tempo skill and then call georgius's number
Ryan:
[33:03] Call georgius's number yeah
Georgios:
[33:06] And like we gave it my number like
Ryan:
[33:08] Your phone number i kid you
Georgios:
[33:10] Not it called me why
Ryan:
[33:11] Did you want this what
Georgios:
[33:13] Did it say like to like see what's going on okay like insane to see that hey like somebody literally just gave it a prompt yeah nothing about tempo it literally so the thing that you said like that you read from the website it's just
Ryan:
[33:28] Called the board
Georgios:
[33:28] So we don't have our own chat interface i don't think we'll like i doubt that we would do one sure so we just copy that problem you put it in your agent your agent could be amp cloud codex open flow whatever whatever you want yeah it will install the tempo skill it will download the tempo wallet in the background without you even knowing about it. It will create a wallet, will tell you to fund it, where you can fund it with Apple Pay, with cross-chain deposits from any chain with a QR code flow, or with a referral code. So afterwards, I can just send you a five or more dollar referral code to play around. And it just goes then and says, hey, what services do I have available? And then just does it on its own. It's really magical. It gives you an incredible...
Ryan:
[34:13] So when this OpenClaw instance was calling you, did it have to go pay for something in order to call you? Did it pay for a service? It paid for a service.
Georgios:
[34:21] It paid for a service that we have integrated, which lets you do text-to-speech to call.
Ryan:
[34:26] Voice over IP type call or text-to-speech type thing? Okay, cool. When it's setting up the wallet, where are the private keys? Are they somewhere on my machine? Yeah.
Georgios:
[34:36] So the thing that we've done in the Tempo wallet is pretty novel, I think, and it deserves a good explanation, which is when you create the Tempo wallet, it kicks you to the website, the website that you were just in. In the website, it tells you to Face ID to use your iOS biometric auth to create a wallet.
Georgios:
[34:58] Now, once that wallet is created, it's using your phone's secure enclave, which means that that is not stored on your phone or your desktop, whatever you want. It's using task keys, which is a well-established technology for doing this thing. It creates the wallet or you sign back into your wallet. So there's never a private key in this case. But when you call into it from your agent, it authorizes, it creates an, call it ephemeral private key, call it a scoped access key. That's the naming that we use. It generates a small private key that can only access up to a certain number of your funds. So if your wallet has, say, a hundred bucks, when you log in and when you do the flow, it takes you to the website and it says, hey, authorize your agent to spend up to 10 bucks or 10 bucks a day or whatever, you know, granularity you want to do, which is, by the way, the same primitive that we use to support subscriptions in MPP. So your agent locally gets a private key that's safe to lose, which is very useful here. It means that if you lose that private key, you're covered. And if somebody steals a private key, the losses are capped. Which is very useful if you think about, hey, I just want to let my agent rip, but my wallet has, you know, $5,000. Like I don't want it to go and lose all my money by accident. So in this case, yeah, you log in, it creates a private key, it authorizes it, and then you can let your agent rip while feeling safe about your phone.
Ryan:
[36:25] So is it accurate to say that kind of the master private key, right? Not the sub kind of pass key or private key that the agent gets. The master private key is like, Like that's passcode secure enclave on my phone? Or does it exist? Do you guys have a copy? Is it shared at all?
Georgios:
[36:42] A copy of it. It's a self-custodial wallet. Okay. It doesn't use any third party. It uses passkeys, which lets you... So in Tempo, we have added a native passkey type. If you have seen all of the account abstraction wallets in the Ethereum world, they all have this feature, but it always requires an extra component called a bundler, a relayer, whatever you want to call it. Versus in Tempo, we pulled that in in the Tempo transaction format, which we have published about, which means that you can use passkeys without an intermediary to, you know, go and transact with the chain. And this passkey is stored on your device. And it's also if you're using iCloud keychain or if you're using 1Password, you can also sync it across devices.
Ryan:
[37:26] Okay, cool. All right. Now I feel like we need to go back to Tempo and the thing that was launched. and we've done episodes on Tempo before, so Bankless listeners may be somewhat familiar, but I think we need a refresh. So this is what? This is a layer one. It is EVM.
Ryan:
[37:45] I'm sure the RETH client is involved somewhere, Georgios, knowing you. Can you just lay out the specs of this thing and what it is and kind of throughput, what assets are on it, just kind of some of the details that somebody from the crypto world would want to know
Georgios:
[38:04] Tempo is a layer one blockchain focused on payments We're making opinionated trade-offs to optimize for the payments use case. Tempo is permissionless, meaning that anybody can run a node and anybody can validate the state transitions of the chain. Tempo is live with 11 validators. Most are operated right now by Tempo and we have some externals running and we're onboarding more validators on this in the next few weeks. Externals from a geo-distributed cluster so that we have in Europe, US, and so on.
David:
[38:39] It's permissionless to run a node, but you are onboarding more validators. How do you square those two things?
Georgios:
[38:45] It's permissionless to run it. Imagine that you're Alchemy, for example. You can just run a node and like serve RPC traffic or you're an indexer like Album and you want to like index the chain or just a normal Cypherpunk user that wants to not trust and verify everything. Running a node is a two command process.
David:
[39:04] So everyone has the ability to listen to tempo, but not everyone has the ability to write to tempo. Is that correct?
Georgios:
[39:13] No, it's similar to, in Ethereum, like to become a validator, you need to stake 32 ETH. Right. Well, we don't have that. And everything else is the same. It means that anybody can run a validator, anybody can run a normal node, anybody can submit transactions, anybody can create a wallet. There's no special casing, no blacklists anywhere.
Ryan:
[39:32] But running a validator is still permissioned in Temporal.
Georgios:
[39:35] Running a validator is still permissioned. We're in the early days, we're still figuring out what the right way to expand this validator set is. I think we have a pretty promising roadmap in the next few weeks, but we need to be very thoughtful about how to execute on this.
Ryan:
[39:49] And there's 11 right now validators
Georgios:
[39:51] That are running. Right now there's 11. And we don't have a dashboard yet on the blog explorer. We need to like make that happen. But there's a validator manager contract in the blockchain that you can check from our docs, which if you query, it gives you the full list.
Ryan:
[40:05] And do the validators have to stake something or they're just kind of whitelisted?
Georgios:
[40:09] We have a multi-sig that controls who the validators are.
Ryan:
[40:13] Okay. Okay. All right. So please continue then. We were just, you know, what else should we know?
Georgios:
[40:19] Yeah. So the node is built on the Reth client, where the Reth client is a project that we've been building at Paradigm for the last four years at this point or more,
Georgios:
[40:33] Where Reth is built as an extensible client, which means that we're simultaneously able to support Ethereum L1 for the normal use cases that people in the Ethereum L1 world use it for normal RPC nodes, for validators, for MEV bots, and all of that. It also supports layer twos like base, for example, which is running Reth underneath, or the rest of the L2 ecosystem. And we're also building tempo with Reth. Why are we doing that? Because we have a stable foundation, which means that we have all of the EVM, JSON-RPC, developer tools, they just work. The node is the highest performance node as of like the last few weeks or months plus something on Ethereum L1. And it's generally very fast, very stable, well tested. It just made sense for us to use that as our foundation. So that's on the execution layer side. The node has features for payments. For example, we have a pre-compile that's an ERC-20 contract with permissions enabled in it. Which is generally targeted at stablecoin issuers, for example, that want to have, say, various rules, or let's say you're Tether or USDC, and you want to, instead of rebuilding all of the permissioning features from Scouts, we'll just have them available out of the box.
Georgios:
[41:56] Then Node also features a thing called the payment lane, where the payment lane is something very valuable. For example, if you remember late last year when one of the big market volatility events happened, fees everywhere spiked. I think everywhere except Solana, which was fantastic. And congrats to them for doing that. And the idea is that you don't want DeFi-related spikes to be affecting your normal payment activity. So what did we do? We just said there's a zone, there's a part of the block that's reserved only for payments transactions. And there's a part of the block that's allowed for anything. And the payments section of the block is going to always have predictable stable fees that you're not going to feel anxious about. versus in the rest of the block, you know, the things that we're familiar with can happen. And a big shout out to the Commonware team by Pat O'Grady and Falch. Tempo is basically combining the best of the Reth project on the execution side and the best of commonware consensus on the consensus side, which is what lets us then like executing this ambitious global validator set roadmap that we're going to be executing in the next few months.
Ryan:
[43:04] How about, you know, throughput transactions per second and also validator node requirements to like run these things?
Georgios:
[43:11] Yeah. So for gas per block, I believe is 500,000. So half a gigahertz per block. And then block time floats from 400 to 600 milliseconds. It depends on networking conditions at any given time.
Ryan:
[43:27] Okay. And so half a gigahertz, does that translate to something like, you know, 5,000 transactions per second? Or are we even able to translate? Does that distinction not matter?
Georgios:
[43:40] Call it 10. 10,000 transactions per second. In our last benchmark, that was what we could hit. So call that like as of last week, I believe that's what our last benchmark. The thing we're going to be doing, we're going to be publishing a bench or perf.tempo.xyz because I think the hard thing about benchmarks is that, you know, we're going to talk about them and then in two weeks, they're going to be outdated because in the background, we have an agent that just looks at all of our stuff and like it continuously optimizes them. You know, it's crazy. It works. Like we just have like an agent that looks, runs benchmarks all the time. And in part, that's why the Reds clan got so much faster in the last like month and a half. Yeah, it's really remarkable. So David, you were saying again, much earlier about urgency, performance, all of that. Like the AI has really transformed how we work on all of this stuff.
Ryan:
[44:34] So really fast block times, as you said, you said 400 milliseconds.
Georgios:
[44:37] Really fast block times, finality, single slot, because we use simple consensus.
Ryan:
[44:44] Half a giga gas throughput,
Georgios:
[44:46] Or sorry, half a giga gas per block, which translates to if you count like 100,000 gas per transaction, like you can do the math. And then what am I missing? Network and node requirements. Well, we are actually super excited about this because we're building it on RETH and we've been optimizing RETH for the Ethereum L1 use case, which is really about like trimming requirements as much as possible. We just published a new minimal mode for Reth where Ethereum mainnet itself is 150 gigabytes. And imagine that Ethereum mainnet has like lots and lots and lots of stuff in it. Running a tempo node on commodity, normal at-home software, node hardware is possible right now. If you open your laptop and you, after this, and you take, say, point it to the docs and run me a tempo node, I'm promising to you it will work. We'll download the snapshot. The snapshot is tiny right now. It will get bigger over time, but it will be small because gas pricing or state growth to be good.
Ryan:
[45:52] But to participate as a validator, like is that going to require some beefy bandwidth? I think I saw a number of like 10 gigabits per second or something like that. Is it still primarily, if you're participating as a validator, let's say, or just like running maybe like a serious node, are you still going to have to run that in a data center? Are you guys trying to get this down to like running your home?
Georgios:
[46:14] Absolutely not. So there's no intention for the, again, I don't think about validators and non-validator is different. I think it's all the same to me, networking-wise and node requirements and so on.
Ryan:
[46:28] The node operators will require,
Georgios:
[46:32] Will work on commodity hardware, normal residential connections, shouldn't need, you know, gigabits upload. In part, like why this is going to be possible is because of the commonware stack, which, you know, you should like go and look at like what they've done recently, where they can use erasure coding to make the networking on a per-node basis to be much lighter. So yeah, I think it's going to be really exciting. I think there's a lot of players that are racing towards bandwidth-efficient consensus that's also very high performance. Big shout-out to the Monad team who has done amazing work on that with RaptorCast. I think the Solan Alpenglow work is also really strong on that. So I think it's going to be, I don't know, I think the world where we thought of low block time, high throughput, and, you know,
Georgios:
[47:18] Again, I think, like, to go to 10,000 TPS, everybody, or maybe, like, a bit more, like, people are covered on a distributed network. I think now the interesting thing in the web payments world is, like, how can we do a million, like, a billion payments per second, things like that. And that is, like, an open question, a lot on how people are going to do that. Our answer is the MPP sessions, where you can bypass going to the chain every time. But I think that for people, few tens of thousands of transactions per second, like in the next, you know, even now, like I think the world is in a good spot. Like whether this Tempo Solana Monad, like I'm saying that off the top of my head is like the high performance, the centralized blockchains. I think like people are covered. I don't think, I think the requirements have gone down so much. I know the operator experience has gotten so much better.
Ryan:
[48:06] It's been a question, I think, since Tempo was announced in sort of the crypto circles as to like why an L1 rather than an L2. And I noticed just yesterday, you tweeted this, Georgios, the thing I learned earliest in my crypto journey is that to really scale a decentralized network, you have to avoid consensus. That's what drove me to L2 scaling. You talked a bit more about state channels and L2s. This seems to indicate that you're still kind of L2-pilled. And yet, as one of the architects and engineers on Tempo, you guys went with an L1. I think a lot of people are probably listening to this and scratching their heads and saying,
Ryan:
[48:47] Why?
Ryan:
[48:48] Why an L1 rather than an L2?
Georgios:
[48:50] I think the simplest answer is just developer velocity and being able to self-express and being able to do the things that we want. That's one. And we didn't want to be bound by the DA that the Ethereum world would provide us. At the same time, of course, we're continuing to serve the RETH project as normal for Ethereum L1s and L2s. But for our ambitions, we felt like we had to kind of be able to self-express in all of these ways. And I think that's the most of it, right? I think being able to just own your fate and own your stack and being able to customize as you want is just so important.
Ryan:
[49:27] What do you think this means for the L2 roadmap that Ethereum has, you know, set off with, you know, four years ago? Is that sort of dead in your mind? Do you think L1s are the way to go? Or do you think there's a future for L2s?
Georgios:
[49:41] I don't think these are intentions. So there's two questions here, right? Like A, like what do I think about like L2s in general? Or maybe for 10.2, what do I think about L2s in the Ethereum context? So I can answer the Ethereum context one first, and then let me tell you what what we've been doing so
Georgios:
[49:59] Spent basically all my career building L2 technologies for the Ethereum world. I think these are the best, the most credible way, like do real scale on Ethereum. Again, even if we make Ethereum scale on L1, we will need to like roll some kind of like L2 technology in it. And it can be done, you know, as a P-Stack Arbitrum base, you know, all of these things, or it can be done in an enshrined way. There's this whole native rollups roadmap that has come out recently. I I think all of these are valid and good. So by no means...
Georgios:
[50:32] No, I don't think L2s are dead. I think L2s are necessary for a decentralized system to scale. I think Ethereum has this very cypherpunk core that is important to protect at all costs. And I love Vitalik's recent cypherpunk warrior arc. I think that's the right thing to be doing. So I think for Ethereum, absolutely, you need L2s. And I would definitely not make changes to that. I think there's things that in Ethereum world we need to figure out around tensions around branding, the whole, oh, are L2s competitive to the L1 and all of that? People disagree. I don't know what the right thing is there, but I would absolutely classify L2s as a necessity in the roll-up sense that people have been doing. In the tempo context, I think there's, again, a few interesting things. Well, A, MPP has two ways to interact with it. One is the charge method. The other one is the stream method or the session. The charge is every transaction goes on chain. which is just a normal payment session or the streaming method is actually opening a one-way payment channel with a server which is like an old school layer two technology from you know before most people heard about a
David:
[51:43] Precursor technology to layer twos.
Georgios:
[51:45] Yeah absolutely it was like kind of like the the grand parent of a lightning network if you want to think about it Like a one hop, one direction payment channel just says, hey, Ryan, I'm opening a tab with you instead of settling with you on every, you know, query, which is a charge method, which is how we're used to doing crypto payments. Write down on your notebook, you know, what is my score? And then I'm going to settle with you at the end when I leave there, when I want to close out the tab. And this means that I can use like an old school layer two technology in Tempo for the specific use case of client server payments, which fits beautifully, right? Yeah, I guess like I just want to sink that in that like the layer two technology that we have, we literally just deployed the layer two technology this morning on Tempo. Why? because even if the chains support whatever many throughput latency constraints, their demand that the AI world is going to bring is just 100x that, or it's just unknown how much it will scale that it's going to just hit the physical. Like to support it, you would need 10 gigabits or 100 gigabits of uplink or whatever. And we're not, I don't think we want to go there.
Ryan:
[52:56] Yeah, I read actually a post on that by Liam Horn, who I think works on the Tempo project as well. And he was kind of the forefather of state channels and kind of exploring that in Ethereum originally. And maybe I want to ask again about that that kind of cypherpunk thing versus what Tempo is doing. Because as you know, Bankless is like primarily a podcast for crypto natives. I think for a lot of crypto natives, they see Tempo and they have mixed feelings.
Ryan:
[53:24] I can say even me going to this episode, I have mixed feelings. On the positive side of things, it's very exciting to see, like obviously a world-class engineering team go execute against this vision towards a real world use case like agents,
Ryan:
[53:40] Start applying that to the eugenic web and kind of winning payments over to blockchain technologies. And do see what you're doing with respect to decentralization. At the same time, for CryptoNative, it's like a lot of our heroes, right? Like Georgios among them. Originally the Loom L2, like scaling Ethereum, Liam Horne, who I mentioned, Dan Robinson, who pioneered Uniswap, And a lot, even Dankrad this morning, he tweeted about Tempo, right? Now he's on the Tempo team. So these are kind of some of the crypto native heroes on this crops type mission, right? Censorship resistance, open source, private and secure, the Ethereum track. And now they're doing all of this in Tempo. And some people are scratching their heads and they're saying, okay, like, is Tempo now just taking the Ethereum vision, let's say, in this whole Cypherpunk vision and corporatizing it or just executing it in a different way. And I don't know if they feel kind of like lost by that or they feel like there's a competitive threat or they feel like it's just the open source thing that we once had and the decentralized thing that we once had and the crops thing that we once had. Now the corporations are here and they're taking over and they're out engineering us and maybe out executing us. And so there's some sense of like,
Ryan:
[55:06] Feeling left behind there.
Ryan:
[55:08] I don't know. It's a jumble of feelings that I think people are probably having as they're listening to this conversation. How do you square these things? How do we think about tempo in the context of Ethereum being cypherpunk, but maybe now it's a lot smaller than we once thought it would. We thought at one time it would take the payment use case and these would be L2s. And now it seems like that's happening outside of the Ethereum ecosystem. Of course, you're still at RETH. You're still co-developing on Ethereum. I'm sure that's going to be open source. So it's not like Ethereum doesn't benefit. And yet, it's not benefiting in the same way as we originally thought it would.
Georgios:
[55:46] And the question is, all of the good stuff that was happening with Ethereum, is it going to be happening not on Ethereum? And how should we... I think a
Ryan:
[55:54] Question that people feel is like, why not do all of this on Ethereum? Right why do this on an l1 chain in tempo with a separate kind of thing separate standards separate maybe future token base why not do this on ethereum instead
Georgios:
[56:11] Well i think would be hard well again look at the niches right like the payments niche just requires so much capacity and so much specialization that and i replied this to someone on twitter also i think earlier today, I think it was Alan, who said that, are there any approaches that you're doing to pre-compiles, for example, which is a technical detail in what we're doing,
Ryan:
[56:34] Ethereum would not do.
Georgios:
[56:35] And I'm like, yeah, of course, because Ethereum is a general purpose platform that empowers developers to just self-express in the most general purpose way and doesn't really discriminate for a particular use case, which is what makes Ethereum so beautiful. At the same time, when you're doing the payments use case, it's not just the performance stuff. There's like all sorts of like specific functionalities that you want to add into the system that just wouldn't happen on Ethereum. They just would never go through the governance post. so it would take too long. So that's one point. I think the other point is that high value DeFi will continue happening on Ethereum, right?
Ryan:
[57:14] Yeah, I guess maybe there's a question which is sort of what you think Tempo's intent is, which is, is Tempo's intent to kind of eat Ethereum's lunch or is it to sort of expand in a different direction?
Georgios:
[57:28] The goal for Tempo is to make the stablecoin native payments world happen. And I don't think that's intention with Ethereum being a large,
David:
[57:38] Successful force for the world. To keep going on the thread, the timing.
David:
[57:44] That has happened with the Tempomaynet launch today, timed with the EF mandate document last week. There's like a tale of two cities here. My read on that document was that Ethereum is like a sanctuary technology. It's a technology to make a sanctuary for people. And also not much more than that. If the Ethereum broader community wants it to be more than that, then the onus is on them to do it. But the EF, as it relates to that, you know, doesn't really, is not interested in doing the agentic commerce, agentic payments thing that is going to change the future because it's not about changing the future. It's about, you know, being a sanctuary for the people that need it the most. And there has been just kind of like a shift, like with some of the people that Ryan has mentioned, like Georgios, Donkrad.
David:
[58:34] Like there's been like a vibe shift of, I think people who are interested in growth is maybe something that I will characterize it as. Like if you were previously in the Ethereum camp, but you were really into economic growth and being more than a sanctuary and doing kind of just like, whatever's on the technological frontier, the AI arms race, you kind of found your way into the Tempo ecosystem. And so there's like, while Tempo isn't in tension with Ethereum, it does seem to represent a different vibes polarity than Ethereum does that is kind of like equal and opposite and an equal there's no question here Giorgio so I'm just wondering if you kind of like agree with the assessment I admit.
Georgios:
[59:20] I did not have the time to read the doc I believe it was 30 pages and
David:
[59:25] There's a lot of anime in it my attention.
Georgios:
[59:27] Span is fried as well I think there is a fair take around the vision tensions, maybe, how different visions rally different people. I think at the same time, I believe in both, which is, well, to me, like it has expressed in a weird way where we've made Reth a library and any changes that we make to make Tempo faster or make Tempo more robust or more feature stable and whatnot, but it just flows naturally to the Reth upstream library foundry as well. So like we announced Tempo, like one of the core points that we wanted to land on people was that all the work that we're doing, yes, it's making Tempo work, but these are the same codebases. So the same people that are working on the new codebase are also working on the last codebase. Why? Because they depend on it, literally. It's like a pin dependency in the codebase. So that's why to me, I don't, while I empathize with what Ryan said, then I totally understand. Yeah, somebody might feel like, hey, like some people that were doing good work are no longer doing that work with our other friends. I don't feel that tension as much because we just do both and it might be hard for everyone to understand, but we're just doing both from our end and we're good in and we think we can do both very well.
David:
[1:00:47] I've got just a handful of questions I just want to ask just as loose ends, it can kind of like be like a lightning round just to like kind of talk about the last little topics as it relates to tempo and agents and all that kind of stuff. Not terribly long ago, ERCA004 was introduced in the Ethereum land. This is agentic identity and reputation.
Ryan:
[1:01:03] What's your take
David:
[1:01:04] On just like the need of reputation for agents? Is that like, talk about the role that that plays in this whole agentic commerce vision.
Brendan:
[1:01:12] I think it's going to be very big. I think we're very excited. Like if you look at what we just last said about the future of Ethereum as this extremely decentralized, reliable place, we think reputation and registry and discovery will live in multiple places. And for the most important, the most decentralized agents, we're really excited about things like 8.004, how can you push reputation back? How can you say, I attest to this thing? But we think there's gonna be multiple and we don't know how that's gonna proliferate. So that's why we're really looking at a lot of things. We're just trying to see what are people building? How do we get the right solution for them? And how do we support as many things? 8.004
David:
[1:01:55] Among that. So like the idea, the concept of reputation for agents, valid idea, important for the future of the internet, totally going to happen, unsure how we get there, but eventually we'll fast forward a few years and then you'd be like, oh, there is totally reputation for agents. You're not in your head in agreement, it sounds like.
Brendan:
[1:02:12] Exactly. I think it's very much inequitable.
Georgios:
[1:02:14] And I believe we're, Brandon, we had that chat with David, right? Around one of the authors on that, where we shared some of our ideas on discovery, some thoughts on ERC 8004 and so on. So we like the idea. We don't know. I don't know if it's like the right way to do it yet because people are just going to do things, but we're talking to the people about it and we're just giving them our thoughts. A quick question on assets.
Ryan:
[1:02:38] So what assets are available right now on Tempo and what do you think that'll look like, say, in 12 months from now?
Brendan:
[1:02:44] So there are really two main types of assets on Tempo. One is our special stablecoin standard that Giorgio has talked to called the TIP-20, which is an extension of ERC-20, but supports a number of things that we just found.
Ryan:
[1:03:02] Stablecoin issuers need, like policies,
Brendan:
[1:03:05] Like controls, you can sponsor gas fees. It also integrates really well with our stablecoin DEX, which is an enshrined primitive in Tempo. So in a lot of cases, and we use this really heavily in MVP, if I'm a server, I am broadcast for service, and I want to get paid in USDC, but an agent might hold some other currency That's also US data dominated. We can handle swaps automatically and do so in a very efficient way that's controlled by the protocol and really just proliferate as many stable coins as possible through tip 20. Also, on top of that, anybody, Tempo is a permissionless network, so you can deploy any token that you want. But we have made, to George's point before, conscientious engineering decisions so that stablecoins via payment lanes via other primitives are more efficient to operate on Tempo than your random run-of-the-mill ERC20.
Ryan:
[1:04:06] Does that also mean that permissionlessness, does that also mean anyone can deploy, you know, some kind of smart contract outside of a token, any kind of DeFi, you know, primitive that they want, they can deploy that?
David:
[1:04:16] MemeCoins.
Ryan:
[1:04:18] Yeah.
David:
[1:04:19] What about offensive meme coins? Gross, disgusting, offensive meme coins. Can I deploy some of those on Tempo?
Georgios:
[1:04:26] There are already, I believe.
David:
[1:04:28] Nice.
Georgios:
[1:04:29] Nice.
Ryan:
[1:04:30] David, you into those? You're really into those? You and Acquire? How about, you mentioned kind of a DEX being sort of part of the core protocol in some way and has some special status. How about identity? Is there kind of a native identity type engine? Is everything AML KYC or how are you handling that piece?
Georgios:
[1:04:50] No AML KYC built into the chain itself. If you're using Chainalysis or whatever else in your app, you're free to do so. We were considering whether we wanted to add a feature that's not for AML KYC, but it's like a co-signer feature on every transaction, which the first use case would be fee sponsorship. So we already have fee sponsorship natively into the transaction, but we were thinking, hey, maybe we can generalize this in a way that you can do not just the sponsorship, but also security scanning, for example, like blockade as part, as a co-signer or your own transaction, or if you wanted to add the chain out or something. We haven't done that yet. So nothing, for sure nothing, like in the chain around KYC, AML,
Georgios:
[1:05:35] it is like app layer concern on a per app basis. But we are thinking about, hey, can we allow people to hook on things more easily to make the app layer better.
Ryan:
[1:05:46] All the existing EVM-based smart contracts, could those be directly deployed to Tempo? And then also kind of the other infrastructure, maybe the user infrastructure. I mean, will MetaMask pretty much work with the right RPC out of the box?
Georgios:
[1:06:02] So all of the smart contracts work out of the box. There is the FITO. So on Metamask, they don't show stable coins as your native balance. So that's something that we're working on with the Metamask team on.
Ryan:
[1:06:19] Okay. But all the EVM wallets, they will shortly work with Tempo then basically.
Georgios:
[1:06:24] They work, but sometimes there's edge cases because they're not programmed to show you your stablecoin balance, right? They're programmed to show you your ETH. And for example, if I have any of the stablecoins that we deploy via the stablecoin factory that people are using and anybody can interact with, anybody can deploy a stablecoin,
Ryan:
[1:06:45] And by the way,
Georgios:
[1:06:46] We're calling it the stablecoin factory, but the tip 20 factory is really just a generalized one, so it can be for any kind of fungible token. So you might deploy something and say it's a stablecoin, but it might not be backed by anything. So just treat it as a normal ERC-20 that's pre-compiled and is more efficient, has hooks on it. But yeah, because wallets just don't take all of the stablecoins that you have configured and they don't just sum it up and say, hey, this is your wallet dollar balance. They are programmed to show, hey, what's your ETH balance? That's the main thing that we've been trying to make work. The other thing is because you're paying fees in any token and Tempo lets you pay from any of your, like for example, if you're transferring USDC, you can pay fees in USDC. If you're transferring USDT, you can pay fees in USDT. And this happens via a baked into the chain primitive that Dan Robinson invented called the FIAMM. It means that wallets have needed to do some custom work to allow you to pay fees in stable coins, which is actually also happening in the Ethereum world because as the account abstraction wallets are gaining traction, they all basically need to expose that primitive to the wallet. And there's some nice ERC standards that we're also following, which are basically such that it works the same, similar to how the Ethereum world is doing the ERC 437 stuff.
David:
[1:08:08] Guys, going back to this very big vision of the future of the internet, between AI, micropayments, agentic payments, cosmetically what the internet looks like is changing potentially changing, potentially critically changing like the idea here is there's going to be like, you know, trillions of microtransactions being passed around, you know, any given day. Money is just going to be flying around the internet much faster. And it kind of makes me think that we're opening up a opportunity, a field of opportunity for brand new types of value to be expressed, brand new types of developers to build apps, new opportunities. So if you were to give advice to some young entrepreneur coming out of college, he's hungry, they're ambitious, they want to go do something and they want to catch some of that money that we think is going to be flying around from all these AI agents. How would you best steer them in the direction of casting the right net in the right location to catch some of the money that's going to be flying around?
Georgios:
[1:09:07] Yeah, I think the
Brendan:
[1:09:08] Most efficient thing that we see is figure out, like one, use the tools a lot and say like, okay, what are they doing? Where are they getting stuck? And wherever you're getting stuck or wherever there's friction, We think there is value to be accrued and just build services that unblock developers, unblock these machines that are increasingly compounding across every single vertical we see now from code deployment to even just people doing administrative work at their office and figure out where they're going to stock. What's the work that they're doing and make that more legible to them. And we think value will naturally flow to those services. And that's why we really developed MPP. We think MPP is the most efficient conduit to broadcast services and to have machines interact with them. I think we've seen this all where it's like Claude is so good and all these agents are so good at just calling the basic emix commands that have existed since the 70s. and we think the same thing applies where
Ryan:
[1:10:16] It is extremely intuitive,
Brendan:
[1:10:17] And we've gotten this in testing, to call the exact same commands HTTP payment auth to perform payments because it is just so baked in and so natural and all the tools exist.
Georgios:
[1:10:29] My take would be just build paid APIs, like figure out, like, for example, like, David, I bet you and Ryan, you guys got like a giant corpus of like interesting files with notes, you know, your prep for all of these like wonderful episodes that you've been doing for so long, you could pretty easily just tell your agent, hey, expose this as a service. And instead of like me and Ryan having to get queried directly by things, why don't these people just talk to my service?
David:
[1:11:03] Wait, you don't want to come on the show anymore?
Georgios:
[1:11:05] No, no, no.
Ryan:
[1:11:07] Is that a paid service, by the way? You just eat.
Georgios:
[1:11:09] You could do it as a paid service.
David:
[1:11:12] You could do it as a.
Georgios:
[1:11:13] You know, You could say, oh, you get like five, three queries and then to go more, you need to do more. I just think there's so many of these, hey, I have like some private knowledge base or some private or whatever information I want to like share the world, but I kind of like don't want to put it out there for free or maybe the hosting costs. I cannot cover the hosting costs and I don't want to talk about them. And like one of our colleagues, Shane, had said that earlier on, which is like, you can just build a service kind of like pays for itself in a very, I think a viral article on Twitter a few weeks ago about that as well. And in a funny way, these are like the self-sustaining, self-building autonomous things where you can just say, hey, here's some data. Okay, start making money on it. When you've made some money on it, then, you know, go build new things with that money. And there's actually a very well-established concept about that. It's called the von Neumann Pro, which is like the spacecraft that goes and builds more versions of itself. Yeah. So I think like basically the advice that I would give would be like get into that mindset of like you got something put it out make a API service figure out how to make it pay for itself and then figure out how to make expand which is literally like starting a business right but instead of being like a business business it's like hey you know take this thing put in an API and it doesn't need to be anything crazy you don't need to like think about it yes it can just be like text files excel spreadsheets templates it's how you think about the world,
David:
[1:12:41] All of these things. So we've got a lot of that. We have like seven or eight Google Docs that are each respectively like 700 pages long. And because these are our notes for our podcast, and then once we realize that like the doc performance is going so slow, we just fork it and make a new one. And we've done that like eight or nine times now.
Georgios:
[1:12:59] Oh, that's why your doc got stuck earlier.
David:
[1:13:00] Yeah, exactly. That's exactly right. So like say we take all the transcripts from the Bankless podcast, we take all those notes, we take everything we've ever produced, we clean it up I don't know what that would look like but we just clean it up and make it just more accessible and we put it behind a paid API and that's our value that we've had the corpus of bankless knowledge which is like you know going on for six years now so a lot of value in there about anything we've ever wanted to talk about ever, How do I tell the agents about it and that it's useful to them and that they should pay me for it? How does discovery work?
Georgios:
[1:13:32] Well, by the way, if you're down to do this, we'll happily call this on install
Georgios:
[1:13:36] everything we need to make it work for you. So we can talk about that offline.
Brendan:
[1:13:41] Yes. Yeah. But if you wanted to make it, we support Way to Discover paid services via MVP. Today, you would just say you register. You could spin up your API and just say, hey, here are the endpoints you can call. This is the cost. These are the ways which you can be like, here's the currency I want to be paid in. And you can do that not only on tempo, but across a variety of payment methods. So it could be like, if you want to put in your card details, put in a bunch of other stuff, you can do that there too. And then we...
Ryan:
[1:14:09] By registering the schema,
Brendan:
[1:14:10] You can make it discoverable to agents. And we see a number of services which are just now even ingesting these schemas to build like, they're calling it like, okay, what is like the page rank for productive work? And to figure out like, okay, how do we build that?
Georgios:
[1:14:24] I think that's also part of like the question, like what's a search engine for agents?
Ryan:
[1:14:29] Yeah, what's a search engine? Also, how do you price this content? It feels like there's a lot of, I mean, like do you
David:
[1:14:35] Want to ingest like thousands and thousands of lines or do you have a small query? Like how do we rank the size of your query?
Ryan:
[1:14:42] I have no idea how much the bankless like knowledge base is actually worth, right? And who's going to pay for that? It's almost like, yeah, not very much probably. It's like, almost like there would need to be a bidding process to almost discover what the actual market price is here, right? I'm sure there's a lot to be built in that area.
Georgios:
[1:15:00] So we had them, so early on, so you know how like OpenAI has like normal pricing and search pricing for the API? You can do all sorts of things where you can just say instead of me figuring out the price what if I just say I'm giving 10 APIs you said Dan Robinson because this is exactly the Dan Robinson problem I got 10 queries I'm selling 10 queries this is like my data go run an auction, figure out the price I can figure out the price that would be valid we haven't done that out of the box yet I think it's a good point You know, the question of the consumer or of the person wanting to pay the API is two questions. It's like, A, how do I price my stuff? How does my stuff get discovered? I think then how do I price my stuff? I think the answer is either you know what the price of the data is and maybe you charge some premium on it for convenience or whatever else. And to the discovery point, I think right now what you can do is like publish your schema. But where do you publish it and what does the search engine look like? I think that's open ground right now. And I think people are going to race for it if that is the next page round. I think that's going to be beautiful. I think there's going to be so many competitors for this brand. And hopefully we can just sit below all of that and let that competition happen.
Ryan:
[1:16:22] So much to build here, guys. Very exciting. The Agenic web, I think,
Ryan:
[1:16:25] is going to be a pretty big deal. And to have, you know, kind of blockchain and crypto be the payments layer for that. It just seems like it's got to be the future. Congrats on Mainnet, Giorgios, Brendan. Thank you so much for joining us today.
Georgios:
[1:16:38] Thank you. Thank you, guys.
Ryan:
[1:16:39] Bankless Nation, got to let you know, of course, crypto is risky. You could lose what you put in, but we are headed west. This is the frontier. It's not for everyone, but we're glad you're with us on the bankless journey. Thanks a lot.