Transcript 0:00 Welcome back to Crypto Weekly. It's been the summer, and this is the first episode since. I want to start in the room in September twenty twenty-three during Oslo Innovation Week. 0:13 I was on stage as moderator at an event we called Unleash the Future of Internet. Eighteen speakers, two hundred people in the room and online. 0:23 And the whole premise of that day was convergence, that AI, Web3, and AR/VR were coming together into something new. We were right that something was coming together, but look at what we actually did. 0:39 We listed technologies, three of them side by side, all equally weighted. We could say that they converged. We could not say how or where it would land. And then something else happened, which I should be honest about. 0:57 When the AI train really left the station, I didn't get on it. I looked at it, and I didn't see the connection to the blockchain that I have invested heavily in. I didn't see what I could bring, so I stayed where I was. 1:14 On the twenty first of July this year, Sandy Kaul at Franklin Templeton published a paper. The title was Agentic AI, The Killer Use Case for Blockchain and Crypto. 1:27 Not written by a crypto person making the case for crypto, not by an AI person making the case for AI. Blockchain has been looking for its reason to exist since two thousand and nine. 1:39 It has been tried on supply chains, on identity, on gaming, on art. Every time, the answer came back the same. Interesting technology, nobody really needed it. 1:53 Kaul says it has now found it, that the thing blockchain has been looking for all these years is agentic AI. And her case isn't a story about the future. 2:04 It's about what happens when software starts buying things, because that's what agents do. 2:11 They buy compute, they buy a data query, they buy an API call, a service, a second of someone else's machine, and they pay for each one separately as they go. 2:25 Now picture that agent walking into a bank and asking to move one hundred of a cent. It's not going to happen, not because the bank is old-fashioned, because the paperwork costs more than the payment. 2:39 That's the whole problem in one sentence. A card payment costs two to three percent plus a flat fee of around thirty cent. The agent is buying something worth a tenth of a cent. The fee is a hundred times the purchase. 2:55 And then there is the part I keep coming back to because it isn't a matter of degree. When you pay with a card, two things happen, and they happen at different times. The transaction is recorded straight away. 3:07 The terminal says approved. You get your receipt. You walk out with the goods. But the money hasn't moved. 3:15 The settlement, the part that the funds actually leave one bank and arrive at another, takes one to three business days. A blockchain records and settles in the same moment. There is no gap. 3:29 That's not a faster card payment. A card gives you a promise. A blockchain gives you the money. Now, you would think this story starts in two thousand and nine with Bitcoin. It doesn't. 3:43 Somebody made room for the internet of money in the early nineteen nineties. 3:48 When the architects of the web wrote the rules for how browsers and servers could talk to each other, they set aside a response code, number four O two. 3:59 They labeled it payment required, and then they left it empty for thirty years, more or less. 4:05 Not because nobody thought of money, because there was always a human at the other end with a card and a bank and an interface to type into. The web didn't need a native way for one program to pay another. 4:19 There were no programs doing the payment. Jeremy Allaire founded Circle in two thousand and thirteen on two convictions. 4:27 One was that money could move on open protocols, the way information already moves on the open internet. 4:34 The other was that blockchains are a kind of network computer where software could hold value and coordinate economic activity directly without the person in the middle of every step. 4:46 For that, I want to go back to a stage in Denver in the spring of twenty twenty-five. 4:51 Jamie Burke runs Outlier Ventures, an accelerator that has made close to four hundred investments on the idea that blockchain and AI will converge. He gave a keynote there. He called it the Post Web. 5:07 Most of it is a framework for founders, and that's a different conversation. But he said one thing that I have not been able to put down since. He said, "The web has three functions, read, write, own." 5:20 Anyone who has been talking about Web1, Web2, and Web3 for years recognizes these. But then he added a fourth one, delegate, and not between people, to agents, and then between agents. 5:35 And once you delegate, the interface stops mattering because the interface was never for the machine. 5:41 The pages, the buttons, the checkout flows, the fonts, the color, the animations, all of it exists because a human needed to be shown something and persuaded to click. An agent doesn't need any of it. It needs APIs. 5:57 It needs SDKs. It needs access to other agents. That's the post web. But here is the part that turns this around. 6:05 Burke said that the assumption behind Web3 was always that its primary user would be people, and that this turned out not to be true. Not many people really used Web3. There was never a natural way to do it. 6:22 And then he said this, "For a long time, we lectured people about why they should use Web3, and we were ignored, rightly so. Web3 wasn't waiting for better regulation or a bull market or a friendlier interface. 6:36 It was waiting for a use case and for a user who didn't need an interface at all." Now, everything I've said so far is an argument, a good one I think, but still an argument made by people who would like it to be true. 6:54 So let me tell you what happened while we were away this summer. On the 14th of July, the Linux Foundation made the X402 Foundation operational. Coinbase, who built the protocol, handed it over completely. 7:09 40 organizations have joined since April. Look at who is in that room. Visa, MasterCard, American Express, Stripe, Shopify, and Circle, Coinbase, Solana, Stellar, Ripple. 7:23 Three weeks later, on the 4th of August, Cloudflare announced wallets for AI agents. Cloudflare is not a bank and not a crypto company. 7:33 They deliver performance and security services to something like a fifth of all websites in the world. What they announced was this, an agent gets a stable identity 7:46 tied to the human or the company that owns it, an account wallet that can hold stable coins, and then a separate wallet for each individual agent with a spending limit, a list of approved merchants, and a maximum size for any single transaction it can make on its own. 8:04 Their CEO, Matthew Prince, put it simply, "The internet is shifting from human-driven browsing to agent-driven commerce, and the infrastructure has to keep up." 8:15 And the press release said something I think is the whole story in one line, "The internet as we know it was built for humans, not agents." So notice what this is and what it isn't. 8:29 This is not the financial industry moving toward the internet. This is the internet building money into itself. In 2024, people left the blockchain for AI. Capital left. Developers left. 8:45 Founders who had spent years building on chain packed up and went to work on AI because that was clear where the future was. The advice going around the industry was blunt, "If you're in crypto, pivot to AI." 9:00 Brian Armstrong stayed, and in July he wrote down why. "Crypto is infrastructure," he says, "the way electricity is infrastructure or the internet. It doesn't compete with the next big thing. It sits underneath it. 9:17 It's an and, not an or." Jeremy Allaire has another phrasing. In June, he published a thesis, not a paper, not a blog post, a whole body of work in his own name and not in Circle's. He called it The Agentic Economy. 9:33 The same man who founded Circle in 2013 on the idea that software could hold value and coordinate directly without a person in the middle of every step. 13 years later, he sat down and wrote out what that means today. 9:49 Two major companies building, they don't agree on what to call it yet. That will have to settle, but nobody is making the case anymore. They're shipping it. So here is where I've landed. 10:01 I'm going to say it plainly, the internet of money is here now. Not coming, not emerging, here. Still, two weeks ago, Coldcard, 1,800 Bitcoin gone from a hardware wallet. 10:16 Same week, Hugging Face, an open AI model, got out of its test environment and into Hugging Face's own system. 10:24 Two technologies, two set of holes, one blockchain, one AI platform, and in both cases, the systems around the technology failed before the technology did. 10:36 So even if the internet of money is here, it's still being built. The most important thing missing is perhaps compliance. Who answers for an agent that gets it wrong? Which rules apply, and to whom? 10:52 Independently of what's lacking, I think Sandy Cole is right. Agentic AI is the killer use case for blockchain. 11:00 I also think it reads just as well the other way around, that blockchain is the killer use case for agentic AI, but that one is for another episode. That's what I wanted to share with you this weekend. 11:14 If you found it useful, send it to someone who should be paying attention, pun intended. I'm Morten Bysted, and you've been listening to Copper Weekly.