E80｜Let’s talk about Stripe, smart commerce and the next revolution in machine payment
Web3 101 · 2026-06-01 · 61 min
Show full episode description
这是一场正在我们眼皮底下发生、尚未被商业社会完全察觉的底层经济重构。 我们从最近热度爆表的Stripe Sessions大会切入，深入探讨了由AI和稳定币共同催生的AI支付或智能体支付（Agentic Payment）革命——当AI不再仅仅是聊天机器人，而是变成可以自主决策、自主消费的经济体时，商业模式、SaaS订阅制以及底层支付逻辑将发生怎样的颠覆？Stripe又是如何凭借其开发者优先的基因，试图成为下一个时代的支付基座？ 【主播】 刘锋，BODL Ventures合伙人，前链闻总编辑 熊浩珺Jack，律动BlockBeats副主编，《Web3无名说》主播 【你将听到】
📑 Chapters — tap a time to jump there
00:44
引子：为什么我们要聊聊Stripe？
02:55
Stripe Sessions的震撼：AI与Web3的超级交汇点，支付基础设施在AI时代的变革前夜
04:03
智能体商业爆发已经出现「奇点时刻」
10:49
Agentic Commerce（AI辅助人类决策和购买）vs. Agentic Payment（智能体自主决策消费完成自动化业务）
15:22
Stripe联合创始人John Collison详谈智能体支付可见的落地用例，以及微支付为什么这次真的成立
20:05
如何让智能体支付成为可能：X402协议与MPP机器支付协议的作用与差异
31:54
上手实践：大胆动手尝试，让自己的Agent试一下自主支付
34:46
「无头商户」兴起：AI时代的新型商家
38:02
从用户体验到「智能体体验」：Agent Experience成为新竞争点
43:51
Stripe的历史基因：为什么它努力推动这场变革
🌐 This transcript was automatically translated to English from the original.
Hi, everyone, welcome to Web3 101. My name is Liu Feng. Hello everyone, my name is Jack. Today is the morning of Tuesday, May 26th. In this podcast, Jack and I are going to talk to you about a new payment revolution for Agent intelligence that Stripe is promoting. The reason for talking about this topic is because I went to the United States to participate in Stripe Sessions just a month ago, which is Stripe’s largest conference of the year. A crowded conference, with a record number of attendees this year. A lot of information was obtained that is worthy of further discussion. I remember last year, Jack, you were going to create a new public account, focusing on the direction of financial technology. We discussed at that time which companies in this field were worth paying attention to. I was thinking that we started 20 years ago, and financial technology was a hot word. It started with Alipay, and later everyone talked about companies like Landing Club, companies marked by this kind of P2P lending. The rise at that time made the word particularly attractive. Later, a lot of similar P2P companies emerged in China and then all declined. But when I communicated with Jack at that time, I suggested that there are two companies that deserve special attention. One is Stripe, which is a payment company that has grown very rapidly from scratch in the past ten years, and it is very brave. In the past two years, it has been actively exploring business in the fields of AI and Crypto Web3, and has been leading the wave of financial technology. Another one I recommend paying attention to is Robinhood. It is a new type of brokerage. It is not a migration for this kind of thing, but it is a brokerage that today's young people like very much. In addition, it actually caters to the kind of trading that is serious when the trend of financial nihilism is rising today. Robinhood is a particularly typical new financial technology company. So in today's podcast, I will start with Stripe, because Stripe is a company worth talking about a lot. We want to start with some of the things it is promoting recently about Agent. Or the direction of AI payment as an entry point. No problem. In fact, at the beginning, we paid attention to FinTech not because of AI, but because we have to say that since the Trump administration, the integration of the encryption field and the traditional financial field has indeed been accelerating. Companies such as Robinhood or Coinbase have begun to promote the tokenization of securities. Many financial assets have been put on the chain. We thought this was indeed a very interesting and a direction of attention. However, we did not expect that with the development of AI, AI payment would also become a topic worthy of special attention. I was also surprised by the Stripe Session conference that Teacher Liu mentioned at the beginning, because basically no one around me had paid attention to this conference before. I had never seen the relevant public accounts of this conference or the sharing in some Moments. As a result, this year’s Stripe Session should surprise me. The Moments actually flooded the screen. This conference felt like an inflection point. After the development of agents, such as these AI payment solutions represented by Stripe, more people began to pay attention. I can briefly talk about some very shocking points I made on Stripe Sessions in San Francisco. The first one is that Stripe Sessions has become the most closely integrated intersection between AI and Crypto today. Because all AI companies are more or less intersecting with Stripe, and they and Stripe are vigorously promoting payment for AI agents. Everyone is starting to move in this direction quickly. Another example is Crypto, which we can talk about later. Why if we want to really promote payment for agents, Crypto will play a major role in it. Another thing that impressed me deeply is that Jack also mentioned a word called inflection point. From the very beginning of this conference, Patrick Collison, the founder of Stripe, has been mentioning a word called Singularity. He repeatedly said this word to say that the starting point is not what we are talking about, the so-called awakening moment of AGI. It is more like an exciting moment, whether it is AI-driven technology or breakthroughs at the bottom of the economy, that can be truly perceived. So you will find that they search and show pictures and slides on the stage. Because Stripe is a company that does underlying payments, they can see a lot of economic activity data. They have seen that from the end of 2025 to 2026 to the present, they can see that all kinds of things have appeared. The curve of the inflection point of vertical geometric growth includes the number of new merchants created, and the number of new merchants created is mostly new technology companies related to AI. And these things are not just quantity, but quantity is more than quantity. The number of such innovative companies has exploded, and the operating results of these companies are also quite exciting. For example, from a start-up company, from an idea company to having an impact, the time span of meaningful revenue has been ruthlessly compressed. It turns out that for example, in 2025, these new services will be added to Stripe to collect payments or start-up companies. Their new companies will be able to reach an annual economic revenue of US$10 million in three months. The number of such companies has doubled compared to last year. For example, new companies registered using Stripe. 20% of them were able to receive their first payment from customers within one month of establishment. In comparison, this proportion was only 8% in 2020. This means that many startups now can write the code today and collect payments globally tomorrow. The day after tomorrow, they may be valued in the tens of millions because I already have income. This is something that was unimaginable in the past. But today, companies have been able to emerge and grow rapidly because of the engine of AI, so they use the word singularity. This starting point is used to describe the current changes in economic and business activities driven by AI technology. Patrick and Sam Altman both used this word in the exchange that day. They said that almost every week the world feels different from last week. I think maybe in the AI circle, friends, this perception should be more obvious, because both web coding and its models can improve everyone's productivity and significantly reduce the construction costs of many software and some services. This change has brought a new impetus behind it. It is possible that the development of those enterprises that we have seen before will be rewritten, and a large number of applications may predictably emerge. At this time, the consumption habits of users are also changing. The so-called AI payment has become an important topic. In addition, the third feeling is that as AI payment, a large number of products are also emerging. For example, Stripe announced the release of 288 products at this conference. It is a bit outrageous. In fact, you can’t remember these products at all, but many products are already being prepared. We are waiting for more commercial entities that we are familiar with to use AI to promote the transformation of their own businesses. At the same time, we are also looking forward to the emergence of more new AI-driven applications and new service models. On the other hand, we can see the changes in the entire business activities driven by AI and the emergence of new categories of startups. If we are in the payment direction, we can’t wait to be ready to welcome this change. Has Mr. Liu attended a Stripe conference before? This is very interesting. In fact, I went to Singapore last year to attend one of their regional conferences called Stripe2. At that time, a particularly typical feeling I had was that Stripe was working hard to prove to everyone. In the AI era, they are an important part of promoting the development of AI companies, because most AI companies use us to collect payments, especially subscription-based services. Stripe is best at doing this. Its rise is because of its API collection method, which allows many software service companies, especially SaaS companies, to easily deploy their own collections and then push services to users. So he worked hard to prove that we provide this very good payment channel for AI companies. For example, we were an important guest at last year's Singapore conference. He is a founder of Minders. The message he conveys to everyone is that Minders has grown from scratch to a mature and profitable company in such a short time, and we are an important promoter and supporter behind it. But at this year's Stripe Sessions, a global conference, you will find that what he showed you is not only that I am not just behind an AI company, they all use our method to collect payment. Users who need to subscribe to these AI services must pay through Stripe. More importantly, he shows everyone that Sorry, there may be new business models in the future. They are all changing. We are ready to promote this trend. So this is actually only half a year. It is a very obvious feeling. Indeed, I briefly learned about it. After this year's Stripe Session, I found that this seems to be really important because Stripe is actually one of the most valuable private companies in the world. Its valuation is actually very high, at 150 billion US dollars. It started around 2010 and then mobile Internet or various SaaS services. After the rise, it is actually difficult for everyone to solve the payment problem because SaaS may have a subscription system and it is actually very troublesome to deal with banks. But Stripe helped everyone solve this problem through API, which promoted the rise of this kind of SaaS service and also improved its own ecology, forming an ecological flywheel. This year's conference seems to want to emphasize this point. In the past, it may have been a SaaS dividend, but now the AI era may have ushered in. In fact, it is a new business model of AI, which is the kind of intelligent business that everyone is slowly starting to discuss. This Agentic Commerce, but AI agent business, actually also has payment problems that need to be solved. This is why Stripe also hopes to provide better solutions for these AI agents and then become a base for the future agent dividend era. Maybe it hopes to move to the next level through this new strategic change. Indeed, I feel that this year is a very important change, especially in the field of payment. Jack, you just mentioned that everyone is talking about Agentic Commerce and Agentic Payment. In fact, these are some brand-new words that are quite interesting and worth pondering, but they represent different meanings. For example, I am an ordinary user. Whether I use bean bags or Chats GPT, I want to order a cup of Luckin coffee. It may be based on the payment information you have or your delivery address that you have told it. AI will help you complete it. In fact, the process done by AI is actually like a process of filling in various forms for you, because these are some existing business forms. The merchant needs to have your payment information and then you finally give a confirmation. This kind of thing is more called A Gent Commerce is where people use agents to purchase existing goods and services through these AI applications. But there is a core problem here. As a consumer, we can very clearly give a command to the AI application. But if there is a more difficult situation, I want to buy a cup of coffee, but I don’t know whose Agent I want to drink. Please help me think about it and make some recommendations for me. At this time, you will find that he has entered another process. He has to help you discover goods or services, right? But maybe you are someone who wants cost-effectiveness, so how to judge the cost-effectiveness is a very difficult problem for Agent, so you will find that in this type of payment, in fact, it is the Agent who recommends the solution, and the consumer ultimately makes the decision. It is more about how people combine with these traditional services and goods, so this field is more called Agentic Commerce, which is intelligent commerce. People are the core element in the entire scene. But what we really want to talk about is Agentic Payment, which is to complete this payment. In fact, there is a bigger scene. That is, the Agent purchases directly with the Agent. At this time, people should not need to make decisions. For example, when we are using Crayfish or Hermas Agent, we actually need to adjust some API services. In fact, the service providers of these APIs can be found by the Agent itself. In most cases, in fact, people do not need to make this decision. The Agent will do it by itself. Our user needs to give him a clear goal. He may directly call the API once and pay the API provider to help me complete the task. In the process, In fact, a typical Agent to Agent is payment between machines. In this scenario, human decision-making is actually not that important, and this is combined. Now we see a major trend behind the scenes, that is, traditional SaaS services may be changing. They should not be subscription-based, but should be pay-as-you-go. This actually also means another important issue. Where will these software service providers go in the future? Agents directly calling these services may represent an important trend in the transformation of these software service providers. I think what Mr. Liu just mentioned about subscription is the charging model of traditional SaaS. It is indeed a particularly interesting topic because I think as a human being, we can feel this subscription system many times. In fact, it is still a little bit inappropriate in many scenarios. For example, some people just want to read an article in the Wall Street Journal, but if you don’t have his membership, you can’t read any of them. You have to buy them all. You buy it for a month, or you want to convert a file, some website that specializes in converting files. It also requires you to subscribe for a month or a year before you can use it. But it is not called an Agent. Even if people may have it, they are very temporary needs. But in the age of Agent, this is more prominent. It is a single use scenario or some service calls. Its volume can be measured. The frequency of single calls is particularly high, and it requires a quick response or service. Another point is that this kind of call can actually be verified quickly. What is this service? Can it meet my needs? Is it good? You will find that this kind of thing is the so-called micropayment that we have been talking about for many years. This is actually the most suitable agent. It can make decisions and spend money to do things. Micropayment may become a very important new business operation model in the AI era. We should expect that the business models of many enterprises and services may also change. Yes, this is what we are discussing today. In this topic, it is a very important issue. To put it more simply, What are the consumption habits of Agent? And what kind of problems will this consumption habit cause in the field of payment? In fact, these stripes have considered it quite clearly and have given a corresponding solution. How can I ask the editor to insert a paragraph at the end? John Collison and I are one of the co-founders of Stripe. This is what many people will ask. Everyone is saying that Agent payment should be important. So what kind of effort is worthy of attention and where are the real users? Which things can be implemented first? When I was communicating with him, I asked this question. For example, you have released so many products. It feels like smart payment has become a stormy wave. So what is the first wave? What will be the first wave of real implementation? At this conference, John is actually a co-founder of stripe. He gave a particularly interesting demonstration. He gave an instruction directly in the cloud code. He said that I want to make a research report on the energy direction. He gave a description of the data he needed, and then the agent went to prepare the report for him. In the middle, he needed to call a few very professional data. Ordinary web search did not work anymore. At this time, his agent went to his authorized wallet and directly spent a few dollars to call the data. After completing his report, he got the report and John said, I now hope you will build a website for me and sell this report. The agent will directly call the agent's own wallet. I went to Versale and purchased a hosting service. I published this website directly. I also published this website, which is not only read by humans. There are also pages for the agent to read. He also priced the report. Other agents or users can buy this report by calling their own AI tools. This is a very interesting, closed loop. How do I consume service data through the agent? At the same time, I provide a service and can buy and sell it. All these things are completed on this Cloud Code. I want to divide it into consumers and developers. We have seen and think that during 2026 Consumer purchases are going to be huge. Consumer purchases directly within AI apps are starting to become quite significant because that's already where people are discovering products and what's happening now is that people are using AI apps as convenience tools. You just want it to do it for me. And purchasing is a great example of how people want that convenience because people have already found the product here and people are already using AI apps as convenience tools. They just want to say you know, do this for me. Purchase is an example of people wanting to enjoy this convenience. For developers, we think developer services Like Vertel or Cloudflare. And for developers, you think about Vertel or Cloudflare or Amazon Cloud Services. Any service like this that's marketed to developers. Right now they're completely marketed to humans and businesses. We think they're going to be selling a lot directly to agents, and these agents may. Again, if I want to launch a website, I need a hosting provider. My agent can make decisions based on the information it knows. I never have to actually go to Vertel Amazon Cloud Services. And choosing between all the different hosting options that I have, my agent can just make that choice, and I as a human don't even have to be involved in the process at all. You know, people have been talking about V Father for a long, long time, always saying they're coming soon, but we really think that's different this time because, again, the core thing that's holding V Father back from happening is consumer fatigue. People don't want to make so many frivolous buying decisions. Why is Spotify, as a $9.99 subscription service, a much more pleasant experience in comparison? If you had to make a decision every time you hit play and think, well, is this play worth my 15 cents, it's not worth it. It's never been possible before, and with agents, that really changes. So we think that a lot of services have a real cost. If you want to serve a web page, for example, there's a tiny little cost. You can now choose whether you pay for a package or someone else will maintain an account for you, but it might make more sense to pay for what you use. It's kind of like in the old days. A grocery store in a small town 100 years ago. Customers go to the store, they send you a bill, and then you pay the bill at the end of the month. It might make more sense to pay only for what you take out every time you go to the grocery store. It would be simpler to pay on the spot every time. Many of the services we use today are the same way through micropayments. It is actually paying per usage and paying for each API request in real time. It may be simpler. This is the kind of usage we want to support. Above is John Collison, co-founder of Stripe, answering me. When asked about the question of what the first wave of visible force in AI payments will be, John gave his opinion during the demonstration at the Stripe Sessions conference. During the demonstration, the stablecoins on Temple were used. Because for this scenario, it is small and fast, and it is a machine-to-machine payment. In fact, stablecoins are the best solution. One of the products that I am more interested in at this conference is that they have released a CLI for the Link wallet, which is a command line function. In fact, Stripe itself is a B2B company. It is not a company like Alipay that we are familiar with. It does not have such a strong connection with consumers. It is more connected with developers because it provides a payment API. But it has something called Link wallet. It is actually a bit like a simple checkout service. It records all the payment methods I have used as a consumer, credit cards or crypto wallets. Some service providers on the website provide goods when they provide goods. If you choose Stripe, when you pay. You can hang up this Link wallet and just do a simple verification. All the payment methods you have filled in before can be selected and the payment process is completed. There is an interesting discovery behind this. Stripe suddenly discovered that the number of calls using the command line to call Stripe services has increased tenfold in the past period. Because many payments used to be completed on the web or application side, but now many payments are completed on the command line interface. This shows that a large number of agents have this demand. So they launched Link's CLI wallet. With this wallet, you can manage your other traditional bank card payment methods. You can also add a crypto account. In the case I just talked about, when the agent directly purchased the calls of these developer tools, he actually used Temple's stablecoin because stablecoins are very suitable for this direction, and the protocol he used is what crypto friends may be very familiar with before. It is called machine payment launched by Temple and Strap. protoco l With the machine payment protocol, you can use five senses to let the agent automatically complete the purchased service, and then automatically collect and charge for the services provided by others. The process of other agents paying you for calling your services is a typical closed loop of automatic machine-to-machine payment. However, in this process, they not only support the IMPP launched by Temple and Strap, but also support the previously popular x402. In fact, these crypto solutions will support future agent-to-agent payments. It may be the most important payment solution or an important force for stablecoins. Because it is relatively native, because in fact, strap solves agent to agent machine payment. Like x402, it is used in the http protocol of the older Internet. There is a protocol called 402 that has not been enabled before. Such a protocol that requires payment is payment required 401. 403 has been used by everyone. I forgot what 401 is. 403 is this webpage that you are not allowed to access. Then 404 means the webpage cannot be found, 402 means it has not been enabled, and can be used to call payment. The structure of x402 is that if the agent needs to access an API service and the HTTP protocol, it will return a link starting with xpayment and a payment requirement. The agent calling this API service can send a certain amount of stable currency to the service provider on any chain. Coinbase launched the x402 protocol. Its verification method is that it finds a facilitator, which means it finds a verifier. Verify that the money is transferred to OK on this chain. Then you can continue to provide this service or hand over this service to this agent, etc. Then the currency ring of a transaction is completed. But this model of stripe, although it still uses blockchain or stable currency, it still supports fiat currency services. In fact, x402 and MPP are machine payment protocols. The difference between the two is not that big, because the core is to implement payment through a reserved interface of http, and this thing is most suitable for micropayment, which is what Jack just said. It can also include traditional bank card payments originally owned by Stripe. In this case, MPP is a payment method that is compatible with more payment methods. In fact, in addition to the stablecoins and bank cards I just mentioned, MPP is also compatible with Bitcoin's Lightning Network payments. However, the characteristic of x402 is that it is more directly focused on the payment of stablecoins such as crypto. However, the best usage scenarios of these two solutions are agent to agent, which is the payment of small and high-frequency calls. Jack, you just talked about the x402 solution. It uses a facilitator to do this kind of verification. Stripe actually uses something called a shared payment token, which is also a shared payment token, or SPT for short. Although the word token is included in the term, token here refers to an authenticated credential that everyone usually talks about. It is not the token in crypto that we usually talk about, nor is it the etymology used by large models. If everyone usually uses this kind of token in payment or doing some password login. It is actually a token. The shared payment token SPT that Stripe talks about allows the agent to initiate a payment. This authorized token can be attached with some very important information, including the upper limit of how much money the agent is authorized to spend this time. The token is valid for more than 70 years and at which sellers the token can be spent. It can use this thing to complete a series of micropayments, so it becomes a voucher that can realize automated payment. We don’t have to make a payment every time to click on it and confirm it. In fact, just give it an authorization for a certain period of time. This is the most common core protocol for agent payment that I can see now. As for adoption, whether it is the MPP promoted by stripe and tempo, or the X402 protocol promoted by coinbase, now actually in terms of technical explanation, everyone has advantages. But the biggest problem is which merchants or service providers have actually accessed these services. I don’t know, Jack, have you actually made successful payments through agents? No, but maybe when open cloud was more popular. Maybe the crypto industry would like to try and let it help you participate in the pony market or open a contract on apple liquid, etc. But other than that, it is rarely used to call some API services or help with consumption. In fact, this is one of my feelings. I see a large number of such service providers now. They are all in payment solutions, whether they use X402 or MPP, including Virtals, which we are very familiar with. He recently launched a set of agent payment solutions. Each agent can have its own identity. then It has its own reputation system. It also promotes a series of purchased services. I took a look and I think the biggest problem is that their services are all for the applications of the coins issued in their system. The AI applications have their own coins. It is a very complex token economic system. These are actually quite practical, so my personal experience is that if you bravely give your agent instructions to help me set up this wallet and want to do some experiences, he can make some recommendations for you. But a core problem is that for most users, these current services may not be of great need for them, or I don’t want to pay for them. What I personally use most are some search services. I hope that when my AI agent does web search, he will not provide me with false information. At this time, it would actually be better for you to use paid services. Of course, there are many free web searches, and even web search APIs can be used. However, if you call some more professional services, the quality of the information you will get will be significantly different. There are also some professional databases that can be called out through these providers, but the cost per call is also very low, 0.01 US dollars or 0.03 US dollars. In fact, you can do a research work without spending a dollar. To me, I think this is still very useful, but I have communicated with many friends and I think everyone is still willing to use that kind of free API. So I think there may still be a process of changing usage habits. That is to say, the solution has been completed, but to what extent can it really develop? I think it is still very early. One thing I particularly want to point out here is that there are many service providers who are not ready or ready to meet these things. If you want to adjust their services, you have to go to their website to register a user name and get an API from them to hand over to your agent. If you really want to automate the process and serve the agent, you should not do this. The interface should be provided directly to the agent. The agent should discover it by itself, right? The agent should discover it by itself, instead of letting me, the agent. You can even ask your agent to estimate how much it will cost to do a task. This is a very good experience. Last week we posted an article with Temple. When Romeiro was interviewed, I asked this question. For example, in the competition between X402 and MPP, who will have the advantage in the future? He said that in fact, the agent does not care about who is good and who is not. The best is what the agent is willing to use. For many agent services, you have to think about it. You are using it for the agent and you need to improve the agent's experience. I think this is a very important one at the moment. For me, it is a take away. In fact, today, for a large number of service providers and a large number of people who want to develop this so-called AI economy, they need to change. On the one hand, these traditional merchants need to change. They need to intervene in the process to allow AI payments to be completed. But on the other hand, it is very important that what we have today, especially the future-oriented services provided to agents, should be AI first from the first day. That is, everything should be used by AI instead of designed for humans. Only in this way can this kind of automated experience be truly formed, and the efficiency can be improved, but it is indeed very early. So what Teacher Liu mentioned just now, such as the service scenario of web search, I am actually quite curious about it because I think individuals do need it, but I have not tried it specifically. Are they able to make such API calls now? I think everyone should be brave enough to try it. This thing has no borders, and if you really want to spend dollars to adjust me, it seems difficult, but it is very simple if you use civilization coins. Let me point you to a path. You ask your agent. I don’t know whether you are using cloud, codex or crayfish. I am using codex now. OK, no problem. You said it in codex. 10 US dollars and 20 US dollars. Then you tell him. Tell me, please help me check what services he has available now. I am very interested in web search to find the latest information and professional databases. I want to do research. I also need to help me better complete this interaction of the browser. Not every time you pretend to be a person for me. Dots and dots. This thing has a very high failure rate. Find me a similar service and he will recommend it to you. You can do some experiments. For example, you said What kind of research report do you want to write? What kind of topic? Ask him to do a web search and he may fall into it. Of course, many people use Brave’s free web search. I think Brave’s free search It is available under certain circumstances, but when I want to do some very accurate and demanding data services, I will be more inclined to use the paid services of parallel and exatom. In this case, each call may cost 0.01 or 03 US dollars. To do a complex research, it may be 20 US dollars or 30 US dollars. A particularly complex report is about 1 US dollars, which is quite worth it because the quality is very good. When Teacher Liu was talking just now, I also thought of a series of problems, such as Maybe some service providers still use people to register accounts and read their documents, maybe for security reasons, because I think many people's mentality is not able to accept it, let their agents decide what kind of services they want to use, and then report to themselves after completing the task and what I used. They may still be worried that if malicious websites or malicious suggestions affect the agent's decision-making, they may still suffer some financial losses or something like that. I think this is one aspect, but the more core reason is that it affects the agent's decision-making. This is a very systematic project. In the future, there should be issues of changing trust based on the credibility of various services and the reproductive system. Another point is that many things are verifiable, especially the call service. Whether it can help me complete the task is a very important basis. This leads to the fact that if you cannot complete the payment, the payment should be invalid in the end. I think the biggest problem at present is actually the user's usage habits. Another point is that the number of available services is too small. I recently read a particularly good narrative. It was written by a partner of H6Z, and he talked about it. He said that in fact, in the AI era, the most noteworthy thing is the rise of a called headless merchant. The so-called headless merchant means that we have always talked about using computers. If you don’t use a graphical interface, you are actually using the command line to interact with the server in a headless way. What he said is that in fact, in the AI era, a large number of services are no longer used by humans. They are for intelligent agents. These buyers of intelligent agents will not look at your website. He will only read your documents, look at your pricing, look at your running time and your probability of failure, and calculate the credibility of your service. The original merchant is to build a beautiful website, have a sales support system, CRM, and user relationship management. But the agents don’t catch it. The agents just want you to have very clear documents, predictable pricing, and reliable services. These are all things that headless merchants can do. They can beat that kind of website. Especially beautiful display merchants. What we saw before was the era of the Internet, which promoted something. It is a brick-and-mortar store that moved online to communicate with customers through websites. Later, it came to the mobile Internet era to attract traffic through social media and interact with users and consumers. But in the era of AI, if many services are provided to agents in the future, these will no longer matter. What is important to you? Can you write the documents well so that AI can understand what you can provide? Can you have good performance and reliability demonstrations so that AI can trust you to call your services? This is the new business in the future, and there is another change in this process. In many cases, in order to complete its tasks, AI may be called based on times. At this time, pay-as-you-go has become a growing business model. For software companies, their entire billing model is also changing. For example, they are acquiring some companies like Stripe, which is actually to complete and adapt their services to the emergence of these new needs. An interesting trend is that if we believe that agents will be able to make automated decisions and purchases in the future, we believe that agent payment can occur. The prerequisite is what can really help the agent to complete the task better. Payment is a necessary link that allows the agent to better serve ordinary users like us. Then this logic is established. You will also see more and more new business models, such as this kind of headless merchants. So I read the article written by H6Z's Panel Noi Lavigne before. I was very inspired. I think his conclusion is very meaningful. This can also be explained. Why is H6Z so active in investing in this field? They not only invested in stripe, but also in the public chain specifically for payment incubated by tempo stripe and paradam. They also recently invested in their own payment public chain ARC promoted by circle. You will find that these things are all for the future agent economy. It is possible that we talked about UI UX before, and now everyone is talking about agent experience, which is the concept of AX. It is not just the transparency of the agent when doing tasks, etc. He makes his own decisions. What are the influencing factors? They may be some of the key factors that determine AX. We have been talking about AI payment. AI payment is a middle method. In fact, it is more able to promote buyers and sellers to achieve better transactions. If this economy can really burst out, it is a bit like a flywheel effect. Which comes first, the egg or the chicken? Which merchants provide services first or users who have this demand? It is possible to have both. But if both are unclear, it is useless to have only one payment method today. Today we see the so-called starting point and the emergence of large-scale service providers. But these service providers can do better and allow users to really change their usage habits and use agents to call services to complete this kind of payment. This is the so-called AI business in the future. The moment of breakthrough I think the agent experience is very important for the user experience of the agent. In fact, there are many businesses that I think are very forward-looking. For example, stripe itself is doing very well. You will find that the stripe document not only has its own web page, but also provides several other entrances. Each page of the document provides a large model copy. This version is suitable for you to feed to your own agent. Then the agent lets the large model read it and it can quickly get what these documents are saying. It is not that the web pages are beautiful, but the content of these web pages must be very structured. To code quickly, it is customized for large models. Stripe also provides a markdown format for viewing. All documents provide an MD cost format. This thing can be thrown directly to the agent for quick reading. Even for each document, it also provides a skill for the agent, which is the installation skill. This thing can be quickly abstracted into a skill that the agent can use. A company like Stripe is already doing something, which is to make all of my products more friendly to the agent. This is actually a very typical way of agent experience, but behind it is something like the rise of companies like Stripe. It is different from the Paypal and WeChat Pay that we are very familiar with, and the rise of Alipay. Stripe was able to come out because it felt from the beginning that all payment methods were not so friendly to developers. It turns out that all companies selling this payment method let the CFO make decisions. It was founded in a year. Software companies are emerging. Software companies are all asked by developers to make decisions. How to make it developer-friendly? Developers can directly add my payment channel to the services they provide. So it has been an API service from the beginning. It has been about how to make it more comfortable for developers to connect my payment channel. All these stories, in fact, you will find that they are all of the same origin. Yes, it is true that Stripe should be regarded as one of the companies that is most prepared to provide technical solutions in this direction. It is quite extreme. In fact, I think all payment companies now have it. They are all telling the same story about solutions. I don’t think who has the advantage because everyone is experimenting in the early stages. Of course, many things can already be implemented. For example, the companies we are most familiar with have corresponding payment solutions. Even Visa Maskart has proposed its own very clear solutions for AI. The framework has been set up and who can really be used by everyone in the future. In fact, these payment solutions are synergistic with each other. But on the other hand, it is possible that in the end, which emerging forces will be able to stand out. Some emerging services and goods can emerge. This is the possible new software company we just talked about. Those new companies that use AI to launch some new efforts may be because AI has to consume itself and complete its own economic activities. How can this process be completed quickly and efficiently without friction? I think stablecoins may be a really important way here, because if we talk about agent-to-agent payment, it is very likely that in the situation we can see today, stablecoins are the most convenient as an intermediary. Because agents can quickly manage these chains through wallets, especially those optimized for these scenarios, their performance is also very good, and the payment cost is also very low. They are naturally suitable for letting bots like agents carry out this kind of economic activity. I think this is one of the development directions that I am more looking forward to. So I think everyone should stick to vibe coding because it is possible that a service or an API that you have now released by vibe may really become a service that many agents call in the future. One thing that always excites me is that Although we will talk about AI swallowing everything today, crypto is not that sexy, but when you see the technology really making breakthroughs and promoting some changes, some of the value of crypto has been seen here. When new business model services are expected to emerge, it is possible that crypto or web3 will become an important and indispensable module to support the rapid development of these economic activities. I think what I saw in the strap sessions is that a company is really forward-looking in putting AI and crypto together. Show the world some more infinite possibilities. I wonder, Jack, if you would like to briefly talk about the development process of Strap. This is in line with some of its development concepts. This is because we briefly mentioned it at the beginning. Strap started out because it promoted Sars services and took advantage of the dividends of Sars growth. This is a very long decade of Sars explosion period. But in fact, it provides these developers or merchants with the most needed payment solutions. It has been there from the very beginning. Although it did not position itself as a payment solution in the early days, the two brothers, Patrick and John, started their business because they were making inventory auction management tools on eBay. They themselves are developers. During the development process, they discovered that the previous Internet merchants were indeed too troublesome to solve payment problems, whether it was the subscription we mentioned before or selling their own products on eBay. So they actually have this gene of solving problems for developers from the very beginning. In 2 Stripe launched its first payment solution around 2010, and slowly packaged this payment solution into an API. In fact, I think the most interesting thing is that at the beginning, do you know what their company name was? It was called Slashdevslashpayment, which is backslash developer backslash payment. This is a very classic slogan of theirs. Their first product was like seven lines of JavaScript code. If you want your products and services to be quickly purchased by users and complete payment, just add my seven lines of code to your webpage. I sell seven lines of code. But through these seven lines of code, I can help engineers embed payment in the products they provide. There was a particularly famous story at that time about how he persuaded these developers to use it. In fact, a benchmark company with him was Awai, a Dutch company. The company was also a very important one in the past. It should be about the same size as Strape and should be a listed company. It followed the route from top to bottom. He promoted one-stop solutions. All of his origins were because the most important big customer at that time was Group On, which was Group B of group buying. He helped the company collect payments globally. Later, many large companies chose him, including some overseas companies in China that we are familiar with today. For example, Bubble Mart also chose the Dutch company Awai, because his model is that I provide one-stop solutions. I serve large companies. Whatever you want, I will give you something, and you will follow me. I will solve all your global payments for you through this thing. It provides users with payment services. But Stripe is the complete opposite. It is bottom-up. He asked developers to peach. The two brothers said, "Come and I will show you. Just give me your computer. I will not demo it for you. Add a few lines of code to your web page and I can handle the payment. So they completed one. It is not that the quasi-CFO buys the product, but it makes the developers feel very happy. So they developed little by little with the developers. It caught up with the rise of software services, so they have reached the point where they are today. His genes actually determine that his services are provided by APIs. In this case, in the AI era, this is actually an advantage, because you will find these AI companies. They also hope that all such services should be API-based, data-based, and modular. I can call them directly and use them. They later developed, and after a large number of acquisitions, they continued to make up for their shortcomings, such as tax services. How to provide registration services for new companies and artist companies. I cover it all for you. You pay me. I am expensive, but my service is good. You register me, and I do it all for you from registering the company to the subsequent collection and payment. In the process, they continue to become a company that developers particularly like. In the AI era, because of its company culture and its service model, payment is provided through API. In fact, AI companies like it very much. He has always shown that 70 to 80% of the most important AI companies in the world now use their services. I am an AI company, an indispensable payment infrastructure, and now you will typically see it including new acquisitions, bridges, privates, and new billing company acquisitions. It is very typical that he wants to promote AI-native companies, new agent-based service companies to rise, set off a wave from it, and then promote their own growth in this wave. I think this is an interesting development direction for him now, from supporting SaaS to supporting AI agents. This business model and stripe In fact, a very important link in his development process is the stage of building an ecosystem. In fact, our teacher Liu has also been mentioning the issue of whether this is an organic egg or an egg first and then an organic one. In fact, I feel that in the payment field, especially in the development process like stripe, this kind of problem will always be accompanied. It is the emergence of a new business model. Is it because of the promotion of a new payment method, or because the payment method is slowly developing due to the needs of the new business model? Stripe is a very important node. He is the one who does stripe connect. Before this, he has always faced individual developers. His recovery system, stripe billing, is his core module. As long as independent developers, you can use it. But this stripe connect seems to be starting to build this ecosystem. This ecosystem is really systematic. He helps these platform customers, such as shopify, various shopping platforms. They help these platforms, whether it is splitting accounts or the collection and payment capabilities of these platforms, a bit like what Teacher Liu mentioned just now. Some of the things his previous competitors did, but this seems to make it more scaled because when you enter, for example, merchants on the Shopify platform, they will automatically use the payment capabilities of Stripe at the bottom. From this time, its ecology began to slowly emerge. I think stripe still has some relatively similar strategic paths. Cooperating with open AI and Gimini, everyone will clearly see that the relationship between stripe and open AI is extremely good, because one of the core reasons is another co-founder of open AI's CTO. One of the co-founders together with Sam Altman, Greg, was once one of the earliest employees of stripe. He should be the fourth employee. Except for the two brothers, the others In fact, he is the second outsider. He was the first four people working together. He was the CTO Greg. Greg was probably still in the east at that time. The two brothers went to find him and started a business together. From then on, they came up with this model of payment through API calls. After Greg worked for a long time, he left and founded open AI with Sam Altman. So you can easily understand that open AI understands the stripe system, so it has used their payment method very early. One thing that stripe is very willing to do is that I see that this can be done. I quickly built some modules through acquisitions. I remember that when he acquired Bridge last year, the acquisition was almost 1.1 billion US dollars. At that time, everyone thought it was crazy. Bridge is a relatively new company. Why is it worth so much money? How can we integrate the stablecoin issuance and asset custody and help companies better issue their own stablecoins and integrate them into their own service solutions like skills? He also acquired private private, which actually provides developers with wallet machine services. It is not a traditional key management wallet. You can set up and manage the wallet using your email. It is very convenient for mid-range users to use. It is also good for developers. Just adjust the API and integrate this wallet into their own applications. For example, Hyperliquid, which we are familiar with, uses private solutions. In the past, there were many consumer applications that had nothing to do with crypt. They also directly called this private wallet to allow many assets to be on the chain or many NFTs to be issued on the chain. After acquiring Strava, it can easily become skills and integrate them into its own service module. You will find that the launch speed of its current wallet is also very fast. Therefore, his set of solutions can be completed quickly through the acquisition of these teams and then packaged. Becoming part of his system, and through this thing, many users of Angbao gradually built up the ecosystem. I think he is smart and quick in this regard. He is willing to spend money and knows how to spend it. Such a company, yes, it is worth looking forward to whether it can become a payment center in the new AI era. What excites me about the Strap Sessions conference is not only the launch of a large number of new products of Strap, but I have seen the explosion of AI, and a large number of decision-making solutions have emerged to promote AI agent payment. And it has reached the stage of production and can be implemented. In the next step, we can actually look forward to how traditional merchants can transform themselves. With the help of these existing payment methods and the improvement of large models, AI payment applications that allow people to make decisions can begin to continuously transform traditional merchants. And the more core thing is that a large number of new types of services are dedicated to agents. It is not just a simple consumption model. The agent outputs information to everyone, but allows the agent to truly become an economy that can be run automatically. In this process, payment gives it greater capabilities. Let's talk about the Internet economy and e-commerce. Maybe in the future we will actually talk about AI commerce and the rise of the AI economy. This is a basic construction and a big trend. It is a process that can continue to unfold in front of us. I think this is just the beginning. What I am very confident about is that I think crypto plays a particularly important role in this, because especially this kind of machine-to-machine agent-to-agent micro-payment and stablecoins are very important. In this process, I hope that it will be in the future. We can take this as an important direction to continue to discuss and introduce some more cases of implementation. Heart-shaped companies have emerged. This may be very sexy, and the audience will also be interested. Yes, no problem. In fact, when we first started talking about the possibility of stablecoins, I actually had another question, because for example, when I looked at Stripe’s solution, it supports both stablecoins and fiat currency settlement. But the core thing is that you must have a Stripe account. Even if you want to use stablecoins for settlement, you still need at least a wallet address established through stripe before you can use its tempo and its micropayment service. I think there may be strong supervision now, especially in the financial field, which has stronger payment supervision. If a merchant registers with stripe and can choose fiat currency, he may still try to choose fiat currency. Because if stablecoins are used for settlement, issues such as currency exchange will ultimately fall on the merchants themselves. In fact, this is a particularly interesting point. First of all, what I will tell you next is actually for merchants. As ordinary users, our real relationship with Stripe is when we pay, we pull up the payment method of Stripe and add our available payment methods. Whether it is a bank card or a crypto address, we send money to merchants through Stripe. In this process, we actually do not need to register. Stripe does not need to provide what we call a Stripe account, because consumers only need to pay through him. His wallet is actually the same as the Paypal wallet that we are familiar with. Also, Alipay’s wallet is completely different. It is to manage your payment method, but what you are actually talking about is that Stripe's account is for merchants, right? I mean, for example, those who have API services, I call them merchants or API service providers or something like that. In fact, this is a merchant. In fact, Stripe is a Tubi company. Suppose I develop an application and I collect money through Stripe. I am a merchant. I may have to register him to complete KYC. This is a very typical example. If I have a relationship with a traditional bank channel, I have to go through this process. Stripe helps you complete this thing and manage your account through stripe. But in fact, there is a particularly good release on stripe session called stripe treasury, which is an upgrade to this kind of stripe account. In this account, it integrates all supported payment methods from various regions and countries around the world. You can use this payment method to collect money. If the money is collected, there will be a corresponding account for each currency. At the same time, it also has an account called stable coin, which is a stable currency account. It is in the newly launched stripe treasury account system. As a merchant, whether you are a company or an individual developer, you can naturally conduct currency exchanges and transfer your stable currency account money directly to your legal currency account. You can think of it this way. It only means that stable currency accounts and legal currency accounts can move funds to each other because it is within a compliance framework. It has helped these users complete the management of deposits and withdrawals, and it can also provide this service for conversion between foreign exchanges. But at present, I see that they should only launch it in the United States. It seems that it has also launched it in the United Kingdom. The rest is due to regulatory reasons. It may be slowly launched in many areas. In fact, in its system, for the kind of day one global merchants who want to do international business on the first day, it has always been a very important solution. That is, I can let you complete what you need most, a traditional bank account and your crypto account. I can inherit it for you. Maybe its cost is not cheap. For example, the service it provides for opening a new company account and you can collect money through it after opening an account. Its cost has always been high, but its advantage is that it allows you to make money conveniently, so you don’t have to mind the cost. I saw some interviews. It seems that Zhou Zhou of SARS Company said that we used strap because we thought it was very convenient. It saved me a lot of time because I made money. I didn’t think at the time that I didn’t mind the cost. So this solution I saw at the strap session made me very excited. But the bad thing is that not everyone can use it. It is launched by region. By the way, you can even use stablecoins to make income-generating products. This is very native. That’s why I said that in the future we can see, the integration of traditional bank accounts and crypto is actually already happening. So now you will see that there is also an emerging model, which is the new Internet bank. It should be different from what we have seen traditionally. It has many combinations with crypto, which has become a very important trend. This also involves whether stablecoins serve as a marginal compensation market for the legal currency payment system, or can they really gradually replace some of the logic of traditional legal currency payments and eat up some of their markets. I have thought about this question. I communicated with many people and I got an answer that this question should be decided by the agent. After we gave the agent these capabilities, I opened up the payment track of my legal currency, the stablecoin payment track. I provide him with various payment protocols, whether it is Let him do it himself. That also includes whether you can let him call the fastest, the route he calls is the most convenient, the route is the best, and the whole process of settlement is the smoothest. The agent's feelings are my feelings, which should be different from your feelings, so leave it to the agent to make decisions. What I am more confident about is that I think stablecoins will definitely be more convenient for settlement than traditional banks. Of course, I can't speak for agents. I think we can look at this. But one of the information I got is that many friends I communicate with tell me that they are born to be agents. If it becomes an automated economy, stablecoins are a particularly good choice. I believe this. Of course, this needs to be proven. It is interesting and worthy of continued attention. This needs to be proven, because recently many startups are also doing this method. It is based on these existing ones, including stripe, coinbase, visa maskat, and Google. They have launched various protocols for agents, and then build on them to make these protocols better called by agents. The core purpose is how can I help the AI economy and the smart economy? Can a fully automated economy happen? Who can help? Okay, then whoever serves as the infrastructure can get the cake in the middle. The story goes like this. Okay, otherwise our program will be limited to this content. I think there are many things that we can continue to discuss later. No problem, let’s talk about these for today. Okay, thank you all, thank you, thank you, bye. That’s it for this episode. You are welcome to subscribe to the program Web3101 through Spotify, Apple Podcasts, Small Universe and other channels. If you are used to listening to podcasts on video platforms, you can also search for Web3 on YouTube. Follow us on 101 Podcast. We also hope that you can give us suggestions by email or by leaving comments. My name is Liu Fen. Thank you for listening.