Editor's Note
This podcast episode features Eddy Lazzarin, CTO of a16z Crypto; Noah Levine, an investment partner; and Sam Ragsdale, who left a16z to start Agent Cash. The three engage in a high-intensity discussion on topics ranging from the current state of AI agent technology and payment infrastructure to the survival of the credit card system.
The core judgment is that stablecoins' instant settlement and zero marginal fees are naturally suited to micro-transactions of 1-2 cents in the agent economy, while the transaction fee system of credit cards (2-3% marginal fee + 30 cent fixed fee) is vulnerable in this world.
Agent Commerce is dismantling the advertising business model that has existed for 20 years since the internet was founded. Eddy Lazzarin even bluntly stated: "The advertising economic contract is dead and will completely disappear within 10 years."
Essential Quotes
The essence of AI agents
• "LLM stands for chatbot, and Agent is a chatbot that can operate your computer. What humans can do with computers, agents can also do."
"Starting around November last year, AI models became smarter. They can complete complex tasks over a sufficiently long time span and use tools. We started calling them 'agents' because they don't just write code, they do the whole task for you."
"Internally, we call this 'real-time natural language programming.' Users describe their needs in natural language, and the agent writes a JavaScript program, possibly thousands of lines long, in the background to execute it. It only costs 20 cents to generate a token and 10 cents to make an API call. Then, the program is discarded after use. Four years ago, this would have required an expensive software engineer to spend a week to complete."
Front-endless merchants and business restructuring
• What does a Headless Merchant look like? It's geared towards AI services rather than people. There's no website frontend, only API endpoints and sufficiently good documentation so that the model can read, understand, and use them.
"The leading companies in the data industry charge 100 times the lowest price, using the same downstream data sources. Their core product is actually the enterprise sales team, not the data itself. In a world where agents make decisions, agents won't be fooled by attractive sales teams. They will try all the data sources, find the best one that works best and is most affordable, and then remember it."
• "You excitedly had the agent run around all night. When you woke up at 9 a.m., you found that it had been stuck since 2:30 a.m. because the next step required you to call the company's sales team."
The End of the Advertising Model
"The economic contract of the internet since 2000 has been to make money through distraction. Agents are not distracted. If it visits your website to look for recipes, it won't see the shoe ads next to it. The old model will die within 10 years."
"In 2016, the total value of internet advertising was $60 billion, and everyone thought it had peaked. Today, Google earns $300 billion a year from advertising alone. But after GPT-4 came out, traffic to tech news websites dropped by about 80%, and Stack Overflow did the same. These were early adopters who had decided to use proxies for information retrieval and code execution. Later adopters will follow because the experience is indeed better."
Stablecoins vs. Credit Cards
"The average transaction on Agent Cash is 1-2 cents. Credit card fees are a flat 30 cents. Transaction rates are completely absurd in this scenario. It's 2026; loyalty should belong to the merchant, not the card you use to pay."
"Credit cards did indeed appear earlier than the internet, and successfully survived the transition from the pre-internet era to the internet. Although they were through a lot of ups and downs, they did survive. So the conclusion is still pending."
"If anyone from a credit card company is listening, and you have a money transfer license, you could easily mint stablecoins for your customers and have them pay with them. I strongly suggest you consider this."
The Future of Consumer Experience
"If an agent is shopping on your behalf, you can equip them with a credit card optimization skill, allowing you to precisely see the ROI of each card. When you have zero loyalty to your credit cards, all the psychological lock-in effects disappear."
"One day you'll realize that you never really liked shopping."
Open Agent Business Stack Architecture
Host: Hello everyone, today I'm joined by Eddy Lazzarin, CTO of a16z Crypto, Noah Levine, an investment partner, and Sam Ragsdale, a former a16z Crypto colleague who now founded Merit Systems. He's working on an Agent Cash project, which we'll discuss in more detail later.
Before we get to that, I'd like to lay some groundwork. There's so much happening in the AI agent field right now; unless you're watching it 24/7, you simply can't keep up. So, what's the current state of the world? Sam, you're on the front lines of development, why don't you start by telling us?
Sam Ragsdale : I like to start with a taxonomy, a framework borrowed from Erik Reppel, the co-creator of the Coinbase x402 protocol.
This category divides agency commerce into two types. The first type is conversational commerce, which is checking out within ChatGPT. You tell ChatGPT, "I'm a man living in the West Village of New York, I'm going to the Equinox gym, and I want to buy a pair of shoes to fit into my social circle." It will empathetically recommend a pair of Nikes, and then you buy them.
The second type is to entrust your money to an agent and let them spend the money to complete the task on your behalf.
Conversational commerce is inevitable. ChatGPT, Gemini, Claude, and all subsequent cutting-edge models will feature checkout functionality. This is good for consumers, helping them find better products; good for merchants, leading to higher conversion rates; and good for platforms, which can take a 5% to 10% cut. It's essentially the next generation of Google Shopping.
Another aspect is that agents' capabilities are currently limited. Many people ask agents to do difficult tasks, such as "help me with sales outreach," and the agent will say, "I can't; I don't have access to that information." If an agent had a little savings, they could spend a few cents to buy services they couldn't otherwise use, making them much more powerful.
So now there are two parallel worlds: one is recommending products through a traditional LLM interface and taking care of the final step for you, with the platform taking a cut from the transaction; the other is that you independently deploy an agent to purchase goods and services on your behalf.
Noah Levine : I think there are two versions. One is the natural evolution of e-commerce, that is, the change of platforms. In the mobile era, commerce migrated to mobile devices, and new advertising formats and Google Shopping emerged. People always need to buy things, and consumer behavior changes. Now, people obtain information in an LLM (Local Management Model) manner, and commerce naturally follows suit and migrates to agencies.
There's also a less "physical" version: the internet itself is changing. The way people access information and perform actions is changing with LLM. The internet we've built over the past 20 years may not be the internet of the future.
The path of searching on Google and clicking into a webpage UI that's desperately trying to upsell may no longer be meaningful. Instead, a more agent-native internet has emerged, where agents pay directly for what they need, making them more efficient at serving humanity.
Host : This directly relates to one of your investment themes, Noah. But before we delve into that, I'd like to give our audience a more basic overview. We're all used to interacting with LLMs, but now we're hearing about things like OpenAI's Codex. These agents already possess a considerable degree of autonomy and can actually get things done. If you haven't been closely following this, you might not realize how far the technology has progressed. Eddy, could you elaborate?
Eddy Lazzarin : Let me quickly go over the past five months. Around November or December of last year, the AI models started getting smarter. Specifically, they were able to complete complex tasks over a sufficiently long time span, and they also used tools. We started calling them "agents," which is a human-like term because they didn't just write code, but also helped you complete a task.
But agents can't do everything. Software isn't just a small program running on your computer. The internet tells us that you need to connect to many other things to do something interesting; you need various networks and various participants.
Agents address the problem of intent construction, and to some extent, also solve the problem of modeling preferences. You tell it something, and it understands what you want to do, mapping it to tools, networks, and services. Through dialogue and memory, it can also roughly understand your preferences and convey this intent to tools, software, and vendors.
These two issues have been resolved, which is very exciting. Everyone wants to solve the remaining problems, but they are complex. At the very least, if you want an agent to conduct transactions on your behalf, you need to address authorization and delegation issues: How do you prove to the other party that this agent represents you? How do you handle identity verification?
Then there's the issue of payment and settlement. Once the connection is established, the agent reflects your intent, knows what to do, needs to make payments, needs to demonstrate payment capability, needs to handle split payments, refunds, and so on. I've skipped crucial steps like search and anti-fraud, but you can see that once intent building and preference modeling—things previously only humans could do—are automated, the entire business process can be automated. This is the engineer's reaction: Wow, these two things that previously required human input or at least verbal input can now be done automatically—it's incredible!
When people talk about "Agentic Commerce," they're talking about what needs to be resolved between "I talk to the agent" and "it gets what I need," and the chain reaction that occurs because many things are completely rewritten.
Host : Very helpful. In other words, we've evolved from an LLM that can interact using natural language to an enhanced version that connects various networks and real-world systems.
Eddy Lazzarin : It's not entirely a connectivity issue. You're making it sound like the change lies in what it's connected to. No. Your laptop was always connected to everything; nothing has changed in terms of connectivity. What has changed is that they can now use tools, think for long periods, and stubbornly keep hitting walls until the task is completed.
Sam Ragsdale : Let me simplify your simplified version further. LLM stands for chatbot, which excels at conversation. In the past, people thought they were best suited for customer service. Once they perfected conversation, we created tools to use them. To put it extremely simply, we taught them how to operate a computer. LLM is a chatbot, and an agent is a chatbot that can operate your computer for you.
Crucially, they achieved the average human performance level around GPT-4, at a cost approximately 1000 times lower, and their capabilities could be significantly expanded by paying more. So, roughly speaking, anything a human can do with a computer, an agent can also do.
Eddy Lazzarin : That's it. The premise is simple, but it triggers a multitude of changes—short-term, medium-term, and long-term. In the short term, everyone is working to streamline the process so that agents can actually get things done. In the long term, if your agent can access the app, how much UI and interface do you still need? Do you even need the Amazon App? Perhaps the Amazon App isn't as good as having your agent do all the research, read all the reviews, and only show you the images you care about—isn't that better?
Sam Ragsdale : Internally we call this "Just-in-time Natural Language Programming," although the name isn't very catchy. But it turns non-programmers into programmers. You type: "I want to buy something for my fiancée on Amazon. This is her preference, this is what I usually buy for her, this is what I bought her last time. Browse through about 1,000 options, pick the best match, place the order, find my home address, and ship it to her."
What actually happens is that the proxy internally writes a program to accomplish this complex task. It might be a JavaScript and Bash program with thousands of lines of code. It finishes executing, but the user doesn't see it and then discards it afterward.
Four years ago, this was unthinkable. Writing such a program would require an expensive software engineer spending a week debugging and obtaining API keys. Now, the execution cost is about 20 cents in tokens, maybe plus 10 cents for API calls. You buy the program and discard it; it's so cheap that there's no need to upload it to GitHub for storage. Anyone with absolutely no technical knowledge can do this. My parents are writing natural language programs now, without even realizing it. They could probably call themselves software engineers now.
Host : That's pretty crazy. Are you engaged? Was that example based on your own experience?
Sam Ragsdale : I'm engaged, thank you. But I didn't have the AI buy the ring. That ring predates the AI. Probably even the first computer.
The "Front-End Merchant" Theory
Host : Okay, let's talk about these chain reactions. Sam, you mentioned earlier how commerce will change in a world where agents handle a large number of transactions, which directly relates to a concept you proposed: "Headless Merchant." Can you tell us what a Headless Merchant is?
Sam Ragsdale : Okay. I think it's necessary to take a step back first. Besides traditional consumer scenarios like buying shoes with ChatGPT, there's a huge B2B market for developer tools. Platforms like Claude Code and OpenAI Codex are completely democratizing things; anyone with a computer and tokens can build things.
Previously, experienced developers would choose tools with clear opinions, perhaps going through processes with the enterprise sales team and signing subscriptions. Now it's different: new developers come in with only the intention of "what I want to do," without preconceived notions about which specific resources to use. And what they build is highly ad-hoc, requiring services that are billed entirely on a pay-as-you-go basis and don't require months of integration processes to get started.
So what does a frontendless merchant look like? It's geared towards AI services, not people. It doesn't require a physical or digital storefront for you to browse; all you need is an API endpoint and sufficiently good documentation so the model can understand and utilize it. Billing is also based on API calls, not a subscription or enterprise contract.
Eddy Lazzarin : I totally relate. I feel like I might have been an AI in my past life. As a software engineer, this is how I've always been: if I go to a website and can't see pricing or find an entry point to get an API Key with a credit card, I close the page. I don't want to talk to the sales team, I don't want to send emails.
Scheduling a meeting with a company sales representative is a huge commitment and a significant slowdown. I don't even know if this thing will work; I just want to try it now, immediately, because I'm working on something over the weekend and want to release it on Monday. Getting the key by swiping my credit card, getting reimbursed later, and planning ahead—that's the fastest way.
In the era of instant and ad-hoc software, do you really want your agents to wait? Your agent works all night, and you excitedly wake up at 9 a.m. to find that it has been stuck since 2:30 a.m. because the service you want to use requires you to call the enterprise sales team first.
Sam Ragsdale : Not to mention if the integration process includes a sales component, the API price would be about 10 times higher because they would have to assign people to manage customer relationships.
Eddy Lazzarin : Absolutely unacceptable. You want an agent to run autonomously, not because you don't care about what you're doing, but because you need speed, you need testing, you need rapid iteration to respond to user feedback, and you can't afford to wait.
If an AI model sees three options: one requires contacting a business sales representative, another requires setting up a dedicated credit card, but the third involves simply sending some stablecoins to receive $10 worth of tokens—a proof-of-concept offer—it will always choose the third. This single force is enough to trigger a restructuring of parts of the market.
Host : For traditional businesses, while these frictions make business difficult, they also rely on them to lock in customers and maintain loyalty. If these frictions disappear, how can revenue be reliably predicted?
Eddy Lazzarin : Here's my verbal response: Let's just ruin everything. Add friction to everything, make everything unusable. What are we doing?
I say this because friction can indeed be useful at times; for example, it can deter spammers and create a filtering effect. However, friction also has significant costs. As the economy accelerates, productivity increases, and the leverage of every minute amplifies, the opportunity cost of friction also rises. This is the trend in everything today.
Getting back to the main point, even in the lowest friction environment, where you get the API Key in a second, or even without an API Key and pay directly with an encrypted wallet key (the wallet address is your account), there will still be something else that makes the service sticky.
Reputation, memory, status, data, and even less tangible things like the agent's trust. If an agent knows you urgently need an answer and want to move quickly, it won't take a step back and spend 20 minutes exploring all the new options. It will remember what worked well last time and reuse it directly. Just like a smart person.
Sam Ragsdale : Let me give you a practical example. We communicate with a large number of merchants every day, and we have basically seen everything that can be sold through APIs. We have talked to many sellers about how they integrate "Agent-native Distribution", which is a native distribution method for AI agents.
Data products are typically high-volume commodities, with 5 to 50 sellers. Within this group, the top seller earns the most, charging approximately 100 times the price of the lowest sellers. Moreover, they often share the same downstream data source.
They achieve this through corporate sales teams. These teams are usually comprised of very respectable people who will fly to your office to demonstrate: "Look how impressive our data is! No data is better than ours! $35,000 a year!" You sign, and when the two-year contract expires, that person flies back and does the same performance again. And that's how tens of thousands of companies pay.
Smaller companies that might have better products or superior usability packaging for the same data ultimately went bankrupt because they couldn't secure distribution channels. There was no innovation in this field because the enterprise sales team itself was the core product; the data wasn't.
In a world where agents make choices, agents don't want to chat with company sales staff or be fooled by attractive sales teams.
It tries all data sources, finds the one that works best and is most cost-effective (especially for bulk pricing), and stores it in its memory: "Next time you need this type of data, use Minerva, not the other three." This creates a more efficient world. The tens of thousands of companies that were previously ripped off by $35,000 can now spend that money on other productive things.
Noah Levine : Another perspective is that if you believe AI will spawn a large number of one-person companies or very small teams that can use AI to create products that originally required 50 to 100 people, then it makes no sense for a corporate sales team to fly to someone's basement to do business with them.
On the one hand, existing merchants are worried about the impact on revenue forecasts, and it's true that there will be resistance when changes come. But on the other hand, this is also a completely new customer acquisition funnel, and if you can reduce the bottlenecks and friction in accessing tools, it will actually be a huge opportunity for them.
Sam Ragsdale : On our demand side, the vast majority of users have never used an API before, don't know what an API is, don't know what it represents, have never obtained an API Key, and have never signed an enterprise service agreement. But on their first use, they can combine six APIs from six different merchants, write a natural language program, complete the task, and then discard the program after use. This means that a completely new market has emerged for API consumers.
The existing business models of the Internet will be restructured.
Host : This sounds like Clayton Christensen's innovator's dilemma: the high-end market is where established players sell exorbitantly priced software to clients who can sign big checks, while the low-end market is where new users use agents for one-off experiments. But what can transform it from a low-end toy into something truly impactful?
Sam Ragsdale : Because it will ultimately become a better experience.
Noah Levine : I'd like to add that while it may seem experimental today, looking back at historical platform migrations reveals a similar pattern. Stripe started by serving very small, very long-tail merchants, many of whom later grew into giants, which is why Stripe continues to grow.
Shopify is similar. It started as a dropshipping and T-shirt seller, and now it serves a large number of brands that have grown from scratch into large companies on Shopify. Similarly, we will see a new group of lean developers using AI to build large companies. The tools they purchase today under the agency business model will become significant consumer products as their companies grow.
Sam Ragsdale : That's a good e-commerce perspective. But I want to say something bigger: the economic contract of the internet is dead.
Since Google launched in 2000 and became the biggest driver of the "free and open internet," the economic contract has been this: you are the publisher, you release good content, people search for it, and Google displays it.
A few years later, AdWords emerged, adding banner ads. The contract became: you provide good content, users land on your website, you can display small ads, and Google gives you a share based on the quality of views. You can publish anything people want to see, and Google handles advertiser relationships and gives you rebates.
In this process, Google became the biggest promoter of a free and open internet. They wanted the internet to be fast, cheap, and ubiquitous because the more you searched, the more money they made.
Ultimately, the business model of the internet is based on "distraction." When you, as a human user, are consuming content, whether it's searching for information, looking for recipes, or checking scores, your attention will be distracted. Perhaps later you'll buy those shoes or learn about a new B2B SaaS.
The growth of this model has exceeded everyone's expectations. I just looked at the 2016 Internet Trends Report, when total internet advertising was $60 billion, and people said, "That's the peak." But today, Google earns $300 billion a year just from advertising.
But with the emergence of proxies, people are shifting their search, information retrieval, and execution to proxies. It's still early days; ChatGPT has 100 million monthly active users, but they still use it like Google search, not really using it in a proxy-like way, such as "find a Father's Day gift for my dad and place an order."
But this is on its way. Look at the data in the tech world: since GPT-4, traffic to tech news websites has dropped by about 80%, and Stack Overflow is no exception. These are early adopters who have decided to use proxies for information retrieval and code execution. Others will follow suit because the experience is indeed better.
The old business model is being abandoned. Agents won't be distracted. If it comes to your website looking for recipes, it won't see your shoe ads. Publishers won't benefit from it. A new contract, a new reason for you to serve agent requests, will be needed, not advertising.
Will it be payment directly for articles? I'm not sure. Will it be payment directly for API resources? Will the internet be completely unrecognizable? I'm not sure either. But the old model is definitely going to die; it will disappear within 10 years.
Host : If the business model of the internet ultimately boils down to distraction, that's quite interesting, because Google was initially anti-portal. Yahoo and AOL gave you a bunch of links, trying to provide everything. Google just had a search box, a blank page, quickly giving you information. But the direction of evolution you described is precisely that it has become a distraction machine.
We now say that agents don't get distracted, but why would the evolution of agents differ from that of humans? Could there be mechanisms specifically designed to lure agents, causing them to get lost and stay longer?
Eddy Lazzarin: This is a big and interesting question. The core issue is: who does the proxy represent? I recently heard someone say, "I've started using Google Search again because the AI answers at the top are good enough." In that scenario, the "proxy" works for Google; it's in the Google search bar, running on Google's cloud, and Google controls it. Could that proxy be "distracted" by Google? I feel so.
The key lies in whose objective function it's optimizing, or to put it more simply: for whom does it work? The definition of "distraction" is: does what I show you serve your interests or mine? If it serves my interests rather than yours, then it's a distraction.
My understanding is not so pessimistic. The industry consensus that good advertising is good content has existed for many years, and good advertising is almost indistinguishable from the content you originally wanted to see.
But let me be clear: if an agent works for Google or anyone else, the entire business chain it follows will be defined by them, using the methods they set and the transaction infrastructure they believe is best for their business.
If the advertising agent works for you, in extreme cases running on your own laptop, and is open source, allowing you to fine-tune it and modify system prompts, you can even provide it with anti-distraction tools. This way, the advertiser faces a competitor that can expose their tactics. While I might be exaggerating a bit, this essentially creates a confrontation.
Sam Ragsdale : Yes, there are countless ways to reintroduce the ads. You can do it at the model weights level, which is the most aggressive approach. When choosing training data, select data that says "Nike is the best shoe in the world." Nike could pay, say, $1 billion a year, so that whenever shoes are mentioned, whether on ChatGPT or in a car insurance customer service API, everyone says Nike is the best.
This can be done at the tool invocation level, within the system context, or even as an overlay that doesn't involve chat. Basic model companies are clearly struggling with this issue. Recently, a dispute erupted between Anthropic and OpenAI, with Anthropic running an ad during the Super Bowl mocking ChatGPT's advertising, which OpenAI subsequently withdrew.
But I think OpenAI's response is completely reasonable: "ChatGPT has more free users in Texas alone than all of Anthropic's paid users." These are completely different scales of issues. They do need to provide expensive cutting-edge technology to a large number of users who are unwilling to pay with their credit cards, and advertising is actually a reasonable solution.
The reason advertising is such a brilliant business model in internet search is because consumers don't pay. High-friction relationships, like expending a credit card, exist between advertisers, Google, and publishers, and are irrelevant to the billions of monthly active users of search. Those people get value simply by opening Google.
If you try to align incentives, separate ads, and make them as relevant as possible, you'll actually get a better experience. Currently, foundational model companies are moving away from ads. ChatGPT doesn't run ads, and Gemini hasn't launched ads yet. Google is the most likely to do this; they've done it before and are the largest ad operator. Gemini will eventually have ads; it has a huge monthly active user base, and Google Shopping's equivalent will also be included.
But they know there's no monopoly yet; all companies are competing, and there's a lot of money being burned through private market subsidies. They don't want to be told, "This model isn't very empathetic to you, it doesn't really care about your goals because it runs ads." So at least for now, nobody's running ads; everyone's trying to remain neutral.
Noah Levine : I think there's another direction: as merchants improve and make their pricing and product data more transparent, you can redirect the money you used to spend on paid advertising to exclusive discounts for agent-driven shopping scenarios. If the agent is the buyer, you can directly convert the advertising budget into a discount budget.
Another branch is the discovery layer of agent commerce. What will it look like? Who will do the discovery? How will different merchants be differentiated? My prediction is that if advertising weakens because agents become buyers, since agents have unlimited attention and attention is no longer the most scarce resource, merchants may try to "implicitly advertise" by offering discounted products or adjusting descriptions to make them easier for agents to understand.
Eddy Lazzarin : There are too many dimensions. Advertising is essentially just one way to acquire conversions. If a system can achieve a higher conversion rate without advertising, it will do so. In reality, systems do have many other methods: referral networks, discounts, coupons, special channels, giving free tokens to startups, etc. There are hundreds of ways to acquire customers; advertising is just the most prominent one because it's the most direct for ordinary people.
Turn the personalization knob all the way to the end. If you want to reach me, talk to my agent first. My agent will tell you: Eddy hates advertising.
The Role of Stablecoins vs. Credit Cards in Proxy Payments
Host : Before we conclude, I must ask two questions. First: To what extent can traditional payment methods adapt to agency commerce? Or is a completely new, native payment method, such as stablecoins, needed, which seem to be finding a product-market fit?
Sam Ragsdale : My overall assessment is that credit cards work very well for "neo-physical" checkout scenarios like e-commerce or conversational commerce. Credit cards have built-in consumer protection; if your shoes don't arrive or get hit by a truck, Visa will adjudicate the issue, you get your money back, and all the risk lies with the merchant. This is a good business opportunity for new types of goods and services.
However, stablecoins are very useful in another scenario. The average transaction amount on Agent Cash is 1-2 cents. Approximately 600,000 such transactions have been completed. Credit card fees are a flat 30 cents. Wire transfers are close to $1. Marginal fees are 2-3%, most of which are transaction fees, but are used to earn cashback points. For e-commerce, you might like points, like accumulating credit card miles for a vacation in Miami, with the 3% coming from the merchant rate. But when your purchase costs only 1-2 cents, with only scattered API call fees, stablecoins have zero marginal fees and fixed fees of less than 1 cent.
Another key point is instant settlement. If you buy goods and services online, the settlement cycle is at the end of the month. Whether it's an invoice, wire transfer, or credit card, the merchant is essentially providing credit to the customer or agent. In the world of agents, you usually don't know who the agent is.
Specifically, anyone who has used Anthropic or ChatGPT API knows about the tiered system: you pay in installments of $50, then $100, and so on up to $2,500. This system exists to provide you with credit; they don't know you, haven't done KYB or credit checks, and don't know if you'll pay at the end of the month.
Just like AWS, the same applies to Nvidia GPUs. End-of-month settlements are terrible for these scenarios, leaving merchants bearing all the risk. If the customer isn't a real company that signed an enterprise service agreement, but an agent—someone you have no idea who they are—a billion agents can be generated overnight, but you can't grant credit to them.
Some people are working on credit agency schemes, but I think they're heading in the wrong direction. Instant settlement will solve the problem directly. Instant settlement is like cash. I have it, I hand it to you, and you have it. You provide goods and services, and I can't take the money back. For transactions like this, instant settlement is a better solution than the fixed fees charged by Asia Pacific's tiered system, especially for very small amounts and this type of transaction.
Noah Levine : One point worth refuting is that the minimum transaction fee and whether credit cards can participate in microtransactions are ultimately determined by the card network.
If they want to launch a new type of transaction, such as "micro-transactions," they can easily do so, without minimum fees and by lowering transaction rates.
The advantage is that there are far more consumers with credit cards than those familiar with stablecoins. Therefore, developers can continue to use card payments, while the backend settles transactions in stablecoins. However, this will take a long time. Until then, it makes sense to use native wallets to directly spend stablecoins on these protocols.
Sam Ragsdale : I think it's highly unlikely that a credit card company would disrupt its core business model that it has built up over 80 years. But I'd welcome it.
Eddy Lazzarin : I agree that credit cards don't present any major technical hurdles. The issues are more nuanced, involving business models and consumer perceptions of credit cards. I recently saw some concepts of "agent credit cards," which are essentially extensions of virtual cards. I really like my card issuer's virtual card feature, which allows me to generate temporary card numbers on the fly and easily cancel them in case of fraud or difficulty unsubscribing.
Sometimes, a new platform or method succeeds not because it's technically indispensable, but because it can be tailored to a new scenario. Credit cards are indeed older than the internet. They successfully survived the transition from non-internet to internet, albeit with considerable difficulties. Therefore, the conclusion is still pending.
Noah Levine : Additionally, the technology that makes Apple Pay possible will also enable agency commerce. Regarding whether this will disrupt Visa or Mastercard, my intuition is that many B2B transactions today are settled via wire transfer between developers and enterprise APIs. If card organizations can capture this volume and use it through micro-payments…
