Stripe, Visa, Mastercard, Microsoft, Meta. All Building The Same Thing.

Stripe announced a pile of new agent commerce products. And I know everybody is talking and buzzing about the agent buying the link and how you can like now tell your agent to go buy you coffee. I've seen it. I've done it. It's amazing. That's not the point. That is not the point. The point is that power in the entire internet economy is shifting from the seller to the buyer for the first time in decades. And if you are anywhere around any kind of economic activity on the internet, which is pretty much all of us, you got to know this because this is the biggest shift we've seen in commerce patterns in two decades. So, we're going to get into it. We're going to get into why Stripe did it, how it works, what the old economy looked like, what the drivers are, why agents are a big part of this, and where we're going next. I want to walk you through three shifts underneath Stripe's announcement. First, why the old funnel was really a machine for making human intent visible. Second, why payment authority starts traveling with the task instead of waiting inside checkout. And third, why the next competitive question for companies is not just whether they use AI. That's kind of done, but whether their business can be called by agents. Because it's not if agents arrive, it's when agents arrive. And when agents arrive, the buying journey is going to be both a customer-driven buying journey and an agentically driven buying journey. And increasingly that duo is going to mean the intent is formed by the time the transaction and the buying experience begin on a seller's website or any kind of property. So the most important part of the Stripe launch is not that an AI agent can buy coffee or book a restaurant or pay for a domain. That's really cool. I did it. It's fun. It's useful. It's very easy to demo and it is going to get the most attention because it looks like the future in a way that is really legible. You ask the agent to buy coffee. The agent goes and buys coffee and the agent pays and it asks for approval and it feels trustworthy and it's done. That's that's the obvious story. That's the headline. But the more interesting story is what becomes true before that starts to work at scale. I said before for a reason. But the more interesting story is what has to become true before that works at scale. For an AI agent to buy something on your behalf, it needs to know what you want. It needs to understand the seller. It needs to know the price, the constraints, the return policy, the timing, the risk, the payment method, and when to come back to you for approval. And the seller on the other side needs to trust that this is a real purchasing attempt and not a fraudbot, a stolen credential or maybe some kind of automated abuse pattern dressed up as an agent. This is a much much bigger problem than checkout. And this is why Stripe sessions announcements are worth reading as one architecture, not some kind of grabag of product news. Stripe announced hundreds of products literally and features. But the center of gravity was very clear. It is building economic infrastructure for the Agentic economy. Links wallet for agents, shared payment tokens, the machine payments protocol, the agentic commerce suite, Radar's token theft defenses, usagebased billing and streaming payments with metronome and tempo, stripe project, stripe signals, treasury, stable coins, issuing individually. Those sound like product announcements, and they are. Together, they point at a commercial stack for a world where the buyer's agent arrives with intent and context and permissions and sometimes payment authority before the seller ever gets a chance to convert anyone. And that is a massive mental model shift for all of us in commerce. The old internet asked, "How do we get the customer into our store?" The next internet asks, "How do we become usable by the customer's agent when the customer never comes to the store at all?" To understand why this matters, it's worth slowing down and looking at what the funnel used to be. We talk about funnels like they're just marketing diagrams. I started in marketing, right? Where there's awareness and consideration and conversion and retention and some version of that, right? But a funnel is not a diagram. It's an institutional arrangement for making human intent observable. Think about the website, the app, the checkout page, the pricing page, the product detail page, the demo request form, the onboarding flow. Those were not just interfaces. They were controlled environments where a business could watch demand take shape. The buyer searched, the buyer clicked, the buyer browsed, the buyer compared, the buyer hesitated, the buyer abandoned a cart, the buyer returned three days later, the buyer upgraded, the buyer churned. The seller could see enough of that behavior to optimize around it. and an entire industry got built around that. There were more than 8,000 companies in Martekch in the 2010s. This is why the internet economy built so much infrastructure around human attention. Search optimization, landing pages, performance marketing, attribution, personalization, life cycle, email, it goes on. Checkout test, conversion rate optimization, pricing tests, all of that made sense because the buyer was a person moving through a seller controlled environment. Stripe's original breakthrough did fit that world really, really well. Stripe did not just make it easier to accept payments online. The deeper thing Stripe did was to make money movement feel native to software. So if you were a developer, you could turn economic intent into code. You could start a SAS company, a marketplace, a paid newsletter, an indie tool, a course, a creator product, or a vertical software company without first becoming a payments institution. And I got to tell you, as someone who built a store before Stripe was a thing, it is inexpressably wonderful how nice it is that payment intents are now code. It used to be so gnarly, guys. It was terrible. And the fact that Stripe fixed it lowered the floor on what a minimum viable company could be. And that really mattered. But the original Stripe API still lived inside a seller hosted world. A buyer had already arrived from somewhere. They were on a pricing page inside an app in a checkout flow at the point of purchase. Stripe just helped the seller finish the transaction once the buyer's intent had become explicit. Agents change where that intent becomes explicit. The user doesn't have to begin with a search query or a product page anymore. The user can just begin with a task. Find coffee I would like. Book the restaurant if a table opens up. Get my mother flowers on her birthday. Provision the services this app needs to go live. Research this vendor with an eight buck budget. Reorder office supplies when prices drop. You get the idea, right? It's not just consumers. Keep our model spend under a threshold. That task may eventually touch a merchant, right? It might touch Stripe. It might create a payment, but the merchant is not the place the decision begins nor the middle of the process. And that's a massive shift. The commercial surface is migrating from the seller's environment to the buyer's agent. The buyers are taking back control. This is also why the simplest version of Agent Commerce is not actually the most revealing version. The obvious examples are commodities. Reorder paper towels, buy the cheapest acceptable HDMI cable, uh find a flight for under 400 bucks. Those will happen. They're useful. They're also way too easy as a mental model, and I think they're really overused for a gentic examples. The much more interesting use case is when the request is human and fuzzy. Let's say I ask my agent to buy authentic coffee. I like coffee. I'm passionate about coffee. And I say authentic. That phrase is a disaster in a normal checkout flow, let alone a normal website where you can search for stuff because you're going to get all kinds of branded disastrous crap. To a search engine, it's just a keyword problem, right? To an ad platform, it's a targeting problem. All they have to do is kind of match you up with what they imagine your intent to be and whatever keyword ranking some marketer person put up there to get the coffee to match. To a merchant, it's a copy problem. Same same issue, right? They've got to make it look like it's relevant. And to a marketplace, it's a ranking problem. If I go on Amazon, authentic coffee is a ranking problem. But to a good agent, that phrase can become a purchasing brief. If the agent knows my preferences, it can translate authentic as shorthand for origin, roast level, processing method, flavor profile, freshness, roaster reputation, whole bean versus ground brewing method, shipping time, price range, prior purchases, dislike styles, and also how long I want the coffee in the mail, and how long it's been since the last roast day. Yes, I'm a nerd. And the human language can stay very vague, and it still works. The agent can make commercial intent precise. It doesn't make it less personal. In some ways, it's a lot more personal, but it's a lot less sellerc controlled. Now, a human shopper lands on a website and lets the seller's interface shape the journey. That's what we call marketing. The agent arrives with a theory of the buyer in hand and strict guidance and guardrails. That's not the same thing. Agent doesn't need to be persuaded in the same way. It needs to know whether the product matches the buyer's intent. That sounds like it may be a small distinction on the surface, but it entirely changes what businesses have to expose. An agent needs structured enough information to reason. It needs confidence that the seller is legitimate. It needs the final price, the delivery window, the return policy, the payment options, the inventory, the constraints, the escalation path, and oh, by the way, it needs intent hooks. It needs ways to translate a wide- ranging intent pathway set into a coherent perspective on whether that particular coffee from Ethiopia that's naturally a process is going to work for me. And so if you're not surfacing an easily readable set of metadata for agents about your product, you're in trouble. Think agent first. And by the way, this is why I think the phrase agentic visibility is kind of incorrect. If agentic visibility just means SEO for agents, that is way too shallow. Search was about being discovered by humans and letting humans be shaped by the seller during the buying journey. Agentic commerce, in fact, that's that's all Google did, right? That's their whole business model. Agentic commerce is about being usable by software acting on behalf of humans. It's actually a much higher bar. A person will tolerate ambiguity. A person will infer from aesthetics. A person will click around. A person will read five reviews and decide something seems fine when maybe it's not or the other way around. And the person can also be emotionally influenced by ads. All an agent needs is for the business to be legible enough to operate against. That doesn't mean every company becomes a sterile API, right? It doesn't mean brand and taste disappear. We'll come back to that because I think that the brand point is actually one of the most misunderstood parts of this whole shift. But before we do that, this is the practical implication. Your website may become less central, but your commercial reality has to be way more explicit. Product cataloges, prices, policies, payment methods, upgrade paths, service levels, usage limits, identity requirements, and fulfillment constraints all need to become a part of the agent-facing surface of the business. Some of that will be exposed through protocols, some through platform integrations, some through feeds, some through APIs, and some through platforms like Stripe itself. But the direction is clear. Businesses will need to describe what they sell in ways agents can act upon. That is what Stripe's aentic commerce suite is really about. It is not just a way to sell inside AI apps. It is a way for merchants to project their sellable reality into the places where buyer intent is starting to form. The store is no longer the only store. And this is also where aentic discovery works differently from search discovery. Search discovery was mostly about ranking a page against a query. You wanted the customer to find you, click you, and enter your funnel, right? Agentic discovery is not that. At least not only that. In the Stripe framing, the more interesting move is that a business can broadcast inventory and commercial capability into the assistant surfaces where buyers are already asking for things. Not just a product title or a landing page, but everything, right? inventory price, payment readiness, fulfillment logic, uh merchandising controls. That means discovery becomes less like winning a blue link and more like becoming an available option inside the agent's decisioning process and looking relevant because you are a software that is usable. The merchant is no longer trying to appeal to a human, right? The merchant is trying to be understood by the agent at the moment the agent is mapping intent to action. This is where Google's worth naming directly, but not because Google proves that checkout inside a chatbot is the end state. By the way, Google matters because Google owns one of the old starting points of internet commerce. We've already referred to it. For years, Google search was where commercial intent became visible. The buyer typed the query. Google sent the traffic and the merchant tried to convert the visit. Now, Google is trying to make AI mode and Gemini commerce surfaces, right? Its universal commerce protocol is meant to work across discovery and buying and post-purchase support. The merchant center work it's doing is adding attributes that go beyond traditional keywords like answers to common product questions, compatible accessories and substitutes. It's a gesture at the agenda hooks I'm talking about. That is the right level of analysis. The question is not really where is the buy button. The question is now where does product understanding form? And this is why instant checkout is the wrong mental model. Wyard reported that Walmart's chat GPT instant checkout test converted three times worse than the products where Chad GPT sent the shopper back to Walmart's own website. Daniel Danker, who oversees product and design for Walmart, called the experience unsatisfying. I bet he called it other words in private, but we won't get into that. Look, I think the problem is structural here. I don't believe that people necessarily want to buy a single item at a time inside a chat window when they already have carts and bundles and loyalty programs and delivery expectations and substitutions and returns and merchant relationships held somewhere else. And OpenAI's own follow-up pointed in that direction. It said that the initial version of instant checkout did not offer the flexibility that it wanted to provide. So it would let merchants use their own checkout experiences while chat GPT focused more on product discovery. So look, the key here is not to immediately dismiss instant checkout. I don't think instant checkout is the way we describe the future of the agentic economy because the whole thesis of this piece, this video is that it's much wider than that. But I do think that the OpenAI team is going to keep experimenting and they're going to figure out a way to tie Checkout in as an adjunct, as an add-on to overall commercial experiences. So, I wouldn't count them out yet, but the product didn't land the first time, and it's worth thinking about why. And that takes us back to why agent discovery is central in the economy of the future. The agent commerce battleground is not simply can chat GPT process a payment. It is can the seller be discovered correctly, represented accurately, compared fairly, called safely, and handed off to the right transaction surface at the right moment. And this also shows why it's not an only Stripe story. Microsoft has pushed shopping inside Copilot. Meta is moving checkout closer to ads. Visa and Mastercard are building agent payment and token systems themselves. PayPal is building a commerce services around wallet trust and merchant protection and product discovery. Open AI and Stripe co-developed the Agenta commerce protocol, but I would treat that less as proof that instant checkout wins and more as early protocol experimentation around that bigger shift as I called out. So the whole market is running toward the same place. Commerce that begins inside the buyer's interface, not the seller store. The second big shift is payment authority. Stripe's most visibly futuristic announcement is links wallet for agents. The basic idea is very simple. A user can give an agent programmatic access to link. The agent can create a spend request. After the user approves, link returns either a one-time use card or a shared payment token. The agent never sees the user's raw payment credentials. Today, Stripe says each request requires the person's review before the credential is shared. They also say they plan to expand the controls so people can set spending limits and decide when agents can act without a fresh approval every time. That's a payment product, right? But strategically, it's a relocation of payment authority. In the old checkout model, payment authority is extracted inside the seller's flow. The customer arrives, the customer chooses the product, the customer enters or selects payment details. The customer passes through checkout. The seller's surface is where the buyer's intent and the buyer's payment instrument finally meet together. In the agent model, the buyer's agent may bring payment authority to the seller. Right? It's like carrying your wallet in and being like, I already know what I'm paying with, guys, and I already know what I'm buying. The payment method is attached to the task. It can be bounded by amount or currency or merchant or credential type or approval state. It might be scoped to a single transaction. It could expire. It could be visible to the user. All of these things are possibilities. But no matter how you shape it, it's not the same as it used to be. The seller does still need to accept the payment. The seller does still need to fulfill the order. And the seller still has to manage refunds and disputes and support, but the commercial decision probably happened somewhere else. The seller may not be receiving a browsing customer. the seller may be receiving an authorized purchasing attempt by a bot. And in fact, more and more I think that's going to be the case. This is why the credential details matter way, way more than they look. A one-time use card lets an agent buy from the web as it exists today. It is like an adapter for the existing commercial internet. A shared payment token points toward a much more machine native world where the seller can accept a scoped payment credential programmatically. Stripe is bridging both at once. That is usually how infrastructure transitions happen. The future doesn't arrive by replacing every old surface in a single clean move. It arrives through adapters. One-time use cards let agents operate against checkout pages built for people. Shared payment tokens and the machine payments protocol point toward a web where agents and services can coordinate payment directly. Most companies are going to live between those worlds for a while and that makes trust a whole lot more complicated. The buyer has to trust the agent. The seller has to trust the transaction. The agent platform has to trust the seller. The wallet has to protect the payment credentials. The payment network has to enforce all of the controls for all of this. And the user has to see enough context to approve the action without being a bottleneck for every task. That is a real product surface. It's not a button that says pay. It's a whole trust chain between buyer and agent and seller and wallet and network and task. Stripe is interesting here because it already sits in the middle of those relationships. It has merchants. It has link. It has fraud systems. It has issuing and billing and treasury and data. It has APIs. It has developer trust. The agent era rewards companies that already turned messy economic interactions into programmable primitives. Stripe did not suddenly become relevant to agents because it made an AI announcement. Don't think that agents made Stripe's older strategy more valuable. Now, there is a temptation to turn this into a very clean replacement story. Cards are old, stable coins are new. Agents are new. Therefore, agents will replace cards with stable coins and only agents will buy. I don't think that's the right frame. The more plausible version is coexistence because the agent economy has very very different jobs. It has to transact with the web as it exists and it has to create payment flows that the old web was never designed to support. Cards are really useful for the first job. A huge amount of the commercial internet does cards. If an agent needs to buy coffee from an ordinary merchant, book a reservation, or anything you're thinking of right now, I'm betting a scoped virtual card is a fantastic adapter. It lets the agent act without receiving the buyer's raw credentials, and it does not require every merchant on Earth to adopt a new protocol before agents are useful. Now, stable coins are more interesting for the second job. Agents are going to create payment patterns that humans rarely created because humans were too slow, too impatient, or too expensive to coordinate. machine-to-achine payments, streaming payments, tiny research budgets, per query data access, autonomous replenishment, API calls that settle as they happen, crossber usage that runs across time zones and services, workflows where money needs to move very continuously instead of in one monthly batch. Traditional guardrails were not built for that shape of internet or that shape of activity. They can be extended. They can be adapted. They can be optimized. But some machine-shaped transactions are simply a better fit for programmable money. And this is why I would read Stripe's card and stable coin work together. Cards give agents access to the human-built web. Stable coins give agents a better rail for machine native transactions. The bridge matters because the transition will be very uneven. The native rail matters because the new behavior will eventually exceed what the bridge was designed to carry. Streaming payments make this a really concrete story. If an AI product incurs cost at the moment tokens are burned and the user is consuming value all the time, waiting until the end of the month to settle creates real risk. The business already paid for compute. The customer might not pay. The fraudster might disappear. The margin might be gone. Stripe streaming payments announcement with metronome for precise usage tracking and tempo for stable coin micro payments is an answer to that timing mismatch. And it sits next to a broader usage-based billing push. dimensional pricing, hybrid pricing, commits, real-time metering, and billing systems that can actually describe AI consumption as it happens. Meter the value as it's created, right? Settle closer to when the cost is incurred. That's a very different world from a normal checkout page. And once you see that, the agentic economy just gets way wider. Some tasks are one-time purchases. Some are scheduled intent. You know, buy the chocolate later, renew the domain before it expires, book the table if it opens up. Some are going to be bounded budgets. Spend up to $100 finding the best supplier. Some are usage based like charges per query or per token or per minute. Some are outcome based charge when the ticket is resolved when the qualified lead is delivered when the itinerary is booked. And some are going to be hybrids. Subscription access plus usage, usage plus outcome, prepaid credits plus topups. The transaction is not just leaving the store. It is now stretching across time. A checkout page is very good at capturing a single moment of intent in time. It is much much worse at representing that kind of ongoing mandate. Agents make mandates more common. Do this when the condition is true. Keep this under budget. Buy this every time it drops below a threshold. Spend a little bit every day until the goal is done. Pay for the output if it passes the eval. These were the kinds of transactions that humans found annoying. Right? We don't like metering when humans have to run it because it's frustrating to us to manage all of the paperwork. And usually that would mean this wouldn't work. And that's why it hasn't been adopted before. But all of the boring pieces that make this possible are pieces that matter when agents are in play and need dependable rails to pay because agents need metering and authorization and fraud and memory and billing and settlements. And Stripe is interested in those boring pieces. And that's why this whole announcement matters because Stripe is the kind of company that follows through on that and actually delivers on all the boring bits with polish so that it actually ships as you expect an agent to act. Now this is the point where the agent conversation gets too clean, right? It becomes a picture of helpful agents and trusted wallets and smooth protocols that it's all happy golucky. The same automation that makes good agents useful also makes bad automation more dangerous. And this is where the radar announcement comes in, right? So, Stripe's radar announcement is fundamentally about making sure that in this world that we're envisioning, fraudsters don't just ruin the pie for everyone. A conventional SAS company could often afford a generous free trial because one more user clicking around the product didn't really generate costs. But in an AI world, one more free user is going to absolutely eat tokens. Especially if they're a fraudster who's trying to burn tokens as part of the fraud. They are consuming the company's costs dollar for dollar, right? They are literally stealing money out of the till by stealing tokens. And so in that world, you have got to be able to get a handle on fraud to take advantage of an agentic economy or the agentic economy will be still born at birth. And so radar is essentially the first play in a larger in a larger strategy that I think Stripe and others will also pursue around how we contain fraud when the world runs very differently when it's not primarily human fraudsters but agent fraudsters. And it's already that way. We already have a few thousand humans in the world running millions of agents to take advantage of businesses to register fraudulent accounts to steal tokens. In that world, you have got to get a handle on those bad actors to avoid poisoning the agentic experience for everyone before it really gets started. And what's interesting here is that Stripe is leveraging its own network to deliver value for everyone because Radar is trained across Stripe. Link gives Stripe a consumer wallet surface. Stripe signals extends risk information beyond transactions processed directly on Stripe. So the companies sees payment behavior, business behavior, signup behavior, and now potentially agent behavior across a large part of the internet economy. John Collison has made the point that one of the best parts of using Stripe is that all the other companies use Stripe. In the agent economy, that is not just a network effects line. It's it's how you build trust. And when commerce leaves the seller's funnel, trust has to come from somewhere else. And that's really what Stripe is setting themselves up to do. They're setting themselves up to be the guarantor of trust in a buyerdriven economy. Now, let's revisit that brand story. There's a claim about Agent Commerce that sounds right at first and then starts to fall apart. And the claim is simple. Agents are rational optimizers. So, brand doesn't matter. Part of that is true. An agent doesn't feel status or nostalgia or aspiration like a person does. It's not moved by the emotional pacing of a landing page. But that doesn't mean brand disappears. It means brand changes location. In the seller controlled web, brand often does its work at the point of persuasion. You land on the site, you absorb the design, you read the language, you see the social proof, etc. And you decide whether the company deserves your money. The seller gets to perform the brand for you every time you arrive. In the agent mediated web, brand increasingly becomes part of the buyer's memory. Your preferences, prior purchases, trust history, your loyalty memberships, your stated dislikes become inputs to the agents decision-making. The agent may not feel brand loyalty, right? But it can carry brand loyalty as a constraint. It can know that I particularly like 49th Parallel Coffee as a coffee roaster. That is true, by the way. I adore them. They're fantastic. And so it can know that and go shopping for coffee with 49th Parallel. But it can also know the opposite. It can know that I avoid this airline. I dislike that marketplace. I choose this vendor because the other one broke my trust. And that's a hard place for sellers to win because they don't get to reset the conversation. They don't get another chance. They don't get to control the surface and tell me, "No, really, trust me. I promise you, I'm good." Brand becomes less like a billboard and more like an entry in the buyer's operating context, an entry in a ledger, if you will. That doesn't make brand less important. It might make it more important, but it's a whole lot less drama. The brands that matter to agents will be the ones that have become reliable preferences. They will have clear data, clear policies, consistent fulfillment, strong reputation, good support, and enough accumulated trust, probably through IRL experiences to survive the comparison. The old brand question was a little bit, how do we make the buyer feel something? Now, in fact, sometimes it was a lot that the new brand question is kind of a lot. How do we become the kind of business the buyer's agent remembers as a good answer to the buyer's question? That shift is really easy to underrate because it is not visible as a checkout butt, but it may be one of the biggest distribution changes on the internet because the seller's persuasion service is kind of disappearing and the buyer's preference layer is replacing it and getting much stronger. And this is where the agent story becomes an economic structure story rather than a shopping story. Agents may become a large class of economic actors, but the larger question is what their existence does to companies that have depended on selling for a long time. If Stripe lowers the cost of transacting and provisioning and trusting across the market, it is real possible that a lot of companies that have existed on persuading people disappear and they have to transform themselves into companies that are actually relevant answers to the agents that buyers intend. If there are brands out there, and I know many of them, who survive not because they have a trusted relationship with the buyer, but because the buyer ends up there when they're tired and frustrated and just needs an answer for something, they're in trouble. Because agents are going to make that moment of pure emotion much less frequent. Long-term, the economy on the internet is going to become more rational. It's going to become a more efficient market. is going to become a place where agents smooth out a lot of the emotional choppiness that has made commerce into a shape that marketers feel really comfortable with. Marketing to agents is going to be different. I've given you some hints. I don't think marketers lose their jobs. I don't think they go away. They have to work extra hard to reach people and real experiences and they also have to connect with agents. It's not the same thing to emotionally persuade an agent. There is no emotion there. You have to make your experience relevant. And so I want to close by asking you some questions for your business, for your store, for your team. If you are interested as a buyer or a seller, you got to think about this. Can you visit a company today and use an agent to transact? Can your business be called by agents? If you own a business or work in one, not scraped, not summarized, called programmatically. Can an agent understand what you do as a business? Can it identify when you are relevant? Can it compare you against alternatives? And by the way, if you're a buyer, can you go shopping for alternatives and feel like you got a great comparison in a particular product area you care about? Can an agent act without human babysitting on the internet? Can an agent distinguish your real capabilities from what's just marketing fluff? And a lot of companies don't want to reveal that, right? That is very buyer friendly. Can an agent read your pricing, but also your terms and your identity and your error handling and your data access and your cancellation and recourse policies? The agent needs a commercially complete path. And this is what a lot of AI discussions have missed to date. Agents do not just need to reason to use the internet. They need institutions around them. Payment systems and trust systems and pricing systems and dispute systems and liability systems. The reason commerce is hard is not that clicking buttons is hard, guys. It's not that taking payments is hard. Commerce is hard because economic action has real consequences. Stripe understands that because Stripe has always lived where the software meets the money. money moved or it didn't. Fraud was blocked or it didn't. A sub subscription renewed or it didn't. A payout arrived or it failed. A business had cash flow or not. That is a different kind of software from a chatbot. And the Asian economy is going to need a lot more of that kind of software. And I'm really excited for that because it means that we are going to start to see a new kind of commerce and an it's a it's a kind of commerce that is buyer friendly. I'm not saying that the website is going to disappear. I don't think human buying is going to disappear. I don't think brands disappear or design disappears. People will still browse. They want to see and feel and compare and decide. But the power is moving into the buyer's hands. And the reason why is simple. As more and more agents get on the internet and start to drive intent over time, the entire infrastructure of the selling funnel is going to crumble. It just won't work the way it used to. You will either have to really invest in highquality experiences for real people and I think IRL marketing is where it's at or you are going to have to invest in clean contracts for buying agents. So they see your entire infrastructure as a company is AI agent friendly and reflective of the intent that the buyer communicated to the agent. So, if the buyer says, "I want authentic coffee." And the agent visits the coffee website, it knows that all of the hooks are in place for it to read the actual qualities of the coffee and map it back to my intent and get me my Ethiopia honeyprocessed coffee now. And that's going to be the future of the internet economy. And it looks a whole lot different from what we grew up in. I think it's going to be pretty cool. I think there's a lot of stuff to build. If you looked at this and you're like, "Well, Stripe built everything." No, you don't have imagination. Stripe built some cool rails, but the world is your oyster when it comes to agentic commerce. The opportunity here is through the roof there. I I named multiple trillion dollar market opportunities as the Asian and internet economy starts to grow. There are massive massive markets here around pieces of the internet economy that are going to be agent-driven, that are human driven today, that need that transformation, that need that fundamental rethink. So, get excited. If you're a builder, it has never been a better time to be a builder. If you're a buyer, it has never been a better time to be a buyer. If you're a seller, you're going to have to rethink some things, and that's why I made this video. Best of luck out there. Yes, I have a whole guide on this on Substack. And of course, subscribe and I'll have more videos, too. Cheers.