Barron's Streetwise - Spooky Agents, Cable’s Second Death, and Stock Picks

Episode Date: October 9, 2026

Jack dives into persistent AI and a broadband land grab, and talks stock picks with Paulina McPadden from Baillie Gifford. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

For AI agents and AI in general to work well, they need an actual source of ground truth. Scraping data off the internet, sure that works fine, but you're going to miss a lot. Hello and welcome to the Barron Streetwise podcast. I'm Jack Howe, and the voice you just heard is Paulina McPadden. She is co-manager of the Bailey Gifford International Concentrated Growth Strategy. That means she buys stocks all around the world and not a lot of them. Just the one she really, really likes, and she'll share some of her favorites with us in a moment. First, we'll say a few words about the cable industry's second death.

And just in time for Halloween, Wall Street has a new hobgoblin. It's called The Persistent Agent. And it's scarier than a wet gremlin. Let's get into it. Listening in is our audio producer, Emily Sommelin. Hi, Emily. Hi, Jack. How much AIing are you?

you doing? Have I asked you this? I'm doing more and more. I mean, are you, are you asking that, that chat bot every day? All your, uh, all your tough questions. You know what I'm asking it, Jack? I'm asking it to get off my screen and my word doc. If you can recall, I'm full pilgrim. I'm, I'm politely abstaining and the quilting and butter churning is hitting its max. It's shocking to me because I'm, as you know, a hundred years old. And, uh, I'm using it at a time. Like, I, I'm, when I was a kid and you wanted to learn about something, you had to go to a library and you had to use something called a card catalog and there was a Dewey decimal system involved. And then I can still remember when they came out with encyclopedias on a disc.

And it was shocking to be able to have something on a disc that used to be on the shelf back at home, a whole shelf on a disc. Then the internet. And now it's like anything you want to know, just take a picture. I want to, I need to, I need to. to power wash my patio and how what kind of PSI am I talking about? And let me take a picture on my patio and send it to you. And it'll just tell you, hey, listen, don't just go by PSI. You got to think about gallons per minute and you got to let's talk nozzles and the thing's got a whole like

education for you on anything you want to know. It's, it's shocking to me. I think I think as a homeowner, you do use it more. As someone who rents, unfortunately, AI cannot get my super Jerry to message me back. That's one thing AI cannot do. Well, it's good to know there's still a place for Jerry out there anyhow. I want to talk about the, not the new twist, but the thing that's driving stock prices recently in AI, which is persistent agents. First, I want to just say a few, few words about cable and cable stocks. These are a couple of columns I wrote recently in Barron's. I don't have a ton to tell you about cable, except that these stocks are now about the cheapest group in the S&P 500.

index. The prices have really collapsed. And remember not too long ago when I told you that Micron was the cheapest stock in the S&P 500 relative to earnings that was because the earnings were growing so quickly, it's no longer near the bottom. That stock is up quite a bit. Since we spoke about Micron, let's see, that was this past April. The episode is called the cheapest stocks in America. And since then, Micron is up 144%. That is not normal. So now the cable stocks are the cheapest in the S&P 500, are they going to do the same thing? Do not count on it. This is a totally different situation. When I wrote about them recently for Barron's Comcast was down 60% over the past five years and Charter, that one's a cable pure play. That was down 84%. And Charter was three times forward

earnings projections, three times and Comcast was six times. They were the lowest and sixth lowest PEs in the S&P 500. And I'm calling this the second death for cable and I think it's going to be little trickier than the first death, which wasn't actually so bad. In a nutshell, the first death was when streaming came for pay TV. Remember all the talk about cutting the cord? A lot of us cut the cord and cable did not mind because when you have pay TV service, the cable company has to go and then pay the networks for airing their content. And they bundle with that cheap broadband. And the broadband, they don't have to pay anyone. They have already laid the cables. As folks would cancel their pay TV service, what the cable companies would do is they would say, okay, but we're going to charge you more for broadband,

which you still need no matter who you're streaming with. You've got to have that fast internet. And they figured out a way to make even more money selling just broadband alone without the pay TV. So those stocks actually kept rising long after investors said they were in trouble. What's going on now is different. Telecom has finished investing to build out their 5G networks. The pricing environment in telecom is pretty benign. In other words, we don't have price wars. These companies are making great money.

They have heaps of free cash flow. They need to put that money to use. And the investment area that makes perfect sense for them is fiber optic broadband. They basically double dip here because fiber is very useful for their cellular networks for something called backhaul, but also they can sell home service to their wireless subscribers. And the money there is good and the churn there is low. In other words, customers never really think about their home broadband. They don't cancel and change often.

But the real miracle of that for telecom companies is that when they bundle wireless service with fiber optic broadband, the churn in wireless falls too. Wireless churn is a challenge. It's pretty easy for customers to hop from carrier to carrier to try to find a better deal. The telecom has figured out a way to keep subscribers around for longer, and that's by bundling. They're loving the bundle. And fiber optic broadband is just better technically than, cable broadband in a lot of ways. I mean, cable was built for television, and that's a pretty much a

download-only medium. So you have good download speeds, lousy upload speeds. And there are things that cable companies can do to upgrade cable broadband. They're trying to buy time and work on it, but it's still not quite as good as fiber. The thing to know about fiber broadband is it still only has a 22% market share that compares with 56% for cable. So cable still has plenty to lose. especially because fiber availability has only reached 65% of households. It's expected to top 90% by 2030. So this is a real challenge for the cable industry. And it's not just fiber broadband.

You've got something called fixed wireless service. That's something telecoms provide too. They basically use excess spectrum to provide broadband that works off a cellular signal. That's grabbed a quick 14% market share. And you've got Starlink satellite service. which is popular in some rural markets, that's got a 4% share. I tell you this just to say, don't be too tempted when you see giant dividend yields on cable stocks or shockingly low price to earnings ratios.

Those numbers are like that for a reason. This is a challenge that we're going to have to keep an eye on over the next several years. Emily, where do you catch that love island that I know you like so much? That's not on cable TV, right? You're using a streaming service for that? I'm guessing it's not PBS, right? Not PBS Masterpiece. No, it's on Hulu.

Okay. And don't worry, they won't let you forget it. Oh, a lot of Hulu mentions on the show. It's almost Hulu-Ween. I know that because they won't let you forget. Oh, I wish I had to heard that. That's awful. Speaking of Halloween, the persistent agent,

which Wall Street has been kind of spooked about the past few weeks. I'm just going to give you a very, quick explainer of what this is and why it's different from the chatbots we've been talking about. There was something called Muse, and that's an app made by meta platforms, and it launched September 8th. Meta, of course, the company behind Facebook and Instagram and creepy bionic eyeglasses. And Musephobia helped send shares of Planet Fitness and Sirius XM Holding sharply lower. Other stocks, too. That was because of fears that agents like Muse will make it easy.

for busy and distracted consumers to cancel forgotten subscriptions. And Airbnb and booking holdings tank too. Mews is pretty good at chisling down travel costs through tireless robo shopping. And Charles Schwab fell because the agent might coax its users to take that money that's sitting in a checking account, earning 0.01% and move it somewhere else. Good for the customers, not so good for Schwab. And meta stock is up a lot since Mews launched. And there are discussions out there about who Mews might help and hurt.

Morgan Stanley theorizes that it could help the cruise line industry. They say if there's going to be a winner in travel, it's going to be a company with differentiated merchandise and relatively limited attention from shoppers and high distribution costs. And cruise lines check all three boxes. Mews is not the only persistent agent. Another early contender is a privately held company called Instinct. That one is invitation only. for now. OpenAI, the company behind ChatGPT, just launched an agent called Dots. Anthropic has

Claude Co-work, and Alphabet has Gemini Spark, and Elon Musk's SpaceX has Grockbot, and Apple has not really that much. I mean, it's got Siri. She, I mean, series of she for me, maybe series of he for you. Siri is not much of a persistent agent. I don't even think she's much of a chat pod, to be honest. But even though Apple doesn't have much here, maybe it has enough. That's an ongoing debate. I'll come to that in just a moment. What exactly is a persistent agent? I do a semi-regular series at Barron's that I call How to Pretend to Understand. In the past I've done, how to pretend to understand Brexit, how to pretend to understand quantum computing. The idea is you give people just enough information so that they can survive an office conversation on a convoluted topic.

And ideally, give them a couple of facts that they can lord over their less informed friends. What good is knowledge if you can't weaponize it? And if it turns out they pick up enough to make some educated guesses about investing, all the better. So here we go. I'll try to do this in a minute or two using small words. That chat bot you've been using seems personable, but it doesn't really chat. What it does is it converts your questions into numbers representing words or word fragments. Those are called tokens.

And then it uses a probability engine to determine the next best number to add again and again until it finishes its response. And then it converts this numerical response back into words. Does that make sense? Is it weird? It's just a giant lightning fast calculation on a huge string of numbers. And the result tells me how to power wash my patio. That's why they need all that computing power. A lot of numbers.

A persistent agent does not do that. It doesn't simply ask itself which token comes next in the answer. What it does is ask which tool it should use or subtask it should perform to get closer to a goal that you have assigned it. And we call it persistent because the agent can keep working even after you've logged off. And it can learn as it goes, just like a human assistant. And for that it needs its own computing sandbox. Something called OpenClaw that launched late last year,

got this whole thing started. That can run locally on user machines. But Muse from Meta uses a virtual machine. That's a pretend computer made with software that runs on an actual computer somewhere. Think of the actual computer as a giant apartment building. And the virtual machine is like a tiny apartment just for you in that building. Agents also use headless tools. When we want to browse websites or fill out forms or place orders or send messages. We need something with a front end, something with a graphical interface, like your web browser. We have to see it in pictures to make sense of it. Your persistent agent does not need that. It can do everything in code without a graphical interface, lightning fast. It can even make phone calls on your behalf using a synthesized voice, and the voice

sounds pretty darn real. Emily, you're shaking your head and making a, making a... That's so spooky. Let's... It is Halloween, because if I was on the other end of that phone call, I would be afraid. If you were to go to Spirit Halloween and shop for a persistent agent costume, what would that costume look like? Oh my gosh.

I think a black morph suit, nothing else. Red eyes, right? And then you've got to run around really fast. There are two last concepts I want everyone to know about in understanding persistent agents. By the way, is that term catching on at all? I mean, that's just what I'm calling them. Where do we call on these things? AI agents.

I don't know. We'll see what term catches on. I'm calling them persistent agents for now. The two concepts are connectors and payment rails. Connectors are things like email, calendars, account, logins, and other sources of personal information. The more you tell an agent about yourself, the more it can do to help you. I mean, hopefully help you. We'll see how it goes, humanity.

A Barron's colleague of mine, Josh Schaefer, he used Mews recently and he wrote about it, used it to find cheap pants, and to book a toenail-related doctor's appointment. I did ask him for a toenel update when we were doing some TV together last week, and all he told me was, it's off, it's off now. I don't know what that, I don't know what was wrong,

but Josh is fine, ladies and gentlemen. The point is he could do those, things with Muse, but it was only when he provided his credit card statement that it showed him a forgotten subscription that he could kill. And that saved him some money. I think he said $25 a month. So these are connectors. You need to tell your persistent agents and things about you. And payment rails allow agents to execute transactions. Mews uses a third party payment company called Stripe. That's for customers who presumably have made peace with meta-founder Mark Zucker. Zuckerberg snooping through their social media, but aren't yet ready to hand him their credit cards.

They use this third party company. Basically, when you want to make a transaction, that third party will generate a one-time kind of credit card number that'll make the transaction for you. So your details stay with them and you don't have to hand them over to meta. As you can see, trust is a big deal when it comes to persistent agents. And that brings us back to Apple. Apple has loads of trusts because it makes money by charging dearly for the things itself. and not by selling customer information. And it has its own payment rail called Apple Pay.

Even a half-century-old wallet carrier like me has Apple Pay. My daughter says it's very important. She uses it to hound me for money. The problem is Siri. Siri was supposed to get a big upgrade with this latest software release from Apple. Has anyone been wowed since then by Siri's new superintelligence? I'm still waiting for the wowing. I'm hitting that Gemini pretty hard in my work stuff.

I'm doing some chat GPT for my personal stuff. The only thing I'm really doing with Siri is the same thing I've always done. Call my wife from the car or find me directions too from the car. Apple's artificial intelligence trump card was supposed to be that tasks would increasingly run on local or edge devices like iPhones. It's just not clear how equipped today's phones are for staying up all night to perform a persistent agent's duties. Now, Metal has a lot less trust, but it does have an early move of. advantage and it has an excellent handle on how and where to get paid on services that don't

have customer fees and that's going to be important here barclays estimates that the cost to host a persistent agent user is around four dollars a month that's 10 times the cost of a chatbot user in other words you might be able to send chat gpt out there for free because it's cheap enough to run but you're not going to want to do that for your persistent agent you got to make money off this thing Amazon, by the way, has banned Mews as a shopper. Why? Well, that's on the belief that if anyone should get paid to steer customers, it should be the store. We'll see how that plays out.

You can expect all the agents that I mentioned earlier to compete on cost and point of access. Do I download an app for this thing or can I just chat with it online? Do I use my weird glasses from meta for it? That stuff that'll also compete on third-party application tie-ins. whether it's consumer or commercial focused, and things like data sovereignty and permission. For example, Muse likes to ask before doing things, while instinct prefers a freer hand. Barclay says that the first agents to lure customers to share connectors will hold the advantage. It says Muse and OpenAI dots are well positioned.

By the way, it calls this the biggest turf war on the internet right now, and it says that ultimately there will be $10 trillion in stock market value up for grabs. That's almost the value of two invidias. Agents are a big deal. B of A security says that Apple can eventually win on trust, but that it will be difficult to do so while releasing major software updates only annually because persistent agents grow more powerful every day. And that I think is about all I have to say on the sub-dust.

Emily, good luck with your costume if you decide to go as a persistent agent this year, or maybe a headless browser. Consider that too. I haven't really gone down this road too far of using these things. I'm not quite sure who I trust yet, but I could definitely see them being useful for planning vacations, maybe finding some savings on the credit card bill from subscriptions you've forgotten about. I've seen people saying, how could you forget about a rerout? How do you have a $25 bill that you don't remember? Folks, you get busy, all right?

There's a lot of money coming and going for some of us, all right? Let's not judge out there. Let's show each other some grace. Emily, if you were going to use a persistent agent for something right now, what do you think it would be? I will say, if the bots are taking all these great reservations in New York City, I'm going toe to toe. I would like to be using the same equipment that all these reservation snappers are taking. You mean like for a restaurant that's hard to get into, they'll call on your behalf and keep calling and keep calling. falling until they get you a table. Is that the idea?

Well, it's more even like online on these open table resi when they drop. Bots snap them up as soon as they become available. A very flawed human finger is always so fast. I don't do hard to get into. I do the kind of restaurants where when you walk in, the owner knows your whole family and he's like relieved to see you. You feel like you might make the difference between a profit or a loss for that month. That's the difference between you and me, Jack.

I want a hostess who is a model most of the time and is disgusted. that I'm even there, even though I have a reservation. That's how you know that food's going to be great. So that's the key to success for these companies. Just sort of... I swear they're hiring off the catwalks. That's the kind of person they need, apparently, to handle their front check-in. I thought you just had to make good food.

I didn't know performative disdain was part of the equation for success. That's good to know if I ever decided to go into the restaurant business. I'm pretty sure all of New York City is performative disdain or actual disdain. Okay, let's take a quick break. And when we come back, we're going to hear from Paulina McPadden about some of her favorite stock picks around the world. You'll hear me mention muse a little bit with her. I've got muse on the brain. That's next after this quick break.

Welcome back. Let me take a quick look on the machine here at what the old VEU is up to at the Vanguard Futsi all-world X-U-S index fund. I use it as kind of a proxy for what is everything but the U.S. market doing. Boy, it was ahead of the U.S. for most of the year, and when I look at it now, it's just, it's a couple of points behind. S&P 500, you've made a little over 15% year-to-date. Everywhere else in the world, you've made a little over 13%. The U.S. market, you're paying somewhere around 22 times earnings, and you're getting shockingly good earnings growth.

Everywhere else, you're paying, let's see, 14 times earnings. The earnings growth is decent, but not otherworldly. Let's hear about some individual stocks that look like good deals outside of the U.S. For that, I recently spoke with Paulina McPadden. She's with Bailey Gifford, and that is based in Scotland. We're in Scotland, Emily? Edinburgh. I'm going to November, and Paulina so kindly gave me some wrecks.

Thank you, Paulina. I needed that insider intel. I have never been to Scotland. I don't know about. I think I have a walking series on my treadmill that will take me on a video tour of Edinburgh. I haven't done it yet. So you've been. So I've been.

So I bet. Yeah, I bet. And with that, let's get to Paulina, who is co-manager of the Bailey Gifford International Concentrated Growth Strategy. Let's hear a part of that conversation now. How do you figure out, I would imagine by the time you identify these businesses that they've already done pretty well. And investors, I mean, you know, with only 20-some stocks, you've got to put a significant

amount of money into each. These are not tiny companies. these are not startups. How do you tell whether the best returns are already in or whether there's a lot more to go? I hear everyone talking about the TAM, the total addressable market. Look at the TAM. The TAM is a smogging. The TAM is a scatillion dollars.

But how do you tell, you know, you're catching the company in the right phase of its growth where there are plenty of good returns to come? It's a good question. And it's always a balance for us between size of opportunity and level of conviction. Because when you're running a concentrated fund, you can't just make a binary bet and hope that it works out, you need a relatively high degree of belief that this company is going to do well. Because if you only have 30 companies in your portfolio in one tanks, that's not ideal. Let's be honest. So we don't tend to buy companies when they're super young,

not because they can't deliver very high returns and in fact they can, but because it's very hard to tell in an early stage company whether it's going to be successful. They've not yet had a chance to actually prove a level of competitive advantage that, again, is extremely important to us. maybe take an example of Mercado Libra, which has been a very successful investment for us. But even before we invested in international concentrated growth, we had known the company for a number of years before that because that's how long it takes us to get to know the management team and make sure that's the right kind of company for this type of portfolio. If I asked you to share with us a few of your favorite names in the portfolio right now,

which ones would you name? So I would definitely name TSM. I just, I love a monopoly. An investable monopoly. What's not to like? But I mean, it's some, but this is not, this is certainly not an undiscovered gem, right? What's, what keeps them safe? What, what prevents someone else from doing what they do?

I mean, I think people are really worried with a lot of these companies. Like, oh, well, we can see what the level of growth in demand is going to be over the next two or three years. But such a small. portion of analysts and market participants look out beyond three years. It's actually very easy to have a degree of insight over even relatively short time periods, which is five years long. And in TSM's case, I think what gives me a lot of conviction in that company is the level of complexity in chip manufacturing is only going up and the level of investment required to keep up with that complexity and enable the manufacturing of the next node, the next. Come on, extreme ultraviolet lithography.

I think those are three of the words. How hard could it be? So it's that, but it's also doing it at scale. It's not just, you know, anybody can show, oh, we've developed a process that makes amazing chips, but actually doing it at scale for thousands, hundreds of thousands of wafers per month, consistently at high yields to be able to deliver a price that your customers are willing to pay. That's not something that you can just replicate overnight. Is there any concern about where we are in the AI buildout? The company spending tons of money, great business for TSM, and maybe let's assume that it's going to maintain its share in its business going forward.

What about the overall level of spending? Do you feel confident about that? I mean, I think what people should think a bit more about when they talk about level of spending is it's not just the absolute level, but it's also where the returns that these companies are seeing from that spending over time. And it's easy to look at, you know, the model builders and say, oh, they're spending so much money on their infrastructure. and the revenue growth is so and so. But if you look at the companies are actually implementing AI in their day-to-day businesses, like they're spending more and more money. Even though token costs are going down over time,

the rate at which these companies are using AI continues to go up because they see the return, whether it's from better advertising, better e-commerce service for consumers. So I think looking at that return on investment and seeing the leaders in AI pulling ahead of the rest of the pack in terms of how they're using it, that gives me quite a lot of confidence that they built out.

This will be a leading question, as they say in the courtroom. Can I infer that because these companies are seeing a positive and an agreeable return on investment, that the outlook for the overall market is okay, that we're not worried about being in an AI spending bubble, that companies are going to make good profits on this money they're putting to work, and we're going to see earnings growth going forward, and it's going to justify the overall market levels we're seeing. That is a leading question. I gave you the answer with the question.

I mean, look, if you call a bubble eventually you're going to be correct. I'm not going to call a bubble because I don't want to have to time when things are going to change precisely. What I want to do is find those companies that are going to be resilient and continue growing regardless. And actually on that bubble fear, I'd almost take a step back. There's a fantastic academic called Carlotta Perez, who's done a lot of work on bubbles historically. And what she found was, yes, you did tend to see high levels of overinvestment. You saw this with the dot-com boom in the 90s and early 2000s, obviously. But what that over-investment does is it lays the foundation for future growth.

So the over-investment in fiber capacity allowed the flourishing of companies like Amazon and Google that were then able to use that capacity cheaply. And I think if when we do see a bubble in AI, you're going to see something similar because the level of investment in infrastructure, whether that's power, or land, even the chips themselves, which actually last a lot longer than you think they do and can be repurposed for other things, that creates the foundation for growth in the future. You end up with good stuff, even if it is a bubble down the road. That's how we got broadband. So you like TSMC. Give us two others, if you were.

Well, I'd actually move away from AI, even though we focused on it in this conversation and say they're almost on the opposite side of the spectrum are companies like Ferrari and Hermes. Now luxury is having a bit of a tough time at the moment people aren't spending as much as they used to. But in luxury, when you find a company, like, say, Hermes, that is able to take that long-term view and continue to create a sort of scarcity and not just flood the market with products and therefore it's able to drive continuous price increases over time, that's really, really valuable. If I was going to pick a single company in the portfolio that I think may well still be around in 100 or 200 years time.

It's probably her a mess. What does that look like? It's not an undiscovered brand. I mean, when I think about how well they already sell and how many rich people there are in the world and future demand for their products, how do you grow a business like that? I mean, the volume growth is very low in the single digits. It's the steady pricing growth on top of that that delivers. But also, when you're investing for the long term, you're able to take a slightly different

perspective because it's not just about the rate of growth. for us. It's also about the duration of growth. And in some ways, the duration and the longevity of that growth is almost even more impressive. If you can compound at 12%, but do that for 50 years, that's tremendous. Okay, so that's one that you like. Is there a third? I'd probably say Shopify. I think Shopify is a really, really exciting business. I mean, it's in e-commerce, obviously, but it focuses on sort of smaller standalone merchants, although also increasing the enterprise merchants as well. And it's been able to drive continuous growth through almost a single-minded focus on making merchants' lives easier, whatever it takes. So a lot of businesses in e-commerce, what they do is

they'll gain a certain level of market share and then they'll start pushing the take rate up because that's how you gain margin. That's how you make more money. Shopify has always taken the approach that, no, if we're in the business of helping merchants, what we need to do is we need to enable them to grow their own businesses. We need to grow their GMV because then our revenues grow as well. So they've kept take rates at about 1.4, 1.5% forever will continually adding new services and new capabilities. The idea of this business, if you're a merchant who's not great at building an entire merchant infrastructure and website, they're going to take care of it for you. Stock the merchandise and they'll take care of handling the transactions for you. Do I have that more or less right?

More or less correct. I mean, I think increasingly it's moving in that direction. Historically, they would provide the tools and the merchant would use those tools to create a website and create a store and start selling to consumers. And the thing that Shopify has always been very good. good at is mediating complexity. You know, you could sell on Amazon, you could sell on eBay, you could sell in person physically in a store, and Shopify enables all of that for you in one sort of package. What AI is doing for them, it's increasingly moving to what you said there, where a merchant may not know anything about how to sell up a store on Shopify. Actually, he can just have a conversation with an AI chat button that does it for him.

I was just reading, where's my stack of analyst research? I was just reading a report here. Boy, this is some dense writing, but this is from an analyst who's trying to explain about why Muse is or is not a threat to Apple. And one of the companies that is mentioned prominently here as being involved in this is Shopify. So is this a company that you see, you know, maybe this is an AI company? Is it a company that gains a larger and larger role in the world of AI agents that are working in the background and figuring out our shopping for us? I think there's a chance that that is the case. Yes. I mean, it's always a sort of an element. outstanding question of does a gentic commerce help or harm e-commerce businesses? And I suspect the answer is going to be, well, it depends. What goes in Shopify's favor that I find very interesting

is they've developed something called the catalog, which is effectively pulling together all the data on the products that their merchants sell, not just things like price and what it is, but also smaller things like, you know, what are the dimensions of this product? How have other people interacted with this product in the past? What are the reviews like? And it's all sort of combined together into this catalog. And the reason that matters is for AI agents and AI in general to work well, they need an actual source of ground truth. Scraping data off the internet, sure that works fine, but you're going to miss a lot.

And also it's very token intensive. And if instead you can go to this database and just know that all of this is true, that results in much better outcomes for customers and for merchants. And Shopify has said, in fact, that the conversion rate of AI searches done using catalog is two times higher than it is for non-catalog, agentic commerce. This has been most informative and most helpful. Paulina McPadden, co-manager of International Concentrated Growth Strategy at Bailey Gippert. Thanks so much for joining me.

Thank you so much for having me. Thank you, Paulina, and thanks to all of you for listening. Emily Sumlin is our producer, and she's going to Scotland next month. So if you have tips for her and you're listening on Apple or Spotify or YouTube-ish-type places, give us some recommendations in the comments. You're going to sound like the landscaper from the Simpsons. I'm not from Edinburgh.

You can subscribe to the podcast. There's a button somewhere, like you didn't know that already. You can leave us a review, give us a persistent agent tip, and if you have a question about investing, what you can do is tape it on your phone, use the voice memo app,

send it to jack. That's h-o-u-g-h, at barrens.com. It could be on a future episode. Imagine the hometown hero you'd be then. Thanks and see you next week.