Google DeepMind Reorgs, Rainforest Discoveries, Monkey Prices Surge | Harley Finkelstein, John J. Giamatteo, Jeremy Allaire, Mackenzie Burnett

22:19Harley Finkelstein:Okay, this is cool. This is very cool. See? Yeah, this is double-decker. Double-decker. See, out in California, we don't know anything about these pontoon boats because these don't work at the beach, right? Correct. But I was sort of correct. Somewhere in between a houseboat. Yeah, yeah, yeah. Aren't you from the land of a thousand lakes, right? Michigan, isn't that the land of a thousand lakes? No, me. Minnesota. Oh, you're from... Okay. Are you a pontoon boat guy? Pontoon boats are huge in Minnesota. Okay. Yeah. How many... They're for lakes. They're for lakes, basically. How many hours have you spent on pontoon boats in your life?

22:5310 ,000. 10 ,000 hours? I'm an expert. The man. Man, putting life's work on the pontoon boat, apparently.

23:03Harley Finkelstein:You're in the wrong job, buddy. I think you got to become a pontoon boat captain at this point. It is crazy to think in another life, Ben, could have been the top pontoon boat salesman in the entire world. This one comes with Apple CarPlay. Don't they need to call it BoatPlay at this point? They got to come with something else, CarPlay. But they're doing navigation, on-water navigation like this. um okay yeah moving over to okay we got what is the story of the day in technology yes this is the story of the day okay hit us john you know what i'm talking about startup billionaire sued over super yacht broker fee everyone's talking about this this is potentially bigger than what's going on at google today uh in london the billionaire founder of one of the world's most valuable startups recently bought a super yacht valued at an estimated$400 million.

24:05Harley Finkelstein:Yeah. Continue. Replete with a glass-bottomed swimming pool and a helipad. Do they make, wait, do they make super pontoon boats? Can you get a$400 million pontoon boat? What does that do for you? Almost certainly. Can you imagine just a 300-foot pontoon boat? The world will create infinite ways to make you part with your hard-earned money. Yeah. And I'm sure we could arrange an almost half a billion dollar pontoon boat. Okay, team, AI images. I want to see what$400 million gets me if I'm going all in on a pontoon boat. I want to really dominate Lake Michigan. Is Lake Michigan too big for a pontoon boat?

24:52no because you said lakes so any lake it works yeah i think the pontoon boat is not for like big waves big waves you can't it can't be like but there's not big waves on on uh on lake michigan you can't surf or anything i guess there's no i don't think so surf yeah anyway so uh it was a 400 million dollar super yacht it had a glass-bottomed swimming pool and a helipad there was just one problem. According to this reporting in the Wall Street Journal, he didn't pay the broker's fee. That is according to a top luxury yacht broker who is now suing Nick Staronsky, a co-founder and chief executive of digital banking giant Revolut, alleging the entrepreneur went behind their back to avoid paying commission on a purchase they had worked on.

25:44Cecil Wright and partners, a luxury yacht broker, filed the legal action in London's high court over Stronsky's acquisition of Nixie, a 335-foot vessel with four terraces. The broker alleged in a court filing that it was negotiating the purchase for Stronsky when his team secretly contacted the yacht seller directly and agreed to buy the boat, avoiding roughly$20 million in standard commission, the amount the dealer now claims it is owed. In January, Stronsky's team allegedly told Cecil Wright to hold off. Let's see. I'm not going to flip through this. Let's see. To hold off. Hold off on contacting the vessel's seller, Canadian billionaire and former hockey player Patrick DiVigi, and said they would talk again the following week according to the filing.

26:40The filing claims that by next week, Stronsky's team had already arranged to buy the boat directly from devogy uh bypassing the broker entirely it was cecil wright that brought the yacht to staronski's attention and introduced the revolute co-founder and the owner matt young says

26:58Harley Finkelstein:unbelievable taking food off the table of that luxury yacht broker's family

27:06a spokesman for the staronski family office said we are aware of the claim it is without merit and it will be defended. He says he's not guilty. Nixie features a private wellness spa, indoor and outdoor cinemas, two movie theaters. Now we're talking. As well as a Japanese teppanyaki grill, according to Boat International.

27:24Harley Finkelstein:Pull up this picture that Tyler just shared. I think this is what Nixie looks like. Yeah, let's see it. When charted, it is equipped with a variety of water toys, including a luxury seabob water sled and inflatable blob launchers. Is this the Bondoon boat? It's a luxury pontoon boat. Oh, sorry. That's the$400 million pontoon boat that Tyler just generated. That's pretty good. I like it. I like it. That just looks like a normal yacht. That doesn't actually look that crazy to me. I don't know. Staronsky is one of Europe's highest profile startup founders. Founded in 2015, Revolut is now the region's most valuable startup with a recent share value.

28:04Share sale valuing the digital bank at$115 billion. Wow, Revolut. What a tear. Originally built around low-fee foreign currency transactions, Revolut is widely used by international travelers and migrants sending remittances home. The company gained a full UK banking license in March after a lengthy battle. Wow. It took them over a decade to get a banking license. That is crazy. But they did it after a lengthy battle. Now they're looking to expand into the United States, having recently applied for a national banking charter. It's crazy. It's crazy that they can't go faster. Like, Erebor got a charter so fast, and then there's been other companies that have explored, like, buying a regional bank and then using that charter.

28:47Harley Finkelstein:It's almost certainly part of why Revolut is now such a massive company. Yeah, they figured it out. They don't have infinite competition in the way that many banks and neobanks have here in the U.S. Earlier this year, Revolut reported record annual profits and a jump in its customer numbers at the same time. Revolut said it remained focused on its goal Of increasing its customer base to 100 million Users by mid-2027 From more than 70 million currently Well, good luck in The luxury yacht battle Hopefully everyone gets what they're owed And everyone is compensated fairly Hate to see people fighting over that Let me tell you about Railway Railway is the all-in-one intelligent cloud provider Use your favorite agent to deploy web app servers Databases and more While Railway automatically takes care of scaling, monitoring And security AI's great reverse bank run.

29:39In today's newsletter, Joe Weisenthal wrote about Leopold's Basilisk and what makes this particular mania feel distinct. People are betting on left tail outcomes without sacrificing the right tail of financial opportunities. Doomers are getting fabulously wealthy. Let's see what he's saying. Here's the deal. Historically, if you want to cut off your left tail, you also have to cut off your right whitetail, or to put it more specifically, if you are up big on a stock position and you want to hedge your gains by buying puts, you can do that. But if you want to maintain your puts over time, you spend a lot of money and erase any upside gains that you might otherwise be making.

30:16Or to put it even more generally, suppose you're afraid of the outbreak of war. Well, you can move to a rural farm and grow your own food, but then you miss out on the economic opportunities afforded by living in the city. People make these kinds of trade-offs all the time, depending on what's worrying them. Pin that thought for just a second. One of the big questions with AI is whether it's just another technology or whether it somehow will be different this time. It's purchasable intelligence, kind of like purchasable electricity, or is it something more like a sci-fi novel where humans get replaced or eliminated in some way?

30:50As businesses, the big American companies selling AI, mostly operating as if AI is just another technology to be developed and solved. But we We know that there's deep anxiety within the companies themselves that AI could play out in the way sci-fi novels have expected for decades. This sometimes is reflected in their governance structures. Here's one stylized fact, though, that might argue for the it's different camp. If you are one of those people who have been reading AI philosophers and message boards for the last decade, familiar with paperclip maximizers since long before ChatGPT came out, you might have deep existential fears about something bad happening right around the corner.

31:26You might also have ended up very, very rich. With AI, there has been no trade-off. The doomers who were early on accumulating GPUs, or better yet, shares in GPU companies, have done extremely well. The put option on humanity has become the call option on the technology. You could bet on the left-tail outcome, not sacrifice anything financially. We've maybe had some shades of this before. For a while, Bitcoin was or had characterized as an end-of-the-world hedge, a form of money for when the nation-state collapsed, and people made a lot of money on that bet. But a trader wasn't driven by the fear that Bitcoin was a world destroyer or take gold, betting on gold at certain times protected you against hyperinflation.

32:05But the fear that came with the bet did not entail gold being the catalyst for doom. Something else would be the catalyst, but you'd be protected. AI is different. The long feared catalyst for doom itself has presented an unbelievably profitable trade, at least so far. What's interesting about this is that he's sort of, he's saying like the doomers got rich. There are plenty of AI doomers who did not join AI labs, did not participate financially, worked in nonprofits, just wrote books, worked on research, did not ever join anything that had economic upside. But I take his point that in general, if you are early to this, oh, well, like things are going to get weird.

32:45You probably were earlier on the train. And a lot of those people that were talking about AI and doom scenarios did wind up working at labs that did very well. So it's a good point. But there are some people that don't fall in this category. Long before we had the transformer architecture of GPT, people have been theorizing about the risks of AI doom. One popular thought experiment is Rocco's Basilisk, a 2010 thought experiment. Wow. When was the last time we got a new thought experiment? I didn't know they were dropping new ones in 2010. That seems kind of recent. Maybe there's like, I don't know.

33:21Someone should come up with a new thought experiment. What's the most recent thought experiment that's really taken hold and been thought-provoking? Golden retriever mode. That's not a thought experiment. That's an axiom. That's a lifestyle. A thought experiment is almost like a paradox, something you can wrestle with. We've got to come up with some new thought experiments.

33:39Harley Finkelstein:I've been wrestling with golden retriever mode. You think it's a thought experiment? I think you can make it one. I don't think it falls in the bucket of thought experiment. What do you think, Tyler? No? What are you guys laughing at? Are you doing more pontoon boats? Crazy text. I'm trying to find the most recently created thought. Okay, Jackson, you are correct, yes. That is the most recent thought experiment. Good stuff. Anyway, let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches. That's very funny. Okay, wait. Quick.

34:15We've got to switch gears. because the price of lab monkeys in China is soaring as demand for preclinical testing outpaces supply. The price of lab monkeys in China is soaring as demand for preclinical testing outpaces supply. Do you think it's time to ape in?

34:36Harley Finkelstein:To ape strong together? Yeah, yeah, go long the monkey trade. So this is the actual news. We got monkeys up here. There we go. That first picture was crazy and really sad. Yep. But it is happening, and it is interesting because it tells you a lot about what's going on in drug testing, animal rights, obviously, and also just the broader economic market. So in June, a Chinese state laboratory paid roughly$26 ,000 per monkey for a purchase of 40 animals, nearly double what similar animals cost a year earlier. The price has doubled in the last year of monkeys. Biotech executives say the same trend is showing up across the industry, driven by a wave of new biologic medicines and next generation therapies that require far more primate testing than existing breeding capacity can currently support.

35:32The shortage could become a bottleneck for China's rapid drug development ecosystem because many experimental medicines must still undergo studies in non-human primates before regulators will allow them to advance. While regulated monkey farms have maintained relatively steady production, demand has accelerated much faster than supply. A similar imbalance emerged during the COVID-19 pandemic when vaccine developments sent prices sharply higher. But this time, the pressure reflects China's emergence as one of the leading centers for innovative drug discovery. The country now generates a third of the global pipeline of novel medicines and has become a major destination for clinical trials and licensing deals with U.S.

36:15and European pharmaceutical companies. So monkeys remain essential for making many preclinical studies because their biology closely resembles that of humans. We know the monkey has remarkable resemblance to our particular makeup.

36:33Harley Finkelstein:Yeah. But broad. Many people have said we have the minds. Yes. Yes. So if they're working on a brain, typically they would test on a golden retriever. But yes. So this makes them one of the best available models for evaluating a drug's safety, how it moves through the body, its effects on different organs, and the potential for serious toxic side effects before human trials begin. While the shortage may slow some research programs, it's also a sign of an unprecedented surge in investment and innovation aimed at developing new potentially life-saving treatments. Tom in the X chat, how long would it take infinite monkeys to make a one-monkey billion-dollar company?

37:13Harley Finkelstein:Yeah, wait. That is an important question. That's a good thought experiment. How long would it take infinite monkeys to make a one-monkey billion-dollar company? That's a new paradox. So there is a thought experiment related to this. It's the infinite monkey theorem. This is a real thing. It states that a monkey pressing keys at random on a typewriter for an infinite time will almost surely type any given text. Deeply relevant to LLMs. Are there any... Okay, get the tinfoil hat. What do you think is going on? Is it possible that our modern AI systems are monkeys just clacking a bunch of keys? And this is just an AI bottleneck.

37:59Yeah. You think that when you go to Chad GPT, it's secretly a monkey on the other side?

38:02Harley Finkelstein:Yeah. Yeah. That it's not a computer? Yeah. Yeah. No one really goes in the data centers. It could just be a bunch of monkeys on computers. You don't know. A bunch of monkeys and bananas. You don't know. Okay. Well, there's also news somewhat related. Are there monkeys in the Amazon rainforest? Almost certainly, right? Almost certainly. Jack in the YouTube chat says, how long until monkeys see? What about monkey do? There's over 50 to 100 species of monkeys in the Amazon rainforest. Okay. Well, a new LIDAR survey of the Amazon rainforest has uncovered evidence of a previously unknown civilization that may have supported as many as 3 million people, challenging decades of assumptions about the region's ancient history.

38:46So we talked to a friend of the show who was doing this, scanning the rainforest with LIDAR, putting new sensor stacks on helicopters, flying them around on planes. But there's new evidence here. So using laser scanning technology capable of penetrating the dense forest canopy, Researchers mapped more than 4 ,500 square kilometers of Brazil's acre state after flying over 4 ,400 kilometers of rainforest. The survey, which was published in Nature, identified 396 previously unknown geoglyphs, massive geometric earthworks made from deep ditches and raised embankments. The discoveries have provided the clearest picture yet of the aquare civilization, which is believed to have emerged around 500 BCE and flourished until roughly 850 CE.

39:45Rather than serving as permanent settlements, the earthworks appear to have functioned as ceremonial and civic centers where people gathered for rituals, political meetings, and other communal events while living in nearby multifamily longhouses and cultivating crops such as corn and squash. Based on the number of earthworks and the labor required to build and maintain them, the researchers estimate that the civilization's population peaked somewhere between 1.25 million and 3 million people. the findings at a growing body of evidence that the ancient Amazon was home to a large interconnected society with engineered landscapes, not just small isolated villages, as many archaeologists once believed.

40:25Even more remarkably, the surveyed area alone contains roughly as many earthworks as earlier estimates suggested existed across the entire Amazon basin. If future LIDAR surveys uncover similar densities elsewhere, scientists may have to dramatically revise estimates of how many people once lived in the rainforest. Archaeologists still do not know why the civilization disappeared, though the abandonment of these ceremonial centers between 850 and 950 CE coincides with the decline of several other major societies in the Americas, including the Maya. Very interesting. I wonder where else people can scan with LiDAR, find interesting things.

41:06Can I go to Antarctica or something? I don't know. Well, in other news, Gwern is retiring from writing to start a company. Gwern is the anonymous online writer. Is it Gwern.org? Gwern.net. Fantastic blog. He's done, I believe, only one podcast in history with Dwarkesh where he was anonymous. They used a 3D rendered version of him, voice changer to keep him anonymous. This is the website of Goran Bronwyn. He writes about AI, psychology, statistics. He also has written about self-experimentation, Bitcoin, all sorts of different stuff. He wrote an amazing piece about nicotine and all the deep dive there.

41:54It was very fascinating. Very formative when I was starting my nicotine company. It was very interesting. But he's starting a new company. so let's dig into what he wrote let's see can i pull this up where is it i lost it gorn gorn gorn gorn he has a locked account so you got to go request to follow and then timeline app open it back up where is it let me tell you about public.com investing for those to take it seriously stocks options bonds crypto treasuries and more with great customer service um guern okay he's retiring from writing to start a company uh the company's first product for writing his his goal is to design is to uh his his design goal is a 100x increase in productivity what would it take for a ga to make me 100 100x more productive as a writer or thinker why 100x specifically, he says, because too aiming at OOMS is enough to hope to keep up for a while.

43:04And this goal will shatter ineffective Band-Aid-style short-term design proposals. But it doesn't require solving the AI alignment problem or value extrapolation. We do not need to be able to oversee millions of super-intelligent AIs working autonomously for subjective millennia, which is good because a GA can probably not do that. What would that look like for me? Roughly, I think it would look like a GA enabling me to produce one to three worthwhile writings a day, which are 100 % written by it without loss of quality. Very interesting.

43:37Harley Finkelstein:I've heard enough. $100 million. Give them a billion dollars, yeah. No, I agree. It is interesting that we're going back and forth on is writing verifiable? I think it's a little less verifiable. but you know with feedback and different algorithms you can get there but it just feels like the progress in mathematics is accelerating faster than the than the progress in writing yes and also i mean in this scenario it's like maybe it's easier to train a model to write like like one specific person true to to write good generally like that's like much more amorphous what does it mean to like write well versus like oh this is in the style of a particular person.

44:19This is how they would think. It also feels like there needs to be a, uh, sort of like, uh, uh, like a whole agent harness or workflow because writing is not just the last step of instantiating the rules. It's all the research that goes into it, all the different references, the, the, all the different links and things in the way he will put together arguments, the way he structures these things. Um, all of that is, uh, is part of the writing process, even just narrowing down, you know, is this an essay or is this a, you know, shorter post, longer post, all of these different decisions. They can probably be built into a system, but it will take time.

44:55So that's what he's working on. Anyway, let me tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in-store, on mobile, on social, on marketplaces, and now with AI agents. And we have the perfect guest to talk about Shopify. Because we have Harley Finkelstein, the president of Shopify, here with us in the TV and Ultradone. Welcome back to the show, Harley. Congratulations. What a quarter. Insane stock chart. Take us through it. What's going on at Shopify? It is, Jordy John, great to be back here. Great to see you. I don't know if the ad role and me coming on together was coordinated.

45:34Completely coincidence. Complete coincidence. Brilliant.

45:43This was a monster quarter. GMB was$116 billion, which is more than the GDP of half the countries on the planet. Revenue was$2.6 billion. Gross profit was$1.7 billion. And free cash was just over$650 million. Most importantly, I think, is that, I mean, this is obviously Shopify's at a huge scale. And we delivered more than 30 % across the board this quarter. And I think it's like our fifth straight quarter of 30 % or more growth in GMV. So Shopify is doing well. And more importantly, the stores on Shopify, the merchants on Shopify are doing incredibly well. Yeah. And we're really, really proud of us.

46:19Harley Finkelstein:People talk about AI victims. And in this situation, I think AI is a victim of Shopify. I mean, you know, we can get into that. I actually think one of the more interesting parts of the call, and we got into a little bit of it, But just around what is happening in agentic that I think the question I got on the Q1 call was, is this real and is this going to continue? And I tried to answer that question clearly, which is, yeah, it is early. And yes, the volume is small in agentic commerce relative to$116 billion. But the trend is real. I mean, and the growth is real. And funny enough, or surprisingly enough, I should say, the merchants that seem to be benefiting most from Agente Commerce are not the big box stores.

47:04It is the long tail of these specialized independent businesses. And just on Agente alone, traffic was up 3x. Like AI-driven traffic to Shopify stores was up 3x. Orders were up 3x. So not just traffic, but the actual orders themselves. And probably the most interesting is like conversion from our catalog, Shopify catalog, was 2x more than sort of general AI scraping. So one, it is real. Two, it continues to grow. And three, it is disproportionately affecting these smaller merchants. So, you know, consumers are searching for like a standing desk for a small apartment or car seats that fit three across the backbench of a sedan.

47:48And I think that is a structural advantage for small businesses, which happens to be a big part of our merchant base.

47:54Harley Finkelstein:Okay, so yesterday we had the most feared lawyer in the world on our show, John Quinn. He was talking about a sort of current lawsuit between Amazon and Perplexity. Amazon sued Perplexity because Perplexity's bot was going on to Amazon.com, finding listings for users, which I think is something that a lot of people want. but Amazon is not excited about that because they have a big ads business and if bots are crawling all over they can't really serve or they don't have a way to serve ads to bots in an effective way. It feels yesterday I think there was a some type of action by the judge I forget the technical term for it but the judges I don't think the case is fully decided yet but the judge is effectively saying like an agent acting on the user's behalf to do something on a website is wildly different than just general like basically the way that we would historically view bots where like a company could say like we don't want your bots on our website but we're entering into this new paradigm it does feel like shopify is in a really great position here because you guys aren't trying to monetize eyeballs in the same way, and you're not trying to monetize that demand.

49:21Harley Finkelstein:You're just trying to get merchants in the position to sell a lot more products. And so how do you see this sort of struggle evolving on a more global level between consumers, which want to use agents to buy products all over the internet, and these sort of big platforms and legacy shopping platforms that want to control attention on their websites? Look, I think I did a search for myself in June. Bailey, our 10-year-old daughter, we wanted to get her screenless phone to use. And historically, I would have gone to a search engine and would have typed in screenless phone. And I probably would have seen something like Best Buy pop up and they'd have like eight different options from some, you know, some big, big, big companies, electronics companies.

50:15Instead, just using chat on my own, I typed in looking for a screen phone for Bailey because it had so much context, because it knows me so well because of all the conversation we've had over the last two and a half years. It actually pointed me right to the tin can phone. Yeah. Tin Can phone is a shop by merchant. It took me right to the store. I have like five of them. They're amazing. They're amazing. Amazing. I never would have found that. I never would have found that product.

50:41Harley Finkelstein:And this is like a can with a string attached to another can? That's the metaphor. Yeah, that's the metaphor. But it's basically, it's Wi-Fi enabled, screenless phone. It's really cool because I don't want my kids on screens. It's amazing. It works incredibly well. That is because agentic commerce is merit-based. It is not based on who is supplying the most amount of ad dollars. And I think it's still – it's really – it is a much better shopping experience, and I still think it could get better. In fact, I sent you guys some graphics here. Can you pull up – did you have those? Can you pull up the first one, which is just the dress?

51:17Yeah. Okay. So I mentioned earlier – just keep this on for a second. I mentioned earlier that right now, if you look at the conversion of our catalog versus general AI search of scraping, our catalog converts 2x. This is what a scraping might actually show, just simply the dress. Go to the next slide. Okay, so this is what we're actually building. This is these taste-driven attributes, whether these articles, these products, is it formal enough for a wedding? Is the fabric breathable? What's the wrinkle tendency if you're using it to travel? Sneakers, are they suitable for an endurance race? This type of structured values agents can actually read, it means that when someone's saying, I want to buy a dress that's good for a summer wedding or a suit that needs to be formal enough or judging its formalness, this is where it is going.

52:11We announced this at.dev two weeks ago, but this is what we are building with catalog. We are building these incredibly rich product attributes that will make the shopping experience from a discovery perspective way better. And then what we're seeing is ultimately the journey compression is insane. So buyers arrive effectively ready to shop. AI referred sessions land on product pages two and a half times more than simply organic search where they just go to the general page and they're perusing and they're browsing. That's not how it works with with with agentic. Actually, they're arriving right on the product page and they're using shop pay to check out.

52:45So the adoption like this is real and it's showing up in our numbers and both in traffic and orders. And I think Agentic is only going to get bigger from here. And I love the fact that these small niche companies like the Tin Can Phone seem to be the ones benefiting more than others. In fact, if you actually just compare 75 % of all AI-attributed purchases in Q2 on Shopify or from outside our top 100 categories. So it is not your I want yoga pants. It is, I want, I don't know, reef safe sunscreen that doesn't leave a white cast on my black leather, whatever it is. It's that type of stuff. And I think that means our merchants, I think, will disproportionately benefit from this.

53:30Because AI is still a little slow, there's a little bit of a cost to going down the agentic search path. I would expect, I have been expecting that we will see disproportional agentic purchases for higher ticket items, more considered purchases. So you're doing research on a car. Yeah, go lead 5.6 pro cook on all your features and figure out everything. But you're ordering just another soda. You're not going to go and down the agentic system for that. at least in the short term. Are you seeing any of that? Do you expect that to be a factor? Yes, so for sure. Products that require more deliberate purchasing intent and more research benefit from it too.

54:22But I would tell you to expand your perspective there, it's not just, it's also products that are very much for, it's a particular product for a particular use case. It is not necessarily, I want to buy a t-shirt. It is I want to buy a high-end T-shirt with a GSM count that is more than – that keeps me warm like when I go out at night in New York City where it knows your location. It knows the weather in your location. And now it actually pulls from your entire conversation history and says you probably want a James Purse T-shirt, which is my favorite T-shirt.

54:56Harley Finkelstein:Yeah, and the other thing I've been doing, I think you maybe discount consumers' willingness to wait or do things in order to save money, right? You don't strike me as the kind of guy that ever was doing credit card point maximization or chasing coupon codes around the internet. But a lot of people are, and a lot of people want to buy something, and then they're thinking, okay, how can I get the best price on this? I saw a pair of sneakers a couple of days ago that I really liked. Yeah. I found them for sale and I just took a screenshot of the, of the product page. And I said, find me this sneaker in my size at the best price.

55:41Oh,

55:41Harley Finkelstein:interesting. And it went and found three different options. So you might land on a shop. And what was interesting is it actually took me to stock X and it was like, Hey, by the way, this is selling for below retail stock X. So you should just get it here. And I ended up saving. whatever 50 something dollars right um and so that even though that was like probably took you know 30 seconds sure like 30 seconds to save 50 bucks like a pretty good trade right yeah i mean especially if you have some sort of like reminder like hey these these are the sneakers i want they're a little too expensive for me right now i would love you to let me know if something else comes up and now it basically runs a query uh you know every couple of days to see whether or not it's being discounted as well.

56:24So even though that's not a deep research product, you actually are doing deep research on the pricing of it as well. But the other thing also that came out of the call today, which I don't know if it necessarily got picked up and it's worth sharing is we've talked about Sidekick on the show before. And I said, ultimately, the goal for Sidekick is that it makes merchants more effective, more efficient, helping them reach their early milestones faster or make better decisions. All the things that when I was selling t-shirts on Shopify or you know, you were selling nicotine, patched dirty, like, you know, that all that stuff, the silly ad spend we made on performance marketing that just didn't convert that ultimately the goal was to to alleviate all that as well.

57:08And what we actually talked about today was when you actually look at the onboarding guidance, when you're just setting up a store on Sidekick on Shopify, we actually saw an 8 % increase in new merchants reaching their fifth order within 15 days. That may not seem like a big deal, but on a massive base of Shopify, new merchants that are constantly coming on, the fact that they're getting to their fifth sale faster is remarkable. And so we think not only is that going to get better for new merchants, but then as merchants also grow in scale, it's not going to be about setup anymore. They're going to be using Sidekick to become more like – to do more things like analytics or reporting or to find new opportunities.

57:46And therefore, all the mistakes that we all made when we were building our own e-commerce businesses 10 years ago simply go away. And so you'll have more people starting and more people growing faster. And as that happens, Psyche just keeps getting smarter and smarter. Yeah, so break that down more for someone who's actually running a Shopify store today. hey, how close are we to the world where you show up, you have a budget, and Shopify can sort of agentically, for me, take a budget and spread it across Facebook and AppLovin and Google Ads and load balance everything, the stuff that you wind up hiring an agency for sometimes, and then maybe they get a little clocked out, then you hire a person and it becomes a job.

58:31And a lot of small merchants might not have that capacity, but you're picking up pennies off the floor if you're doing it properly. Yeah, so that's already happening to some merchants some of the time. We have something called Autopilot, which we announced in our last edition, which effectively allows you to give us your ad budget, and we will go and find the most effective uses for you. Yep. But in terms of like, so there's different stages of usage on Sidekick. For new merchants, the first job is getting that first sale and then the next sale. And you guys know this, the first few orders, it creates incredible momentum.

59:04It gives you confidence. You become an entrepreneur when you have your fifth sale and it's not your mom or your neighbor. It's like some stranger who found your business. As the business then grows, we've noticed obviously setup accounts go down. Now they're actually using it to do things like reporting analytics. And over time, what we're doing, Sidekick is actually proposing things like bundles, for example, or SEO optimization driven product creation. Then we have something called Pulse that lives inside Sidekick, which is effectively saying this is what's happening right now on your business.

59:34And these are some proactive ideas you should consider, like useful recommendations. You can accept them. And if you do, Sidekick will implement them for you. It could be something as simple as moving your product page around or moving the buy button, you know, the shop pay button further up the stack a little bit. But this idea is we want to deliver the best practices across, you know, hundreds of billions of dollars of GMV everywhere that goes to Shopify to anyone using the platform. And then they can decide on their own if they want it. The reason that this is possible is because it understands, unlike any other AI agent, it understands a merchant's products, customers, transactions, history.

1:00:13And that context allows us to give this incredible grounded advice in the reality of a merchant's business. And it's an advantage, actually, that just continues to compound. As more people use Sidekick, more merchants are able to get value from it as well. And I think the path forward is Sidekick will take on more of the work, not just with ads, but also selecting a 3PL or figuring out, doing line skipping to figure out how to have the best deal on shipping. But this is really – I mean it's not just science fiction. This is really happening where you can put more and more of your business on autopilot.

1:00:45It's great. What a great time to start a company.

1:00:47Harley Finkelstein:In two minutes, I want you to talk about the year 2035. And let's assume that you have a widespread use of robotics that can make anything that you could possibly imagine spread out all over the world that can make custom products or one-off products or small batches of products. I have so many different products in my camera roll that I just made with ChatGPT images. and I would love to be able to make like just five of them or ten of them or make one for myself and throw up a product listing, right? It could be something like sculptural or a piece of furniture or even something in consumer packaged goods.

1:01:30And I personally have a theory that as robotics bring down the cost of making anything,

1:01:37Harley Finkelstein:that there's so many people on the planet that have ideas that would create a Shopify store And if they could make, you know, 10 or 20 or 50 or 100 of products really easily, they would become, you know, basically entrepreneurs overnight. You're just about reducing that friction all the way. So, like, 30 years ago, it was quite a bit harder to sell products online. It was possible. But, again, it was, like, clunky. Now it's easy to sell products online. it's still incredibly hard to make high quality products, especially as somebody that's a new entrepreneur that doesn't have a lot of capital.

1:02:14Harley Finkelstein:So how are you thinking about the, like the year 2035, assuming, you know, a takeoff in, in robotics? Can you guys pull up gamers.town gamers, like the gamer, like someone who's a gamer dot T O W N. Just pull that up for a second. I want to show you something. I just saw this before the call, but this is fascinating. This is very cool. Is this on Shopify? Yeah, so I get it. This is so interesting. So can you pull it up here? Okay, okay. So this is an entrepreneur who basically took our catalog API. So a billion products, right? Like Aloe is in here, Viore is in here, Mattel is in here, Cannon Goose is in here, everything is in here.

1:02:54And they also looked at every single store that, obviously, this entrepreneur is a gamer, and said, I want to pick across a billion products the most relevant, interesting merchandise for all these games like Minecraft. And effectively, they took the catalog API as the access point, and now these developers can effectively just build their own stores with other people's products. So I think, and it's beautifully done. You can see it's like this great constellation type of shopping experience. So the fact that even before the year 2035, where maybe robotics help us with manufacturing, I actually think now that allowing anyone with some ambition to take the Shopify catalog and create their own interesting ideas, in this case, it's for gaming, but there's so many other ones as well.

1:03:41I think that will happen not in the next 10 years. I think that will happen in the next 10 days. This is the first one I found. I'm getting a full Counter-Strike costume for sure. This site is amazing. Thank you so much for coming on the show. Very cool. Always fantastic chatting about all this stuff. And congratulations. What a quarter. Great stuff. We'll talk to you soon. Thank you guys for the best. Goodbye. Let me tell you about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI. Own the data platform that powers it. Our next guest is John Giamatteo from BlackBerry.

1:04:19He is the CEO, and we're very excited to talk to him because BlackBerry has evolved many times. And the company's been on a fantastic run. Thank you so much for taking the time to come chat with us. Would you mind kicking us off with a little bit of an introduction on yourself? And I want to sort of set the table on the shape of the business now for anyone who's been living under a data center and doesn't understand the current shape of the business. Hey, thanks for having me on the program, guys. It's an honor to come out and talk a little bit about BlackBerry, 42-year-old iconic Canadian company.

1:04:57A lot of deep roots in technology and a lot of different transformations over those 42 years. You know, so many people know us from the old smartphone days and the devices, but our transformation and pivot into embedded software. And we could talk a lot more about that, the things that we're powering around the world and secure communications where we provide mission critical mission, critical communication capabilities. that's really what the new BlackBerry is all about. Those two categories in terms of serving the market. Yeah. On automotive software, obviously, it's a very high stakes, in some cases, life and death scenario based on the security of that software.

1:05:49At the same time, the theme of 2026 has been cybersecurity with advanced AI models showing really powerful capabilities there. What have automotive leaders been coming to you to solve? How are they thinking about solving the problem of keeping automotive software secure in the modern era? Yeah, great question, especially now as there's so much more technology going into the vehicles than ever before. The need for a rock solid, mission critical, secure operating system. And when you think about BlackBerry and our QNX platform, think of it really as the operating system of the car, the windows of the car, the way windows is to the PCs around the world.

1:06:36The QMX operating system is really the operating system that powers all things that go on in the car, particularly on safety-critical types of functionalities, from digital cockpit to ADAS to body control. All of these different domains of technology that goes into the car and the CPUs that power them run on top of a BlackBerry operating system. So we're proud to be powering over 275 million cars around the world today and growing millions over the years. And we got a tremendous amount of progress all around the world in Germany, Japan, Korea, obviously here in North America. China is one of our fastest growing segments.

1:07:28So a lot of exciting things as more technology gets introduced to the vehicle. How has you as a CEO been grappling with this sort of story of token maxing? You get these new models. They're incredibly powerful. You have the opportunity to deploy them across the entire workforce. Rewrite every line of code if you want. At the same time, this comes with a cost. You have a serious business to run with a bottom line to manage. How have you been deploying AI internally in a way that maintains safety and security, but also doesn't balloon cost, but does drive efficiency? Great, great, great question.

1:08:10And it's something we think about every single day. We are, the way I would describe it, we're pretty pragmatic about AI and how we use it. Because like you said, the applications, the car, failure is not an option in a vehicle. So we've got to be really careful about how widely we deploy AI as we build and enhance our existing operating systems and new products. So, you know, one of the things that's in our industry, you have to meet the highest levels of certifications. And in the car industry, there's a particular certification of ISO 26262. Very high standard. You can't even think about being deployed into a vehicle unless you deploy that.

1:09:00For an AI agent to try to meet that type of standard, it's a non-trivial type of exercise. So we are thoughtful about it. We deploy AI to help us maybe test our products. We deploy it to help service our customers. We deploy it in our go-to-market activities. But in the actual core development, the code that goes in it, we're very thoughtful and limited about because of those certifications and the importance of it. BlackBerry has been on such a fascinating historical journey. What is the area or cultural value or discipline or source, center of excellence that you think has remained the most true throughout all of that?

1:09:52What is the narrative arc that actually brings coherence to a company that's operated in multiple markets over multiple decades? Yeah, yeah. Yeah, that's, you know, for a 42-year-old company that has evolved in so many different ways, you know, what I always tell people, BlackBerry has always been about security, trust, and innovation. And even though the products and services that we sell to our customers today can't be held in the palm of your hand the way they might have been with a device, that security, trust and innovation of everything that we do from the products that we deploy to the services that we provide to the support that we provide governments around the world.

1:10:38You know, eight out of the top 10 G10 countries around the world are still deploying BlackBerry's secure communications capabilities from mission critical types of applications. So if it needs to be safe, secure and certified, that's really our value proposition. Our our 2000 plus employees and contractors around the world are big hubs in Ottawa and Waterloo, Canada is where a lot of this technology gets developed. That's the pride of what we stand for. And those are the values of what we stand for in terms of security, trust and innovation. So I imagine that although different technological solutions, different products have evolved, there's always been a really, really solid operating bar for building trust with customers, go to market, sales, actually partnering really deeply with the companies that you work with.

1:11:41How has that informed your M &A strategy going forward? Do you have a particular philosophy of how you can bring a new product or company into the fold successfully? Yeah, yeah. That's one of the things we do now that I don't know, we did it to a degree in the past, but I think we've we've really doubled down on it over the last few years is we really work very closely with the ecosystem to bring a lot of this capability to the market. So, for instance, our biggest partners in the automotive sector, NVIDIA, Qualcomm, ARM, Texas Instruments, Intel, big tier one operators like Bosch, take our technology, integrate it into their solution and deliver it to their customer.

1:12:34So whether it's the chip makers, the operating system maker like us, tier one providers or the end customer like the auto manufacturer, having a comprehensive ecosystem approach has really been the way that we've we've gone about it. So when we think about M &A, we look at what piece of the puzzle might be missing, and we kind of put it through that filter. But I got to be honest with you. It's right now with the organic growth that we're seeing, the opportunities in physical AI with robots and medical instruments, there's a lot of organic kind of runway for us to enjoy. So from an M &A perspective, if it fills a gap, if it helps us accelerate a certain category, I think we'll absolutely think about it.

1:13:31But right now, you know, the runway that we have, we have a$950 million backlog in our UNX operating system that's going to be deployed over the next five, seven years. So these are the types of things we think about when we think about M &A.

1:13:47Harley Finkelstein:uh jordy i want you to make a guess for us considering uh how embedded blackberry is with you know automotive manufacturers and and uh other suppliers uh what at uh at what point do you think that uh every auto manufacturer in the united states that sells let's say more than 5 000 cars will have a functioning autopilot out of the box in every single one of their vehicles? What year?

1:14:24Interesting. You know, there's 90 million cars get made a year. 75 % of those cars have basic fundamental technology, like the media type of software that operates. 25 % have more advanced technology. Along the lines of what I talked about. So taking it to the next level around an autopilot and, you know, autonomous applications. And, you know, I think that's well into, you know, the next decade of the 2030s. Early innings, there's a lot more technology going on. And we'll see more and more of it get deployed. but I do think it's a little bit more of a journey than an instant thing that we're going to see in the next, you know, 18 to 24 months.

1:15:15Harley Finkelstein:Yeah, yeah. No, that's what I was, my estimate would be 2035. Yeah. What are you seeing on the industrial automation side, the industrial either robotics or just industrial equipment side? I mean, we've seen Caterpillar is doing really well and it just feels like America is starting to re-industrialize. There's reshoring. There's also the data center build out, the AI build out. There's just more construction happening, it feels like. Is that showing up in your business? Are you seeing that with your partners? What is the nature of the changing landscape of industrial autonomy? Absolutely. It's the general embedded space for us, which includes things like industrial automation, robotics, robotics, medical instrumentation, surgical robotic arms that go on in the operating rooms today.

1:16:12These are all, while we've got a great footprint and presence and growth trajectory on the automotive side, these other use cases around robotics. So, for instance, AMI, AMR rather, automated mobile robots. These are robots that are doing pick, pack, and ship in B-scale manufacturing. Those are applications that would need a QNX operating system based on the precision and the safety levels of things required. So it's the fastest growing segment. It's our fastest growing pipeline. Oh, wow. And we think it's going to help us fuel that next generation of growth here at BlackBerry. Last question, and we'll let you get back to your busy day.

1:16:59I'm interested to know how you've processed the story of Palantir's forward deployed engineering strategy. This, in some ways, has always existed. There's always been deep partnerships with enterprise companies where employees of one company might be embedded in another organization as they're deploying a piece of software. Palantir took it sort of to the extreme. How have you thought about deeply integrating your actual workforce and solutions engineering into the companies that you're working with? Yeah, I follow the Palantir. It's obviously working for them. They're just killing it right now.

1:17:42And it helps them, I think, with their cycles and bringing more products and solutions to the market. I think for us, we're probably a little bit slower to adopt something like that, if I'm honest. We're a little bit more, again, back to some of the safety certifications. There's dozens of them all around the world that we have to comply with. We take that really seriously. And as a result, we try to do as much of that in-house as we possibly can. And we got to a point where scale was really affecting our ability to get some of these solutions out there in a timely manner. I think we'd take a look at it.

1:18:23But at this point in time, it's a little bit more of a made in Canada approach. I love this. Made in Canada. Well, thank you so much for taking the time to come chat with us. Great to meet you. Congratulations on all the recent success. Really impressive. And really appreciate the time chatting with you. Have a great day. Thanks so much for having me on. We'll talk to you soon. Have a good one. Goodbye. Let me tell you about Cisco. Critical infrastructure for the AI era unlocks seamless real-time experiences and new value with Cisco.

1:18:51Harley Finkelstein:We missed an opportunity to pitch, John, on rebranding back to Research in Motion. Oh, yeah. Because that is one of the best names in history. It is. And it feels even more fitting for the current business. Maybe, maybe. But everyone knows BlackBerry. David in the chat wanted to ask, has BlackBerry thought of renaming their company? It's now B2B a totally different offering than what people. I do think the current name probably holds them back. Really? I think it accelerates them a ton because even though BlackBerry was popular in consumer, in enterprise it was known as a much deeper solution that you would buy and deploy across your entire organization.

1:19:38Harley Finkelstein:But I would say that the next generation of enterprise buyers associate the BlackBerry name with just losing smartphone marketing. Oh, yeah, yeah, yeah, sure. Yeah. Oh, well. Well, our next guest is in the waiting room already. We have Jamie Nellar from Circle. He's been on the show before, and he's back in the TV bin. How are you doing, Jamie? Great to see you. Great to see you. Absolutely. Thank you so much for taking the time to come chat with us. Give us the update on Circle. Give us the update on stablecoins broadly. I want to go into sort of the knock-on effects, how the AI era is helping.

1:20:15But let's start with the state of the business. Yeah, absolutely. We reported our Q2 earnings today. I won't go through all the numbers. But I think one of the things I talked about at the beginning of that is we're going through a pretty interesting moment right now. and that's really being driven by the fact that stablecoins are now becoming part of the actual dollar financial system. So the Genius Act passed, this new law is going into effect, and really the big trend that we're seeing is that all of the traditional players in the financial system, from the big technology companies to the payment companies to financial infrastructure companies, they're all starting to build on this.

1:20:54And so we're really poised, obviously, for that moment. USDC is today overwhelmingly the most widely used regulated digital dollar in the world. Visa reported at the end of June 70 % of real-world stablecoin transactions happening with USDC. We saw record numbers in terms of minting and redemption of USDC. And so it's sort of the backdrop is like stablecoins are working their way into kind of everything. And at the beginning of next year, it's sort of officially part of the financial system. So, you know, that's been going on. And I think the other big shift that's been happening, and we talked about this earlier as well, is that we're kind of exiting the, you know, what I like to think of as like the speculative phase of crypto.

1:21:39And we're moving into like a more mainstream phase. And actually, there's a tipping point data point that happened last week, which is that close to 75 percent of the volume traded on Hyperliquid was actually real world assets, tokenized stocks, tokenized commodities, other things. And it's no longer people who are like, you know, kind of going leveraged betting on Bitcoin. It's actually this infrastructure being used for traditional assets. And that's obviously, I think, a pretty interesting moment in time as well. Interesting. Wow. How have you been processing the moves in the traditional finance rails world?

1:22:16There's this talk of a new payments network that might compete with Visa. A bunch of the banks are working on this. Are they fighting the last war? Is this not a threat to you? Is this irrelevant? Is this maybe beneficial to you? Yeah, I mean, like, I think our philosophy, you know, basically for the last 13 years is like all of the financial systems going to be moving to being run by software on the Internet. It's going to run on open networks. It's going to be open source infrastructure. It's going to be machines that are that are running that. And as obviously we see this convergence with AI, it's actually machines building machines that are running the financial system on the Internet.

1:22:54So I think that's looking backward. I think looking forward, this is about programmable money all on open infrastructure. These new software created money like stable coins itself is actually software generated money. So I think that's where the world is going and where the velocity is going to happen as we go forward. how are you thinking about the i mean software created money uh the uh the ai agents there's been a whole bunch of different sort of concepts we were just talking to harley at shopify about the uh like the growth in agentic shopping but it's mostly research and then it's handed off at shopify um and i'm wondering uh do you have any visibility into where the near-term killer use case within a very broad AI world that can do everything from probably figure out how to mail you a gold bar if you try hard enough to microtransaction streaming at USDC.

1:23:57There's such a broad remit there. Totally. I mean, this is a huge focus for us. I know we talked about it in the past, but right now, there's these agent payment protocols like X402 and MPP, machine payments protocol. Right now, over 99 % of the transactions that are happening with USDC, are happening on blockchains. So we're leaning pretty hard into that. And lots of companies from Amazon to Cloudflare to Shopify to Stripe and others are building on this. And so that's sort of happening. But I think our conceptual model, and I actually talked about this earlier today, we have a whole roadmap we're publishing on this, is we actually think that the real opportunity is not necessarily agents that are out doing shopping.

1:24:38It's actually agents as actual units of labor. And so really looking at agents as cognitive workers and looking at agents that are actually the supply side of the agentic economy, which is essentially agents that you can discover, that other agents can discover, that can provide and conduct work. And really, that's a whole infrastructure that has to be developed. And that includes things like identity and reputation and marketplaces where these can be discovered. We've launched a bunch of pieces of this. There's over 900 services now that are available. It's across a huge range of things on the internet that you can now do with agents and agentic payments.

1:25:18And so I think from my perspective, the agentic economy in some respects is about the transformation of cognitive work into agentic execution and then how all that gets orchestrated, not just inside a firm. There's lots of people talking about agent harnesses and orchestration, but it's like cross-firm boundaries. If I'm building a startup and I've got two or three people and I want to employ agents to actually work on my behalf and build things, we need an infrastructure to do that. And so I think it's actually – that's the looking forward model. The looking backward model is like, yeah, there's research and shopping.

1:25:49The looking forward model is there's work that's executed. It's a labor market for agents that has to be established. That's what we're excited about.

1:25:57Harley Finkelstein:What does stablecoin adoption actually look like in a large financial institution? Because these institutions are using them in some ways, but there's all these different flows of money. And I imagine that when you're talking to CEOs or your BD team is talking to management teams or individuals at these companies, you're probably identifying different flows and saying, like, this is a good use case for stablecoins. But I imagine there's a lot of basically payment flows that have been set up and maybe running for like 20 years. And there's a system that works. And so like our how much of adoption is like net new flows that are being set up stable coin native versus like historical, you know, basically payment processes that are getting, you know, turned over to stables.

1:26:52Yeah. I mean, it's really interesting. I mean, I think we're sort of seeing this like incremental upgrading that's happening. And, you know, you see this with like pretty much every payment company and every neobank in the world now has some kind of stablecoin strategy. They're integrating it into how they can receive pay-ins or make payouts. You're seeing that with, you know, companies like Meta who are, you know, like making payouts to creators in far-flung places around the world. you're seeing global treasurers who are starting to realize like, wait a minute, I can actually move my capital around all around the world 24-7.

1:27:24I can make payments into hard-to-reach places a little bit more efficiently. We're seeing big ERP systems and treasury management systems adding stablecoin rails as an additional option. So I think we're basically seeing the rails get added in all over the place. And a lot of where it's starting is sort of where it's been hard, you know, internationally. cross-border, and where there's demand for dollars, right? I think there's also this sort of shift that's taking place. Digital dollars are very attractive. People want to hold them. And so you see more demand side, people who want to receive funds this way and want to keep it in that as well.

1:28:01And then within the financial infrastructure, the financial markets, et cetera, I mean, basically, what we've seen happen is digital asset markets got all the big Wall street trading firms beginning to trade crypto. And then they learned about 24-7, 365 collateral and the ability to move and settle stuff instantly around the world. Now that's starting to penetrate into traditional markets. And so you're seeing more and more the DTCC, the big kind of clearing infrastructure for securities, standing up infrastructure where the cash settlement is actually happening with USDC because it's sort of an on-chain system.

1:28:38So we're seeing it work its way into the core of the financial system, and we're seeing it out at the periphery of the way global payments happen. My view is obviously this is a superior medium of exchange. It's programmable money, and the unit economics are better, the user experience is better. So eventually, we think this will obviously absorb, sort of like digital media slowly absorbed and became a much, much larger thing compared to non-digital media. I feel like you're using infrastructure, financial infrastructure, in sort of the ephemeral sense and the metaphorical sense, not in the literal, like we need server racks that are running the rails, right?

1:29:15Yeah. But I'm just interested, even if it isn't the most crucible moment for the business, what does the physical infrastructure side of the business look like? Is that fairly turnkey at this point? Has most of that engineering been done or is there still work to be done to keep the network even moving faster and even more uptime? It's a great question. So, I mean, so two things. So, one is like Circle is a software company, basically. Like, we're a bunch of people with laptops. We make software. We deploy it in clouds. It runs and it powers like all this stuff. But critically, though, you know, a lot of this money and the smart contracts, they run on these blockchain network operating systems.

1:29:58So, you do have this new layer, which is these blockchain network operating systems. And actually today, we announced with ARK, which is the new blockchain network operating system that we've been building and launching, that essentially the actual physical infrastructure, which are often called the validators, these are the firms that run the nodes that actually run the compute, run the transactions, store the data. The validators are actually, we had a number of them around 10 or 11, which included the largest asset manager in the world, BlackRock, the two largest retail payment systems, Visa and MasterCard, the largest securities clearing and custody firm, DTCC, Global Banks, a whole bunch of others basically are running that infrastructure alongside us.

1:30:40That is actual infrastructure. These are data centers. These data centers have SLAs. They have uptime requirements. They have information security requirements. And so these new networks, these new network computers, which are these blockchain network operating systems, you have to run that infrastructure. And so we've actually built a model where there's like very high service levels where actually that infrastructure can kind of withstand the scrutiny of like a central bank that's saying, is this real money? Are these real financial market transactions that are happening and kind of meet the kind of assurance that is required to do that?

1:31:13And that is going to continue continually evolve. We're pumping a lot of transaction volume and stuff through this, but with Agentic, with money velocity growing much, much larger, we're going to need to continue to see that scale. But it actually is like there is a physical infrastructure layer to support these new operating systems as well.

1:31:31Harley Finkelstein:What concerns do you have for the blockchain and crypto industry broadly with new powerful AI models coming online that have cyber capabilities open and closed? I imagine that you guys have been super, super on top of this as Circle, but there's so many small companies and blockchain organizations that have already been the target of various cyber attacks. There was some CEO that put 100 Bitcoin in a wallet and said, like, AI Labs, go hack me. Come get it. It was like a challenge. This is, you know, crypto is cryptography, right? And so at the end of the day, like, the whole concept is in code we trust.

1:32:13I mean, that's literally the motto of crypto is like we're trusting cryptography, we're entrusting mathematics and smart contracts and the execution environments are very, very sensitive types of code because they actually intermediate money. And so it is like high stakes. It is super high stakes. We saw, you know, a hardware wallet that had not actually, you know, had not used the best encryption methodology for its seed phrases actually has been drained. People thought it was in a hardware wallet. It's been drained over the last week over$100 million of losses. So it's a very serious issue. We're taking it very, very seriously, not just for our own operations, but as we bring ARK online and we put out this new operating system for economic activity, it needs to be hardened in a different way.

1:33:01The attack surface, the attack vectors, the kind of security that, say, I'm a startup builder, I'm developing an app that's going to deploy on one of these networks, what do they need to be able to have some level of assurance that the code that they're deploying on these new types of network operating systems is going to work and be secure in light of these kind of new cyber capabilities. So it's like a whole new problem space. We're very actively working on that. Like it's now like a critical part of our engineering and what we think actually is needed for builders too. Like builders, you know, have to kind of take this into account and And it's a very different world.

1:33:37And that's changed. Yeah.

1:33:38Harley Finkelstein:And I do think that, you know, the industry's overall problems effectively become your problem as a regular Bitcoin provider. Because it's just, you know, the more hacks that you have, the more trust issues that people have with the technology. Even if it's on a completely different architecture. Yeah, completely different chain. It could have even been a social engineering hack. Yeah. It's still related badly. I mean, we've seen, you know, a real increase in attacks. And even the social engineering attacks are often AI orchestrated now because, you know, it used to be you'd get these shitty emails that, like, had broken English.

1:34:16But now the emails are actually well written and they have a better contextual understanding of who they're targeting. And so everything is just a different risk surface. And, yeah, we think about the whole of the ecosystem, not just our own infrastructure, because we kind of have this broader role that we play. If I go back to sort of the early Bitcoin days, there were a fight, the small blockers versus the big blockers. I don't really remember all these details, but the Ethereum wing, you know, all these different people with really, really strong philosophical differences. Yeah. Huge economic interests at stake.

1:34:55It feels like over the last 12, 18, 24 months, the crypto community has sort of come together loosely and there's been some regulation that's been laid out. What do you think the AI industry can learn from the journey that the crypto industry has been on from a regulatory perspective? Yeah, I mean, look, I think, you know, monetary infrastructure, financial infrastructure has always been pretty heavily regulated. I think, you know, we've always tried to walk the line between, you know, open permissionless infrastructure, open source technology. All this stuff is built around those themes and that thesis that's fundamental to the DNA of crypto is open networks, open protocols, open source software, all of that, doing things out in the open.

1:35:51And so that's just sort of philosophically there. I actually think it ultimately helps from a regulatory perspective, the more the sort of more open that you are, the more that everyone can kind of see what's going on. I think, you know, you know, so so I think that's certainly something that can be learned. And that has to do with this sort of open weights discussion vis-a-vis, you know, kind of kind of labs and other things like that. But, you know, I have my own views on on AI and regulation. And I think, you know, very clearly, like, these are foundational utilities for society. And they're now cyber weapons, if not other types of weapons.

1:36:30And so the regulation, there's going to have to be regulation and supervision and registration and other stuff because the stakes are so high for society right now. Yeah, it's interesting that you mentioned the openness being a, like, it sort of correlated with, like, a bigger tent. And we saw that with Jensen from NVIDIA putting out that open letter and getting so many names to sign on in a way that that has not happened with other little regulated ideas of, oh, should this state law or federal preemption, all those little things that the labs have been doing. Jensen was able to just come out with something that was very broad.

1:37:08Some people disagreed. Some people didn't. But a lot of people sort of were like, OK, we're going to band together for this particular vision, which is interesting. Yeah, it does correlate to what you do. just said. Love it. Definitely. Well, thank you so much for taking the time to come chat. Of course. Awesome progress. Always great hanging. We'll talk to you soon. Yeah, great to see you. Great to see you guys too. Thank you. Cheers. Cheers. Goodbye. Let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. Just do it. A couple quick hits before we bring in our next guest.

1:37:39Google, we missed it with the Demis news. That's obviously steamrolling the timeline, but there's also a story in Business insider, Google is in talks to buy a coding agent startup called Mechanize for$1.5 billion. Do you know that?

1:37:53Harley Finkelstein:Yeah. So it's an RL environment company. So they basically build a bunch of these environments and sell them to labs. Yeah. So Sean Kai, Sai says the first RL environment, major acquisition outcome, Mechanize for$1.5 billion, makes a lot of sense when it's been in the data discourse for a long time that GDM's synthetic capabilities weren't up to OpenAI and Anthropics. Many of the recent departures from GDM and the other labs cited this. Certainly mechanized as one of the four players in RL environment markets that I would even consider to have some modicum of good synthetic data scaling. And so, you know, identifying a problem with the organization, with the build out, going and open up the checkbook to the tune of 1.5 billion to potentially solve it.

1:38:39Still in the rumor phase. the other the other news we didn't really get to yet is ilia satskiver's company safe super intelligence is allegedly rumored to maybe be releasing a model in august that's this month and chubby here says i took a look at their website they stated very clearly their only goal their only mission is to develop super intelligence that's their sole focus in that sense this model release will be far more than just another model release in some form they They must have achieved super intelligence if we take their statements seriously. So very interesting to see what happens there.

1:39:16If it's pressure is on or if it's something that's just sort of sits within all the other benchmarks. But we'll we'll see. Good luck. Yeah. I mean, this was like very briefly touched on in the Gavin Baker and wrestling best episode. Yeah. So it's like unclear. You know, we'll see what actually. Yeah. It's not like Ilya came out and said, like, we're launching. Get ready. they did announce some fundraising news with nvidia and uh he did say that it's time to scale the the sort of the age of research might be over and so we're thinking that maybe a product is coming maybe a demo maybe some evaluations maybe some benchmarks we love benchmarks maybe we'll be able to move the goalposts because i want super duper intelligence i'm not settling for that's

1:40:01Harley Finkelstein:right intelligence well let's not keep our next guest waiting too long yes let's bring in mckenna Lindsay Burnett from Ambrook. Back on the show. Welcome back to the show. Lindsay, how are you doing? Good. How are you guys doing? We're doing fantastic. Congratulations. Give us the news. What happened? So excited to announce that we just raised$30 million Series E. Congratulations. We're lucky groom. Yeah. Locky groom. What a great partner here. Jordy? I think it's been somewhere between six and 12 months since the last time you were on the show. Maybe on the shorter end. Talk about the progress.

1:40:44Harley Finkelstein:What have you learned about the state of American agriculture, the state of the industry that you're operating in? The physical economy. Have you narrowed the scope of the go-to-market motion or have you actually broadened it? Yeah, so since we last chatted, we were last summer around 2 ,500 businesses using Ambrook. Now we have over 8 ,000 businesses using Ambrook. And a lot of them are in agriculture. But actually, we raised the B in order to start to expand to other verticals in the real economy. Because we started to see a lot of natural adoption. with we have over a thousand truckers on the platform, general contractors, property managers, nonprofits.

1:41:29Ag is super diverse. So a lot of those customers started pulling us into their side businesses or their neighbor's businesses or family members. So that's a big part of the focus. Are customers excited about potentially not needing to work with like some behemoth legacy an ERP system that might be less agile in the AI era versus maybe you have a smaller solution now, but if you just look at the growth rate of the technology and your team and your product, you're just going to naturally solve a lot of problems that come up over the next few years for them. Yeah, I think our customers are already seeing that Ambrick is a lot more tailored to the complexity of their operations.

1:42:12So there's kind of this messy middle that hasn't been venture backable for a long time, which is if you're a business that's too complex for QuickBooks or spreadsheets, but you don't have the team or resources to onboard to a Sage or NetSuite. You know, I would say probably about, we estimate about 25 % to 30 % of America, American small businesses kind of fall in that messy middle where they're multi P &L, they have complex balance sheets, but they don't, you know, I think there's, there's some talking X about like building for finance teams of one, we're building for finance teams of zero. And that is bringing AI CFOs to folks where we might actually be their first business software.

1:42:53Like 50 % of our customers are moving off a pen and paper. What's needed to build and to help them leverage AI is a pretty different way of building than if you're building for a professional desk worker. Okay, so historically there was fintech companies that would make vertical solutions for things like construction and agriculture and trucking.

1:43:16Harley Finkelstein:And I'm sure you're now competing with some of these players. How much is AI allowing you guys as Ambrook to go into these new verticals and create the specific features or workflows that are more relevant within these sub-industries? Yeah, it's making it a lot easier to come with the right context layer, I think, for what folks need. So you can think of, I think, Amberg is not vertical SaaS. We're more vertical go-to-market. So under the hood, we actually look a lot more like a much more user-friendly next suite or stage. 30 % of our customers access Amberg exclusively via our mobile app. We definitely have the best mobile app in the industry when it comes to accounting software.

1:44:01because we had to build for users that were not by their desk most of the time. We kind of call it the deskless economy. And so there's a lot of workflows that we had to build for. But you can kind of think of the ERP, which we spent the past couple of years building, as that agentic context layer. Like if you ever had to build agents, the first thing you have to do is actually clean up all the underlying data. That's what we've done really well. And now we're building so that both that whole data layer is both human legible and agent legible. And those agents make it a lot easier to personalize or to otherwise like speak the language of someone's business, even if the underlying components are primitives are actually quite similar industry to industry, which is why we've been able to expand so quickly without having to like build a lot of custom features for, for these different verticals.

1:44:50Harley Finkelstein:What is sentiment around AI for your customers? How do they feel about AI generally? I noticed on the website you guys are certainly not leading the here's the new hot tool that's going to revolutionize your business with AI, even though you're using it under the hood, I'm sure, in a bunch of different ways. But how do they feel about it? Do they feel like it's a useful tool? Are they generally against it? What's your read? I think our customers are generally open to it. I think there's just a lot of really healthy skepticism about whether AI is actually useful, which is actually useful to them as a business owner, which is actually quite similar to the conversation I think is happening on X right now, which is you have some folks that have been able to really understand how to build, but it requires a lot of infrastructure to be able to get there.

1:45:45And our customers don't have teams of engineers to be able to, you know, build all the things that they need in order to actually build their first agents. They're not going to be rolling their own, you know, agents on Eve or something like that. So there's there's a lot of work that we're doing to kind of meet people where they are. But I actually have been talking to a lot of our customers, just asking them about general sentiment toward AI. And there is just a broad openness. It is just a little bit.

1:46:13Harley Finkelstein:Yeah, it's more like show me, show me. Yeah, exactly. And there's a lot of things where I think if you build it right, they will come. But a lot of people don't know how to build it right, I would say. Totally. Parker Conrad, when he launched Rippling, had this idea of the compound startup, this idea that he had to build multiple products and not a single point solution. Do you like the term compound startup? Do you think it applies here? Do you think it's a relevant philosophy for the modern era? Yeah, I think there's a lot of learnings in talking to, one, a lot of admiration. for like Parker and the way that he thought about building that company.

1:47:05I think talking to a lot of friends at Rippling, I think there are some learnings to take from that. And there are other learnings, such as in a compound startup, you end up seeing a lot of duplication of work across like real. So there is actually some inefficiency within the company itself. And so when we've been thinking about Ambrook, I've been trying to take the elements of that, that I think make a lot of sense, which is you, you build a platform that can enable sort of an emergent ecosystem on top and that ecosystem can either be opportunities within the company, or it can be things that our customers are actually able to build themselves.

1:47:38So one thing we decided to do a lot earlier, I think than a lot of other players in the space was just to, was just to give our customers access to an external MCT. And so we have customers now that are rolling their own, like with lovable or base 44 or something, rolling their own apps for like hunting bookings for like one of our Texas ranchers, for example. And then they're pulling in context from Ambrook into that. And we're, you know, we're, we're now letting people write back, you know, we'll be letting people write back into the self. And I think that that's, it, it allows for a lot more creativity outside the company, not just in the company, but if you think of yourself as a really good platform, then I think you can, you can get some of the benefit, a lot of the benefits of the compound startup.

1:48:21Yeah. That makes a lot of sense. You said you're vertical in go-to market. What is the playbook for entering a new market? Like if you're saying we're going to go after hunting organizations or trucking organizations, are you just dialing in the targeted ads and the messaging? Or are you going to conferences and hiring a whole new sales team?

1:48:46Harley Finkelstein:She's booking the hunting trip. Yeah, yeah. The hunting trip is for fun. Yeah. No, you're like, we're here for two days. Let me tell you about Ambro. Maybe that's it. Yeah, I know. So the way that we've been finding this playbook starting to work and replicate across these different verticals are first, you have to do a lot of brand awareness. That's the push. And that can be Facebook. Facebook is actually super, it still has a lot of art left within non-SaaS, non-tech industries. um so a look at brand awareness on facebook and then what you start to do is once you get enough of a um a set of customers that are really happy you start to do case studies and then take those case studies and you start to do build partnerships with local trade associations and then you start getting third-party credibility rolling um we also go on a lot of uh really uh niche podcasts this is not the only podcast but i've also you know gone like the wealthy cowboy or returns or, you know, and that has been really effective.

1:49:53And we also help on those. And so it helps them with brand awareness. And then you start to build it. And then you start to get word of mouth. And then you start to see the regional density kick in. So in the counties where we have more than 1 % to 2 % market penetration, our CAC is half of the counties where we have maybe 10 to 25 bits of market penetration. And so that spread is just social proof and word of mouth. So you have to kick the flywheel going. How old is the company? So we started in 2020. Okay. Yeah, because it feels very, this feels like a very mature operating philosophy. Yeah, but we launched the product in 2024.

1:50:35Yeah, okay. So we spent a couple of years, we spent one or two years kind of figuring out what product we wanted to build. It's keeping the team really lean. And then we spent two years just sitting in a private beta with a couple of dozen of design partners. Interesting. Interesting. Very cool. Well, congratulations on the round. Congratulations on the progress. Yeah.

1:50:53Harley Finkelstein:Good alpha. Yeah. Good alpha on go to market. Yeah. I like it. Lots of people taking notes. Thank you so much for coming on the show. Yeah. Great to see you. Have a great weekend. We'll talk to you soon. Cheers. Goodbye. Let me tell you about Codex. Codex is a powerful workspace for getting work done with AI agents. Whether you're writing code, analyzing data, creating content, or automating business workflows, Codex helps you move projects forward from start to finish. Last story that I want to talk about, Jordy, if you got anything, let me know. But the founder of ByteDance apparently just banned his employees from distilling U.S.

1:51:27frontier models, which begs the question, what were they doing before? but uh according to this post fears distilling u.s models would trigger washington that there might be pushback in washington dc over this obviously a lot of labs they don't like getting distilled we've seen all the complaints um and it might put tiktok at risk again anthropic has accused deep seek moonshot minimax uh zai alibaba all of distilling claude but they haven't i think

1:51:57Harley Finkelstein:haven't accused i think he's doing a little bit of comms and marketing he said we we should be willing to sacrifice some short-term gains for longer-term goals. No distillation allowed, even if we lose. Interesting story. Yeah, we'll see. We'll see. Maybe the distillation attacks will stop. Maybe they will continue unabated. Maybe superintelligence can weed it out if you're trying to distill. Anything else in the timeline? From SpaceX earnings yesterday, Buko Capital says, Elon on the SpaceX call, it's probably also worth mentioning that our internal projections for reaching$1 trillion in revenue have moved up from 2031 to 2030.

1:52:40Harley Finkelstein:And Ryan Peterson says, F it, we're making our 2031 goals, our 2030 goals now too. I love it. Well, there's another model that launched Muse Code from Meta and Superintelligence, Mark Zuckerberg, posted again on X. the everything app. Wait, I got to follow him. Why am I not following him? Because it's just a new poster now. Releasing that Muse code in beta today. It's a terminal coding agent that takes on complete software engineering tasks powered by MuseSpark 1.2, a coding agent, a coding focused model update. And so I'm sure people will be benchmarking this, seeing where it stacks up, how it prices out, where it's good, where it's bad, all those things.

1:53:23Over in Microsoft land, They're going all in on GPT 5.6 Sol. A Microsoft exec just sent an internal memo that's designed to end token maxing and makes OpenAI's GPT 5.6 Sol the default model for internal use. So that was line one on the Ed Zitron to-do list for Sam Altman. Make sure Microsoft likes the model. And it seems like mission accomplished. Any other news you want to go through today, Jordy? We can talk about the New York Times tomorrow.

1:53:58Harley Finkelstein:We'll see you tomorrow, folks. Have a good day. Have a good Thursday. Is it Wednesday? Have a good Wednesday. We'll see you tomorrow. Have the best August 5th of your life. Five stars on Apple Podcasts and Spotify. Sign up for our newsletter at tbpn.com. We love you. And we love you. Goodbye. Peace. you