LIVE CLIPS
EpisodeĀ 1-22-2026
Gets me better. But when I noticed that, it's like, okay, I am more at risk perhaps, than I thought. What have you been building? Yeah. So I'm very obsessed for years with this idea that people write and speak fundamentally differently. And when we give speeches, we use various things, we rhyme and we speak in rhythm, and we use mnemonic aids, and we say first and you. And all these different linguistic things that have just been a personal interest of mine. And I sort of think that as digital media has saturated our lives over the last 15, 20 years, even our written text comes to resemble more spoken text. And so I thought, well, it would be fun if you could build a piece of software that actually empirically tested this question. Like, does a piece of writing in a newspaper today resemble more written word than, say, that same column or whatever 20 years ago? I'd like to be able to test this out. And there are various lexical markers that linguists and others have used to work on these things. So I thought, okay, it'd be fun to see if I could build this. And so I've actually been able to build it. People go, check it out, Havelock AI. And you just drop a piece of text in, and then it says, this sentence looks literate. This sentence looks like the spoken word, and so forth. And it almost works. Like. I'm actually astonished. It's not perfect by any stretch. I still have a lot of work to do, but basically, I sort of was able to vibe code this sort of machine learning model, where I trained a model to detect this is rhyme, this is not, et cetera, and then it breaks out every sentence and then sort of scores the whole piece, and it works. And I'm sort of astounded. And you've been testing it on Davos? Yeah, like, so, you know, I tested on the Mark Carney speech and the Trump speech, and the difference is remarkable. It's every sentence, it sort of classifies one as, this resembles the spoken word, this resembles the written word. And the difference, you know, we hear this, right? You hear the two speeches. You're like, okay, they're speaking at a very different register. This feels very different. And it's cool that, like, I've been, you know, with two weeks of hobby coding, been able to build this model, and I actually, I trained. You know, I learned how to. I don't know what I learned, but I was able to build this little machine learning model that identifies each sentence and categorize in one. And I'm like, this is crazy. It actually kind of works. And I don't know anything about code. And this is something that I've wanted to exist for a long time and I was able to just. Yeah, and it's cool because it's one shot. It's something. It's something that, like, I'm sure you could find a way to turn it into a business, but it's actually a tool that just should exist. Like, there were so many tools that didn't exist because there was no business justification to devote hundreds of hours. Exactly that. Right. Like, so this is an interesting thing, but who's actually going to take the many, many sort of hours that it would take to actually code this? Because I don't think there is like some obvious path to riches here, but it was a. You know, I did like an hour or two a day for a few days at five in the morning before my kids got up. And it is a properly working piece of software that I can use. I still need to fine tune a little. There's still mistakes and so forth. But it actually works. You have to turn around. You need to turn into like a vibe coding slash mindset influencer. Every morning it's like it's 5:00am I'm. What? Are you still at my. I'm up at my desk. Yeah. Everyone else is sleeping. Jocko Willink. It's the photo of the watch. Every day. It's 2am on the west coast. Every. All my enemies are sleeping. Yeah, that's right. That's right. I'm going to. I just ran my most recent. I just ran my most recent essay through Havelock AI. I got a 76% orality score. I write very, very loosely. Strongly oral. Part of that is that we do. I started the essay. Trump Greenland, Elon, blah, blah, blah. Let's focus on the important thing. Because I deliberately write exactly like I talk. Yeah. But I don't know. I'm pretty happy about that. It sounds like it scored it pretty well. 100%. That also means that you are writing in a way that connects with the modern, you know, the modern audience. And so I think that's great. And you could test that. And now you can know. And is this in the contemporary register? And I don't know, like high score, low score, better or worse. I just think it's cool that this thing can actually sort of figure it out. I love it. I love it. Anything else, Jordi? No. Super fun. Thank you so much for taking the time. Thanks for having me. Always. Great. Can't talk to you soon. Great to hang. Have a good one. Talk soon, Joe. Have fun at goodbye figure.
Needs to be deeply, deeply kind of reevaluated just with how good these tools are getting. And we're excited to see what's your. Personal framework internally and kind of like guidance or how are you and the team thinking about when it makes. Right now I feel like there's this obviously so much excitement. You can build an internal tool in an hour or something like that, but every time you build a tool, you're creating some type of ongoing maintenance burden, something like that. Tech debt? Yeah, tech debt. So what's your framework of like, hey, this should be something that we just should build versus finding a point solution or even using something like a Google sheet or at the kind of low end. Yeah. So we use notion a lot internally and I think lots of deeply flexible tools are going to be really, really key because now that the agent can kind of go in almost like orchestrate these things in general, you can kind compartmentalize very, very flexible tools together. Right. And so I think that when you're thinking about kind of the build versus buy boundary. Right. It's a lot on kind of pushing the context into the tool so that you can access a lot of the context and take advantage of it. Right. So I think the burden for almost buy goes up a lot because the ambitions and the capability of you being able to build something that's purpose built internally is actually so trivial now. Right. Like historically I've built a lot of internal tooling to allow us to do project management, releases, etceter. And now I can actually just kind of push the boundary on a lot of those things. One of my hot takes is that like, you know, ERP software, right? Yeah, of course, right. We love ERP software. I was born in ERP software. ERP software, right. Only useful at scale to build an organization, to manage resources or whatever. Right now you can kind of build those like organizational structures way earlier. Right. So the way that railway runs has historically been different than organizations. And so I think what you're going to see is almost this Cambrian explosion of various different ways that you can run your organization, which I think is super, super cool. So unfortunately like I'm an engineer, so I have engineering brain damage. I think everybody should be building kind of their own stuff, you know. So yeah. How is your.
And we think that is washing through the system. We are very positive on bitcoin for the three reasons I always say, you know it's a new technology, really important. It is a new and a first rules based global private meaning no government oversight, monetary system. That's huge. That is huge. That is its calling card. Big hedge against inflation and a hedge against deflation. If we end up in a banking crisis like 0809 and it is the first the leader of a new asset class, many people are looking at gold and they're saying well wait a minute, gold has doubled in the last year or so. What? Bitcoin's way behind. Well if you, if you start the clock at the end of the bear market in 22 gold is up I think 170%. And these are maybe dated by a couple of weeks and Bitcoin by 360%. So we think that bitcoin. Oh what happened? That was just a little air horn we use. You said 360%. It was a nice sound effects over here. Not to be jarring. They didn't prepare me for this. Anyway. Okay, so do they brief you about our flashbang? We will save that one for another time. But yes, sorry, continue. Yeah, so I'll just say it's very interesting. Over the full market cycle from 20 to now, gold and Bitcoin's correlation. Hardly. Hardly anything. Non existent. Yeah, hardly anything. But we think bitcoin is, is especially as intergenerational wealth transfer takes place in the next five to 10 years. Yeah, yep it is the digital gold and its supply growth is lower than gold's will be. And especially now that the gold price is up. Miners can go out there, they're being paid to find even more. So so that can't happen with Bitcoin. I think in a big ideas you'll see we expect Bitcoin to scale to $16 trillion in market cap by 2,030 and it's below 2 now. Can you talk about the name of the.
Like, prompt it. And so, yeah, I think you're obviously a student of history. We talked about the books behind you. I've read them all. What anecdotes can you share about previous technologies rolling out in music? Backlash to electronic drum machines. Backlash to electronic music. Give me some more. Like, what are you drawing on for its core? Okay, so, I mean, the story that I think becomes really. It looms large in a lot of people mind Dylan going electric, right? This is an era of folk music where everybody's like, you know, what's authentic is me. I don't smell great. I'm strumming on an acoustic guitar. I perform at four cafes a night in the Village. We all hang out and write together and have sex with each other. And that's the scene. And then Dylan sees what's happening with technology and sees what's happening especially with the guitar. And the electric guitar has become a creative tool for an entire legion of rock musicians. And Dylan brings it to the Newport Folk Festival and says, I'm plugging in. I don't care what you guys think. And of course, people call him Judas and say that he's doing heresy. And the fact is, he's just following. He was a follower at that point. I also mentioned Autotune earlier. You may remember Jay Z put out a song called DOA Death of Autotune. This is at the peak of T Pain. T Pain is dominating radio with essentially computer music that sounds a lot like Zap and Roger from the late 70s into the early 80s. That kind of like synthetic funk music that was really popular after, like, the Earth, Wind and Fires of the world. T Pain comes along, formalizes it becomes a go to songwriter, producer and collaborator. And all of a sudden, everything on the radio sounds like a computer. Jay Z comes along with Death of Auto Tune and says, actually, stop all that. And then people are like, that's a nice old man. Like, like, let's keep it pushing. And Auto Tune is now widely accepted in every genre. You hear it in country music, you hear it in pop, you hear it in reggaeton, you hear it in hip hop. Those. These questions that were sort of like anxious, critical. Anxious critical discourses when they emerged. Nobody really gives two thoughts to anymore. I think that's probably what's going to happen 12 to 24 months from now with A.I. Yeah, I want to go so much deeper and have you back to spend way more time with all this. I'm interested in your thoughts on the downstream.
But we're super excited to have you on here. Thank you so much for taking the time. Yesterday was kind of a funny experience. Somebody sent me the video that you posted with the rugby, and we only made, like, a hundred of those. We gave them out to friends. We would have happily sent you one if we knew you were in the market. But of course, I was tapped in. I was tapped in. Yeah, you found it. You were tapped in. You found it online. So I'm glad. Even though it was bootleg, I'm glad that it looked great. It looked great in the game. It looked great. Yeah. There it is. There it is. Well, we have sent out a. An official TBP in the mail. It should be writing any day. And what did it smell like when it arrived from Canada? Okay, so can I be candid? Can I be candid with you? Yes, of course. I feel like we can be candid. Of course. When this package arrived in my package room in my building, it looked like it had been assembled by a small child. It was. It was packed into, like, a mailer. Like a. Like a regular mailer that was tattered at the edges. No, it had a. A label from Canada. Yeah. And there was a full retinue of packing tape layers, like four or five layers on top of the mailing label. I opened the package. I had ordered one. I got two. You got two? Okay, I got two. So there's a bonus. One of you guys. Okay. Thank you. Smelled like cigarettes. No. Smelled like cigarettes. Wait, so you think a child would smoke cigarettes? Selling bootleg. Child is selling merch. Non fumar. Yeah. And. But I will say, as a. As a clothing connoisseur, when I saw the images that you guys had posted of your polo, I was, like, incredibly high quality material. I could see it in the photos. I knew that the embroidery would be, like, robust to the touch. I knew. I get this. Yeah, this looks like it was made. From, you know, you get, like, a 10 pack of Terry cloths. Not good. Yeah, See, this was an issue. A lot of people. When I first I posted that, that we had that there was bootleg merch floating around, and a lot of people are like, oh, what's the issue? Like, you guys are podcasts. It's not your core business. We don't make merch to make money. Right. But the issue was like, I knew people would buy it and they'd get it, and they'd be like, this is low quality, of course, and we care a lot about quality. Yeah. I was like. But on the. On the flip side, you guys are popping and you're getting bootlegged, so mazel tov. Yeah, it's a good interview. Thank you. Anyway, sorry. For your first time on the show, please introduce yourself for or might not be familiar. Sure. My name is John Caramonica.
Cheers, Jay. And in case you've been living under a data center, Railway Railway simplifies software deployment, web apps, servers and databases run in one place with scaling, monitoring and security built in. We're very happy to be partnered with Railway. We have.
You know, they're less like obtrusive people generally don't notice them. I mean we do a lot of deliveries, people's homes where they didn't even realize the delivery happened. Do you have a timeline for me or you actually getting into an evtol and the true flying car vision? Maybe it's not like, let's say Waymo level adoption of a human in a, in a vertical takeoff and landing. Not a helicopter, not a single rotor, not a plane, but something like that. Are we five years away from that? Longer? Is there something fundamental? Because this has been something that's been promised and I've been excited about my entire life. And it feels like there's a lot of serious companies working on it, but we're not quite there yet. How are you thinking about that market generally? I'm also really excited and it's a question that I'm very interested in. I think that, I think a lot of the technology has made probably faster progress than people realize. Like these, you know, you can build vehicles that achieve that are electric vertical takeoff and landing, fixed wing hybrids that like achieve a lot of the core performance characteristics. I think two challenges that I would point out. One is on the autonomy side, like those companies are still putting human pilots in the cockpit. And it's not going to be like Uber if you have to hire an FAA certified pilot to come to your Uber ride to get you to work in the morning. Like that doesn't scale. It's a little too expensive. Yeah. So I think from an autonomy perspective, who is going to design that autonomy layer? It's always seemed kind of obvious to me that the company that does that is going to do billions of autonomous deliveries delivering non human things before you want to carry humans. I think humans will be more comfortable with that. The other thing I would kind of point out is a huge part of this actually has to do with like integration, which is like you can totally do the flight. There are vehicles again flying those kinds of mission profiles today, but where are you landing? I mean there are no heliports in cities. They are getting shut down rather than built. And I think that it's all about the, it's all about integration. You know, if you want to get design a vehicle that is more than just like a, an electric version of a helicopter that can do like 20% of missions that a normal helicopter can do. I think if you're really talking about flying car, the key is to solve the integration problem. Like can you actually get picked up directly from your home and delivered directly to your office for your daily commute. That's the core question. Yeah. At what point do you think having a curry.
You know, this could happen. I don't know. Look, as Josh Wolf likes to say, chips on shoulders, but chips in pockets. You're putting a chip on his pocket. You are. Exactly, exactly. The thing that I'll also sort of point out, and this is more of like a, you know, sort of, you know, general company building, you know, sort of macro point. But it's an analogy that I learned from Keith back in the day, which is building companies is a lot like laying cement in the early days. You have the ability to like, you know, manipulated shape it, etc. Right. It is, you know, sort of soft and malleable the moment that it starts to harden, though, trying to change company culture requires like a jackhammer. It's loud, it's painful, it's disruptive, it breaks your structural sort of integrity. And so I think the thing that I was kind of listening in to, you know, sort of Pedro's interview, and it's a little bit of a sort of joker quip, but it's still funny to me that throughout the entire interview, there was never a time where he really deeply was focused on how this is impacting his customers in particular, is it going to save them time and money? Right. And I think that just comes down to what the early days of the two companies were like in terms of focus areas. The early days of Brex were focused on celebrating go to market, celebrating, basically like signing up logos, celebrating wins. I'm not saying that's like necessarily a bad thing. I'm just saying that is what the culture was. They celebrated the champagne campaign, they set up the sort of South Park Cafe. And those types of micro decisions feed into what employees want to work on because they see what the sort of founder celebrates versus, as you guys know, what ramp celebrates from the get go has always been saving customers time and money through better product, better automation, better our savings. And that's what ultimately the individual employee online, you know, celebrates and thinks about every single day because they see that that's what the founder is celebrating. And so I think as you see this, like, you know, the two sides of the story basically play out, I actually think it comes down to those early, early cultural micro decisions. And I think companies and founders should be thinking about that a lot, even in completely unrelated, you know, sort of categories of like, what is going to be the penultimate, you know, sort of winning culture. You know, when I think about the space industry, are you relativity space and you're celebrating like renders and capital raising? Or are, are you SpaceX and you're celebrating, like, technical accomplishments and actually, like, landing rockets. And ultimately that's going to be what sort of plays out. Relativity goes and raises three and a half billion dollars, never launches a rocket. SpaceX obviously launches their first rocket on, like, I think, $60 million total raise.
That. The thing about the New York Times is, you know, other than there's a handful of others perhaps, and I'd like, Bloomberg is one of them, but, you know, it is a brand. It is a lifestyle brand. Right. In a way that very few news outlets of any sort have ever been able to achieve. So, you know, they go into cooking and they have cooking shows. They go into, you know, simple games on the phone. There are not many news entities of any sort that could pull that off. And. And I've always been for years, sort of, not literally, but a New York Times bull, because it's just been so obvious to me that there is this one news brand that is also this big lifestyle brand and their core subscribers will follow them into all of these new ventures. And it's just something, you know, like, I don't want to name names, but like, other, like big publications, similar national general news publications do not have that same cachet where if they get into some totally new line this, that they'll have subscribers who will pay up all that money. So the New York Times is sort of a sort of a special entity. Totally, totally. You know, I completely agree. It's also interesting.
Around it. Well, I think you might be surprised. I mean, you know, similar for what I just said, you know, with flying cars. Integration is the thing that Zipline is most focused on. You know, we're launching all these amazing new partners. I mean, you know, a lot of the statistics that we're seeing in the cities, if you look at Dallas, I mean, there are municipalities in Dallas where more than 50% of homes are ordering from zipline. We had an all time new record on Sunday. We blew away the previous record by 25%, which had been set a week previously. And we had 10% of all homes in a municipality place an order with Zipline on Sunday. Wow, was that, was that football? Is that like, does that spike with like football? It is a. Holiday weekend. But, you know, these are like shocking statistics. I mean, we have, you know, a lot of the restaurants that we serve. Zipline is a huge percentage and a majority of deliveries happening from those restaurants today. And so I think that the usage is way different than what we were originally expecting. And the key is integration. You know, we want to be able to add so many different. And this is why we announced Zipping points. This is really simple new kind of ground infrastructure that we can install for free next to any partner, whether it's a hospital, health facility, retailer, restaurant, or even eventually someone's home. You know, you can almost just think of it like a new kind of mailbox for autonomous logistics.
Love it. Back to AI and how it's changing. Software development, database selection, database migration, switching something is ingrained in an organization as a database has historically been very difficult. Are AI agents making it easier? Are you actually seeing companies move around from one database to another, maybe based on cost, maybe based on features? Is the fact that a particular LLM is in love with a Redis or a MongoDB, is that going to going to be relevant in the future? So I'm unfortunately a little bit like too Dario pilled in terms of like the fundamental laws of physics. And so I think it's actually probably the hardest thing to change about your system is the database, roughly, right, because you have that storage, you have all of those other things, right? Like, this is fundamentally also why we build data centers, right? One, you need more production roughly in the world, and we can start with data centers, and that's wonderful. But two, it's actually a really, really solid mode in terms of, you know, you can't just will into existence a data center, right? Like, and at some point, potentially down the line, like, those things will get more and more expensive just because of the cost of labor, but like, the cost of capital is decreasing, right? And so I think the similar thing kind of applies roughly with data centers if you kind of move roughly up the stack. And so we actually started building or, sorry, databases. We actually started building databases like we were originally just a, you know, click a button, get a postgres instance or Redis instance or anything else like that. And we've slowly worked our way back up, right up to building, you know, web apps. You know, we'll launch a CDN at some point in the future, stuff like that. So I think that when you look at that lens of fundamental law of physics, I think the database is a very, very important part of your kind of rough calculus. And I think the agents are making it easier to go and almost manage those migrations. But when I was Talking about those 5,000 lines of code, you have to think about what you are roughly building. And so you're still going to need to audit it. And that's why I say moving from engineers to architects, right, where you're thinking about that kind of bigger, bigger vision. And so, you know, as Jordi said earlier, like, I don't think anybody can roughly pick, predict the future for a lot of these things, but I do still think that fundamentally there are going to be requirements for how do you integrate the things that exist in the physical realm so roughly around those databases to storage, to data Centers and with kind of your expected reality, basically, right. And the hope over time is that Railway can actually help you do that. Right? So we do this on a software level and we've built tooling internally to help build data centers. And so we've actually virtualized a bunch of those racks. So now you can almost drag and drop those servers into them. And at some point we can get to a point where we're working directly with the Super Micro API. You can drag and say, hey, I want this data center of this size in General, you'll create PoS for that and say, okay, cool, hit a button, it'll cost you 25 mil or something like that. And one click checkout for data centers. One click checkout for data centers and stuff like that. And moving towards a world where you can almost. Because it's easier to build things on the Internet, right? It's a fundamental fact, right? If you can simulate it and simulate that small section of it and you can create almost APIs to reconcile. This is what Terraform did really, really well, like expected state versus desired state. And if we can do that with your intention, as you kind of move through time and give you those safe kind of primitives to say, okay, cool, I want to make this change. Let me fork my entire environment, get copy on, write, storage, everything, make some small change. I make it in production, it merges and you deploy it instantly. Then you're kind of moving at agent speed, right? And that's what we think a lot about in terms of building the next fundamental layer of software development and then physical production beyond.
So, yeah, Zipline, I mean, we've always had this attitude that like, this technology can only scale if it is wider, more serene for neighborhoods than the traditional delivery mechanisms of cars. And so, you know, Zipline has a big team of aerodynamics engineers and aerodynamic aeroacoustics engineers who have been focusing on this question of like, how do you design something that's really, really quiet for the last five years? And by the way, you can't. Planes aren't like cars. You can just put a muffler on the back of it and make it quieter. Like, the design of the aircraft has to be from the ground up. Made with aero acoustics in mind, Zipline is. Despite the fact that we operate bigger, more complex vehicles, Zipline is six times quieter than the next closest competitor when we're making a delivery to someone's home. And we'll actually make significant improvements to the aerodynam aero acoustic profile of the vehicle this year. We actually think we're going to get about another 8 decibels of improvement this year, which for non engineers, you know, think of 8 decibels as being like about half as loud to the human ear. So having again, yeah, that's awesome.
How do you see it progressing? I think focus really matters, right? Like we have a like, you know, sort of core ICP which is just like, you know, you know, the all basically enterprises of America. And I think like there's so much more that we can do to serve our customers even better. You know, sort of seeing us, you know, basically spin up, you know, sort of Ramp labs and you know, today we just, you know, they launched, you know, a new version of, you know, sort of RAMP budget. And I think any additional energy spent on just continuing to develop next generation products, especially ones that obviously utilize AI that is focused on the core ICP for Ramp is a far better utilization of our time. And you know, to take it back to like the Peter Thiel philosophy, we believe in building monopoly, right? And ultimately anything that isn't serving that core ICP is a distraction from the potential to build a monopoly. I think what I find particularly exciting about today and about the news about Brex is that it is probably the strongest monopoly thesis for ramp that has existed since the company's basically start. If you remember when we funded it, you know, and it like, you know, came out and launched in 2019 and 20, Amex still had like a really big enterprise business. Brex was a multibillion dollar company. Divi was still super sort of fast growing. Bill.com was still super sort of fast growing, expensive. I was fast growing. Like there were so many players in the market. You fast forward today, Divi basically got acquired, is barely ran anymore. Brex obviously, you know, sort of got acquired. Amex has almost entirely shifted their focus to their consumer business, right? If you look at how much of their gross profit basically comes from consumer business and how rapidly that's growing relative to enterprise. It's all the CEO is basically talking about on quarterly earnings calls. Their entire basically like, you know, you know, fmv basically growth over the past three years is entirely based on the consumer business. And so enterprise, corporate spend card, spend management, budgeting, just the like smart CFO suite suite that smart CFOs use. Dude, the like field is somehow more open today than it was five years ago. There's less funding going into it, less like super competent teams that are, you know, sort of focused on it. And so why would you ever like give up focus on just continue to maximize that when like, you know, we're still only 1% of the way there. There's so much more basically to do. And so yeah, would I as a consumer like love to have that level of power like in my hands sure, but, like, as a ramp shareholder, why would we not give up the fact that we literally have the best monopoly opportunity since the company got started, that we need to continue to go capitalize on Jordan?
Maybe that's what's happening at the low. End to give them some credibility or at the lower end there we're in the camp. And you'll find this in Big Ideas as well. And we've been saying this for a. While and at first when we said it, probably said it three years ago for the first time, we think there's. Going to be a step function change. Here in real GDP growth, just as there is with every technology super cycle. So just a illustrate that in the. 400 years between 1500 and 1900, Brett Winton, working with academia found out or learned or. Yeah. That real GDP growth was 0.6%. Then we had railroads, internal combustion engine, telephone, electricity, and it went to 3% for the next 125 years. Here we are, we think it's going north of 7%. And that's conservative. That's conservative. I think it's conservative. Really Productivity driven. Yes. And, and we think that, that this is one of the greatest times to start a business. When I started ark, there was no chat GBT to say, okay, do a deck for my startup because this is. What I want to do. There wasn't anything like that. I had to create everything. Now you can look at an unmet. Need out there and say, hey chatgpt. Or hey Grok, come help me build this business. Right. So we think that fears of unemployment, sure, there's displacement short term, but think. About what's also happening. 1.3 million baby boomers are retiring every year. Right. And we've just had or we're in the midst of an exodus of immigrants and you know, little in migration. So that I think is helping keep the unemployment rate low. But entry level jobs can be done by AI. And so that's what we're seeing at the younger level. And so what I say to young. People is you all in your mind. Have thought of a new business that would really fulfill an unmet need. Something that frustrates you. Why don't you do it and use. AI to do it, to plan it. And just get going while you also. Look for a job if you're unemployed. I think we're going to see the. Biggest entrepreneurial explosion in history. I love it. I love it. Thank you. We got to hit the gong for 7% GDP growth. We have a gong around here. I love the sound effects. We have real sound effects in addition to the soundboard. Thank you so much for joining. Yes, please, can I just give you one more stat? One more stat? Yes, please. We think, we think inflation will go negative. Whoa. I love that technology. This is amazing. This is a very optimistic picture of the future, and I'd love to hear it. Thank you so much for taking the time to come on the show and. Explain all of that.
Big tech world. Let's head over to Mark Gurman's latest report. The germanator Apple is revamping Siri as a built in iPhone, Mac and Chatbot to fend off OpenAI. And there's an image here. What is this? The Gemini Google. Google Gemini AI chatbot on a Galaxy S25 Ultra smartphone. So Apple plans to revamp Siri later this year by turning the digital assistant into the company's first artificial intelligence chatbot, thrusting the iPhone maker into a generative AI race dominated by OpenAI and Google. The chatbot, codenamed Campos, will be embedded deeply into the iPhone, iPad and Mac operating systems and replace the current Siri interface. According to people familiar with the plan, users will be able to summon the new service the same way they open Siri now by speaking the Siri command or holding down the side button of their iPhone or iPad. The new approach will go well beyond the capabilities of current Siri for even a long promised update that's coming earlier in 2026 today. John, say Siri again really loud. Clearly for anybody that's watching on their TV at home or in their office. The feature is a central piece of Apple's turnaround plan for the AI market where has lagged behind Silicon Valley peers. The Apple intelligence platform had a Rocky rollout in 2024 with features that were underwhelming or slow to arrive. Shares of Apple gained on the Chatbot news, climbing as much as 1.7%. The previously promised non Chatbot update to Siri, retaining the current interfaces planned for iOS 26.4 due in the coming months. The idea behind that upgrade is to add features unveiled in 2024, including the ability to analyze on screen content and tap into personal data. It will also be better at searching the web. Chatbot's capabilities will come later in the year, according to people who asked not to be identified. The company plans to unveil that technology in June at WWC Campos, which will have both voice and typing based models. Tyler, what's that sound like? Typing? You're going to be able to type to Siri back and forth. Oh, now let's go. You can you type to Siri? No. Right now you can't. I thought it opens. I haven't used. Maybe, but you don't have an app. You don't have a place to store all your conversations, which was my take. Okay, cool. But I still care about. I still care about the model being implemented way more than I care about being able to say yeah, yeah, yeah. Yeah, of course, like where, where, where are. I mean, like, the model is going to be good. Like, they're using Gemini. We know that, but like, what specifically are you looking for on the Jimmy Boys? Jimmy Boys, Jimmy Boys. Tim Cook's a Jimmy boy. But where are you. What are you worried about them not implementing correctly? You want an agentic harness so it can do Claude code style work for a long time on your phone directly. Yeah, I don't know. There's just like, the current Apple intelligence is so bad. I just like, have very low. Like, I think there's a real possibility that they just like make it so every time you have to talk to Siri, you have to like press an extra button or something. There's just like random stuff like that. Yeah, but what if it's just you press and hold the button and then you're just submitting a Gemini prompt and then it just returns exactly what Gemini would have returned? That's like not implementation. That's just like, normal. Yeah, but like, why? I mean, they could have already done that. You think they might like an open source model? Like, they could have done that? No, they needed a deal. They can't just. Apple can't just go and implement anything. I think it's. I think it's going to be tricky for them, like, specifically. But I'm Llama. That would be good. So specifically, Llama has. In the, in the terms of service, like, you cannot use this, it's open source, unless you have more than like 800 million users or something like that. Like, there's some specific fear about opening. I trained their open source model for like $5 million. Like Apple could do it. 100 million. You think so? Yeah. People always talk about how amazing the open source model is. Okay. It's way better than the current Apple intelligence. That's true. That's true. Yeah. Okay. Like, open source models are pretty solid. So you're saying you want Apple to make the phone with Foxconn and run Deepseek? Yeah, come on. They're not doing that. They're not putting deep seek on there. Other than the Chatbot interface, the operating systems aren't getting big changes this year. Of course, we got liquid glass. So Apple is more focused on improving performance and fixing bugs. Last year, it rolled out a major design overhaul, unifying the look and feel of its operating systems. Internally, Apple is testing the Chatbot technology as a standalone Siri app. Let's go. Similar to the ChatGPT and Gemini options available in the App Store, the company doesn't plan to offer that version to customers, though. So we might not get an app. I'm on a roller coaster here. Instead, it will integrate the software across its operating systems like the Siri of today. I still think they need an app because I'm going to hit it, I'm going to ask it a question, I want to come back to it. How am I going to do that? So. So I don't know. Hey, Siri, go back to the recent. Yeah, maybe, but it's gotta be good then. It is. Then the implementation has to be good, because even In Gemini and ChatGPT, there are often times when I will fire off a query, it will come back with like thousands of words. And if I search a keyword in the search box of previous chats, I have so many previous chats that I can't necessarily find the exact one that I'm thinking of. Okay, so one implementation thing that I think is important is like, if it's super long, if the response is super long and it reads it out to you, you should be able to pause or something like that. Like in Claude. You can't do that right now. Yeah. You can't even fast forward five seconds. That is an important thing. There's all these implementations that materially affects. How good it is to use. Yeah, yeah, yeah, that's good. Embracing the chatbot approach represents a strategic shift for Apple, which has long downplayed the conversational AI tools popularized by OpenAI. Google and Microsoft executives have argued that users prefer having AI woven directly into features, something Apple has done with its writing tool, Genmoji emoji generator, and notification summaries rather than standalone chat experiences. That's not an unreasonable take. I do see why they landed there. And in the long term, I wouldn't be surprised if the AI woven in the features, the AI that's in the Photos app is better than, okay, I gotta open up the chat app and ask it to edit this photo and import the photo from my photo. No, you just want to be able to go into the Photos app, have the, you know, change the color temperature, change the brightness, draw on it, add text, and then also have AI features in there, I would imagine. So. I am bullish on Apple diffusing AI into all the other apps, but it's taking a while. Yeah. What do you think? Yeah, I just hope the team specifically that designed the new version of the Photos app from the ground up takes the lead on Apple Intelligence. Because you're not a fan. It's been getting better. It's not perfect, but it has been getting better. I've been learning how to do it. I know there. I know the quirks. Is it getting better or are you getting. I'm getting better. I'm getting better. It can't really just kill it. They should have just come out and said, hey, we hear that you guys. Don'T have this photo app, skill issue, billboard.
Particularly bitcoin. So we are not worried. We are seeing whales. So the original, the OGs, they are shifting bitcoin from wallet to wallet because of quantum fears. So that's happening. But it's also created some FUD for, for the markets. Are they selling because they know something we don't know? Are they selling simply because they don't want to go through the downside of another four year cycle? Have we hit the low down only 35% this time instead of 85%. So we're in that kind of a market and I do think we're in a basing period which really started on 1010, the flash crash. It was the software glitch at Binance caused a lot of auto deleveraging. It was kind of crazy to learn about that kind of deleveraging. Meaning firms who thought they were hedged if they were on two different exchanges and Binance was one of them, they were auto deleveraged on one side of a trade. And so they were totally exposed in a way they never expected. That caused we think $28 billion of carnage. And we think that is washing through the system. We are very positive on bitcoin for the three reasons I always say, you know, it's a new technology, really important. It is a new and a first, rules based global private, meaning no government oversight, monetary system. That's huge. That is huge. That is its calling card.
That's huge. That is huge. That is its calling card. Big hedge against inflation and a hedge against deflation. If we end up in a banking crisis like 0809 and it is the first the leader of a new asset class, many people are looking at gold and they're saying well wait a minute, gold has doubled in the last year or so. What? Bitcoin's way behind. Well if you, if you start the clock at the end of the bear market in 22, gold is up I think 170% and these are maybe dated by a couple of weeks and Bitcoin by 360%. So we think that bitcoin. Oh what happened? That was just a little air horn we used. You said 360%. It was a nice sound effects over here. Not to be jarring. They didn't prepare me for this. Anyway. Okay, so did they brief you about our flashbang? We will save that one for another time. But yes, sorry, continue. Yeah, so I'll just say it's very interesting. Over the full market cycle from 20 to now, gold and bitcoins correlation, hardly anything. Non existent. Yeah, hardly anything. But we think bitcoin is, is especially as intergenerational wealth transfer takes place in the next five to ten years. Yeah, yep, it is the digital gold and its supply growth is lower than gold's will be. And especially now that the gold price is up. Miners can go out there, they're being paid to find even more. So that can't happen with bitcoin coin. I think in a big idea. As you'll see, we expect Bitcoin to scale to $16 trillion in market cap by 2030 and it's below 2 now. Can you talk about the name of the.
Scale. Are you one shotted by vibe coding? Yeah. You know what? I have, like, full on AI psychosis. And I'll tell you, here's the evidence that I realized. Complete victory for tac. Yeah. So here's. I'm going to admit this here, but here was the moment that I knew I had full on one shotted psychosis, which was whatever term you want to use. So I was like vibe coding. I was using cloud code, and I was very impressed. And then I opened up a second Claude code window because I wanted to be able to work on a couple aspects of the project in parallel. And then the second window, it made a suggestion for what I work out. That was actually a very bad suggestion. It was the first. Had I actually said yes to the suggestion, it would have really set me back quite a bit. And I'm glad I paid attention. Overrode it. And my first thought in that moment was like, the original window would have never made that mistake. You know, it's like the original window, that window really got me. And I was like, I'm sticking with the window on the right side. And in that moment when I had that thought that, oh, this window doesn't get me the way the other window, I was like, joe, okay, I got to step away from the computer a little bit. Those are early signs, I think I'm like, you know, I'm very proud of myself. I think I avoid psychosis of all sorts. I've been on Twitter for 15 years and I haven't gone crazy. And so I was like, okay, I saw that in myself. It's like, step away from the computer. It's just a computer. There's no personality. There's not one screen that gets me better. But when I noticed that is like, okay, I am. I'm more at risk, perhaps, than I thought. What have you been building? Yeah, so I'm very obsessed for years with this idea.
Was from a sort of geopolitical standpoint, the highlight of the whole event. The other one, the other speech that I thought was really remarkable or very important was the Chinese vice premier there. Talking about how China doesn't want to just be the factory to the world, it also wants to buy your stuff. Because for all the anxiety that we. Have here in the us, every other. Country is feeling that stress as well. Their manufacturing sector is getting hollowed out by Chinese competition. So it's clearly a source of, I would say, a different an anxiety for China that every country around the world looks at them and thinks, wait, are. You going to kill my domestic champions as well as you've done to European car makers, automakers and so forth? Yeah, it's great. It's great that they're saying that, but does anyone actually believe it? I mean, ask like Elon, do you think, do the Chinese really want to buy Teslas? No. I mean, so I wrote about this and my take take is that, like, they have been saying this for years. This is not the first time. And they're like, oh, we're going to lift domestic consumption and that's going to create imports, then that'll benefit the rest. Of the world and so forth. They've been saying this for years. And I think there is this view that it's like, well, this is just lip service. It's just they just say it every year and then they say it again the next year. I think, however, in their defense, what I would say is the structure of. The sort of domestic Chinese political economy makes it very hard. And if I could just get technical for a second, they have this very. Decentralized approach to fiscal policy, which is all of the different provinces basically run. Their own economies, and the provincial leaders. They try to achieve goals that are set by Beijing. And basically this is how you get this crazy race to the bottom, which. Is every province wants to have their byd, their battery maker, et cetera. And so the issue isn't so much whether Chinese leaders in Beijing are serious. About wanting to boost imports. The question is whether their system will. Ever allow them to get out of the sort of game theory equilibrium where all of the provincial leaders feel incentivized to just export as much as possible. And it is probably a tricky thing to turn around that political system in the same way that we have difficulties that are very difficult to turn around, even if someone in Washington, D.C. says otherwise. What's the sweet spot for China?
From, so we took from a top down point of view, two years ago in Big Ideas, we took the position that because we were seeing it play out in the numbers, that the tech stack would change, that you have the infrastructure layer, which of course very strong. You have the platform as a service layer. Poster child for that. Palantir was gaining share, significant share. And then you had application layers, software as a service growing but losing share of incremental growth. We put that chart in and we had no idea how quickly this was all going to happen. Software stocks, more traditional software stocks last year were terrible and even this year they're still suffering. So AI is happening faster, I think, than anyone could have imagined. Anyone who thought they had time to adapt, I think has been mistaken. And you know, I even see it with large enterprises when we go in there and ask them, okay, what's your AI strategy? You know, they, they treated it in the beginning like, yeah, we'll let our, our teams experiment with all these new tools coming up. ChatGPT. No, no, no, no, no. That's not the way this is going to work. It's, there's, there's going to have to be complete restructuring around AI or else you're probably going to fail. Maybe not outright, but you'll become irrelevant. So, so I think even ARC, for example, we have brought in Palantir. We're only 70 people, but we've brought in Palantir. And I can see we brought it into research first. I wanted to see how could we transform our research process. And we're already seeing incredible results in terms of productivity gains by the research team. But I can see that in the rest of the firm it is going to take. Well, first of all, you have to have someone other than the cto. And that's true even in our case, saying, we're doing this, we're doing this, our research is telling us this is the way the world's going. We cannot be caught flat footed. And so I really think you need the CEO, not the cto, CEO, CFO to drive this change. And I think, you know, many people think, oh, the results are so immediate and they are with our writing and editing perhaps. But when you're trying to transform an enterprise, it takes, it's hard work. You have to collect all the data, even data that you didn't know existed. So you have to figure that out. You have to clean it up, you have to integrate it, you have to map out workflows in excruciating detail. And that all takes time and a lot of people in this world are impatient. They don't see results right away. But I think the results are going to pay off. I think the way Palantir is doing, it just put its software on top of legacy, no rip and replace. And over time it will usurp the role of all of those that inefficient legacy software. So I think it's going to take time in the enterprise. And I know there was the MIT study that said, hey, this isn't happening. This is a figment of Silicon Valley's imagination. And we disagree mightily on that. How.
Triple legs. Let's just roll right. Market clearing order inbound. You're surrounded by journalists. All keep. Strike 1. Strike 2, Activate. Go. Golden Retriever mode. Trust. Market clearing order inbound. 5p. On the horizon. You're watching TVPN. Today's Thursday, January 22, 2026. We are live from the tvn trough. The temple of technology. The trough of technology. The fortress of finance. The capital. The capital of capital. Ramp.com Time is money. Save both. He's used corporate cards, bill pay, accounting and a whole lot more. Thank you again to Arik Garazi for coming on the show yesterday, breaking it down for us. That was a lot of fun. And thank you to Marc Benioff. Benioff. We have acquired a one of one truly. He says, aloha One Master Puppets record signed. Signed by Hawaii resident. He was keeping us guessing when we had him on the show. We were like, can you send that to us? He was like, no way. It's incredible. Look at this thing. Look at this thing. We got it. This is going in the business hall of fame here. The Museum of Business. Museum of Business. TVP and UltraDome. That's right. Let's pull up the linear run of show. Meet the system for modern software development. 70% of enterprise workspaces on linear are using agents and you should be too. We're going to take you through the story, talk about Davos, talk about a whole bunch of news in the timeline. Then we have Keller from Zipline coming on, Jake from Railway, Joe Wiesenthal's returning. That's right. We're very excited to get a whole bunch of updates. I mean, look at, look at the range here. We're going with Cold Healing. The Exonon is coming on. John Karamica is coming on. Brother Joe, big time. Jake announcing a big round first. Davos is winding down. We're finally getting like the final surprise. Guess Elon is the big one. Elon had to come in. He had to come in last. Buzzer beater gesture match. How do you think that works? How do you think that Elon do you think he just has like an open invite and then he makes the call like a few hours before and just says, oh, it's going well. I'll be there. And then they just throw him on? Or do you think he said. He said yes a week ago. And then I think, don't announce. I think Elon calls up Jake how says, hey, save me a seat, save me a seat. And then it could be they figured out there, okay, I Think. I think when you're tracking to be the world's first teen air, you got an opening. They just have a stage set up. Yeah. Say come. Apparently the events go really, really long. Some of the interviews go at 10pm, some start. It's like a full, full tilt across multiple, multiple event spaces. There's different houses all in was doing something at the Freedom House. They had a really cool interview with Satya Nadella that I watched. Tech was out in full force. And that's what I wrote about today in the TVPN newsletter. Before we go through a little bit of this, let me tell you about Vanta Automate Compliance and Security. Vanta is the leading AI trust management platform. So fun fact. Davos makes half a billion dollars a year in revenue. You think of. Yes. Tyler's clapping. It's good news. It's a nonprofit, an ngo, a think tank. But you thinking what I'm thinking, Jordy for profit conversion. That's right. Let's make it happen. That's right, let's make it happen. But it's an old, old organization. Klaus Schwab, who you probably know as the face of it, he was the founder. He started Davos in 1971, the World Economic Forum. He was pushed out last year. Yeah. So it's been on the ropes. My thesis is that Davos is back. It was a really good. I think tech is going to double down on this. But the first decade, it just sounded like it was an amazing party. One person in 1981 told Time magazine that the forum offers a delightful vacation on the expense account. What a great, What a great line. Time was really doing great reporting. Also Time owned by Marc Benioff, friend of the show. So it's been a rough decade for Davos. Liz Hoffman teased the concept of an inverse Davos index in semaphore, which I thought was sort of funny. And she sort of wrapped up a few times when the consensus at Davos wound up being, like, woefully wrong. And so she went through a few things. First was that everyone got together at Davos. There was an economist who went on stage in 2008 and said, it is inconceivable, repeat, inconceivable, to get a world recession. It's like the one thing, like, not mincing words, not hedging at all. Just like my prediction is that we will never see a world recession. And of course we did. The Davos crew, they also missed Brexit and the rise of MAGA. And then in 2020, they had a forum that Davos in 2020. It's always in January in 2020 it was January 21st to 24th and there was really no talk of a global pandemic or being worried about that. There was one panel about sort of risks from antibiotics or something, but they certainly weren't talking about the coronavirus. And just a week later Biology posted that famous going viral post on January 30 saying like, Hey, I' this, I've been tracking this for a while and this seems really serious and a lot of people in tech were already tracking it. So there was this disconnect between like. Like you go to Davos and you hear like the number one thing this year is the metaverse. Apparently they did a big metaverse push and then the year plays out and it's a wildly different story. I personally stand behind the idea of Davos in the metaverse. I would love the all of the global elites coming together to create ultimate VR skills simulated. Has anybody built the World Economic Forum in Roblox? They should Minecraft, something like that. You know, one to one replica. Yeah, it would be good. There's been a bunch of interesting like behind the scenes videos because obviously the big, the big tentpole media groups, Bloomberg is there, CNBC is there. But then there's also a few just folks I follow on Instagram who went and are doing sort of walk and talk vlogs. And that honestly gives you more of like the lay of the land, like what it's like getting around. Pretty interesting. But before I continue, let me tell you about console. Console builds AI agents that automate 70% of it. HR and finance support, giving employees instant resolution for access requests and password resets. It seems like there may be the Internet. I certainly wouldn't say the Internet has forgotten about the iconic line. You'll own nothing and be happy. But I haven't seen it so much over the last week. A lot more focus on. Yeah, it's interesting. That line was all about just financialization and was really. I like to think it was about enterprise software. Yes, we're moving away from box software. Exactly. To SaaS. You're going to own nothing. Yeah, you don't own this particular version of TurboPuffer. You pay for consumption. Exactly. That's what it was about anyway. TurboPuffer of serverless vector and full text search built from first principles and object storage. Fast 10x cheaper and extremely scalable. So this feels like this year feels like a big turnaround. Dario Amadeh Demis Hassabis, Satya Nadella All Their interviews have been going viral on the timeline in California. There's been a bunch of articles written about what they've said at Davos. And there hasn't necessarily been a huge reversal in the positions of these tech leaders. They're saying things that they've been saying for a year or two, sometimes more. But there's some feeling like when you say a line like we could have 20% unemployment in two years, it just sounds different when you say it directly to a bunch of world leaders, a bunch of global elites, a bunch of international executives, than when you say it on a tech podcast that's a little more insidery, a little more inside baseball. So it just hits different. And then it's also an interesting reflection to hear. Like the Satya Nadella all in interview had had an interesting, like, almost like high, low aspect to it where Satya was talking about very specific details of copilot implementation, what the future of the workstation could look like, whether people will be running, you know, Sidecar. He's a product guy. Yeah. So he's going in a lot of detail there. And then he's also zooming out and reiterating what he told other business leaders just about the general wave of AI adoption that's still pretty early on the global stage. And so it's felt like there's two different conferences going on. We've read this from this take from a couple of different people where tech leaders are talking about research, progress, employment impacts, sovereign AI data center buildouts, and then the politicians are talking about Greenland, Venezuela and trade deals. But even though there's like this to this bifurcated vibe going on, it still feels like a win win because the tech is getting their messages into the global stage now, where in an audience that will ultimately determine how quickly this technology rolls out and how it will be regulated. And I imagine that the tech industry is driving a lot of growth for the World Economic Forum because every company you've seen, the Palantir House, there's a lot of different tech companies that have really gone big on sponsorship, gone big on their presence. And I imagine this will get bigger. OpenAI was sort of notably absent or not absent, didn't take over the stage. And you would imagine that next year, given the vibes around this one, that there will be more of a presence from basically all the big tech companies. We had an opportunity to have a presence there and we didn't think that it fully made sense for this year just given historical. Yeah, travel was a big factor, but. Yeah, I did like the fact that Demis and Dario presented this, like, friendly and cooperative united front for what American AI progress looks like. It was particularly in stark contrast to last week when there was this incredibly messy AI drama, the opposite of cooperation between Elon and Sam Altman. And so last week we were like, oh, no one. Or between tinky machines and yeah, yeah, with the talent mix up, that's been. A big drama story. And then you see Demis and Dario sort of hanging out, talking about how friendly they are and how much they want to work together. And it feels a little bit like faith in humanity restored. It really captures the spirit of this global cooperation, which I think is good. And so I think Davos is back. I think it's time to ski. And congratulations to everyone that's been having fun and making waves in Davos. There's a whole bunch of videos that we should go through. First, let me tell you about Okta. Okta helps you assign every AI agent a trusted identity so you get the power of AI without the risk. Secure every agent. Secure any agent. Did you want to watch this clip from Midnight Capital? Jensen, pull it up. Wong is Midnight Capital. He's given us some instructions. He says, or they say, listen to this. Then watch the second video. And this host over at Fox Business has given us a run for our money in the costume department. Maria Bartiromo knows how to dress. This is fantastic outfit. Let's play. Jensen. Wong, what are you doing? And trying to compete. Is the space getting more competitive for you today? Our space is incredibly competitive. I've got a lot of competitors. It's hard to. It's hard to imagine. Google has been here for a while, and they're excellent AI company for a very long time. And. And they're incredible company, but we have competition from. From all sides and, well, we just have to run fast. You know, Nvidia is the only. You're doing and trying to compete. And then Midnight Capital says, watch the. You know, a company has an insane monopoly. If they're trying to convince you that. I think that's where we're going with this. You're giving away the punchline. Play the teal clip at the London School of Economics. Oh, I didn't even know we had a. Scroll down. It's right here. This is the whole one, two punch that Midnight Capital's serving up for us. Here we go. The people who have monopolies pretend not to have them, and the people who don't have monopolies pretend to have them. And so it gets kind of confusing. If you're Google, you will never say that you're a search engine. You will say you're a technology company and technology is a vast, incredibly competitive space. If you're in a completely competitive business, let's say you're trying to open a restaurant in London and you will say, well, this is totally different from any other restaurant. It will be the only British Nepalese fusion cuisine within a five block radius of the lse. The people who have monopolies pretend not to have them and the people who don't have monopolies. This is a very popular point. I studied. Yeah. Jensen studied. And so. And you know what? I'm incredible. It's incredibly competitive. Of course he's also going through an acquisition with Groq and it seems like the GROK deal will go through, but. It'S like I have so much competition I had to spend $20 billion on a team. Yeah, yeah. And yeah, please define. My market is also including CPU and also including Asics and also including like everything else that could possibly compete with Nvidia. Yeah. Because the Journal had a good summary of kind of some of Jensen's points of view. What do you say out of Davos? Nvidia CEO says AI needs more investment in defiance of bubble fears, Jensen called for higher investments to further spread the technology across developed and emerging economies. Speaking at the end, Davos, Huang described AI as a five layer cake consisting of energy, chips, cloud infrastructure, models and application. He said AI's applications, how the technology is used in a specific industry is the most critical layer of that cake, as it is where the economic benefits lie. Sectors like energy or semis, both key to developing and harnessing the technology, are already growing thanks to AI. But he said more investments are needed to ensure that benefits of the technology spread to more industries across both developed and emerging economies. The AI bubble comes about because the investments are large and the investments are large because we have to build the infrastructure necessary for all of the AI layers. I think the opportunity is really quite extraordinary and everybody ought to get involved. It's open for business, folks. Yeah. On the topic of just do we need more investment? There was an interesting detail in the TSMC earnings this week that Ben Thompson highlighted. Basically, tsmc, they had this earnings call and they announced the planned capital expenditure of 52 billion to 56 billion, up 27 to 37% from last year. Last year. And there was some pushback from an analyst. So I'm not sure that coming in at the top of the predicted range qualifies as quote, higher level of capex spending. An analyst pushed CEO CC way on this point, noting that TSMC's revenue has grown 50% since 2022, but capex has only grown 10%. There weren't many ants, there weren't too many answers from TSMC about this, which is understandable given that they won't announce years capex numbers until next quarter. And so Ben is sort of pointing out that TSMC is potentially the biggest bottleneck. TSMC has reiterated this that we're not bottlenecked on energy, we're bottlenecked on chip production. We're going to scale up, but at the same time they don't want to be caught holding the bag. So Ben Thompson says the implication of cloud service providers showing huge financial gains from AI now, combined with the admission that there is an insufficient supply of silicon, is that foregone revenue isn't a theoretical risk. It's being foregone right now. And you've heard this from all the labs. If we had more chips, we could actually deliver more services, charge more, make more money. We are bottlenecked on compute. And yet at the same time, notice that Wei led his answer, admitting to his fear. TSMC doesn't want to get stuck holding the bag, spending billions of dollars for demand that might disappear. This is a legitimate concern for TSMC, but it's a big problem for TSMC's customers. The only way they're going to overcome that fear is by helping bring competition to bear, such that TSMC worries more about losing business than they do about investing too much. And so who will be the competition? Maybe it's intel, maybe Samsung. There's a number of other foundries that could potentially ramp up and put the screws to tsmc. But the Nvidia tpu, not Grok, but Cerebras and a few other of the chip makers are all sort of still constrained by tsmc. And so there's still some worry that if we don't expand just raw fab capacity that that will be a bottleneck for AI. But anyway, before we move on, let me tell you about Gusto, the unified platform for payroll benefits in hr, built to evolve with modern small and medium sized businesses. Let's go to who is this? We have a clip here from. Well, Philip Johnson's having fun at Davos, but before we do that, let's head over to David Phill. Me defending my vibe coded CRM I made in one shot with GPT4O. Sometimes it's too slow for sure and needs to be Reformed for sure, but which is predictable. Loyal sometimes. That's a great speech. He's on one right now. What was the story behind the glasses? Somebody. I saw some meme that I think he, he suffered an injury or something and some people would say do an eye patch, but he went with the sunglasses. It's a fantastic rock star. Yeah, he looks like a rock star. Elon Musk has, has hit the timeline on Davos. I love how, I love how Philip. Is just like in the front row. Of everything with his phone camera. It's incredible, incredible citizen journalism here. Let's play this clip from Elon Musk at Davos. What's he saying? Why did he go to Davos? What is his message? What is he going to talk? Is that the lowest cost place to put AI will be space? And that'll be true within two years, maybe three. Three at the latest. So looking 10 or 20 years out, how would you describe success? Three at the latest. Three at the latest. That's an aggressive timeline. Yeah, for that. Blue Origin launched satellite Internet to rival SpaceX and specifically to rival Starlink. Bezos backed Blue Origin is launching a satellite network for enterprise, data center and government customers. TerraWave. The company aims to begin deploying the first of 5,408 satellites in the fourth quarter of 2027. So a little under two years, the service will compete with Starlink and it will also compete with Amazon operated leo, which is interesting. I guess they'll maybe work together at some point. The network, called Terrawave, is targeted for enterprise, data center and government users. The company said it will provide speeds of up to 6 terabits per second from satellites positioned in LEO and medium Earth orbit regions of space that are between 100 miles and 21,000 miles from Earth's surface. Blue Origin says it expects deploying the constellation in the fourth quarter of 2027. Bezos is entering an increasingly crowded satellite Internet market that's currently dominated by Starlink. Starlink has more than 9,000 satellites in orbit and roughly 9 million customers. Amazon, which Bezos founded in 1994, has also ramped up its own offering. That service recently rebranded from project Cooper to LEO. The company has set up 180 satellites since last April through a series of rocket launches handled by partners such as ula and actually SpaceX. Several future deployments are expected to be handled by Blue Origin. Amazon aims to build a constellation of 3,236 low earth satellites that will serve business, governments and consumers. Last November, the company opened up an Enterprise Preview to select users Ahead of a broader commercial launch. Bezos predicted in 2024 that blue origin would one day be a bigger company than Amazon. That's a wild prediction. He founded Blue Origin in 2000 and Dave Limp, Amazon's former device boss, serves as CEO. Quote, I think it's going to be the best business I've ever been involved in, but it's going to take a while. He set a deal book in 2024. Blue Origin is primarily a rocket launch company, flying tourists and research to the edge of space on short trips. Last January, the startup notched a major milestone when it successfully launched its towering New Glenn rocket for the first time. Though it was unable to return the rocket booster back to a barge for reuse, they nailed the landing of the New Glenn rocket last November following a successful launch of a pair of NASA spacecraft. So good progress over there at Blue Origin. We got to do some more research on the Chinese Starlink equivalence. They have Guo Wang, which is their like state led effort trying to get to 13,000 satellites by 2030. Currently they have around 150. And then there's Qian Fan which currently has around 100 also trying to get. They're trying to get to 15,000 by 2030 and they're more commercial focused. So interesting to track. And you would think that the Blue Origin news, the news that Starlink has a second competitor, Blue Origin, Jeff Bezos, major players. Well, what does that do for AST Space Mobile? The stock is up 14% today and it's up 24% over the last five days, 35% over the last month and 100%. Space Mobile can't keep going. There's no way over the last six months. It's at all time highs. It is at all time highs and the market cap is now $42 billion. I have some friends who are obsessed with AST space mobile and fascinating run. They keep, they keep running. Anyway, moving on. Elon also gave some timelines for Optimus humanoid robots at the World Economic Forum. He said by the end of next year we will be selling humanoid robots to the public. To the public, not just enterprises. You're going to be able to just buy one of these the way you buy a Cybertruck or a Model S plaid or anything else. That is a very, very aggressive timeline. But he has been buying the parts to actually manufacture them a lot. What does selling to the public mean? Does it mean 1,000 deliveries, 10,000 deliveries? I don't know. But he's definitely pushing that project forward. Having an Optimus that is just on staff here that can just go hit. Like if I can hit a button and have the Optimus hit the gong physically, that would be fantastic. Yeah, priceless. Really? Yeah, priceless. There's really no amount of money that we wouldn't pay to be able to get that kind of experience here in the Ultra dome. Yeah, I'm excited. I mean, we've seen like Boston Dynamics, we've seen with the Chinese humanoid companies. There's like, you can do cool things with these, with 1x and whatnot. Even if it's teleoperated, even if it's prescripted, the technology does work. It's more just like, how impactful will it be? How expensive is it? How reliable is it? Is the battery one hour or more? Has he spoken about teleoperation? No, I don't. Seems like he'd just be generally against it. Yeah, I mean, on principle, I know that. Don't the Tesla Robotaxis do some teleoperation? I think in the test zones there was like the ability to take over on that early demo. So I don't think he's dogmatic about it, but he certainly is. You know, if you're against lidar, you. Got to be against Tragically Sane says robots replacing humans. Hitting the gong. Okay, never mind. Shut it down. I'm full luddite now. Don't worry. Great. That will be the last job in this studio. Let's move on to this clip from Satya Nadella at the all in AI Summit in Davos. First, let me tell you about Vibe Co, where D2C brands, B2B startups and AI companies advertise on streaming TV, pick channels, target audiences and measure sales, just like on Meta. So let's play this clip from Satya Nadella explaining how he's thinking about. Jobs had the best line, I would say for PCs or computers was to say if you. It's a bicycle for the mind. Bill had a line which I liked as well, which was it's information at your fingertips. We kind of need now a new concept metaphor for how we use computers in the AI age. And you have one. And the one I like actually came from the CEO of Notion, which I like. You know that manager of incredible product. Yeah. You haven't bought it yet. But management, you know, basically a manager of infinite minds. That's a nice way to think about it. Right. If you remember, Jobs had the best line, I would say for vc, the. Manager of infinite minds. That is a good framing. I like that. It's nice that how Many minds. Are you managing Tyler right now? Right now it's probably like four, because I have every LLM actively running. You got to get those numbers up. Yeah. I've never really used multiple cloud instances yet. Okay. But, yeah, good framing and good to see. Yeah, it's interesting. Like the manager of Infinite Minds concept, obviously, that essay from Ivan, CEO of Notion, it went viral over the holidays. I think a lot of people in tech read it, but it did not necessarily break through to the Davos community. And so having Satya Nadella there just further popularize that concept, bake it down into a repeatable phrase that can prepare people for what's coming. Adapt. That is valuable. And that's what we're seeing here anyway. It'll feel like you're playing Starcraft now. Managing the Starcraft analogy is real. Zerg Army. Let's pull up this video that was posted by the New York Times Popcast. John Karamica posted this video yesterday. Somebody sent it to me. Actually Leaf at public sent it to me. And I was so confused because there's only about a hundred of these rugbys out in the world. Yeah. And we kind of knew exact. We kind of. Yeah. Like, everybody on the team, like, kind of sent them out. Yeah. So I was. I first thought this video was AI generated. It did. And it's about AI. So let's play the video. And I think we maybe don't want to do sound. Okay. So, yeah, it's about Bruno Mars and. And if we. If we. I don't know if we can scroll back to the beginning. Is that possible on Instagram reels? If you re. Post. But yeah, if you zoom in here, he's dancing and the dancing. This part, I was like, maybe this part's AI. But then when he sits down, you can see that he has a microphone, a lav mic clipped to his collar. And so everything about this says, this is not AI generated. This is real. And the hand motions are so accurate. And they're occluding, and he just looks extremely cool. Yeah, it's great. And so he goes on to talk, and we're gonna have him on the show later. Anyway, so Dylan messaged John, and we figure out that John accidentally purchased fake TVPN merch online. Be careful. But at least he's still looking great. Yeah. So, yeah, it works. Silver Line. But yeah, I mean, our saga to battle the fraudulent merch, the knockoff merch has been incredible. This site popped up and we assumed, okay, they're not going to take payments. Certainly they're definitely not going to print and ship anything and we'll be able to get this taken down in two seconds. And I actually sent them an email saying like, hey, hey, I assume you're a fan. Just so you know, we don't want this out there. Like I'd be happy to talk to you about maybe working together or something, but can you please just not use our brand and our trademarks and all this stuff out there on the Internet without talking to us first? Like let's have a conversation about this. Did not get a response. Not a fan. And I think some other people on the team were like, yeah John, that was never going to work. And they were right. You tried to go golden retriever. I did, I did. I was like, I'm just going to assume the best. I'm just going to assume this an over eager fan or someone who just is entrepreneurial kid who's just trying to. Just now this is a Canadian. We know it's a Canadian. We know it's Canadian. This is a Canadian. Yeah, anything could happen up there. We need a dot in the chat says you guys have a good line to Shopify. I assume most of them are hosted there. No, it's not hosted on Shopify. It's not. And at one point we've already gotten them taken down from, from a number. Of platforms where we know the CEOs and we'll say, hey, we have this fake site, can you take this down? They will and they'll pop up on a different site. And so it's been a true game of whack a mole. We've sent a number of like, you know, sort of, you know, demand letters. At this point we've really pushed it a lot but they, they keep finding a way. And at this point they are in fact taking people's money and shipping fake product now and fooling people that are discerning. And apparently it smells kind of funny, which is a big, which is an issue. The real merch is coming. The real merch is. The real merch is coming. We do have, we've been sampling a bunch of different stuff. Of course, Shopify, we need to put. Out like a maduro style bounty. Put a stack on there. Shopify of course 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. Trey says Mark Carney is selling fake TVPN merch. Crazy. He heard, he saw your post about him. His cortisol spiking his cortisol spiking and he was like you know what, I'm not doing anything anything about this Canadians. It's really concerning the number of people that didn't understand that my post was a joke. You're very online and you're very early to a lot of these things. You were trying to make aura happen back like over a year. Q1 of last year didn't happen. It was flopping a long time. Then you tried to make motion happen. I still think people are just waking up to motion. This will happen but you got to let it simmer. I mean I've seen this multiple times. But I just the idea of processing Davos as though it's like a bunch of kick streamers. It is very funny to look at the aesthetics and the whole kick clipping economy is crazy. I found one of those clips and then I went and looked at the account and it's like dozens of posts that are the same format where it's like they have a word in all caps and then they have an emoji and then they have this and it's very sensationalist. It feels like paparazzi on top of a kickstream or something. Very, very odd phenomenon. The thing that is real is the drama that's going on at Davos is not that different. You see Besant taking shots. There might be some spikes. They're definitely spiking each other's cortisol getting each other to crash out. I still don't even. I still don't even know what cortisol spiking means. Is it just like getting angry? What does it actually mean? If you get into any type of. Verbal physical altercation, it's like the fight or flight. Yeah, the cortisol floods your brain. Just makes you okay. Well yeah, gets your Anyway, private credit. Investors cortisol has been spiking because they are cashing out in droves. Redemptions by individual investors in funds soared at the end of 2025 after performance declined reviving questions about suitability. This is from Matt Woorz in the Wall Street Journal. He says for the first time since the start of the private credit boom, large numbers of individual investors are trying to get their money out. Several of the big funds eligible to wealthy individuals receive requests from about out 5% of shareholders to cash out at the end of last year, well above the normal volume. According to the SEC. One managed by Blue Owl got redemptions for about 15% of its shares. This was something where people were nervous about the data center build out cyber credit the expansion here, apparently what's driving it primarily was Asian clients asking for redemptions from Blue Owl. Specifically, the rising redemptions come at an awkward time for private credit fund managers and for the Trump administration as they push for new rules that would democratize private markets by encouraging their inclusion in 401k retirement plans for all Americans. Private fund managers, including Apollo Global and Blue Owl blame fear mongering about a recent spate of corporate bankruptcies. This is the whole cockroaches back and forth with Jamie Dimon and some of the folks in the private credit world, Blackstone, Apollo, Blue Owl because there was an automotive supplier that went bankrupt, bankrupt first Brands we talked about and there was a surge of withdrawals. Analysts say there could be a simpler explanation. Individual investors are falling into a similar pattern, a familiar pattern of selling out when an asset class underperforms expectations. These investors got really surprised when their dividends went down, said Robert Dodd, an analyst at Raymond James BDCs, the business development companies. They typically make high interest loans to mid sized corporations with junk credit ratings, using the interest income from those loans to pay dividends. A handful of these funds have cut dividends because the yield on their loans are falling in lockstep with benchmark interest rates for more dividend reductions will follow, Dodd said, likely prompting more redemptions. Total returns from five of the large, largest private credit funds aimed at individual investors declined to an average of about 6.22% in the first nine months months of 2025, compared with 8.76% in the same period 2024 and back in 2023. If you were in these funds you were making 11.4%, not bad. Money managed by BDCs has tripled since 2020 to about 450 billion. So there's been a big growth. And the funds still took in more new money from new investors than they paid out in their most recent quarter, a sign that they are still popular among investments, investors and advisors. Let's move on, but first let me tell you about phantom cash. Fund your wallet without exchanges or middlemen and spend with the phantom card. The Kobse letter breaking SpaceX is set to hire bank of America, Goldman Sachs, JP Morgan, Morgan Stanley to lead its IPO. That's a whole host of murderers. The IPO is expected to be valued as much as 1.5 trillion, making it the largest IPO in history. But the community notes put it in the truth zone because Saudi Aramco's IPO in 2019 sported a valuation of 1.7 trillion. Interesting very different. It raised 25 billion. 25.6 billion. Oh, yeah. Okay. So anyways, obviously Elon targeting more than that on an actual dollar basis. Well, there's other news about Anthropic. The revenue run rate for the end of 2025 was 9 billion, up from 4 billion in July 2025. What incredible growth. Iconic Lightspeed and Menlo are set to join the new funding round from techmeme here and Bloomberg reporting. This is maybe a deceleration. I was debating this with Tyler. I mean, they went from 100 million to 1 billion. They 10x, then they were going to 10x it again. So everyone was expecting 10 billion and 10x it again. And they landed at 9 on run rate. Yeah, I was trying to find. I'm seeing deceleration. Did they predict 10 billion? I don't know. I actually don't know the exact quote. I think Dario was pretty loose about it. He was just saying that we've seen order magnitude. Yeah, he was saying we went a full order of magnitude from 100 to a billion. We're going to do a full order of magnitude again. Is it a full order of magnitude to go from one to nine? And he said he only expects, like. You got around three or four more. You know, ten X's. He said that? No, no, no, no. I specifically remember him saying, like, we cannot keep that level of growth going. He said it would be pretty crazy if that. Yeah, I don't think he said we can. Oh, okay. He said it'd be pretty crazy. Yeah. Okay. Well, obviously he said if that continues, then, like, you're really in. The biggest companies in the world are on that scale. Yeah, I mean, at a certain point, there's no company that's doing a trillion dollars a year in revenue and you're only two orders of magnitude away from that. So we'll see. We'll see. But obviously Anthropic's been on a tear and the new round seems to be coming together nicely. There is other news from the information. Amir says that Anthropic's inference costs on Google and Amazon servers were 23% higher than the company projected. Of course, when you're growing so fast, you're probably willing to pay more for inference just to make sure that everything stays online. Better to service the demand and not get that frustration. People hitting rate limits, et cetera. Yeah, I think later in the summer there's a lot of news about how the Anthropic API is always down for issues. Yeah, there was that and then there were Also some FUD type articles about negative gross margins, about margins being really, really bad at these labs. Where's the value accruing? Is it all going to accrue to Nvidia? Last month month Anthropic projected it would generate around 40% gross margins from selling AI to businesses and application developers. But I think that gross margin came in a little bit lower. But you know, there's so much more that they can do to optimize. They're buying TPUs now, they're going to build new data centers. And also there's, it does feel like we're going to enter a world where inference is load balanced across a variety of, of semiconductor stacks. And so for really fast things you might be going to a GROK or Cerebras or you might, for more basic stuff you might be going to a legacy model that's cheaper to inference and all of that might be blended together into something that's more profitable. And then also the dynamic of the equilibrium in the ecosystem could just be higher prices. If it's delivering a lot of value. People are really happy with the value that they're getting from this. And Beth Jesos posted a real picture of Dario. Fact check. True that. You can pull up here. Zario Amog Da. The jawline is crazy. AI is good. Claude's gonna be able to do this to people. No, no, Claude will not. They're not launching an image model. Claude will have. No, no, I'm just saying you're gonna be able to talk to Claude and say like turn me into a, into a mogger in real life. It will, you know, orchestrate the jaw. Okay, got it. Got the double jaw surgery, leg lengthening surgery, purchase it all for you and just create a schedule for you. Yeah, call the WAYMO to take you to the, you know, to the dentist or whatever to do the surgery. Well, if, if you're in San Francisco on February 3rd, you need to call a Waymo and head over to the Cisco AI Summit. It's bringing together leaders from Nvidia, OpenAI, AWS and more to discuss the future of the AI economy. The whole thing will be live streamed and we'll be there for the stream. Hope to see you there. Apparently Google DeepMind is signing a licensing deal. It really is a licensing deal economy that we're in. Last year was the press release economy, this year very, very much the licensing deal economy. Humai, which builds emotionally intelligent voice interfaces and they're going to hire CEO Alan Cohen and seven of their engineers. Is this the same ghost ship story that we've seen so many times before. It's essentially an acquisition. Is that what we're reading this or. Yeah, that's. That's major licensing. I mean, that's. Yeah. The CEO's joining DeepMind. Interesting. I never thought of Google as particularly behind in voice interfaces. It doesn't. I'm wondering what else is going on here. I mean, certainly it's an important part of the stack if you have of an app that people are chatting with and talking to. Obviously NotebookLM was sort of a viral success, but I do really wonder about the longevity and the retention. It feels like they were maybe like, I don't know, 80 or 90% there in terms of people. Maybe it's just a speed thing. Maybe it's the actual cadence. I've talked to some people that. Didn't that team leave and start their own? They did, they did. But I mean, you gotta be able to backfill that if you're googling teams. Like, you have a great product with a lot of people talking about NotebookLM very positively. I think a lot of the NotebookLM functionality eventually got baked into the Gemini app. And so. But this. But there still is just. It feels like when you have a latency step where you ask a question, okay, prepare a deep research report and turn it into a podcast. Read it to me on the Roman Empire. And you have the option to just go to YouTube and just find a video of.
Out there. That's where I came from. So we took from a top down point of view, two years ago in Big Ideas, we took the position that because we were seeing it play out in the numbers, that the tech stack would change, that you have the infrastructure layer, which of course very strong. You have the platform as a service layer. Poster child for that. Palantir was gaining share, significant share. And then you had application layers, software as a service growing but losing share of incremental growth. We put that chart in and we had no idea how quickly this was all going to happen. You know, software stocks, more traditional software stocks last year were terrible and even this year they're still suffering. So AI is happening faster, I think, than anyone could have imagined. Anyone who thought they had time to adapt, I think has been mistaken. And, you know, I even see it with large enterprises when we go in there and ask them, okay, what's your AI strategy? You know, they, they treated it in the beginning like, yeah, we'll let our, our teams experiment with all these new tools coming up. ChatGPT. No, no, no, no, no. That's not the way this is going to work. It's, there's, there's going to have to be complete restructuring around AI or else you're probably going to fail. Maybe not outright, but you'll become irrelevant. So I think even ARC, for example, we have brought in Palantir. We're only 70 people, but we've brought in Palantir. And I can see we brought it into research first. I wanted to see how could we transform our research process. And we're already seeing incredible results in terms of productivity gains by the research team. But I can see that in the rest of the firm it is going to take, well, first of all, you have to have someone other than the cto. And that's true even in our case saying, we're doing this, we're doing this, our research is telling us this is the way the world's going. We cannot be caught flat footed. And so I really think you need the CEO, not the cto, CEO, CFO to drive this change. And I think, you know, many people think, oh, the results are so immediate and they are with our writing and editing, perhaps. But when you're trying to transform an enterprise, it takes, it's hard work. You have to collect all the data, even data that you didn't know existed. So, so you have to figure that out, you have to clean it up, you have to integrate it, you have to map out workflows in excruciating detail. And that all takes time. And a lot of people in this world are impatient. They don't see results right away. But I think the results are going to pay off. I think the way Palantir is doing, it just put its software on top of legacy, no rip and replace. And over time, it will usurp the role of all of those that inefficient legacy software. So I think it's going to take time in the enterprise. And I know there was the MIT study that said, hey, this isn't happening. This is a figment of Silicon Valley's imagination. And we disagree mightily on that. How.