TBPN - Marc Andreessen on M&A, AI Ads & His Advice for Young VCs, Shopify's Harley Post-Earnings, Google's Genie 3 Potential & More | Marc Andreessen, Harley Finkelstein, Anton Osika, Thomas Dohmke, Alex Jacobson, Nicolas Kopp

Episode Date: August 6, 2025

(01:41) - Timeline (04:28) - How Serious is Google's Genie 3 Model? (22:42) - The 5-Level Framework for Robotics (40:35) - Cerebras Proves it Can Actually Work (48:36) - Timeline (01:26:53) - Marc Andreessen is an American entrepreneur, software engineer, and venture capitalist. He is best known as the co-creator of Mosaic, the first widely used web browser, and co-founder of Netscape Communications. Andreessen later co-founded the influential venture capital firm Andreessen Horowitz (a16z), which has backed companies like Facebook, Airbnb, Coinbase, and OpenAI. A prominent voice in Silicon Valley, he's known for his essays on technology, including the viral "Software is Eating the World." (02:02:28) - Harley Finkelstein, an entrepreneur and lawyer, is the President of Shopify, a leading global commerce company. In the conversation, he discusses Shopify's strong Q2 performance, highlighting an 87 billion GMV and 2.7 billion in revenue, and emphasizes the benefits of going public, noting that it has enhanced the company's transparency and operational discipline. He also touches on the integration of AI in commerce, introducing tools like catalog search APIs and universal cart features to enhance shopping experiences. (02:33:55) - Anton Osika, co-founder and CEO of Lovable, discusses the platform's rapid growth, achieving $10 million in annual recurring revenue within two months with a 15-person team. He highlights Lovable's mission to empower users to transform ideas into production-ready applications swiftly, emphasizing its appeal to both entrepreneurs and larger companies. Osika also touches on the future of web development, suggesting that websites will become more personalized and optimized through advanced algorithms, while acknowledging a countertrend towards minimalistic design. (02:42:50) - Thomas Dohmke, CEO of GitHub, has been leading the company since 2021, following his tenure as Chief Product Officer. In the conversation, he discusses GitHub's significant growth, including surpassing 150 million developers and over a billion repositories, the widespread adoption of GitHub Copilot, and the company's commitment to offering developers a choice of AI models to enhance their coding experience. (02:52:14) - Alex Jacobson, co-founder and managing partner of 137 Ventures, discusses the advantages of private companies offering structured liquidity to employees, highlighting SpaceX's model of regular tenders at controlled prices to provide predictable equity value growth. He contrasts this with public markets, where stock prices are more volatile and less predictable, emphasizing that private companies can offer employees greater certainty regarding their equity's future worth. Jacobson also expresses skepticism about initiatives aiming to provide retail investors access to private company shares, noting that such efforts may undermine the benefits of being private, such as controlling shareholder composition and pricing. (03:10:55) - Nicolas Kopp is the founder and CEO of Rillet, an accounting platform for software businesses, and previously served as the U.S. CEO of N26. In the transcript, he announces Rillet's $70 million Series B funding co-led by Andreessen Horowitz and Iconic, discusses the company's growth and customer base, and highlights Rillet's AI-driven integrations and reporting capabilities that distinguish it from legacy ERP systems. He also emphasizes the importance of human oversight in AI processes and outlines plans to enhance product features and customer success initiatives to meet increasing demand. (03:19:41) - Timeline TBPN.com is made possible by: Ramp - https://ramp.comFigma - https://figma.comVanta - https://vanta.comLinear - https://linear.appEight Sleep - https://eightsleep.com/tbpnWander - https://wander.com/tbpnPublic - https://public.comAdQuick - https://adquick.comBezel - https://getbezel.com Numeral - https://www.numeralhq.comPolymarket - https://polymarket.comAttio - https://attio.com/tbpnFin - https://fin.ai/tbpnGraphite - https://graphite.devRestream - https://restream.ioProfound - https://tryprofound.comFollow TBPN: https://TBPN.comhttps://x.com/tbpnhttps://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235https://www.youtube.com/@TBPNLive

You're watching TVPN. Today is Wednesday, August 6th, 2025. We are live from the TBPN Ultradome, the Temple of Technology, the Portures of Finance, the Capital of Capital. Jordi, how are you doing? Have you been sleeping okay? You were up late, partying, hanging out with some of the greatest, the greatest capital allocators of generation. We won't say who you were hanging out with. But give me the redacted read.

How did it update your AGI timelines? How did it update your worldview? you hanging out with some of the most powerful people in Silicon Valley last night. I don't really know what I can say on this front. I feel like it was all set in confidence. Did it make you more optimistic? Did it make you more pessimistic? Did it make you feel like they're top?

Is this a top? Is this a local bottom? Do we have a long way to go? Are you more excited about technology? Are you more excited about capitalism? I left extremely bullish on enterprise SaaS. That's fantastic to hear. Even in the age of AGI.

Okay. I also listened to a friend of mine say that he wanted to drink the blood of his enemy. Wild guess who that is. I wonder who that is. Wild guess. But anyways, it was very, it was very inspirational. That's great. And I think everybody should have a friend that gets so fired up about winning in their market.

Yeah. That they have the sort of bloodlust. Yes. Well, you know who's fired up about winning in their market. Eric climbing at Ramp. Go to Ramp.com. Time is money saved both. Easy use.

Corporate cards, bill payments, accounting and a whole lot more all in one place. That is right. I wanted to highlight this post from Nikita Beer, who said, getting the Algo just right, and he shares a beautiful photo of the vibe aliner himself, the Tetragrammaton host. All these AI researchers talking about fine-tuning. Yes. We care about fine-tuning the X algorithm. I think people have it.

noticed how good it's been because you only notice when it's bad. I feel like the ex-algo has been incredible lately. What have you been thinking? Have you been satisfied? I think it's good. I think it's good, but we. Oh, Tyler's got, Tyler's got something to say. What takes? We train. Okay. Tyler's already and then we'll go to Tyler. So for us, yeah, yeah. We constantly are sending our favorite post to each other. Yep. And so we're just like manually training. Crazy reward signal to to, to show us more of what we like. Also, it's very good at that. I've noticed that like, I really like retweeting stuff. If I just see, if I see any post I like, I'll like it. But if I see any post that I think, oh, that's my friend and they're just trying to like mention something,

even if I don't think it's the most incredible, hilarious post, I'll repost it just to amplify it. Just because I'm like, yeah, yeah, show some love. So I've been more liberal with the like button, more liberal with the retweet button or the repost button. And I think that's also amplified the algos. So I feel like my algos dialed, but Tyler, what do you think? I feel like I'm getting like massive amounts of Normie slop. Dude. Are you engaging with Normie Slop? No, I always press.

Not beating the allegations. You know that these algorithms are reflections of yourself. You just, you just told on yourself. Normie. Normie. Normie. What is an example of a Normie post? Like, are we talking thread boy stuff?

Are we talking like thread boy tier? Like here's how to start your first company. You're like here's the secret. No, like just like not anything that has to do with technology or business. I never engage. It's like about like some like dating thing or well. I swear I don't engage. I always, I negatively engage.

I say I'm not interested. You say not interested. I say I mute this account. Okay. And they just feed me more and more stuff. It's just because it'll be like 100,000 likes on some random tweet. It's like I am getting I am getting some of that. I try to not give it any of my attention.

So I try to move past quickly. I don't want the algorithm to I don't want to send the signal that I want to more of it but it's tough yeah well I don't know maybe it's a skill issue maybe it's just Nikita beer working his magic in the algo but thank you for your service Nikita we know it's a hard fight getting the algo just right and I think uh I personally think you've been doing a great job so thank you and you know who else has been doing a great job restream one live stream 30 plus destinations multi-stream and reach your audience wherever they are go to restream to get started show wouldn't be possible without that

truly would not be possible without it. We talked about Genie 3 yesterday. I wanted to noodle on it more. So Genie 3, if you're not familiar, is a new frontier world model from Google. And we played around it with a little bit yesterday. We saw some demo videos. We don't have access, right, Tyler?

No, like no one has access. Just like engineers. Yeah, so this is at what, like this is Bard tier software. We never got access to Bard. And then Gemini was the consumer application. And then there was Palm, too. Palm was their big multibillion parameter large language model that Google did not release to the public, but they started writing papers about.

And even the transformer paper, that was just an architecture. There was no product associated with it. It wasn't until February or March of 2023 that Google came out with their first consumer chatbot for public consumption, correct? Yeah, I think it's mainly, there's probably a bunch of safety stuff they got to figure out, Especially Google, they're pretty safe about that kind of stuff. That's a good point. That's what I want to debate today. So, interestingly, Google creates the transformer.

They create Palm, they create Bard, they have all of these incredible AI researchers, DeepMind's fantastic research lab. They're scaling up large language models. Transformer-based large language models. They are on the correct path in the tech tree, and yet they don't release the consumer product fast enough. ChatGPT releases in I believe November, I think it was November 30th of 2022, maybe December 1st. So you can think about it at the beginning of December 2022. Google takes three months to respond.

They respond with a product that's pretty close. And yet consumer adoption is power law distributed and chat GPT has been the runaway success in consumer. And so the lesson there is you've got to be first. if you have the technology, you've got to get it out in the public, even if it's messy. And I mean, maybe this is apocryphal. Maybe this is like the wrong way to frame it. But like there is a story out there that people are telling in the media in books that basically says Sam Altman got fired because he launched Chatsypti. Right?

They say the board, he didn't go to the board. He didn't go to the board and say, hey, we, I want to launch Chachabit. He didn't tell the board he was launching Chachypt. SaaS product. Yeah, I want to make the next consumer internet company. I want to make something that makes a trillion dollars a month. Is that okay? I want to launch an app. I want to launch an app. Is that okay? Is that okay? That's basically, that's basically what they do. So, yeah, not a fan of the board here. But the, the, the, the interesting point is that is that that move fast, break things mentality. Like, when chaty bd launch, people forget how bad the first version was because, I mean,

it was amazing. It passed the touring test, but it was built on chat GPT, or it was built on GPD 3.5-002 DaVinci, right? And so the model was hallucinating constantly. It had no access to tools that. This crazy knowledge cutoff. So you couldn't ask it anything about the last six months or the, or what's going on today. It was, it was like pretty useless. It was magical and it was useful for some things, but for like 99% of prompts that we hit chat, with today, they would completely flop on the V1 of chat GPT. Yeah, and I remember when it launched, there had been a bunch of companies that had been building on GPT doing like text generation for different use cases. There were a number of companies that were generating text for marketing assets or emails or things like that or essays.

And those companies had absolutely ripped. They had insane revenue ramps. And when ChatGPT launched, there were still some people that thought, oh, you're still going to be able to build a business here around just like generating text for the like niche use cases. Yeah. And then like a month or two in, it was very clear that like all those companies had gone ramped and then like basically, you know, right back to where they started. Totally.

I mean, I talked to a YouTuber at the time, someone with like, I think many millions of followers in the, uh, in the true crime horror genre. And his whole, his whole channel was designed around finding an interesting, uh, case or, you know, uh, crime. and then telling the story. An interesting crime. Yeah. I mean, it's a true crime podcast.

We saw odd lots. Like true crime podcasts are the number one and the number two biggest podcast in the world right now above Joe Rogan, at least in the charts. And so the like true crime, it's all about storytelling. It's all about just writing, getting the facts together, assembling them into a script that makes sense on YouTube. This guy, he's not like the most insane writer. He's just like somebody who figured out how to write for YouTube and do the, the video production and he has a good voice for it. He's a great voice for it actually.

And he has good lighting and he does all this, all this great stuff to make the stuff work really well. He's good with thumbnails and titles. And so he'd grown to the point where he could just put up a million view video like every day. And he was like, I'm gonna use Chiatupit to write all my scripts. And he did that for like a week. And then he just completely quit doing YouTube videos entirely

and switched to an interview show. This is the guy who interviewed Andrew Tate. He interviewed a bunch of people, but there's a couple like super viral, like, clips of like hustlepreneurs that he's the other guy interviewing them. He's the reason that he, that like this clip exists on the internet. He's a really nice guy. But it was just funny because he went like so hard into this and the product wasn't there yet. And now maybe it is, but still people, people still can tell the flavor of text written from GPT, 4.5 even.

But at the same time, the actual use case was information retrieval, transformation of information, translation, all the things that we use chat Chepti for for 30 minutes a day. I'm not telling it, write me a novel. I want to read a great novel. But I am saying, you know, pull all this data from the internet and from all your resources and put it together in a table, teach me this thing. It's super helpful for knowledge retrieval. And when you look at Fiji Simo's five things that Chachapit is going to focus on. like writing new books was not one of them. And I think that's by design,

because that's not where the models excel. We're in this age of spiky intelligence. And so you want to do, you want to let the models cook on what they're good at, and they're good at knowledge retrieval and they're good at, hey, my knee hurts, here's some symptoms. Let's go look up with another knowledge retrieval task. And they're also good at just chatting back and forth, having a conversation and that's the therapy angle.

Yeah. But my question is, Jeannie III, Google is clearly, at the very front of this from a research perspective. Yeah. And so it's super high fidelity. It looks HD. It's rendering it 20 to 24 frames per second. You can prompt it.

It seems super fun. There's this world memory. They've clearly gotten a number of fundamental improvements. Like, you know, you watch what has happened in the LLM world with adding reinforcement learning, adding memory, adding tool use, and a web browser and a Python repel and a bunch of other tools that have made it great. Clearly in this Genie 3 model, there's a bunch of those tools. Like, you know, they figured out how to do memory. It's probably not just one simple algorithm. It's a bunch of things working together. They're crushing it. Deep Mind is doing what they do best. They've done

the research and they've created something awesome. The question is, what are they going to do with it? They can't get in the chat GPT world again. They can't let another lab do the same research and then productize it first and then be playing catch up. So three months is enough to lose the in the chat world? Yeah, I think one question I have, is this is this the kind of tool that's sort of like a Google Earth that I remember as a kid discovering Google Earth and realizing that you could just explore the real world from your computer and it was this like magical experience and I definitely spent a bunch of time just like going around, bop it around different places

being like, I'm going to drop into this continent, I'm going to drop into this city. And that was really cool. I think the question of Genie 3 is like, is this something that people are just going to be you know, prompting themselves, generating worlds. Is this the kind of thing that video games will adapt and build on top of? Is it more of a developer-focused product? Unclear so far. We should probably have Logan on the show to break it down

because I think there's a lot of people, game developers specifically that would look at this and realize just the incredible potential of this if you can integrate some type of game engine into this and rules. Are people going to be doing design? Can you imagine? I mean, it's an interesting thing. When you think of like first person shooter games, when I think of my memories playing those games,

there's like a few key maps and a few key games that were just like iconic. And like people loved playing in those maps. And I think the interesting thing is when you have infinite optionality, you can imagine a world in the future where Call of Duty lets you just prompt battlefields, prompt, you know, different kind of arenas. settings for different gameplay, but it'll be interesting if people still gravitate towards certain maps just because it's fun to have consistency with certain games. Completely agree. There will definitely be a power law and the only way to

actually realize that power law and find out what's at the long tail of that power law is to get it in the hands of millions of people. Yeah. And that's the same thing that happened with chat GPT. And so beautiful website by the way. I wonder if they designed it in fig, but think bigger, build faster. Figma helps design and development teams build great products together, get started for free. So my point is, is that I agree with everything you're saying. Is this going to be just like this meditative hangout space? Will it be something that people put on in the background and then they're studying or will they be, you know,

maybe people will be using this as a development tool to then go and hard code a new video game or they'll use as an exploration canvas in the same way that people use, you know, A lot of people use generative imagery as I was listening to a talk about this where someone was like, like, if you're an artist, you probably can't just just turn in a prompt as your final work. But you're going to be on behance and you're going to be on, there's a bunch of these services that like. Midjury now. Yeah, you're going to be on, before you would like scroll Instagram for references. Yeah.

And you'd put that into a mood board and you'd be like, okay, this is the brand that we're going. going for. Now we're going to create our own and it's not going to be derivative of any one thing. We give you look at like, I don't know if we, I don't know if we ever actually did a mood board for TPPN, but you could imagine there's like, you know, there's the, there's the hardwood aesthetic, the mahogany, the official wood of business. Then there's some racing livery. There's the green that you really liked that you were like, let's do green and that was great. And then there's, you know, a little bit of sci-fi. There's a gong for some reason. You mix all these up and all of a sudden it's not just, we're not just doing

a racing team schick. It's like racing team plus gong plus hardwood plus all these different things suits and you add all that up and you get something new and people used to do that just by pulling imagery on on Pinterest and Instagram and B-Hance an arena was one of them

I think and now people use generative art to do that and you can see that happening there but the main thing is that like there's zero chance that we figure that out in the lab I think you can only figure it out by actually getting it in the consumer's hands and being very responsive to it.

But yes, there are safety concerns, but this is a game where the most risk-taking person, the most founder mode company wins. And that's exactly what happened. The two use cases for Jeannie that I'm interested, like I would be interested in playing around with. One is like a voice prompting functionality where you can just be sitting with the screen

and just be live prompting. Like no hands on the keyboard. No hands on the keyboard. That's interesting. you're saying a dolphin just skateboarded by and a dolphin is going to skateboard by, right? Then you're like, oh, it's like, and there's a La Ferrari that has wings. And it's like taking off into the sky, right? So you're just like giving people the ability to like have this almost like lucid dream effectively.

Yep. And then what was the other thing I was thinking? Blanking Tyler, did you have a thought? Yeah, I mean, I think obviously it's like super cool as a consumer use case. like I want to play with it. But I mean, I think you could definitely make the case that the main value of this is it being used as training data for like robotics.

Like if you train a humanoid robot off this, that's like everyone says like the main issue, like why don't we have humanoids yet? Because it's a data problem. We don't have like insane videos of first person like doing every single task, right? I wonder, yeah, I mean, I don't know enough about it, but you'd think that you'd be able to generate

endless procedural worlds with just traditional game development pipelines on Real Engine. Houdini, you can define endless corridors. There's that game, No Man's Sky. Have you heard of that? It's like every single planet you go to is procedurally generated.

There's an unlimited canvas. And so you'd think you'd be able to walk around with that and then just do some style transfer on top of it to get to something that is trainable. But maybe this is better. I don't know. We need to talk to some some robotics people to see if this is actually like a bunch of breakthrough.

Yeah, when we talked to, we talked with the team from physical intelligence and they said they have effectively they're like generating a bunch of live real world data. The question is how valuable is this synthetic data? I mean, Waymo, Tesla and Cruz all had very serious simulation teams that were developing. They would they would go and map the world, take all the pictures, reconstitute that into essentially a video game and then the cars could virtually drive around. And then they they did have the trouble of like, they kind of needed hallucinations because the first time that like

it wasn't just like a dog running out and thing, but like what happens how much train data is there on like a cow going in the middle of the road? Like that happens like one in a hundred billion vehicle miles traveled. So you could have like petabytes of data

of cars mapping the road and maybe the big Tesla data set has like one image of it and so you can train on it But in a simulation, you're going to get a lot more crazy hallucinations and then that might be valuable. I guess my point, Tyler, is that like you're describing like the B2B use case. And I think that there might be a consumer use case. Here's a consumer use case. And the consumer use case will be more monopolistic.

So, so yes, like Open AI Anthropic are duking it out in B2B, right? And it looks like that will be a lot less, a lot less winner take all than chat. like consumer chat like the the interface the the new consumer company with all the aggregation that chat chpt has is going to be incredibly valuable and then all the hyperscalers will have a frontier model that they can sell at you know somewhat competitive rates and they'll make a lot of money and they'll do great and it's a big new market so that could that that could happen here the thing is that if they if they start selling this to robotics training companies what's stopping someone else

from training a similar model and then selling it to another robotics company, it becomes more oligopolistic, I would think. Yeah. Well, isn't it like, okay, if you're super AGI-I-pilled, then- as you are? Then, like, consumer doesn't matter as much because they're not doing as, like, economically valuable things as a, like, as B-to-B would, right? So in that case, then B-to-B is, like, way more important, even if it's not, like, as monopolistic. No, Tyler, so much of the economy is raw consumption. It's the will of the American.

Does consumer matter in an AGI? So here's the other use case I was thinking. I think it would be cool if you could upload some images and some video and like recreate a scene that you're in. So think about like your oldest like first birthday. If you could take like three images from that like, you know, celebration like cutting the cake or whatever and just like drop it into the genie and just like recreate the scene and be able to just like witness it again. Yep. in virtual reality, that would be an absolutely wild experience. And I think a lot of people would do that when, you know, just reminiscing on.

And that feels entertaining and important whether or not I have a job. Like if the robots take all the jobs in this ASI future, I still want something to do with my time, right? So like, why not go hang out in Jeannie 3? Tyler? That's like pretty blackpilling, right? Why? It's like wireheading all the time if everyone lives in the simulation you don't need to be wireheading You can just you can just watch a movie like is that wireheading? Yeah you have more free time you can you can watch dunkirk

You Jordi's gonna have the best time in the singularity He's gonna be like I've never seen a single movie I've so many movies I'm gonna be like I need a generative slop I've seen everything And George's gonna just been building up an incredible backlog Incredible backlog almost a 30 year backlog it's gonna be amazing you have you have you have you have have a decade worth of films to watch and enjoy and discuss. It's going to be a fantastic post-singularity life for Jordi. Well, until we hit the singularity, you're going to need to stay compliant. You're going to need to get on Vanta, automate compliance, manage risk, prove trust.

Continuously, Vantus Trust Management Platform takes the manual work out of your security and compliance process and replaces it with continuous automation, whether you're pursuing your first framework or managing a complex program. So speaking of robotics, Tyler, I want to talk about the five-level of robotic autonomy from semi-analysis. This article dropped when we were in New York last week. We didn't get a chance to cover it, but I thought it was pretty interesting because just like with autonomous cars,

they're, like we had the five levels of autonomous cars, like level zero is basically no autonomy. It's your Carrera GT, fully, full driver engagement, no, no, no, no traction control. No traction control. Is that level one autonomy? No. Level one was like cruise control basically, lane keep assist, the basic stuff. Level two was it does lane centering and adaptive cruise control.

So it will break if a car is slowing down in front of you. It'll speed up. But you have to be very engaged and you have to be prepared to take over at any time. Level three is the car is basically driving itself fully stops at stoplights, does everything, But it can tell you, hey, you got to get back in the driver's seat. And then level four is the human doesn't need to be involved at all. And level five, I think like no steering wheel, something along those lines. But there was always this question.

You know, the humanoid robotics discussion has become really, really focused on like level five, the most insane. We're a GI pill. It's going to be humanoids that can do everything. They can do plumbing. They can move perfectly. But clearly there's going to be a rollout here. And so semi-analysis did a great job of kind of breaking down like where we are because there's actually a Continuum of robotic capabilities and you'll see some of those in that image

So level zero is scripted motion. This is high accuracy, high repeatability But this this unlocks 24-7 automation and high throughput. So this is when you see the car factory with the robotic arm that just moves the windshield picks it up puts it on the car moves the windshield puts it back like that is perfectly scripted in a computer simulation, like the key-framed animation essentially, does the same thing every time. There are no cameras on that robot whatsoever. And so if you are next to it and it decides to go like this,

it will hit you and it might kill you. So you have to be very careful. So there's typically like fencing around these and they have these. Have you ever been to Hadrian and seen the light curtain? Have you seen this thing? So the light curtain is a bunch of lasers. And if you break the beam,

it will immediately trigger stop everything on the line. So it just knows if you go. Don't kill the humans. Exactly. Exactly. But there's no cameras involved. There's no machine learning. It's purely just a big powerful arm that is scripted to do exactly what it can.

But that's extremely useful. And we use tons of these. And these robots have been around for, you know, decades at this point. And they are definitely out there. They're expensive. And you need constant oversight. And interestingly, even though they do.

in theory enable 24-7 automation, they often don't run 24-7 because they need a person managing them all the time. And so, like, there are tons and tons of facilities that actually just straight up shut down the robots when they go to lunch, which is kind of crazy, but that's just the nature of these things they're so precise. And then we've talked to other folks about, like, you've got to change out the motors, change out the grease. You've got to put the oil can in the in the joints, basically. Yeah, and the big, you know, we've talked to a few people around trying to get a sense for where humanoids will be valuable and people that run factories with traditional robots,

you know, these sort of, the arms that you're describing, talk about just how often they're having to replace parts and motors. And when you have a humanoid, you're like adding it, you're basically like multiplying the number of motors that need to be functioning properly in order for the robot to be like online and productive. and that's just going to be one of the challenges that anybody building humanoids is going to have to deal with. Techno Chief 2000 in the chat says, good morning, brothers, good morning, Techno Chief. I hope you're on graphite code review for the age of AI.

Graphite helps teams on GitHub ship higher quality software faster. Now back to the semi-analysis breakdown. Level one is intelligent pick in place, and this is important because this is also something that we've had. So you put motors on a gantry, X and Y access, and then you put a little grabber. everything from like a suction cup to a little like actuator and all of a sudden if you just say there is the Diet Coke move this over pick it up it's like the the claw game you know the claw game at the carnival so you put a camera some object detection and then you map that and this is like very basic open CV computer vision very I mean it's amazing it's miracle but it's still like it's very doable at this

point there's plenty of like open source packages there's a bunch of companies that have done this And it's important to define this as like, this counts as level one autonomy. There is machine learning in the process. There is computer vision at work. And because of that, this is a tool that can be used and has benefits and costs. But is it end to end? I mean, it doesn't matter. But sometimes it can be.

Sometimes it isn't. But I think the more important thing that you're getting to is that there are a lot of people out there who are showing humanoid robotic demos doing things. And they're saying, hey, we have a level four fully autonomous. humanoid robot, it can do anything, and they show it doing a level one task. And it's like, I remember seeing some, some demo of, of a humanoid robot just taking parts from one place and putting them to the other side. And it's like, well, for that, you just need a robotic arm with scripted motion.

You need level zero autonomy, actually. Like, you haven't, like, you're not demoing it in the way that would actually necessitate that. And so even if you did it, and even if you are successful, it's a waste of money. because you'd be better off using a simpler robot for that. And the analogy here is that, like, LLMs are clearly incredible to the point where I believe the OpenAI IMO model was purely text-based. Like, it didn't have access to Python. It didn't have tools, right?

And so, but at the same time, no one cares. Like, why not given LLM tools? Like, obviously chat GPS is better when it has a browser, when it has a Python repel, when it has, you know, access to a database, when it has just things that it can actually look up a calculator. Like, it should, there's no reason to artificially constrain it just because, like, oh, that's more impressive from an AI research perspective.

And so in the same way, a humanoid robot should also have access to a standard robotic arm if it needs to. And like, you should, if you're running a factory, you should say, hey, for this task, we don't want to use the humanoid robot. In the near term, John, we want to use conveyor.

Potential job opportunity for humanoid robots. to just turn the regular robot arm on and off. Just be the guy that presses the button. You're joking, but it's real. It's 100% real. Just like the LLM can try and memorize everything on the web. That was the first version of ChatGPT. Didn't have access to a web browser.

Now, if I ask it, you know, how many, how much sodium in a can of Diet of Coke, it probably has that memorized, but also it can just Google it and look at the results. Well, maybe not Google it. Maybe Google it. We don't know. We don't know how they use it. We don't know. We don't know. I don't think they're Googling it. I think Google does not allow other people to use like Google searches and API. But who knows? Big question. I never know. I can use Google. Why can't I tell chatGBT agent to use Google. Yeah. Open up Google.com. ChatGPTGPT agent. Go to Google.com. Download, hit query Gemini. Search this for me. Search this for me. And so there are other levels and I think these other levels from two to four are going to become increasingly more important because these are the ones that are more on the frontier and they're more and they require more and more and more and machine learning. So level two is autonomous mobility, scene understanding, higher order planning, long horizon reasoning, agile movement capabilities are open world navigation and traversing. and traversal. So these are like the Boston Dynamics robots.

So when you have that like Boston Dynamics dog that runs around, that dog is very useful for, let me see. So level two is autonomous mobility, robots that can understand the open world navigate and traverse various terrains. And so early production phases for inspection and data collection rules,

send a drone out somewhere, construction sites, oil and gas refineries, critical, infrastructure. The default example is like you got the nuclear power plant and you want the robot dog to run around and take images of everything. But there's a bunch of different places and there are a bunch of different companies that essentially offer level to autonomously mobile robots now. And it's real. It's like it's a real thing. It's just a narrow use case. They can't do everything. But if you want to send the dog out on patrol, you can. and of course

occasionally you'll want a humanoid not a dog for certain things but this is not it can't do everything but it can get over rough terrain it can climb stairs it can do things that aren't pre-programmed where it has to understand

okay there's a stair coming I gotta move my paws it's a Chinese company that's been showing off a robot that can crawl, swim and fly they're gonna just like walk it can walk around

like it's like a spider and then it can jump into the water and swim through the water like actually go under and then come up to the surface and actually take off into the air. Can it do sales tax or does it need to use numeralhq.com? It will have to use numeral. Okay. Well, we'll tell the robot, put your sales tax on autopilot, spend less than five minutes per month on sales tax compliance. Go to numeralhq.com.

So that's level two, autonomous mobility, the robot dog. Level three is low skill manipulation. These are getting more humanoid-like robots that can perform basic, non-critical, low-skill tasks. The unlock here is generalizable manipulation, advanced pick in place as a capability and mobile manipulation. So you want a robot to go and pick up a box and move it across. It's not a defined zone where the pick in place robot is going to sit. You can kind of set it up anywhere, give it some basic rules, but you're still going to be, you know, somewhat piloting it, giving it general, general like, you know, overview of what it needs to be doing.

And so this is actually already deployed in some pilot stages in kitchens, laundromats, manufacturing, and logistics. And these are like bipedal humanoid robots. Now they don't have five fingers, five toes. They don't look perfectly human. Yeah. But they can actually carry box. Yeah. And a big thing is even a remotely operated humanoid that had the ability that had fingers that could do traditional, you know, activities that were more complicated. Yep.

Even a remotely operated humanoid could be completely valuable. right because you could have you could have somebody somewhere else say like pick up every single leaf leaf in my backyard and put it in the green bin and there's people and there's like problem there's like a bunch of like a bunch of use cases you can think of i mean the big debate is around um what is the level of human involvement in waymos right that the allegedly uh there's there's at least one person kind of overseeing every active waymo remotely. I think that number is blended across the entire org. So like at a given time, there's probably like one human looking at a screen with four Waymo viewpoints on it and they can kind of jump back and forth. But then there are other people like monitoring the network and checking on the tires. I guess my point is that like having somebody remotely observed the activity of a robot while we're trying to get to the point of full autonomy is totally worthwhile and is what people should expect right this we yeah this should be

pretty searchable Tyler can you look up the number of employees at Waymo and then the number of Waymo's on the road and get the ratio there because the real test of robotic leverage should be more vehicles on the road than employees at the company but it's It seems like it might be met in neck. So there's 2,500 employees around, and there's only 1,500 Waymo's. Yeah, so that's the number that we're hearing. When people say like 1.4, 1.5, I think that's it.

I'm sorry. These could easily be contractors that aren't technically employees that are hired to work around the clock for every single, you know, active Waymo. So. But still, you would expect, I mean, the bottom line is that is that. It's still a massive step up from not having somebody that's just. just sitting in the vehicle observing it real time. Yeah.

But even if we assume that there are zero contractors in there, like the job displacement narrative, at least at this moment in time, is completely debunked because we put 1,500 taxis on the road and it took 2,500 people to do that. Right? Like now, obviously the technological vision is that you'll have 2,000 employees.

$1,000. Yeah, yeah. Just like, you know, you don't need, You don't need, just like you used to have, you know, one paper boy delivering a newspaper to every city block. Now you have, you know, a number of engineers that serve, you know, social networks and they just kind of go out and they're distributed across the internet. So, but that is an interesting, that is an interesting benchmark. And something that should be, like the ratio of vehicles on the road for Waymo to employees at Waymo is an interesting stat to follow. I agree that the contractor thing is important.

Anyway, let me tell you about profound. Get your brand mentioned by ChatGPT, reach millions of consumers who are using AI to discover new products and brands. Profound is our latest sponsor, and we're very happy to tell you about Profound today. Get on it if you're running a business or a fund or anything like it.

You should probably know how people are understanding your business through LLMs. Profound makes it super easy. So the last level of a robotic. autonomy. Level four, force dependent tasks. These are things that must be done very precisely. Robots can perform delicate tasks at this level that require force and weight understanding, i.e. binding a phone in a pocket sounds easy, very difficult if your robot. Doesn't sound that easy. I think everybody here is experienced. Yeah, and you actually drop your

phone a bunch. It's one of the harder things you do. Driving a screw in on the correct threads. So this is purely in the research phase. The capability is that this will unlock are delicate, forced dependent tasks, fine grain manipulation, and this is purely in the research. So there are different, and one of the examples that they give is like place the soft bread on a plate. So you can imagine that being valuable in the kitchen context. Or another example. The robot's just crushing the bread every time.

Another example, trying to pass steak to your friend that are sitting in a different part of the plane. Yes, yes, yes. That requires a force dependent task. Definitely. Because you might get tackled. And when the flight attendant comes up to you, you have to put your hand on their chest, not fully threateningly so that you get on the no-flight list, just gently. So they don't.

But you want to touch them. You want to place your hand on their chest. Just a little bit just to let them know, this could go one of two ways. Probably shoulder. Yeah, probably shoulders more respectful. But you want to let them know that you're ready to die. I took custody of that steak. Yeah.

Yeah. That this is happening one way or the other. This can go down in one or two ways, but you don't want to actually cross the threshold. So, you know, if you're passing, if you're passing steak to your buddies and economy, don't use a level three robot. And I have a tough time. Because the robot will just shove right through them. Other examples include unstacking single plastic cups from a stack.

Any fraternity member should be familiar with that, but the robot is still working. So there's still full employment for the solo cup distributors out there saying I'd love to see a robot Set up set up a stack solo cups What's the beer pong? Yeah, robot's You know not gonna be able to set up beer pong for a while, but they're researching it Yeah Place an egg in a pot place a bag of chips on a plate twist and lift a bottle cap literally like open a bottle of medicine Actually one of the more useful things very hard for children do very hard for children's very hard for children

They're working on it Yeah. Yeah, what's the arc AGI of robotic tasks? That's what we need. Maybe it's taking the phone out of the pocket, something that anyone can do, even a child, but it'll be the last task solved by a humanoid robot. But I mean, you can imagine, like, twisting a bottle cap is actually, you know, we have an aging population. We need more help in the elder care market in hospitals and nurses. in the nursing contexts. And if you have a robot that can dispense medicine effectively,

you can imagine that being very valuable. But right now that would require an entire rebuild of the way we distribute medicine because we distribute it in child safe and also robot safe bottles. So anyway, let me tell you about linear. Linear is a purpose-built tool for planning and building products. All of these companies that are going to develop level four,

forced dependent task robots, they're going to need to be on linear. It probably is the system for modern software development It allows you to streamline issues, projects and product roadmaps. So moving on to the next story. There's a lot going on. Well, the big thing, we should talk about Cerebrus, because this was a company that was kind of dogged for a while.

Are you familiar with this company? Yeah. So Cerebrus makes a wafer scale, GPU, a wafer scale chip. When semiconductors are made, you go and you want an Apple, you know, Apple silicon chip or an Nvidia GPU, there is a wafer that is grown of silicon, and then the wafer is etched with all the different transistors on there, the three nanometer, the four nanometer, the six nanometer transistors are put onto the chip, and then the chip is the wafers sliced up into individual chips, and the benefit of that is that if one of the chips is bad, one of the

transistors is off, that chip's not working, you can dump that one and the rest are going to be fine. But with wafer scale computing, with wafer scale chip manufacturing, if there's even one defect anywhere on that wafer, you've got to throw the whole thing out. So everyone was very worried that the yields would never get high enough. And they were saying cerebrus, also there's other tradeoffs with when you're at wafer scale with memory and inference. But what we've heard and what Andrew Feldman, the CEO is saying is that a lot of the crazy bets that they place, they're paying off and it's working very well. So Andrew Feldman says in 2016, Sam Altman and Andrew met.

Open AI was a vision. Cerebris Systems was a PowerPoint. Sam and the Open AI founders became one of the early investors in Cerebris Systems. Good to see Sam getting a markup. He needs it. He needed that. He really needed that.

Badly. In the following years, the Cerebris and Open A.I. Frequently met. The teams frequently met to explore working together, But the timing was too early. Large language models hadn't been invented yet. This is what being early looks like. Truly.

Like developing a foundational technology. That's dependent on another foundational technology that doesn't exist yet. Insane. Today, the story comes full circle. Open AI just released its most powerful open weight reasoning model. And it runs fastest on Cerebrus Systems. Not a little bit faster than the competition. It smokes the competition.

Running our third generation wafer scale engine GPTOSS-120B runs it up to 3,000 tokens per second, the fastest speed achieved by an open AI model in production. Reasoning that takes minutes on Nvidia GPUs can take a single second on Cerebus systems. Good things take time. Congratulations to the Cerebus team. An overnight success. Been in the trenches for a while. But I think this is interesting.

We've talked to some folks who have started using Cerebrus to speed up inference for specific use cases within larger AI systems. I think that's very interesting. And I think that we are, we're, like, it's very easy to collapse the narrative of AI into, you know, there's going to be one single technology, one single path, one single implementation, but as we're seeing with the development of the new GPT models and the new open AI, chat GPT functionality, It's like you squeeze out performance in a bunch of different ways. You give it a browser.

You give it reasoning. You give it, you know, a Python repel. You give it the ability to write code. And like you add all that up and you get something great. And it's the same thing on the inference side. Sometimes you want to do things on device. Apple's, you know, kind of steering towards doing that with Apple intelligence. Sometimes you want to do it on a really big cluster of Nvidia H-100s or big GPU cluster.

And sometimes you need to just run something extremely fast, and that's what Cerebrus is powering. So I'm interested to understand what 3,000 tokens a second on the locks. Yeah. Because I was thinking about it in the context of that famous Amazon quote. I mean, an example being like right now if you want to use, like if I open up chat GPT and I want to use it as functionally like knowledge retrieval, Google search, then doing 4-0 and just getting

something like relatively instantly makes sense. And then, but I know I would get a better result with O3 Pro. Yep. And, but if you're using O3 Pro for like raw knowledge retrieval in terms of like, I want to understand this fact or this person is just not a great experience, right? Because I don't want to figure it out in like a few minutes. I want to figure it out now, right? And so theoretically you could get to the point where you could get O3 Pro quality answers.

Yep. relatively instantly. I want two things. The first thing is I want you to sign up for fin.a.I. The number one agent for customer service, number one in performance benchmarks, number one in competitive bakeoffs, number one ranking on G2. The second thing I want

is I want when I go to chat GPT and I type in a query, I've been extremely frustrated by the model picker because I'm constantly going back and forth between 4O and 03 pro and one takes five seconds, the other takes 15 minutes, and sometimes I

accidentally hit something with 03 that I should have gone to 040 with and then I have to open a new one and then I copy my prompt over and put it there and then I'm like actually yeah that's a great one like this was Google a bull and so it just pulled it into the nice format what I would love is is a system where I can where I can hit it with a prompt and it goes to Cerebrus and it gives me just hey off the top of my head here's what I know here's the basics of what I what I'm pretty confident about But I'm going to keep working on it. So I kicked off an O3 Pro heavy duty reasoning. I'm searching, I'm collecting data.

I'm going to get you the best possible answer on this. But while you're waiting, here's the thing that I can turn over in two seconds. Here's what I know right now. Yeah. And I think that that would be amazing. Yeah, and that's the same experience you might have working with somebody where you ask them a question and they're like, I'm 90% sure it's this. Give me 30 minutes.

I'll come back to you with confirmation. We'll figure out the plan from there. Exactly. Exactly. Exactly. And that also allows you, if somebody says, I'm 90% sure that it's this direction, that allows the person that you're interacting with to go and think about that path. And now there's a 90% chance that the work that they do. Without necessarily doing something mission critical that you can't make mistakes. Yeah. And then if they need to adjust course, they haven't burned a bunch of times. So I'm excited for more advanced model routing. We've talked about the mixture of experts. we, Tyler, what was the GROC of evolution? It was like a mixture of models. Mixer of models of experts.

Well, yeah, what was the latest GROC heavy architecture was something else? It was like, it was a mixture of something on top of a mix. It was a bunch of mixture of experts models. And then it was a bunch of mixture of models. And then we were going to do mixture of models on top of models. Mixed your companies. Oh, yeah, yeah. Because we were going to make a router on top of that, which would prompt.

Every time you prompt, it prompts all of them. Yeah, and then there's a mixture of models of experts. Yeah, and then it, and then there's a scoring that gives you the final best. Okay. But I think apparently people, I mean, this is all like leaked stuff on Twitter, so it's like probably not true. So people are saying like GPD5 is like supposed to be kind of what I described. Yeah, yeah, some kind of routing thing.

Let's go. Let's go. Put me on the team. I just predicted it two days before it comes out or something. I'm good. Yeah. It took me a while to get there, but I think I think I'm excited for that. Yeah, I'm excited. Anyway, let me tell you about Adio, customer relationship magic.

Adio is the AI native CRM that builds scales and grows your company to the next level. Get it. Adio. Did you see this Hunter Biden clip? We stay away for politics, but we pull up the video because this is this video. This was very, very insightful as to how people think about artificial intelligence and automation. So this is from Channel 5, Hunter Biden on AI automation and whether not you should tip it in and out and growing Daniel has a very insightful post here says when he says

AI he really means software in simple robotics but still I'm actually very surprised by how smart by Hunter Biden is like my friends are smarter but I thought he'd be way dumber after all the crack but listen to what he has to say it is fine you don't tip it in and out I do what you tip it in and out yeah they have to wear those hats you know what that feels like did you did serious conversation I met somebody that is in the, you know, owns fast food franchises. His McDonald's employees 55 people, okay? He went all AI. It's about a $2 million investment. And they just show you a humanoid robot. It's like, he only employs five people now. So his margins go up by like 27%. It'll make back the $2 million investment in under like 18

months. Think about this. There's about like 13,500 McDonald's in the United States of America. I did the math. If every one of them went down to five from 55, lost 50, on average 50 jobs, it's 600 and like 70,000 jobs in America. And that's like it's a certainty. And that's just McDonald's. You think that Burger King, Wendy's, Taco Bell, I mean, across the board, you're talking what, like probably three and a half million jobs in the fast food industry alone that will AI easily within the next five years, if not three years. So I have flipped my thinking about this whole AI thing. You're either going to have a massive extinction event or it's going to be really good

with a rough patch in the middle. Because if between supercomputing power, which we're, we're there already, okay, like quantum computer. We're not quite there on quantum computer. Put that in the truth zone. I think we are generating exactly zero tokens. Fusion within the next like five years. And if we do that, a limitless source of power.

I have some friends that work on fusion. I've seen enough. Give him a hard tech fund. Yeah. Yeah. Yeah. Well, he doesn't even need to be in the private markets. There's plenty of public.

names he could get into if he's bullish. Fusion, quantum. Yeah, I'm sure that'll work out great for him. He gets in this. Be rough. Anyway, the interesting thing here, yeah, he's talking about automation. I actually know an entrepreneur who built a, you know, you take a TV, you turn it horizontally, you touch screen and you can order at McDonald's. I don't know if he actually had McDonald's as a client, but he had a number of those types of customers where you shop on a screen and check out on the screen. And that's something that we had the slowest takeoff ever for that technology truly like when do you think it would have been capable from you know from first principles just the capability of software and hardware

to show up click buttons pay and then and then send the order to a human cook who's still cooking and then just getting even even even at grocery store grocery store check out 2005 level technology in my opinion a lot of grocery I mean Air One still doesn't have like self-checkout. But I just actively avoid self-checkout. And you wonder, you know, I think a lot of people feel the exact same way. It's like being able to sort of outsource the scanning.

Yeah. Well, people do the self-checkout because it's a thrill. Because like it's this negotiation, like will they catch you for how much you're stealing? Because it's like they're not really watching, but there's someone that's kind of watching and you could get into altercation. And so it's about like,

It's about getting your pulse up. It's about going risk on. Exactly. It's about going risk on. Yeah, it's like the girls that still from Sephora. To encourage thrill seekers should actually start having like arm security there. Yeah. It needs to be high stakes.

Standing there like this. Exactly. High stakes. And so if you mess up and they catch you, you're getting tackled. It is hilarious how, yeah, they kind of got rid of like the scanner, but they definitely need to watch that stuff because people will just like fake scan and then check out. I don't endorse doing that. Chris from Pace has good takes here. where it's like in areas where there is.

Yes, yes, yes, this is a great take. There is. And Aaron Ginnettin too. Yeah, basically this idea that like, I think that 20 years from now, I will still go to restaurants and have a human waiter. Yep. Because it's a great human waiter improves the experience.

Yes. It's enjoyable to have somebody that understands the restaurant and you're hanging out with across. But the person will look more like an educator and an entertainer, than like a plate carrier. So I wonder if anybody, by the way, I wonder if anybody's building.

You will be someone who talks about the food. It gives you advice. So here's something. I wonder if anybody's building like a friend pendant style device for waiters. You just walk up to the table, have a conversation, but it's just actively like submitting them

through like the restaurants like order management system. So the waiters just hanging out there. They don't have to be like writing down stuff or in a phone. They can just have a conversation be like, you want this, no cheese, you want fries with that, you want a side salad? Great, great, great, great. And that's just automatically in the system and starts getting made. You might like this idea.

Usually when like consumer hardware goes to the enterprise, it's like where it goes to die, like the original HoloLens and the Google Glass, like as soon as they said like, hey, we're actually going to be focusing on niche enterprise use cases, usually a bad sign. But that could be kind of like a toast type, you know, vertical SaaS outcome. I don't know. I'm more bullish on that idea than at least hearing it for 10 seconds. But I don't know. At the same time, if you go to a really nice restaurant, I would be leaning even more in the pure human world and being like, I would like to go to the restaurant that doesn't even have phones. And there's actually a great post in here by HIPCity Reggie. We will read through because I want your take on this. So Reggie James, friend of the show, says, screenshot essay on wearable AIs. Will all AIs leave the

the room. Sitting down at breakfast, Yatu reveals he's wearing an AI pendant product on me. I've never heard of that one. Very cool. There's a bunch. Sean and Jackson freeze. A couple days later, I'm wearing a friend necklace, shout out Avi, and talking to my wife and I stopped to look at it wondering if it should be included in the goings-on of my home life. I think it's clear that we will start to create a set of social norms around AI that's extremely explicit. it, we will state things like, can all AIs leave the room? Spaces will have machinery running that renders connected electronics useless. So yeah, we should have a Faraday cage that we can go hang out in.

At the core, it's a question of which minds authority and priority in a given space. It's probably a good idea to assert the human ones. What do you think? So Jackson Dahl had a review of friend. I guess you got an early one from Avi. The review was generally positive. He said, clearly feels different to talk to a thing you're wearing and must touch to get a response from. With friend, you put voice in and text out, which is like kind of an interesting feature.

So it's just hanging out with you, but then it's messaging you on your phone about your day, which is kind of a new, it's just a new format. He says, All Said feels like Avi have built something truly different, a hardware device that embodies a new set of values. I suspect they will find an especially young group of people who quickly find themselves in daily communion with their new friend. And again, this seems to just continue to be incredibly controversial, and I'm sure it will for a while. Daniel says, how do you feel about recording everyone you meet without their consent? Jackson says, bad. Lots to figure out societally with these types of things.

Yeah, the voice in, text out thing is interesting. feel like there's there's a ton of different surface area to explore outside of just give someone a blank text box like that worked for Google with the 10 blue links it works for chat GPT but I think now the new service area is figure out how to push stuff to the user and this is a way to do that you're just ambiently recording and then I feel like the retention is going to be way higher than other devices because as long as you put it on it's going to be sending you messages. Yep.

And so I think that's going to be really, really good for retention. Yeah, there's this, there's, do you get push notifications from chat GPT at all? Only if only post prompt. Yeah. So if I fire off deep research, it will give me a push notification telling me, hey, we're done with that. Yeah. But I've never just.

But there's a world in the future, for example, where chat GPT has access to your calendar. Yep. And it's like, hey, I saw your meeting with this person this morning. Yep. Do you want me to run a deep research report on what they've been up to in the last like few months. It's like, oh, they shared this update on LinkedIn and they were in this article and they went on this podcast and talked about this thing. Right. You can see glimpses of it. More proactive. If you go to the chat GPT app, it will actually populate ideas for prompts. And so for me right now,

tells you a lot about me. It says Hollywood movies, AI and filmmaking, clearly because I was searching about Disney. Stock market, Mag 7 performance, I can just click that and get an update. And it says, how have the Mag 7 stocks performed over the past decade, what factors have driven their growth, and it just boom, pulls that prompt up. Media industry, luxury watches, automotive industry, F1 schedule, venture capital, tech business. And so some of those, if there was something triggering in the system that,

hey, there's this interesting thing going on, we should just generate this prompt and send it to John. That's probably an interesting feature that they'll probably explore at some point. but friend seems to have at least be exploring this type of thing. But voice in, text out, very exciting. Sleep in, rest out, also exciting. 8Sleep.com, pod five, five-year warranty, 30-night risk free trial, free returns, free shipping. Go check out.

I was riding a high Monday night. I put up a 95. Let's go. And I'm down in the dumps again, John. That's rough. I got a little cocky going on last thing. Well, you know what's not in the dumps, the future of nuclear uranium. enrichment because general matter is bringing uranium enrichments back to the United States

starting at the site where the U.S. enrichment industry was born. We've signed a lease with the Department of Energy to establish the nation's first U.S. owned, privately developed uranium enrichment facility at the former Paducah gaseous diffusion plant. 75 years ago, the U.S. Atomic Energy Commission selected Paducah to help lead the nation's original enrichment efforts. Today, we are proud to return to and rebuild this historic site. to power a new era of energy independence.

This is Scott Nolan's company. I hung out with him while I was at Founders Fund. He's an absolute dog. An absolute dog. One of the greatest to ever do it. And he has been on absolute tear. Spent, I think over a decade at Founders Fund, helped Peter Thiel with the original zero to one lectures while at Stanford.

Helped, you know, ideate and bounce ideas around for that book. Then invested at Founders Fund for a decade and was, kind of thinking about maybe starting a company but didn't want to force it and just waited, waited, waited until the perfect opportunity came around and really put together the perfect team, perfect partners and has accelerated way faster than I think he expected. When General Matter first announced, my first reaction, which is always a positive reaction is nobody, I haven't seen a pitch for this anywhere, but it's incredibly obvious that this company needs to exist. Yep, yep. And so it's It seems like it's working.

They won a contract with the government very early, earlier than they expected. And so they are moving timelines up and now they say we will enrich uranium by the end of the decade. US leadership and enrichment will allow us to lead once again in nuclear energy. This lets us lead in everything downstream of safe, clean, baseload power, AI manufacturing economy.

Today marks the beginning of America's restored leadership in nuclear enrichment. We thank our partners in Kentucky. Kentucky and at the DOE for supporting us. And you can see a beautiful render of the future facility that is absolutely massive. But, you know, they're thinking big and they're going for it. There's also a great post in PirateWire,

which you should subscribe to, digging into this from a different perspective and profiling the company, which is very exciting. Anyway, if you are interested in investing, go to public.com. investing for those that take it seriously but they have multi-asset investing. You better take it seriously. Industry leading yields. Not financial advice. And they're trusted by millions,

not for casuals. Taking investment, investing seriously is not financial advice. Let's go to Adam, person of swag. He says he started a startup to impress a girl, starts seeing product market fit, gets more compliments from male users, girls don't care,

keep trying to get more MRR. Now I spend every day in a dark room coding with four men. What is even the point of this? and the best photo you could imagine. Well, I would love to see Adam ring, hit the gong with us at the New York Stock Exchange and say, Adam, why did you start this company? And he just says, I just did it to impress.

You know what's funny about this? This is the apocryphal tale of the founding of Facebook. Like in the Aaron Sorkin social network movie, the whole premise, I know you haven't seen the movie. I have seen that one. What? What, Jordy Hayes has seen a movie? Incredible.

But the story that Aaron Sorkin decided to tell was basically that Mark Zuckerberg started Facebook as like a way to get a date or something like that. And then Mark was like fact check false like I was dating Priscilla happily for years before starting Facebook. So your whole your whole. Nor the truth. Your whole plot is fake. But he was kind of just like whatever, you know, people like telling stories and maybe a better story that way. But the social network too will be dropping soon. If you have a way to get us in that movie, make it happen.

Hollywood, you're listening. We're ready to be extras. Get Tyler in this movie. I think it would be so good if we could be extras just in the background. I think it's pretty doable. We're in Hollywood already. Make some calls. This is my number one like manifest it, put it on the vision board.

because a very small crew of people, those of you in the chat, bizarre, the game, you guys would see the movie, you'd see us and be like, wow, they pulled it off. So we'll have to make that happen. And then we'll have to buy a billboard

on adquick.com, out-of-home advertising, easy and measurable. Say goodbye to the headaches of out-of-home advertising. Only ad-quick combines technology, out-of-home expertise and data to enable efficient,

seamless ad-bibald. Adam, DeAngelo says, I love the fact that all the billboards on 101 are for AI products. And certainly, many of them are for regular enterprise SaaS companies announcing AI products. What is this? Oh, interesting.

Hmm. Do you see this response from Jasper? He is advertising and he says, cheap on-demand GPU clusters. Hyperbolic.a.i. Hyperbolic is an on-trend term right now. A hyperbolic can refer to several different concepts. In language, it means exaggerated. In geometry, it's related to a hyperbola or hyperbolic space, and in functions, it's related to hyperbolic curves.

The irony here that hyperbolic AI, you could read it as exaggerated AI, and then every billboard on the 101 is just, it's just like over-promising. Yeah, yeah, yeah, yeah, the nominate determinism is that they're over-promising. But I think in this case, it's like a neoclass. Love Jasper. No hate on Jasper. Good luck to him. I hope he can deliver

and I hope he can go exponential. This feels like something that can be delivered quite well today. Yes. But a hyperbolic function is something that goes up into the right and reaches an asymptote and goes to infinity essentially.

So it is a good example and it is something that is important in AI as we scale to infinity and beyond. David George has a post. He says, what do Roblox, and roll, crowd strike, and apple all have in common? None of them look like obviously.

winners from the start, but they built something far bigger than anyone expected. We call these companies model busters. Great. Named. Model busters either reveal a market that's much larger than anticipated or expand into new product lines so effectively that they break out of their original category entirely. We're talking about model busters now because we believe AI is creating many more of them. AI will maybe the defining platform shift of our time. It's changing how products are built, how they're distributed, and how buying decisions get made. It's creating new consumer experiences, new workflows, and entirely new business models, platform shifts, often create new winners.

AI is no exception. Today's most ambitious companies will grow faster and become bigger than anything we've seen before. Example here, if you were an investor modeling out Figma in the early days and you thought it would always be a design tool, again, this is not how Dylan was pitching the business, but if you thought it was going to be a design tool, you could simply look and see, all right, how many designers are in the world. And if you could get 5%, 10%, 20% of them using it,

would this be a big company? Still would have been a big company, but wouldn't have been a $50 billion public company, most likely. And it ultimately kind of like broke out of that tam by being collaboration tool across entire companies. We should have David back on. I really enjoyed talking to him during Dresen LP Day. I do wonder, like, is it a tautology that a model buster cannot be modeled?

Like, are we talking about an ineffable quality that cannot be defined? Like, is there a pattern for finding model busters? Because, like, definitionally, the model busters are things that cannot be modeled. It's funny because he's also on the growth team, and it probably does, like, the most modeling of anyone at Anderson. Well, Roblox is an example. So Roblox has 20.6 million DAUs in the U.S. in Canada last quarter. Yeah. And there's, I think that I don't know how many young people under 18 there are in Canada,

but in the U.S. it's like something like 70 million. And so that is like a meeting, like if you were evaluating Roblox and saying like more than a quarter of people under the age of 18 are going to be playing this game every single day, by 2025, you would have sounded a little bit crazy because I don't know that there's that many video games that I don't know how many video games have ever achieved that level of daily active usage in a cohort like that. Video games in general have been like a model buster in the sense that people were expecting it to be comped to the film industry and I believe the video game industry is like an order of magnitude bigger than the film and television industry. It is a significant

expansion of growth. But my question about like, how much time should you spend trying to codify model busters or should it just be something that we refer to looking backwards? Because my question is like, if you go back and you look at Danny Rimer in the C is seed series A of Figma or Mamun at KP doing the B or Andrew Reed doing the C, like at what point did it become important to win that deal to identify Figma as a model buster versus just say the metrics are really good. Dylan Fields a killer. I'm investing on that. And then yes, maybe there's a second act. But if I get a bunch of those ultra high quality entrepreneur, ultra high quality product,

ultra high quality KPIs, I'm good enough. And I don't even need to really predict that this will you know, have a second act or expand out and the Tam will moon. Like it's not. And you can see the same thing about like, you know, Anderol Palmer Lucky. Like you just look at the team and like the market's big enough to justify the investment in that phase. And then you just keep doubling down. And then it busts your model. But you're in it for different reasons. You're not in it because you predicted the model would be bust. So I wonder what it teaches you about like the philosophy of of early stage investing. Yeah, I mean, I think you could have, you could have underwrote SpaceX for a long time, ignoring, ignoring, uh, the opportunity for Starlink, right? And it still would probably be,

it'd be interesting to see, you know, how SpaceX would be valued into its first deck. Yeah, like its first model. Yeah. Probably 10 years into the business. I, I have to imagine it was not in any of the series A through C. Yeah. People are like, I thought I was investing in a rocket ship company. accidentally invested in the internet company internet company the accidental internet company again and again and again the internet is the it is the category and that every company makes that money yeah and I still I still remember as a kid when when I grew up in an Apple household I never had a like I never used Microsoft products really ever as a kid and I remember it was still at a time where you were

kind of like weird if you didn't have a like a traditional PC at home. Yeah. And then 10 years later, everybody has an iPhone, right? So it was impossible to like underwrite having a breakout product of that caliber. Kind of like luxury watches. Not many people wear them right now, but in the future, it'll be weird if you don't have one. So you got to go to get bezel.com. Your bezel concierge is available now to source you any watch on the planet.

Seriously, any watch. Someone in the chat yesterday said they got a moon swatch on bezel. No way. Yeah. Amazing. I didn't even know that those would be on there. That's amazing. Moon watch or moon swatch?

They've had different variation. Okay, cool. Very cool. Well, congrats to them. Hope they enjoy it. Send us a picture. We have some breaking news about Sirius XM. They are canceling the Howard Stern show.

Can you call it a cancellation when the guy's 71 years old and he's been doing it for 20 years? They say it's no longer worth the investment. They've been paying him $100 million a year. That is a huge salary. And that's what three, four times, Colbert? Wow, that is, that's big. That's power, power of radio, power of, you know, the power law. He's done fantastically.

So congrats to him. Howard, if you're looking for a new gig, you're welcome to come and hang out at the Ultradome. You can sit at the intern table next to Tyler and we'll bounce ideas off you. Yeah, it's wild. I really wonder where SiriusXM business goes. I know where this goes. Post AGI, you're going to listen to every hour of Howard Stern. I know you've listened to zero hours, but there's probably 20,000 hours in the catalog

or something like that of Howard Stern content. You could listen to it from the beginning. Do people listen to the back catalog ever? Absolutely not. But in the future. Sirius XM is still a $7 billion company. Wow. That's bigger than I would have thought.

I wonder, yeah, revenue, like how much of that cost. I mean, it makes sense to give him a huge slice of that. It is a talent-driven business, and he could go to Spotify. He could go somewhere else. My question is like, he is old. Will he retire or will he do a podcast or do something independent? Is there news? Now, they apparently did how much?

Okay, you look that up while I tell you about Wander. Find your happy place. Book of Wander with inspiring views, hotel-grade amenities, dreamy beds, top tier cleaning 24-7 concierge service. It's a vacation home bed better. So they did 8.6 billion in 2024 revenue. So they're trading at less than one X revenue, which says when you have a shrinking business, it is a rough place to be in.

Not a growth stock, I suppose. Yeah, I wonder how much of that is getting eaten up by talent. This is a trillion dollar stock if they have. Yeah, I wonder how much, I wonder what percent of serious XM content is like power, law, celebrity, host, red, high salary contract versus essentially programmatic or AI. content because it's probably not even AI but if you're just like there's a there's a station on serious X-M that's just the Grateful Dead and it just plays them the whole time like I don't think you need someone making a hundred million dollars to like randomly play

grateful dead tracks like there's probably some Grateful Dead fan who manages that and picks the songs and orders them but like that could essentially be pseudo random maybe one day you go through the the back catalog and then the new stuff and and then you mix it up and then you play the hits or something. I don't even know if they have hits in that way. I know it's kind of a jam band. But I wonder of their content, of their tonnage. Is that the term?

Tonnage is like the amount of content on the network. I wonder how much of it is driven by these high dollar deals. And I wonder if they'll get a new host in the seat. I wonder who this generation's Howard Stern is. it's like Tim Dillon or something, some irreverent comic essentially. I think the thing is as content is on demand, fewer and fewer people are just turning on the radio or turning on the television and listening to whatever they, whatever just happens to be playing.

Yep. And so if you take people that were subscribed to Howard Stern, we're like, you liked Howard Stern. We're now just going to play this other person through his channel. Yeah. They're just going to be, they're just going to ask what's going on here. Yeah. I mean, certainly if you have a specific car that has Sirius XM and it doesn't have Spotify

and Howard Stern's new show is on Spotify because it's a podcast, you might stick around on Sirius. Maybe they get someone new in the seat who's, you know, almost as good or can build a relationship, but it does seem like a challenge to fill that. But also, you know, it's a lot of money to pay. So clearly it wasn't penciling out. So they had to move on. Anyway, X is now noticing when you, you know,

you take screenshots. Do you see this? I've seen this in other apps. It's usually pretty annoying, honestly, but when you take a screenshot, it triggers a UI element that says, hey, do you want to actually just copy the link to this post instead of taking our content elsewhere? It's a little bit of a retention hack. I don't know if Nikita Beers involved in this, but it's, it's just like a classic UI thing. I wonder what the actual benefit of this is. So many people still just screenshot a post and share it in like a group. Yeah, because posts get deleted and then posts, it's like, oh, that person blocked me or that person is a locked account. And so the screenshot is just the universal language of like, you know, sharing information and content.

And a lot of people still share links, but the, like, the screenshot is super Lindy and I don't see it going anywhere. So, I mean, the good thing is that this doesn't block your screenshot. Like, your screenshot is still added to your camera roll, but then you're prompted to, hey, do you want to actually show the link? I haven't actually run into this. I don't. I take a lot of screenshots of posts, but I do it on my computer. on my phone. So anyway, and Preston continues and says, Nikita is going to look like a hero for putting features on Twitter that have existed on TikTok for a decade. I don't know enough about

TikTok features to know what he would port over, but I know he's a student of algorithms and student of social apps, so I'm sure he'll find something. TikTok has, is very smart about all these type, all the different ways that people are like diverting attention and hacking it. It would make a lot of sense to just copy best practices from there. So I think this is a great, good take. Yeah. Well, we have Mark Andreessen joining in about 10 minutes. If you have questions from Mark, throw them in the chat.

And we'll try and get to them during our interview with him. Ray Sullivan says, once the summer is over and Mestral is back to work, I'm sure they're going to be dropping some cool stuff. 1.6K likes. How do you even fight that? That's so rough. Just Europe is down bad in the meme in the memetic war. I would be very interested to actually understand.

We should have somebody on from Mr. All to talk about it. There are that the meme that European founders and teams take off eight weeks in the summer is extremely real. And it's not always a full eight weeks. We have taken this summer, we've taken one day off, July 4th. Yeah. And it's like extremely evident to what's going on over here. in America. We run a media company. And it's like you're in the most, yeah, you're in the most aggressive fight ever. Not to say the Mr. All's taking days off. I actually don't know about

that company specifically, but, but I have seen some, some friends in the show proudly take time off. I know, I have friends that are European founders that have massive companies that are very successful. Yeah. And they, their businesses allow them to take time off. They can go. They can have a nice vacation with family, friends, whatever it is. They can maybe just dial down meetings quite a lot. And their businesses are going to be fine. They're probably still going to have great quarters. They're going to put up great numbers. The issue with Mistral is you are competing with the entire world, an entire world where if you're not at the frontier, you're not going to get usage. And so I wouldn't be surprised if Mistral said, hey, no, no Euro summer, this summer.

But we don't know. I mean, we don't know that they took a summer. We don't know that Mr. I don't know. I'm saying I wouldn't. But yeah, I mean, there's also like the, so there's the, there's the facts of the matter, like you probably shouldn't take a huge vacation in the middle of the biggest fight over new technology in the past decade or two. Not that they are, but the other interesting thing is like the vibes and the aura farming that comes from doing things that show that you're working extra hard.

So this is the Elon, you know, taking a meeting really late at night with a journalist in the room and then that gets out and you see that, oh, he really was sleeping the picture of the sleeping bag. The X-A-I team having tents in the office. Another example is Mark Zuckerberg. Meta dropped Lama on a Saturday or a Sunday or something. And somebody asked, Mark, why'd you drop it on a Sunday? Because that's when he was ready.

And it was a little bit rough because Lama wasn't like fully ready to the full extent. But it just showed that they were trying to move as fast as possible very clearly. And then similarly, Sam Altman and the Open AI team, they dropped that IMO stuff at like two. am on a Saturday. And it was like, okay, there's like all this debate over, you know, were they inside the stadium, where they run in the parking lot outside, you know, wasn't signed off.

Was it signed off or whatever? But you can't say that they weren't up late working. Like, that's true. That's the one thing you can't say. You can't say they didn't have that dog in that. Yeah, yeah, exactly. You can't say they didn't have that dog with them. They might not have been nice with it, but they definitely have the dog in them, right?

That's true. Like, it wasn't very nice with it to, you know, kind of like, frustrated. Krishna here says the irony is that mistrawl, the meteorological, the meteorological, The meteorological event the company was named after mostly occurs in the winter and spring. That's wild. The nominative determinism strikes again. T.J. Parker says the best Amazon leadership principle is right a lot and it's not particularly close. Quote, leaders are right a lot. They have strong judgment and good instincts.

They seek diverse perspectives and work to disconfirm their beliefs. I thought this was interesting. I thought it was a typo. A typo. Because Amazon famously. writes a lot, W-R. But this is R-I-G-H-T. They are correct a lot.

Interesting. Stewart says, kind of true with VCs, too. Feels like the best ones just have good sense broadly and pick really well. T-J says, yep. T-J has been right a lot himself.

Created Pillpack, sold it to Amazon for a billion dollars. So it pays to be right. He's also been right a lot about his selection of timepieces. He's a fantastic collection of watches. Some of them are on display from time to time. Every now, yeah.

Andrew Reed says sector, focus VC firm called specific catalysts. Okay. Andrew's been, you know, yeah, yeah. He took Figma out at IPO. He's been on a generational run. He's back like a dozen deck of horns, potentially greatest allocator of this generation doing very great. Getting a little cocky on the timeline.

coming for general catalyst making fun of their name coming for a 16 z you remember this he said a 16 z the 16 doesn't count the space in the middle so he's taking shots from his high perch heavy is the head that wears the crown andrew be careful or else people are going to start asking questions about why you don't own sequoia.com and you don't want that to happen whoa you know this yes no but i think he's having fun i don't think across the lines but it is It's funny because he's just like, hmm, I'm having a good time. Let me just fire off some jokes at the expense of my competitors. But he's having fun.

This is hard. This is fun. I love this. Yep, there's certainly plenty of VC firms formed to capitalize on a specific technological trend. For sure. And maybe one of them should be called highly specific catalyst capital. And maybe we'll ask Mark to settle the debate. Is it a 16 Z or is it a 17 Z?

Or is it like Wilmer Hale where Wilmer Hale, the law firm, two names, there's no space in between Wilmer and Hale. And so if Wilmer Hale were to do the A16Z thing and abbreviate it, they would be correct. So you never know. Maybe Andrewsson Horowitz doesn't have a space in the middle. Post here from Dan to me, quote, I don't do drugs.

Lame, weak, beta, childish. I don't touch the stuff. Refined, respectable, aged, wise. It does sound like, yeah, you have experience. I don't touch the stuff. Trump's line has never been big into that whole world. Yeah, yeah, yeah. It really, really hits way, way differently when you say that.

So, yeah, kids, just say no. Just say you don't touch the stuff. It's great. Post here from Carl Rivera over Shopify. Being an angel investor is just a different way of subscribing to extremely expensive email newsletters. Yeah, you're oftentimes, you know, basically giving somebody 25K for five to 10 years of monthly sends.

Don't we know an angel investor that writes like, $1,000 checks to get the email updates? Just to get that. Any LPs and funds too? And then gets all of that information. He's just like a master of information collection. We do. We certainly do.

I had a portfolio company that I hadn't heard from for years. I got updates for a while after I invested. And I got pinged to sign some docs. and they and I assume that the company was shutting shutting down and they got a massive acquisition oh acquisition that's great yeah was it was it extremely pleasant pleasant surprise like normally if you just the updates trail off yeah things are going that well but every once in a while you get a really nice surprise that's great reads with Ravi says great leaders think like farmers think like a farmer

Don't shout at the crops. Don't blame the crops for not growing fast enough. Don't uproot crops before they've had a chance to grow. Choose the best plants for the soil. Irrigate and fertilize. Remove weeds. Remember, you will have good seasons and bad seasons. You can't control the weather.

Only be prepared for it. Very nice. Philosophy. Think like a farmer. I wonder. Think like an aura farmer. Think like an aura farmer. Don't shout at the aura.

We still need to have this debate. with Tyler around ORA farming. Well, I'm trying to get him to coin it, cause groves law. I'm still writing it. You're still writing the post? I've got to get it perfect. Okay, yeah.

Because the tweet in and of itself is aura farming. Exactly. Right, so it is. I have recursive in that way. Same with Coagin's law. Cougain's law is a recursive coinage. Exactly. Yeah, I think this could do really well.

This has 1K likes written all over it. It's happening. And endless citations of Cosgrove's law. So we will work on that. Merrill from graphite says it's a massive week for his customers. Figma IPOed at 36.3 billion. Ramp raised at 22 billion and Clay raised at 3.1 billion. The future is being built with graphite. Congratulations to all those companies.

Congratulations to graphite. You got an absolute murderer's row of clients. Congratulations. Well, we have Mark Andreessen joining us. He's live from the TBPN Ultraljum. Welcome to the stream. How you doing, Mark? Hey, what's happening?

Great to see you. Yeah, you too. A lot. It's a little bit of a slow news day, but exciting stuff with GPT open source. It's not a slow August. It's not a slow August. We're glad. We were just reflecting that we've taken exactly one day off this summer that was July 4th,

and we're showing the Europeans how American companies work. American work, we're setting an example. And we have proof of work because we exist on the internet and you can see us live every day. So we're setting an example. How are you doing? How's your summer going? Fantastic, going really well. So how long is it going to be until you guys put up avatars that make claims that you're working hard all through the summer?

But it turns out here on the beach. You might have caught us. I think you'll know better than us as to when the technology gets there. We've been demoing some of the stuff. People have been doing a lot of deep fakes of us. And fortunately, all of them have been clockable. So it doesn't feel like a brand risk. But they're getting closer and closer.

And I know that there's going to be a moment when we have to say, hey, that's actually using our name and likeness to endorse something that we don't necessarily endorse. Can you please take that down? So we're approaching the touring test, the uncanny valley. We're escaping the uncanny valley. I think a question, looking back over the, you know, maybe 10 or 15 years, was what moments did you feel like there just was not a lot of action happening? Because this summer is just the pace from so many different teams has been absolutely insane. everybody's like trying to keep up and it didn't used to feel that way at least from my point of view

so my view it always is there's like these there's this these disconnected you know kind of patterns or trends there's there's sort of the the sort of day-to-day phenomenon where like engineers show up every day and they make things a little bit better and then every once in a while you know you get a technical breakthrough or a new platform and and and that process kind of this you know kind of of sawtooth kind of up to the right kind of process kind of plays out over time kind of regardless of what else is happening in the world. And so it keeps happening through recessions and depressions and wars and like all kinds of crazy, crazy stuff that's happening. But basically, you know, the technology keeps getting better.

So there's kind of that curve. And then there's the sort of enthusiasm curve and then the adoption curve, you know, which is basically like, when do these things actually show up in the world? And then by the way, when are people actually ready, you know, for the new thing? Like if you talk to people who worked on language, I'm sure you guys have talked to people who work on language models, they will tell you that they were surprised the chat GPT was the breakthrough moment because they thought everybody already knew what these models could do for, you know, three years before that. And so they were, you know, they were shocked that it was the chatbot interface that made the thing go. And so there's somewhat of a sort of arbitrary disconnection between what's actually happening in the substance and then what people are seeing and failing. And so it's just, it's really hard to predict when these things pop.

But also if you're in this day to day, it's really hard to tell, you know, when things are going to be hot or not because it doesn't necessarily map to how much that took. We were just talking about that in the context of Google's new world model. It's this generative video game that you can kind of move around in. And it feels like deep mind is just absolutely crushing at the AI research frontier. They have the best world model simulator that you can walk around in. The question is like if they let another lab do the chat GPT thing and just get it out into the consumer three months earlier, they might wind up kind of chasing and trying to catch up if somebody actually figures out how to make it like a dominant consumer product. Now in the enterprise,

it's more oligopolistic, but consumer seems to be winner take all. I guess the question is like how much value do you place right now in the AI race to just like moving fast, breaking things, you know, dealing, having like the thick skin to deal with like the safety constraints and all of the different stuff, obviously not being irresponsible, but just speeding up the organization as much as possible. It feels like now is the time to really push on that. Yeah, well, first of all, I need to correct you. It's moving fast and making things. Making things, that's right.

I don't even know where it came from. I have no idea what you've never heard. Never heard of it. I mean, Chad GPD didn't really break anything. I think that's a good point. It really did just move fast and make things. The first things it made were weird, but that was fine. And it failed and it hallucinated a ton, but it didn't really break anything.

I don't know. Yeah, I believe in this case, total deaths attributable to Chachupitia are still zero. Zero. So notwithstanding all of it. now we're standing all the all the caterwalling but yeah so look I think the AI industry in particular has a very acute version of the of the of the sort of challenge that you identified with and and you know and I don't say this negatively just an observation which is that they're you know like in sort of a normal technology company you've kind of got engineers who make products and then you've got

you know kind of sales people or marketing people who sell them you know in the in the air companies you have this third tier of you know the quote unquote researchers yeah right and so you know which is which has worked out incredibly well I mean the researchers have done you know they've just done like amazing breakthroughs at these companies. But, you know, the handoff, you know, there's not necessarily clean handoff from the researchers to the market. And so it kind of raises this question of like, okay, like, is there, are these companies therefore kind of three, you know, kind of three segment companies where they have research

and then they have product development and then they have go to market. And I think that's a really open issue. I mean, if you, you know, Google's kind of a case study of this, you know, you alluded to DeepMind, but even more broadly, Google, you know, Google developed a transformer in 2017. And then they basically let it sit on the shelf, right? because it was a research project. They didn't productize it. They were very worried about, you know, from people I've talked to, they were very worried about the, you know, brand issues and safety issues, you know, kind of all these, all these, they had all these reasons to not productize it. I talked to somebody senior who was there at the time, and I asked them, you know, when could you have had chat GPT with GPT four level output? If you had just got, you know, gone, gone flat out starting in 2017, and they said by 2019, you know, they already knew how to do it. And then, you know, they've now caught up, but it took, it took an extra five years, five years to catch up. And so I think a lot of these companies kind of have that challenge. Elon, as usual, of course, is provoking this question.

I'm sure you guys talked about, but he has now, you know, with an XAI, he's now collapsed, you know, he's eliminated the distinction between research and product. And so, you know, of course, he's pushing this as hard as he can. And I think it's a good question for a lot of other companies, kind of how hard they want to push on actually getting these things in fully productized form out to the market. Yeah, yeah. On Elon's like distinction, it feels like there is more research to be done. done, but it feels like we're entering like a new cycle of, you know, just focus on the engineering, focus on the deployment, the applications, let's get all this technology out into the world,

let's reap all that benefit. And yes, there will be a different track of fundamental research that's happening somewhere, but it's really, really hard to predict. And so if you have something that's working, just double down and just go really aggressive on it. I'm wondering more on that, but also on Apple strategy. It feels like Apple's been kind of like, you know, people have been maligning them for not, for missing the AI opportunity. And Tim Cook's just there on the earnings call being like, look, we acquired a couple small companies and seven this year. Seven companies.

But then it seems like they're taking more of like an American dynamism approach. Like there was news today in the journal that they, that they're investing $100 million in American manufacturing. They're certainly doing stuff. They're just not chasing the, you know, the shiny tennis ball. The headline $100 billion cap. So I'm wondering about your thoughts on when you have a, you know, when you have a platform, how hard is it to resist chasing the new shiny object?

Is that the right move? Or are there any other things that you think Apple should be, you know, changing their strategy on? Yeah. So look, Apple's always had this, you know, very clearly defined strategy that, you know, Steve and Tim, you know, working together figured out a long time ago, which is, you know, they, I forget the exact term, But it's something like basically, they invest deeply into the core of what they do. You know, they'll basically work internally on things for many years. They only actually release things when they feel like they're kind of fully baked.

Right. And so as a consequence, they have this thing where, and Jim says this, right? You know, they're rarely first to market with new technologies. You know, they're more often in the category of what, you know, Peter, Peter Thiel calls last to market. You know, they'll come out whatever, three years later, whatever, five years later. You know, there were tablets for years before the iPad. There were, you know, smartphones for years before the iPhone. Folding phones. They're about to do a folding phone. It's like 10 years into that technology.

I'm sure if they do it, they'll hit it. The last mover. The last mover. Yeah, yeah, yeah. The last mover. I guess, yeah, what I would say is like, look, that clearly works if you're Apple, right? And so it clearly works if you're Apple. But I would say there's a fine line between that strategy and just simply becoming obsolete, right? And so the problem is like if you're not Apple and you don't have all the other kind of super strengths and, you know, kind of now the market position that Apple has, you know, do you really want to be a company, you know, if you're not Apple? Do you really want to be a company that basically sits there and says, yeah, the world's moving and we're very deliberately not going to lean as hard as we can into it?

And so I think there's a lot of survivorship bias in these kinds of strategy discussions where people look at the one company that's able to pull this off and they don't look at the 50 other companies that are in the graveyard because they didn't adapt. I mean, you know, all the other smartphone companies when the iPhone came out, they were like, oh, yeah, well, we could do touch too, right? You know, we'll just, you know, we'll get to it, right? And, you know, they're gone. What do you think? The black, very bold, I remember. like an iPhone knockoff. What do you think, you know, right now people are,

or variety of, you know, shareholders are annoyed at Apple around their reaction to AI, LLMs, John's annoyed around just like transcription generally, just like super basic stuff. But it doesn't feel like the core businesses immediately threatened today. It feels like it's still on the horizon around these sort of like, you know, eyeware-based computing, you know, potentially net new devices that were,

that we'll see from, you know, companies like Open AI over time. But where do you, like, how real is the threat, you know, this year versus 10 years from today? And kind of what's your framework? Yeah, well, look, I mean, I think the biggest ultimate danger, I mean, the biggest ultimate danger is very clear, which is just like at what point do you not carry around a pane of glass in your hand, you know, called a phone, you know, because other things have superseded it. And, you know, look, everything, you know, everything becomes obsolete at the point. So there will come sometime in the picture when we're not carrying phones around and we'll watch movies or people have phones and we'll be like, yeah, look at how primitive they were.

Right, because we'll have moved on to other things and whether those things are eye-based or, you know, the other kinds of wearables or whether it's just kind of, you know, computing happening in the environment or just, you know, entirely voice-based or, you know, who knows what it is. But, you know, there will come a time when that happens, you know, is that time three years from now because there's like some, you know, huge breakthrough, you know, from some company that figures out the product that. absolutely is the phone right away, or is that 20 years from now, because the phone is just, you know, such a standard platform for everything that we do in our lives and everything else, you know, kind of remains a peripheral to the phone. I mean, that, you know, that's, you know, that's the game of elephants that's playing out there. You know, obviously, I think, you know, I think it's highly likely that we'll have a phone for a very long time. Having said that, it is exciting that there are companies that are going directly at that challenge,

and, you know, whoever cracks the code on that will be the next Apple. And by the way, that may end the fullness of time be Apple itself, you know, they may be the company that figures Is that out? Yeah, I remember being at a board meeting at Andrews and Horowitz maybe a decade ago or something and Chris Dixon showed me the hollow lens and I was like, okay, we're one year away from this being everywhere. And I feel like today I'm still in the like, yeah, VR, it's definitely one year away. The next quest I'm going to be wearing daily.

And it feels like we're always there, but it does feel like Apple did a lot of work on the, on the fundamental, you know, pixel density of the resolution of the display and then And meta's been doing a ton of work on just getting it light and affordable. Like it feels closer than ever. But, you know, you always got to wait until you see the churn numbers until you really call the game, right? Well, you say, but, you know, I think that's true. But you'd also say, you know, I'm on the meta board. So I'm kind of a dog on this one.

But like the meta, rayband glasses are a big hit. Oh, totally. Right? Like they're big, you know, so I think we now have a form factor that we know works, you know, for, for iBased wearables. You know, there's not VR. And then VR, you know, on top of that. But, you know, just the, you know, the glasses, you know, and then the glasses of camera, you know, sort of integrated camera, integrated microphone, integrate a speaker. You know, that's a very interesting platform.

You know, the watch clearly works, by the way, which Apple, of course, you know, is played a significant role in making happen. You know, that now sells in huge volume. You know, so that's the second data point. And then, you know, look, I think these, you know, these, I think some form of AI pin is going to work. I also think, you know, headphones are going to get a lot more sophisticated, which is already happening. And so, you know, you do have these, you know, kind of data points coming out. And then, yeah, look, the trillion-dollar question ultimately is are these peripherals to the phone, which is what they are today, or are these replacements for the phone?

And, you know, we, yeah, I would say we, you know, we have, we, I think we have a lot of invention coming, both from new companies and from the incumbents who are going to try to figure that out. Yeah, I always think about the value of, like, narrowing the aperture on these new technologies. Like, with, with the meta-ray bands, I feel like the fact that they aren't also trying to be a screen is actually a feature, not a bug. And I always go back to the iPhone. like it was first and foremost a phone and people bought it because it could make calls and then it

could make text messages and then it was an iPod but I do you disagree with that please well you guys you guys might be too young the first iPhone actually was a bad phone how so then she guys for the first two years it couldn't reliably make phone calls I I had I had like the third one and a friend had one but I feel like it was still like people were carrying cell phones and that was the at least of the expectation but yeah I mean I guess you're right So for the first two years, it was a classic apple story because the first, for the first two years, the thing couldn't make fun reliably make phone calls. And then it turned out there was an issue with the antenna and with how you held it. And there was a famous Steve jobs email.

Yeah, you heard it and you would disconnect it. You could basically brick the device. Yeah. Based on how you held it. And somebody emailed, this is when Steve would respond to emails from random people. And somebody emailed Steve saying, if I, you know, hold the phone this way, doesn't make phone calls. And he's like, well, don't hold it that way. Yeah.

Right. Yeah. Yeah. So even there, it was like, yeah. And people, you know, people forget. It took like five years. the iPhone to find its footing. It took like two years to get the thing. I remember also the original iPhone didn't have, it didn't have broadband data. It was on the old 2G. It was called the AT&T

Edge Network. So it didn't have broadband data. And then of course it didn't have an app store. Right. It was completely locked out. The challenges for Apple now is that people are so used to perfection with the device that launching a product that isn't perfect, like is embarrassing, right? Like you look at the Vision Pro and it's like, well, the battery's big. Steve would have hated this, right? Like, how he never would have shipped this and that being constrained and not being able to innovate because you're tied to this like impossible standard of being on whatever generation 17 of the iPhone and perfecting every element is a real challenge. So I would say there's a corollary to that. One of the things I've observed over the years is I think algae products become

obsolete at the precise moment they become perfect. And what I mean by perfect basically is like, yeah, it's like the perfect, idealized, complete product. Like, it does everything you could possibly ever imagine. Everything a customer could imagine, everything you as the technology developer can imagine, it's absolutely perfect. And there's been tons of examples of this over the last 50 years where it's like the absolute perfect permanent, it seems to be the permanent version of that product. And then it just turns out that's actually the point of obsolescence because it means

creativity is no longer being applied right into that platform. You're just like, there's just nothing else to do. You're just like, you're done. The product has been realized. And then the cycle is what has been realized? happens to your point, the cyclist other people come in with completely different approaches, completely different kinds of products that are broken and weird in all kinds of ways, you know, but are fundamentally different. And so, you know, that is one of the time

modern traditions. And, you know, one of the, you know, things you could say about, you know, Tim is, you know, his willingness to kind of break the mold of Apple only shifts perfect products, but, you know, being willing to ship the, you know, the Vision Pro, you know, shows a level of determination to kind of stay in the innovation game, which I think is very positive. Yeah. Yeah, yeah, yeah, yeah. That's great. updated thinking on open source since we last talked uh there's there's a lot that's been opening i is an open source company yesterday yes opening eyes open again yes yeah yeah look very encouraging you know a year ago i was very you know i was getting very distressed about open

you know whether open source say i was going to be allowed uh right it was even going to be legal and so and i think you know we're basically through that at this point uh we're i'm supposed to say we're through that in the u.s um you know we'll we'll see about we'll see about the rest of the world And then look, you know, the U.S.-China thing is obviously a big deal, but, you know, I think it's been net positive for the world that China has been so enthusiastic about open source, yeah, coming out of China, which has been great. And then, yeah, look, opening eye, leaning hard into this, you know, and releasing what, you know, what they did is, as I think fantastic, both because of what they released, which is great, but also just the fact that they are now, you know, willing to do that. And then Elon reconfirmed overnight that he's going to, you know, open source, you know, start open source in previous versions of GROC. And so, yeah, so we, you know, we seem to be in the timeline where open source. AI is going to happen. Right now, you know, I think what you would say is it kind of lags the

leading edge proprietary implementations by, you know, six months or something like that. But I think that, you know, that's a good, if that's the status quo that continues, I think that would be a very good status quo. What are the rough edges that we need to kind of sand down when we're thinking about Chinese open source model specifically? Is it we need to do some fine tuning on top of them to add back free speech or do we need to watch for back doors, say it's phone and home if it runs into this specific thing. The Chinese open source thing, it was remarkable because I feel like it really does accelerate

the pace of innovation because everyone gets to see, oh, this is how reasoning works. I think that's great. At the same time, it made me much more appreciative of AI safety research and capability research and actually being able to interpret what's going on and say definitively, this model is going to behave weird in this weird way, like the Manchurian candidate problem.

We haven't found any of that, but it certainly seems like something we'd want to keep an eye on. But from your perspective, of like what are the risks that we need to be aware of going into a world where China is really pushing hard into open source? Yeah, there's two, there's two and you identify them, but let's let's talk about both of them. So the phone home thing is the easy one, which is you can put a, you know, you can packet sniff, you know, a network and you can tell when the thing is doing that. And you can, and plus you can go in the code and you can see what it's doing that. And so you can validate, you can validate that that's either happening or not happening.

And I think that, you know, that's important. But, you know, I think people are going to, people are going to, are going to, are going to, are to figure that out. You can kind of gate that problem practically. The bigger issue is we have this term in the field right now called open weights. And open weights is a loaded term. It uses the open term from open source. But of course, with open source, the thing is you can actually read the code. You know, with open weights, you have, you know, just a giant file full of numbers. As you said, that you can't really interpret. And then what you don't have, what most of the open source, open weights models don't

including deep seek specifically what they don't have is they don't have open data right or open corpus right so you you can't actually see the training data that went into them and of course you know most of the people building models are kind of obscuring what that you know what that training data is in various ways and and so when you get an open weight model you know the good news is the the software source is open the good news is you can run under your new machine you can verify that doesn't phone home but you don't actually know what's happening inside the weights and so i i think that that is going to be a bigger and bigger issue which is like okay how how the thing behaves, like, yeah, what has it actually been trained to do?

And what restrictions or directives has it been given in the training, you know, that are embedded in the weights that you need to be able to see? You know, this is, I would say, this is coming up as sort of, I would say, a global issue, you know, which, you know, we worry about when these models come from China. Other countries worry when these models come from the U.S., right? Which is, right, so one of the phrases you'll hear when you talk to people kind of outside the U.S. is kind of this phrase people are kicking around, which is not my weight's, not my culture. Okay.

Right, right, or by the way, for that matter, not my weight. not my laws, right? Which is like, okay, like, what actually is this thing going to do? Right? And to your point, the Chinese models, for example, might, you know, never criticize, you know, communism or something. I can tell you, the American models have all kinds of constraints also, right, implemented, you know, usually by a very specific kind of person at a very specific location of the U.S. And so, you know, I think that this is a general issue.

And we're going to have to see basically people's tolerance levels, being willing to run open weights models. where they don't fundamentally have access to the data. And then correspondingly, I think what we'll see is more open source developers also doing open corpus, open data. So you can see what's actually in them. Yeah. Obviously, open source is very important in terms of just distributing intelligence broadly,

giving people the ability to run their own models and really fine-tune them and have control. There's also the big push just to make frontier models and high-capability models free. One model is you charge for the premium, you give the free away. It's a freemium model. That's what we're seeing at most of the labs right now.

There's also this kind of specter on the horizon of potentially putting ads in LLMs and what that would do to the world. Jordy got in a little dust up with Mark Cuban on the timeline deciding whether or not it would be a net good to put advertising in LLMs, what might happen that might be bad there. What do you have to take?

Yeah, my point broadly was that ads have been an incredible way to make a variety of products. products and services online free and just saying like default, just no ads would potentially, you know, be incredibly destructive. But yeah, curious, your framework. Yeah, so I just start by saying like whenever I personally is an internet service, I always try to buy the premium version of it that doesn't have ads. Right. And so if I can like live personally inside an ad for universe and pay for it, like that's great. And I'll free, I'll freely admit, you know, whatever level of, you know, hypocrisy or incongruence,

you know, kind of, kind of results from that. No, the point is choice. The point is choice. Well, the point is exactly what you said. It's affordability. So the problem is, if you really want to get to five, if you want to get to a billion and then five billion people, you can't do that with a paid offering. Like it just, you at any sort of reasonable price point, it's just not possible. The, you know, global per capita GDP is not high enough for that. People don't have enough income for that, at least today. And so if you want to get to, you know, if you want

the Google search engine or the Facebook social app or the whatever, AI, you know, Frontier AI model to be available to five billion people for free. You need to have a business model. You need to have an indirect business model and ads is the obvious one. And so I do think if, you know, if you take some principle stand against ads, I think you, unfortunately, are also taking a stand against broad access just in the way the world works today. And then look, the other really salient question is, you know, the same question that the

companies on Google and Facebook have been dealing with for a long time, which is, are ads purely destructive or negative to the user experience or are they actually, if done properly, are they actually either neutral or even positive? And this was something that, you know, Google, I think, to their credit figured out very early, which is, you know, a well-targeted ad and a specifically relevant point in time is actually content. Like, it actually enhances the experience, right? Because it's the obvious case. You're searching on a product. There's an ad. You can buy the product. You click you buy the product. That was actually a useful piece of functionality. And so, you know, can you have ads or other things that are like ads?

look like ads, you know, different kinds of referrals, you know, mechanisms or whatever, can you have them in such a way that they're actually additive to the product experience? And you can imagine, just like with search and with social networking, you can imagine lots of examples of that. People will, you know, people will, you know, they'll whiner and in lots of different ways. But I think, you know, I think that hasn't been a bad outcome overall. And I think that, I think it's entirely possible that that's what happens with these models as well. Yeah. So kind of similar kind of question.

what should be legal, kind of trying to create legal frameworks on a number of issues with AI. There's been a number of IP cases that have been working their way through the courts, what can labs use to train models, et cetera. There's been some good outcomes recently. Sam also was talking about how a lot of people are using AI as like a confidant, like a, you know, a friend, things like that. And he mentioned that currently your chats are not privileged. They can be used in a lot of.

lawsuit or or other situations. How how optimistic are you that our sort of legal system in the U.S. can get some of these issues right where maybe it can't just be, you know, total free markets, kind of lawless, whatever goes? Yeah. So in the case of training data, I think that there, I mean, there's a bunch of these copyright, you know, kind of lawsuits happening right now. There's, you know, the big New York Times opening I want and there's, you know, been a bunch of others. I think in that, for that particular problem, my guess is that problem ultimately has to be solved through legislation. It's ultimately a legislative question. The reason is because it goes to the nature of copyright law itself, you know, which is legislation.

And of course, you know, the content industry is already claiming that, of course, you know, using copyrighted data to train, you know, without permission and without paying is sort of, you know, they believe illegal on his face, you know, due to violation copyright law. The counter argument to that, which, you know, which we believe is, well, it's not copying, right? there's a distinction between training and copying, just like in the real world, there's a distinction between reading a book and copying the book, you know, as a person. And so there's going to need, I think, you know, the courts are trying to grapple with that. There's a whole bunch of cases. There's jurisdictional questions. You know, probably ultimately Congress is going to have to figure out a, you know, figure out an answer on that.

And by the way, the president is kind of, you know, thrown down that gauntlet in his, I think, the speech he gave last week or two weeks ago, you know, where he said of, you know, Washington probably needs to deal with that as an issue. So that's one. On the on the on the on the on the privacy thing, I think that one feels like it's a Supreme Court thing to me. It feels like that's the kind of issue said it's Supreme Court. And in other words, like whether for example your transcripts are considered your property and whether they're protected against, you know, warrantless search and seizure. And the observation I would make there is if you look at the march of technology over time. So the Constitution has like very clear, you know, fourth, fifth amendments, you know, very specific rights around the, you know, the, you know, the things that are yours, you know, such as, you know, your home, you know, being in your home, you know, by the way, the thoughts in your head, right? You know, that the government can't just, like, come in and take. They can't, you know, they can't just come in and search your house without a warrant.

You know, they can't, like, you know, put you in a jail cell and beat you until you fess up. Like, you know, there are, you know, we have constitutes of protections against the government being able to basically, you know, take information, you know, fundamentally, you know, as well as possessions. And then basically what happens is every time there's a new technology that creates a new kind of sort of, you know, thing that you own, you know, thing that's yours, thing that you would consider it to be private, thing that you wouldn't want the government to be able to take without a warrant, you know, out of the gate, law enforcement agencies just naturally go try to

get those things because there are ways to solve crimes and, you know, it feels like that that's a legal thing to do. And then basically the courts come in later and they, you know, rule one way or the other and basically say, no, that, that actually is also a thing that is protected against, you know, warrantless, for example, warrantless wiretapping. And so I, I feel like this is the latest of probably, I don't know, 20 of those over the last 100 years. And, you know, I don't know which way it'll go, but I think it's going to be a key thing because, as you know, people are already telling these models, you know, lots of things that they're, you know, that are very personal. Okay. Lightning round, quick questions. We're letting you get out of here in a couple minutes. We're in this age of spiky intelligence. Models are great at some things and then terrible at others. Where are you actually getting value out of AI right now? Where is it falling down for you? Where are you? How are you using AI day to day?

Yeah, so I have two kind of, I don't know, barbell approach. One is for serious stuff, I love the deep research capabilities. And so, and I'm doing this in a bunch of models, but like the ability to basically say, I'm interested in this topic. And I just felt like write me a book. And I, you know, I'm kind of hoping for the longest book I can get. I always tell it like, go longer, go longer, more sophisticated. You know, but the leading edge models now, they're getting up to like 30-page PDFs, you know,

that are like completely well-form, you know, basically long-form essays, you know, just like incredible richness in depth. And, you know, if it's 30 pages today, I'm sort of crossing my fingers, it'll get to, you know, 300 pages coming up here in the next few years. And so, you know, I'm able to basically have the thing generate enormous amounts of reading material with just like, I think, incredible richness and depth and complexity. And then on the other side of the barbell is humor. And I've posted some of these to my X-Feed over the last couple of years.

But I think these models are already much funnier than people give them credit for. Really? Yeah. I think they're actually quite highly entertaining. a while ago I posted I had specific specific formats like you know that B, chatting back and forth, be me, B Mark Andreessen, you know, that format's great. Take a dip in my pool in my office.

They're really good. So they're really good at green text. That works really well. But the, for some reason, the ones I find historical are the, I have it right screenplays, you know, for like TV shows or plays or movies. And I posted it, I had it right, new season of the HBO Silicon Valley, you know, set 10 years later. Yep. And I had it write like an entire day. I had it right like 10 to 10 scripts for a complete season.

And of course, I just said, you know, make it like Silicon Valley, except, you know, it's happening. It's in 2021, it kind of peak woke. And I thought it was just, I think it's just, you know, I'll sit there two in the morning. You're just like laughing my ass off at how funny this thing is. And so I think these things are actually already like extremely funny. They're extremely entertaining when they're, you know, when they're used in that way. And I do enjoy that a lot. And I generate a lot of those that I don't post.

It's stay in the group chats. It's probably good idea. Your property. Yeah. Hopefully the Fourth Amendment holds on these. Yeah. I have one last question. Go for it.

And then I've got one more. How do you get a job as a venture capitalist in 2025? So I mean, look, the best way to do it is to have a track record early as somebody who is like in the loop specifically on new product development. And so somebody who, you know, be like deeply in the trenches at one of these new companies in one of these spaces, you know, participate in the creation. of a great new product and a great new company and, you know, really demonstrate that you know how to do that. You know, there's, you know, there are, there are great VCs who have not done that, but, you know, I think that is sort of a foundational skill set, you know, for working with the kinds of founders that you want to work with who are going to want you to have, you know, kind of very

interesting things to say on that, as I think, you know, still the best way to do it. Yeah, like feel the growth, be, immerse yourself in the growth, the, the, the, the, the, the aggressive growth environment, and then you'll be able to identify it when you see it from far. Yeah, that's right. Last question for me, state of M&A in your mind. How are you advising, you know, companies where you're on the board or just the portfolio broadly around what they should expect now and in the near future? You mean in terms of whether you can get things approved? Basically, yeah. Yeah, yes. So look, approval still, approval is not a slam doc. There was, there was a, you know, there was a, I just saw there was a medical device company this morning,

you know, where the acquisition was not allowed by the FTC. So, you know, look, there, still scrutiny. It's obviously a very different political regime in Washington, but you know, this is, this is not an minute, you know, by their own statements, this is not an administration that believes it's in total laissez-faire. MNA, and it definitely wants to, you know, in their view, maintain a very healthy level of market competition. Yeah. How many, do you expect certain companies to be negatively impacted by the Figma story, right? You have this deal gets blocked, successful, you know, IPO, Lina Khan is taking a victory lap. You know, many people are responding and joking saying, you know, someone, Lena cuts off the arm of a pianist and they endure and can create a masterpiece. And so I expect, and then you look at the example with, you know, Roomba, I think it was where Roomba had a deal with Amazon. It was blocked and the company has just been a shambles ever since.

So my concern is that people look at Figma and say, you should be independent. You just figure it out. Nothing can go wrong. Yes. Yeah, it's kind of taking a victory lap was very disconcerting. And for exactly the reason you said, which is survivorship bias, right, which is you pick the one that worked out. And then, you know, it's the airplane, the red dots, the airplane. You know, you ignore the 50 that are in the ground that you've never heard of.

And so that was very disconcerting because that, you know, it's sort of the central planning fallacy, which is like we make centrally planned economic decisions. We have one example. You know, it's like in Europe. It's like, yeah, well, the bottle caps actually don't fall off the bottle, right? like, you know, it works. Right? It's like, okay. But do you want to live in an economic regime in which the government is dictated

bottle cap design? The answer is clearly no. Because the downside consequences are thousands of weight. Or even looking at that, you know, the Chinese model, which is, you know, people can say they're picking winners, but to get to maybe picking a winner, you have this intense blood bath of competition where, you know, teams need to rise to the top and sort of, you know, prove themselves before they get any of that real, like, you know, meaningful state benefit. Yeah, that's right. And so you just, you just, yeah, you just, you just have this adverse

selection, survivorship bias thing where you just, you don't pay attention to all the collateral damage. So I do think that mentality is like super, super dangerous. And so, yeah, look, I think companies just have to be very thoughtful about this, both acquirers and the acquires, you know, and the big thing is if you're selling a company, like, you just need to anticipate that you might, you might not get it through. And if you don't, there's sort of, are they like, okay, number one, is there like a big enough breakup fee, right? Are you going to get, you know, paid for the, you know, paid for the, you know, the damage that you're going through, you know, and how is that structured on the one hand? And then two is, yeah, look, do you

have the kind of company culture that's going to be able to withstand that? And is your business, you know, strong enough to be able to get through that. And it is a real risk and something worth, you know, taking very seriously. Yeah. And that's, that's why it felt emotion. We were at NICC last week. It felt emotional this, that, that the, the, the Figma team was, was able to like effectively just like restart the business and say like we're we're taking this all the way to think about it if you talk to any really successful company what they'll tell you is yeah over the years we have these like crucible moments in which like we almost died right but we like pulled together and we pulled it off and then that became like you know one of these central kind

of mythical events in the history of the company that we always referred to and like my god we got through that and we're so strong and tough and we've been forged in fire and now we can do anything and it's like yeah that's great and then there's 50 other companies that's at those critical moments blew up and died. Right. And so, yeah, it's all of the quote lessons learned on this stuff. They're all conditional on life survival. And so these things need to be taking incredibly seriously, you know, which the great CEOs do. Well, thanks so much for joining. We'll let you get back to your day.

We already give you five minutes over. Next time we have to book five hours because this is fantastic. I got 10% of the way through. We'll do the first 24-hour TVP. Yeah, we would love to have you again. Marathon. Enjoy the rest of your day. We'll talk to you soon, Mark. Have a good. Bye.

Thanks good. Thank you guys. Thank you. Up next, we have Harley for Shopify coming in the Temple of Technology, the Fortress of Finance. The Capital of Capital, welcome to the stream. Harley, get that gong ready, Jordy. What happened? Talk to us.

The update, Harley. How you doing? The update's good. I just finished watching you guys with Mark. It was an amazing interview. It was really great. He's fantastic. Before I get into Shopify, you guys mentioned the emotion of being a,

at the NICU with Figma yesterday, I think, right? Yeah. So we are 41 quarters now, and maybe 42 quarters post-IPO. I think it's emotional for anybody. I mean, obviously, that story is incredibly emotional because of what happened with Adobe and all that, but I think it's emotional for anyone who goes there. Although I do think, you know, you'd said that

is that going to, you know, create some momentum for more companies to do it independently? I don't know if that's going to be the main catalyst or not, But I said this last time when I was on your show, I want to say it again, for those companies that are out there, there is this perception that, like, the public markets are something to avoid as much as possible. 42 quarters in, let me just say it, like,

one of the best things Shopify did was go public. It's made us a better company. It's allowed us to be a lot more transparent. I think there's, like, this hygiene thing that these quarterly reports do. So I don't know where else to do other than a show like this, but this is my endorsement that if your company is ready, the team is ready, the business is ready, accessing the public markets should not be this thing that is like, if I don't get acquired, I guess I'll have to take the IPO route. It's been an amazing

experience for us. That's awesome. Forty one quarters. Yeah, amazing run. Advice for Dylan Field. How can he make the first 41 quarters of Figma, their public debut, a success? Oh, wow. I know Dillian, he's an amazing founder, a great, great, entrepreneur so I don't know if I much to teach them other than to say that you know we looked at the we looked at the IPO as sort of like we actually called it game day like we were graduating from the minor leagues to the major leagues and I actually think of that that metaphor actually works really well it's it's not like you're done in fact it just like you're just sort of graduating to the next step the

thing that I think I I on the earnings call today I spent a lot of time talking about this and get into the results in a moment this idea of like providing like breadcrumbs to the public to the street to your you know you have this huge book of investors including retail but you have these 10 or 20 funds that if you're lucky and we've been lucky I assume Dylan will have the same fate they kind of hold your stock you know for a very long time and making sure you leave enough breadcrumbs so that they can anticipate where things are going to provide some consistency I think is really really important so I don't know

like becoming a trusted version of Figma in the public markets is a much better way to, I think, do things than just becoming like a different version of Figma. I think the reason that Figma is so successful is because they understand their culture, their product, their customer base. I think this is just sort of the next phase for them as opposed to like, okay, we're done. Let's now change the company, become something different. And in sort of this era of founder-led companies and founders being allowed and permissible to run their companies over the long period of time, I think that works much better. One thing that stood out to me is Figma was launching new features on the day of the IPO,

which to me just sent this signal that were still the same company that care. Dylan was replying to customer support posts on the IPO day. It's amazing. I mean, look, that is like we do the same thing. There's something different about founder-led companies. And with that, less about Figma, more about Shopify. Yes. Okay.

So obviously, you know, we had results this morning. Give us the news. The news is Q2 GMV was 87 billion, that's up 31%. Let's go. Revenue was 2.7 billion, it was 422 million, which is 6% of revenue. So back to the consistency point, we've now had 11 consecutive quarters of positive free cash flow, eight consecutive quarters of double-digit free cash flow, and it's, you know, this is Shopify operating on all cylinders.

Amazing. And yeah, I think that it was a really good quarter. I think there was a bunch of these interesting uncertainties that the street had. Tariffs, de minimis, macro. But our merchants did disproportionately better than the overall e-commerce market, which is really cool. And then I got to announce some amazing brands, you know, Michael Corr's, Canada Goose, Starbucks, these incredible companies that are now coming to shop at Burton came, melee came. So really cool to also be able to talk about some of the big brands that are,

now joining Shoplight too. Break down some of those concerns kind of one by one. I'm sure you did, but tariffs, de minimis, you know, the market broadly, how have you guys been kind of approaching that? And, and, yeah, how do you think you got through it? Is it just that it was always, you know, kind of like a lot of these political changes, it feels like really crazy and then it rolls back and it's actually fine? Is that the right narrative? Or is it the actual like adaptation and agility of both Shopify and the merchants to actually work around a changing environment. I mean, it's a team effort, too, because each individual brand has to say, hey, we're facing some headwinds here, and we have to figure this out. We have to navigate this and, you know,

just find a way. Yeah, yeah. It actually carries all the way to consumer because the consumer also is like, well, what if I lose my job? What if I, you know, have, have less than supposed to one come. Now do I have to select, do I have to make choices and tradeoffs of what I buy, what I don't? I mean, you know, just to kind of at the high level, we are not seeing signs of slowdown. I'll just start with that. We can look at data actually through early August. We're August 6 today, so not obviously not half the month, but certainly the first week of August. And generally, we're not seeing any slowdown.

The factors we monitor at Shopify are consumer spending, household savings, tariffs, foreign exchange trends, and then supply chains. And generally, we're not seeing nearly what. what I think a lot of us were concerned about. The other thing that I think is interesting is that I mentioned, like, you know, Starbucks and Burton, Cana Goose joining Shopify. One of the interesting things that happened that I don't think we fully anticipate, I didn't anticipate this, is that because of this uncertainty, a lot of these bigger companies were beginning to be like, they were rethinking whether or not their tech stack was future-proof.

Like, am I spending too much money on technology? Is my technology partner and commerce, in our case, platform? Are they future-proofed? Like, I'm hearing all, you know, you think about, like, you know, think about like a metaphorical board meeting in one of these very large retailers. Someone is going to raise around and say, how are you guys thinking about AI inside the company? And what about like agentic commerce? And so I think one of the neat parts for us as act as a tailwind has been that a lot of the big brands

that historically said, we, our stack is not great, but we're fine with it, are now looking and saying, like, all right, this is ridiculous. It's like duct taped together. It's not future-proof. We're not able to, you know, we don't even. like our provider doesn't even know what agenda commerce is and that's leading a lot of brands to come to Shopify. The thing that we've, I think, you know, we were around in 2010 and so sort of at the tail end of the global financial crisis. Certainly pandemic obviously was had a lot of changes

too. What we try to do is not necessarily, you know, forecast what any, any organization, any administration is going to do from a policy perspective, but rather figure out, okay, if any of these things happen, how do we set up merchants on Shopify so that they're better off than merchants that are not? So in the pandemic, for example, like, immediately, we didn't know how long the pandemic was going to last, but we immediately were like, okay, there's a bunch of these, like, restaurants, for example, that are now going to have to do some sort of, like, delivery service. We're not, we don't, restaurants is not one of the core verticals that Shopify has ever been in, but, like, let's just make sure that we help them. And perhaps

they end up going to accessories later on and they'll eventually stay with Shopify. Or, you know, a bunch of physical retailers now have to move online very quickly. What can we do to actually help them do so at this incredible clip? So that's kind of how we look at these times of uncertainty, which is just simply prepare merchants on Shopify. So maybe they don't want to do a tariff calculator. Maybe they don't need to do tariff calculator. But if they decide to do so and it's valuable to them, let's make sure it's embedded in the product

and they can simply just opt into it. So that's kind of the way we do that as opposed to, you know, You can read as much information. You read every paper and listen to every interview from the administration as possible. You still may not know what's happening. And so rather than do that, let's just anticipate all these things could happen. Let's just make sure our merchants are prepared. I'm not sure if I'm like over storytelling here.

But Burton, is that a full circle moment for the company? Hell yeah, man. Tell me about the significance of that. The significance of it is that, you know, it takes him so long. What were they doing? Oh, man. Okay, so Canada Goose, Danny, who's incredible, you guys know, Canada Goose, right? Yeah, yeah.

Okay, so Danny Reese, an incredible, incredible entrepreneur, one of the greatest retail entrepreneurs on the planet. He's a friend of mine, he's a Canadian guy, Canada Goose and Shopify are two Canadian, you know, stories, success stories. And I've been trying to get them on for a while. They finally came on, too. But Burden is really relevant because the history of Shopify is that when Toby moved to Canada in 2004, he couldn't get a job because he was a new immigrant and no one would hire. him so he ended up deciding to start a business which was you know if you're an immigrant that's okay and he decided he would sell snowboards on the internet

because he was in Canada and he left a snowboard and he couldn't find any good software and so he wrote he wrote the software to sell these snow words the store was called Snow Devil and that software that he wrote this for Snow Devil would become what is now Shopify and so the fact that Burton one of the most important snowboard companies in the planet is now using Shopify is really really cool. But I love the, I mean, I spend, if you listen to the earnings calls, I spend quite a bit of time on these calls talking about some of the larger brands coming on. Part of it is that I'm very excited by these brands. Like, you know, I love the fact that these, it's weird, but like Hunter

Douglas, Burkinstock, Mattel. These are brands. Hunter Douglas was created, I think, in 1919. Burkandstock was created in the 1700s. And, and I think, yeah, so Mattel was created in 1945. I'd like that these very large, iconic. retailers are selecting Shopify. The other reason I like to talk about these brands. And it's significant too because there's certainly some people living under a rock and early on Shopify was loved by small merchants that were just starting out. And I'm sure, I'm sure you went through a bunch of calls with investors and things like that over the years, which was, yeah, it's great, but like you guys just have the like long tail of enterprise.

You're never going to, you're never going to really dominate with the retailers that matter. And like clearly that's, that hasn't been true for a while, but you still kind of have to just like say it over and over and over and over. Yeah. So I'm just, I'm just repeating it over and over again. The other thing that was that we announced today was that we now have 12% of e-commerce market share in the U.S. So if you think about from a checkout perspective, yeah, that's, thank you. That's really good advice for entrepreneurs over there. Just go get 12% of the entire market. You probably be good.

A lot of people say I just, I just, all I want is 1%. No. I'm a little bit higher. Go for 12%. It's pretty good. Go for 12%. I think makes us the second largest checkout in the U.S. on the internet after Amazon. Congratulations.

How do you think about sales cycles? Because for someone like a Burton, I imagine the first conversations there. That feels like this whale that is just a natural fit for the product, but it's not the kind of thing that was closed in like a quarter. So like how do you kind of even work with the team on that front to understand that like, we're actually playing. And, you know, ideally we close all the customers we want to close next quarter,

but realistically, sometimes great partnerships take time to come together. And you guys almost have the luxury of like your 41 quarters in. You're probably thinking about, you know, you're thinking,

you know, actually able to think, you know, long term. But in some cases, it's actually, it's not necessarily based on size or like GMV band.

But we have merchants on, Shopify that are doing hundreds of millions of dollars of GMV and they have like I'm not joking like they have like 12 employees. Wow. And then we have merchants that have you know thousands of employees and are doing far less in GMV so it's not necessarily about the size of merchant it has to do I think with two things. One is complexity in terms how much you know technical debt and baggage do they have meaning how much duct tape do they have to undo to migrate over and then the second part is like you know what how

How motivated are they to move over? I remember when Emily Weiss left Glossier, Kyle Leahy replaced her. Kyle's an incredible CEO. Kyle came in to run Glossier, and Kyle called and said, you know, we're a cosmetics company. We have this incredibly, you know, complex technology stack.

We have tons of people running e-commerce. Do we need this? And we looked at it and were like, no, we can help you. And I think the Glossier migration, Again, that was a homegrown stack. And we said, no, we can help you know. Well, that was at an era where investors, by the way. Well, yes, but also investors at that time wanted to invest in consumer brands that had an insane CTO.

You know, it was like, it was a badge of honor to be like, no, we're not using off the shelf software. Like, we're running on our own tech stack. We're a technology company. Give us a technology multiple. Well, and exactly. You got, I mean, you said it, right? It's a technology multiple. They're like, well, we're not really.

this word with this other thing as well like sorry garlay i can't switch because if we don't have our own tech stack like my multiple is going to get cut i talked to one cTO at a d to c c company that they didn't just build their own e-commerce stack they built their own sales force they built their own c rm email email management and the good thing is that everybody realized that you can build your own tech stack and you're still not going to get a software tech multiple no and your cost is going to eventually and not just just not your cost to Just talk about your efficiency. I mean,

Oh, totally. Back to the 12% market share in e-commerce. There is no checkout that is more performant than Shopify's check-out, simply because we have more data in which to make better decisions with. So even if you do have the most technical team building the most incredible technology stack, you still don't have the economy of scale that you can get by being part of it. That's the weird part that people miss about Shopify. That's the reason why the 12% number is, I think, valuable.

If you sort of put together, if you were to assume for a second, that we were one single retailer instead of millions of, of individual stores, we would be the second largest online retailer in America. As part of that, you are entitled to incredible economies of scale. So even if you can do it yourself, even if you can do it in a cost-effective way, you're still missing out on, like, you know, we announced the agentic stuff, which we should talk about yesterday. You can go to your own deals with every single agent that's available, or we can just

do it for you. We're already doing it for millions of others, might as well throw in with us. Your point though on the multiple is interesting. Something shifted, I think. At some point, I don't know what year had happened, but at some point the flex of, oh, I built my own e-commerce stack, stop being a flex and started to look like, why did you do that? I mean, why? 2016, unit level write downs. That's what happened. You basically start looking at it. It's like, wait, you're spending $3 million a year on this, like, technology team and that is like cutting that could be cutting I don't know safely like 30 million dollars off of your like I was running a new commerce company in that era and I went around and I went

and I did uh I looked at all the different companies all the leading DTC companies looked at their head count on LinkedIn how many people they have in their technical organization what's the average comp and I found out the amount of money that they were spending as a function of revenue and some of the companies were spending 10% of revenue on just building e-commerce software. And I was like, and so at that time, I was like, oh, you know, we were on Shopify Plus and I was like benchmarking. And I was like, okay, so we're like an order of magnitude more efficient, but we could be even more efficient. We need even less people. Like it was just like we could we should be focusing on the marketing. We should be focusing on the brand. And there's this great.

PE firm, you guys probably know it called L. Catterton. Oh yeah. Yeah. Yeah. So El Cajon is amazing. Michael Chew runs it. It's like Michael Chew and and, uh, the R&O family. Bernard Rode was involved there too. And Michael is an incredible investor, but he's acquired a bunch of merchants, some on Shopify, some or not. And one of the things he's recently told me is that in some of those meetings

where he's about to, you know, where he brings, sort of their first meeting once they've been acquired by them, he says, okay, well, now when are you migrating over to Shopify? Like at some point, it flipped from being, oh, I have all this opportunity

because I can build everything myself to, I actually think you should be, like, do what you do best, which is you're an incredible cosmetics brand, and go do that really well. Yep, totally. Did Stables come up on the earnings call at all? I brought them up a lot many times.

What's your updated thinking? How do they fit into the kind of Shopify ecosystem in the near term, long term? Sorry, you said Staples or stable coins? Sorry, sorry, stable coins. Oh, if I just staples. Staples is a merchant on Shopify. Oh, yeah.

Yeah, yeah. So, you know, the big problem. It feels like we're in the, They're not legal. There's a public company. They're out there. But when are we going to see people paying for it? I'm sure that there's integrations.

I mean, a decade ago, you could pay for a shop. You could do a shop on a checkout with Bitcoin with the Coinbase integration and a couple of others. Since 2012, I remember because that was when Coinbase started in YC. Exactly. 50 cents had SMS by 50. 50 cent was trying to compete. No way. Beats by Dre.

No way. It was like it was a big battle. Anyways, we had both, I think I'd done we had both of them on. And 50 cent actually. You have a payment that work? I think we're one of the first merchants to actually accept Bitcoin. I don't know if anyone did it. But look, I think in that era, that was really about speculation, not about detail.

The way that we think about stable coins in general, and I'll get into USC in particular, is that anything we can do that can bring more flexibility to commerce is a very good thing. I think stable coins, U.S.D.C. in particular, it just gives merchants more choices, more security, and it offers much faster settlement. Now, the other thing it solves, which is not getting discussed, but it's important, is that it solves a real problem, which is cross-border. Cross-border is now becoming this, it's not becoming a feature of modern retail. Like, default global is how most modern, the best merchants, the best companies I know are default global. They don't necessarily look at these geographic physical bounds as different markets.

It's like, well, if you sell to the US, why wouldn't you also sell it to Canada? It's a same continent. And so I think with stable coins, this idea of offering a payment method that combines, like, the transparency with a blockchain with the speed and a price stability of a major currency is really, really valuable. And then in terms of like, you know, doing it with a trusted partner with us, it was Coinbase, which we love, it means that you can actually bring all these, like, you know, very familiar your commerce features like authorization and refunds, but you don't actually need a new wallet.

Part of what I think is scary about this type of stable coin is like, well, do I have to now create a new account? And like, what is the friction involved? And I think the way that we're looking at it is that merchants get paid in dollars. There's no new wallet created. There's no added friction, but it's a seamless, flexible, safe transaction. So again, it is not a about, it's not a speculation, it's more about utility. I think once, I think it'll be a slow sort of increase in penetration. And then eventually there'll be a cohort of consumers and a cohort of merchants that are just like, this is so much better. Can we shift to AI? I know that you have some stuff that we should review. I also want to ask about shopping in LLMs and then how

that flows through. There's the potential of ads and there's potential of the checkout happening, you know, or, you know, I send my agent out to purchase it. Just give me the updated thinking on how AI is changing or how AI is changing commerce. So if you think about, if you think about it from a sort of a model perspective, think about Shopify as being the hub in the middle. This is like retail operating system where you have your inventory, you have your analytics, you fulfill orders from there, you have your customer data. Everything is sort of at the sub. And you sort of think about like the main spoke, that hub has been e-commerce for us.

And then a second spoke, obviously, was like point of sale at physical retail, which is one of our largest growing segments. And then you can sort of think about new spokes. Another spoke, for example, is like social commerce. We integrate with like Instagram. We integrate with YouTube. We have a Roblox integration, a Spotify integration.

So one of the things that we think a lot about is, the future of retail is not gonna be this weird binary online versus offline. It's going to be where every surface area where consumers spend their time. And for most merchants, most of the time, you know, the Roblox integration is not going to be their main thing. Although, you know, Fenty Beauty, one of our brands, is doing really well with this Roblox integration.

They have figured out that their consumers are spending time. And there's actually a physical, they have a story, like it looks like a physical store inside of the Roblox universe. Roblox, by the way, is like hundreds of millions of monthly active. Yeah, we were talking about earlier. It's model. I think it's, we talked about earlier, 20 plus million DAUs in the U.S. That's unbelievable.

Under 18. Under 18. It's like all, every, everyone. Okay, so like think about it from like, you know, the perspective of a direct-to-consumer cosmetics brand where their core demographic is that age group. Now it's a new place and there's no competition because no one else is like, like Fenty created a store inside of Roblox. You look next door unlike their store in Soho, there's no one else.

So we think about this idea of like these different sort of spokes being different channels. Yep. We think that agenda commerce is, is. very likely going to be a new one of those spokes. And so in the last 12 months or so, we began to build infrastructure that allows these AI conversation, basically to bring native shopping inside of these AI conversations. So we launched three things. And I'll go through all three because you guys are technical and you have a technical crowd. So the first one is catalog. We launched it in this past quarter. So catalog effectively helps agents to search and to sort of surface exactly

what customers want in seconds. It looks across all of Shopify's products, every skew on Shopify, and then it uses large language models to categorize them based on like, you know, metadata and the types of products they are. It also creates sort of this like standard product data at these massive volumes. And then we package it in these like basically these search APIs. So if you are, if you, if you have an agent, you simply can ingest the search API and have the entire catalog of Shopify merchants. And is this tech buying an MCP server or is this AI that's just accessible? So you can do an MCP server with like UI components or you can do it through API.

Okay, got it. The second piece is universal cart. This is new. So it's part of what we call checkout, K, which I'll get into a second, but effectively it holds items from multiple stores in one single spot. So as you're sort of having a conversation with your agent, let's see you're going on a camping trip, for example, you may want a tent, but you may want to on a sleeping bag, those are different stores. You may not be ready to check out immediately, but you want to hold all these things

in one single cart. And then that all feeds back in the checkout kit, which you launched last year, which lets partners embed merchants checkout directly in their agent. And it has shop pay built into it, but it's the best converting checkout on the internet, which is ours, and now we're effectively giving it

to these agents. What we announced yesterday was that now partners, any agent can actually theme the checkout kit so that it matches the applications look and feel. So it doesn't feel like some weird, you know, eye frame that pops out that looks like it's a third party. So it actually looks natively integrated. And then, you know, merchants get the tools that they need. They'll get our adage verification. They get fulfillment. So and actually in the case of checkout kit, Microsoft's co-pilot is already using it. So the benefit is sort of threefold. The first is for

consumers. Now they can get these like personalized conversational-led shopping experience. From the merchant perspective, merchants on Shopify are now going to have a new place to get discovered across all these AI platforms. But from the partner perspective, which I think is the most interesting, is that they don't have to build the complex parts of commerce. They get access to millions of these merchants through this MCP catalog, and they also get the best converting checkout. And we think if this does become a place that is a predominant or a popular surface area for commerce, it means that merchants on Shopify will be very well positioned. And what's neat about it is if you think about how a lot of those, you know, like if you were trying to go on a camping trip now,

you'd probably do it in some sort of search, Google or something like that, the products you're going to find are likely going to be based on, like a lot of them at least, may be sponsored, where the more you pay, the more likely you are to show up. Whereas in sort of the agent world, in sort of agenda commerce world, it's actually more of a relevancy. It has all this history of everything you've ever talked to it and said to it.

It knows that you're price sensitive for certain things, but not for other things. So it may turn out to be a really new sort of vector for commerce and we want to be at the center of it. How do you, what's your kind of looking out into the future commerce within LLMs? Are you bullish on paid placements? Are you bullish on referrals? I was obviously like in the same thing as, you know, in the influencer world, like consumers deserve to know what's an ad and what's kind of an organic mention and all that kind of thing.

There have been a debate recently around, you know, should ads be banned entirely from LLMs? And I, you know, generally don't agree with that just because I think, you know, these tools should be broadly accessible and ads are potentially a way to do that. But the other thing would be ultimately, like, do LLMs, would LMs at some point earn referral fees from merchants? I mean, all this stuff is, like, extremely complicated and can have, you know, positive effects, negative effects. But what's your framework?

That I think is the $64 question. I'm not sure. I don't know exactly. Because part of the reason that I think agenic shopping is so interesting to so many people is because it feels. more democratized. It feels more based on it has a great understanding of who you are. Therefore, you know, you guys started the show with me today saying, like, I'm going to have my agent go and buy me a bunch of stuff. I want my agent to go, if that is how I am purchasing in the future, I want my agent to go by things that I actually really like, not the thing that

my agent is getting paid for. Yeah. Yeah, when I think about the use case here is like I have certain brands and I only buy their t-shirts. And if I could just be in an LLM and say like, hey, I want to get some new t-shirts. Can you like? And they're like, cool, here's the T-shirts you normally buy. Would you like? And I'll be like, I'll take five white t-shirts and five black t-shirts. And it's just done. Like, that's amazing because I don't want to be navigating around. And, you know, it's just like removing that friction, which is kind of the, I mean, has been the history of Shopify. So all these new products. If you think about, you know, I think Toby told me this, that like, if you want to look at,

you look out to where technology is going, look at what, like, rich people currently do, and then everyone else be able to get that feature. So you go back to this idea of, like, personal shoppers. If your personal shopper is constantly selling you the thing or trying to sell you the thing that they make the most commission on, it may work once or twice. Eventually, you'll stop going to that store.

You're going to look in the mirror. You're like, dude, this is, like, some gold chain. Like, I wear a black t-shirt. Like, I know you're getting paid more for this gold t-shirt, but I wear a James. purse black t-shirt every day of my life and James purse is great chopper by merchant and a great entrepreneur himself there we go he's never he's advertised doesn't advertise so like he's so like he's not going to participate in that but if it knows that i really love James purse black t-shirts

it has to be highly contextualized so that that is a really good question I think the way that we look at is is like this there is a there is a there is a decent chance that agentic shopping will become very relevant for some segment of the market. Therefore, if we want to qualify and re-qualified to be the merchant that retail operating system for all these incredible brands, we have to be there in the same way that we have to be in social commerce.

And social commerce, for some people, it is a really, really important channel for them. You know, my mom is, I don't think my mom has ever bought something on social commerce. She still goes into a physical store. So we're not going to predict, like our job is to make it

so that wherever you want to sell, you should be able to do so really easily. It's going to be absolutely wild when you think about, I mean, this experience where you're chatting with a model and it surfaces you a product and you can ask it, well, what do people like me think about it? And it automatically pulls, you know, reviews from relevant people. It pulls what your favorite influencer thought about the product. It surfaces other brands. Like it's going to be incredibly powerful. And I think at best, it's like a personal shopper that's aligned with you, not, you know, trying to maximize every person.

purchase or a great sales rep who's if you're working with you know a sales rep at a retail store they're not fixated on like how do I maximize this the value of this cart right they're thinking about this sort of long-term relationship with the customer so very exciting I'm excited to see what people build well congratulations thank you so much for hopping on and chatting with us always fun to be always so much fantastic work you guys are the best I love your show love what you guys are doing and I like that this is becoming a new routine that after we'll see in three months we'll Four times a year. We love you, too.

Thank you. Great work. Cheers, I'll see you guys. We'll talk to you soon. Up next, we have Anton from Lovable coming in the studio. Get that gong ready. Jordan. Shopify.

Oh, what? 20%. 20%. Yeah. I was trying to tell him that on the stream because I wanted the eyebrow raise, like the, you know, from the Brian Chesky when the IPO and Emily Chang tells him how much the stock is up.

And he goes, eh. It's like one of the most iconic moments in tech history. Anyway, we got Anton from Loveable. How you doing? What's happening? Great to meet you finally. Yeah, yeah, great to meet you too. We were going back and forth with folks in and around your orbit.

Every time we talked, there was a new massive number. There was, you know, certain amount of people on the platform, certain amount of websites. Give me the latest headline number that's shocking. Isn't it like some insane number of websites on the internet are generated every day? That is true. What are you holding? I'm holding a gong, a mall to hit the gong to celebrate. I just learned today.

We have 100,000 new projects built every single day. Congratulations. That's fantastic news. How'd you do it? What's the key to success? What are people using lovable for more than anything else? When you think about just a web page can be anything. It can be everything from Facebook, a billion users, a trillion dollar company, to, you know, a little homepage for my, you know, my, like my resume.

What are people using it for? Yeah, look lovable, the fastest way to go from an idea to a production-ready application. And a lot of people make that into real business. And that's something we, that's in our mission to empower much more people to do. But apart from the entrepreneurs, the solepreneurs like building for the first time or building for the end of time without but by themselves, you have a lot of people in larger companies now. seeing that use case growing very, very fast, like percentage wise much faster. And then they go from like, I have this idea.

My engineers should build it, but they don't understand me. So I need to create a full working version of it. And then they're going to be like, okay, I get it. It looks great. Let's build it. How are you thinking about tool use right now? I feel like there's so much open source software there. LLMs are incredible.

They could probably rewrite Linux from scratch. They could rewrite a database from scratch. but you want to be provisioning and standing on the shoulders of giants. But how often are you focused on allowing the like the like to grab different open source projects and tools? Like if you want to build an e-commerce software, you probably don't need to rewrite a catalog from scratch. It's probably a tool for that.

Yeah. Now, how we think about making the product as good as possible. Thought about it for a long time is like we have an agent that can do things. You can browse the web, generate images, but specifically create applications and the website, personal website for you, whatever. We are giving this agent more capabilities. We're adding new things. Okay, so for e-commerce specifically, we're taking the best of breed to build e-commerce as a tool.

And more to, I'm spoiling some, I'm teasing some things there on the e-commerce side, actually. Giving it more tools and then we're making the agent itself smarter. And those are the two things we do. Give me portals, make it smart. What's the key to high retention? I feel like there's a world where you use a tool lovable, you build something, and then you're like, OK, we're ready to go and stand up a whole team of engineers to actually build this thing and take it to the next level.

How are you fighting that urge to just use this as a prototyping tool? Yeah. Like if you have a high performing engineering team with a low-performing team with a large existing system, then there's always this balance of like, do I take this thing I built in Loveable and I connect it to my system? Many people do that. Or do I just use it as a design? And then I know exactly how to build it, so it's going to be super fast with my engineers. Both are fine. Or do I even edit the code as an engineer level also popular? I think regardless of all of those, you can have very, very high retention because if you want to just prototype in Loveable,

it's the easiest way to do it. You just open the browser and then boom, you have your prototype. And it has access to your design philosophy. It has access to integrations and different tools that you might want to use. So there's a high retention. There's a lot of high retention on that use case. But how I think about it going forward

is that we want to just build a platform where you never want to leave this platform because it provides you so much value. And what large language models have done now, that they generate code really well. But that's just one tiny part of the entire life cycle of building and maintaining and operating and the products and growing a business on top of that. I'm not sure you know Elina on her team.

She's like, I've been doing growth for 20 years. Now I'm so excited to not have to do the boring growth, and just do the innovative growth, and AI is going to handle all the growth things. So that's on the horizon. And when you have a platform like that, you should be getting so much value. that you never want to live.

That makes sense. How do you, what do you think is the future state of, you know, what do you think the average startup's website will look like five, ten years from today? I feel like right now we're in this era where AI is changing everything. It's changing workflows. It's changing the way we work with software. And yet startups, you know, still buy, they try to buy their one word. Dot com domain.

And the typical website is like incredibly, you know, static and tries to reach a odd audience and I can imagine a world in the future where websites are being almost generated in real time depending on the user and the type of customer they are. More parallels. We're getting more parallel. Knowing the trend to date, it will be more of that. But I'm curious, you know, kind of if you have. I think the websites of the future are going to be better at hacking the human brain and

like it's addicting the nature, addictive nature and so on. And what that means exactly, I don't know. We're like, you let the algorithms figure out partly. But, no, I think is going to hack your brain, everybody. Yeah, I mean, do you think that starts with like A, B testing? Because you can generate more pages quicker that you're just kind of optimizing for retention or conversion. And then because you're just generating the entire site instead of just, you know,

testing two different headlines. testing the entire concept of what the website is. Yeah, I think the algorithm, like there's more and more be algorithms to figure out how things should be optimized for us. And then I think what like what you're also seeing now already is a bit of counterculture to that with a human coming in and being like, no, I want it to be much more mean than least to like a new type of. Yeah, this is the Nat Friedman style that Mark Zuckerberg.

Yeah, exactly. Exactly. Times to Rome and just raw HTML. There's a time and a place for both. Time and place for both. I guess we'll see like some complex evolution of different ways of doing things. And generally, I think the future has always been more and more diverse. There's like many different websites and applications.

And I think that trend is going to continue. It's not like it will converge to just one way of doing things. Last question. From my side, what's next? what's next on the horizon? Are you focused on taking the product that you have? Clearly a product market fit. Are you focused on sales marketing, ramping up, expanding what you have or expanding the portfolio to different areas and different markets, different products? Yeah. I'll answer it like this. So in the order of like the sites of the use cases,

founders are taking their ideas and they're building businesses, software businesses. as designers and product managers and others in companies, they're taking their ideas and building concepts to communicate with their team up and like a hundred times faster than building out the concept in the past. And then everyone else is building their personal or business websites much, much faster. And we're doing all of these things. And LaMable does all of these things at the same time. And it becomes, as you use it, it becomes better and better.

at helping you. So then that's what we're going for. Amazing. Well, congratulations on all the fantastic growth, truly staggering daily use and numbers. Barely a week goes by without a big lovable number hitting the timeline. Yeah, another lovable number has hit the timeline. Congratulations. Thank you so much for stopping by.

We'll talk to you soon. Have a great rest of your day. And up next, we're shifting over to Thomas from GitHub, another little post-earnings breakdown, get the update from Microsoft and GitHub. Nat Friedman was running the organization. Thomas has taken over and they've been on an absolute tear. Just crossed to follow. Just crossed 20 million users. Its growth is serving 90% of the Fortune 100. Congratulations.

Kick us off with an introduction. Thank you so much for having me on the show. Great to be here. Thanks so much for joining. And yeah, I'm Thomas and I've been running GitHub, for the last four years. I've been with GitHub for seven years since we acquired it in 2018, and it's been a tremendous run that GitHub had as part of Microsoft.

Is it fair to describe the run as growth in overall users or growth in GitHub co-pilot revenue or, you know, this 90% of the Fortune 100 is using the product now? Has that been the win case? Has it been a little bit of all of them? Well, the real question, what's the other 10% doing? Yeah. Well, that 10% is not being a check on it, I think.

Yeah, this is the idea of principle. The last 10% is going to take you 10 times as long to get. We got to figure out the 10% and short them. I mean, but seriously, if you're a Fortune 500 company and you're not building some software in your organization. There are other companies that provide similar services that do well. You know, we're not going to talk too much trash about the competitors, but, you know, we are big fans. But yeah, over your tenure, what have been the big initiatives that you've shipped and then maybe open a bottle of Dom Peringian champagne or just had a pizza with the team to celebrate?

You know, we always keep pushing further. So there's celebrations every now and then what's really like about what's next, what's next, what next. If I look back to 2018 when we did the deal, we really had three principles. Put developers first, and I think we nailed that one. Get Up is still GitHub. You know, we are talking about developers. We're building for developers. We are meeting with them all the time.

And we have preserved that spirit of what GitHub is all about. The second one was that Microsoft helps us to accelerate. Microsoft, if you will, as an investor, into GitHub. And I think that's, you know, the 150 million developers on the platform. Recently, we hit over a billion repositories, both original repositories and Fox. You mentioned the 20 million copilot users. Microsoft's virtual studio family, so VS, the IDE, and VS, code together has 50 million active users.

So you can also see how much room there is still for co-pilot to grow its market share. And a year ago, we passed 2 billion in ARR annual revenue runner aid. And I think all these metrics show that both the business itself is healthy and keeps growing the network effects that GitHub as a platform has. And then the third one was that GIP helps to accelerate Microsoft. And the most obvious thing obvious is that Microsoft developers are using GitHub and are using co-pileged. So Microsoft itself benefits from the same productivity gains

that our customers benefit from. And then we're sharing a lot of the AI knowledge across Microsoft, you know, within the core AI division, Azure AI foundry, in all these tools. I'm hearing glimpses of self-improving artificial intelligence. Microsoft is using GitHub to improve GitHub. I'd love to see it. Where else are people using GitHub copilot?

Where are the, what are the biggest, like, win, conditions or win use cases in those Fortune 100 companies, where's the biggest value being derived right now? Believe it or not, I think the biggest value is still the core scenario, which is you have your IDE open, you know, V.S. code or JetBrains, even Apple X code. And you have writing code. And instead of switching between the IDE and the browser, where you have, you know, all these tabs open and you watch this show. And there's all the distractions, you know, in your browser and you try to find the answer to problem you're trying to solve you just

have it right there a value building so whether that's code completions that you know predicts to you something that you might have not even thought about in that moment or whether there's the ability to go into chat and ask the question about that code and find a bug or I was working on my blog on Sunday and I couldn't figure out how I couldn't remember the mock-down syntax to an embed an image and just like give me that a bit quick right and then all the way to agentic scenarios And I know you talked with Anton right before. It's this ability to give it a task and then what should do these things and call tools.

And it figures out how to install this NPM package. And then it looks at the error message. And it sees that there's a test case missing. So I think this end-to-end spectrum where I'm still the pilot and I'm still in charge because I think and I strongly believe that the majority of developers actually want to build something. They don't want to offload their drop to somebody else, like an agent or, that set of agents, they want to use these agents to do what they love doing most. Yeah, can you talk to me about like where the boundaries of GitHub or Microsoft's agnosticism

land? Like, you don't make everyone right C sharp, you don't make everyone use VS code. But how are you thinking about integrating different models at Microsoft Build? Satcha Nadella was talking about the importance of Azure being a place where you could get everything from GPT4 to Deepseek to Lama to it sounds like XAI is coming on as well. How are you thinking about integration with a broad suite of new tools as they come up? Choice is crucial for any developer tool to win the market. You can't be in the space of selling developer tools and not offer developers choice. Because if you don't do that, they go somewhere else and they will find the choice somewhere else.

So since last year, we're offering what we call multi-model choice. So we're not only having the open-EI models, which are still great and many people use as their default. We also have Anthropics models. We have Google's models. We actually have, you know, bring your own models so you can connect from co-pilot to OpenRouder or OLama. And from there, you can go to, you know, every model imaginable as long as it has an API and a key that you have access to. And so we think that is crucial. and developers will not want us to tell them what model to use.

Like that's just like, you know, intrusion in my, you know, personal freedom. And they know better what's best for them in the same way that at GitHub, we wouldn't tell you use this open source library or, you know, use this programming language. Like you wouldn't, GitHub wouldn't be what it is today if it had only, you know, one ecosystem, let's say JavaScript, right, and all its libraries. Now obviously there's boundaries there because we have only so many people, people at GitHub at Microsoft that can integrate all these models and there's a new model every single day sometimes multiple times a day yes yesterday there were two new models

three I think you count the two open source ones and all these models need GPU capacity and you know responsible AI testing and and red teaming and all that and so we believe it's a mix of we provide models out of the box and you bring your own model um uh through open router olama and what have you fantastic jrida last question last question And you have over 100 million developers on GitHub. Have you tried to estimate how many of them are not using any AI tools that are just, you know, hanging back saying, I'll make my code by hand, please. Handmade. None of that AI for me.

It's over 150 million. I think it's still, you know, less than 50% that use AI on a daily basis. And that's just the nature of things, right? If you look at what are these 150 million developers, there's lots of students there where the professor may not allow the use. of AI or they haven't made that step because they're really just trying to solve that, you know, task from their homework or from the test. But at the same time, we see more and more developers making their jump. Software development has always been a spectrum, right? Those that are still working on co-borne mainframes. In Germany, there are still companies.

I think the train system is looking for people that have Windows 95 experience in 2025, right? Like, that's the, like, any of you new grads, no Windows 95 for when from when you were? But believe it or not, even if you are new grad or post-grat and you join a company and you have to work on a COBO project, as long as you can then also use GitHub and copilot and AI tools, that might not be the worst job in the world, right? Because you can combine modern technology and apply to that old stuff. But what I'm going to say is that we have this huge spectrum between the people that are most ahead on the curve, you know, those watching your show or talking on the show here,

and then those that are still having to maintain the same the systems that power the world. And our mission at GitHub is to cover all these developers and enable them to collaborate between humans and humans and humans and humans and agents with each other. That's where the platform is going. Well, thank you for your service. And congratulations on all the growth.

It's been great chatting with you. We'll let you get back to your day. Super insightful. Have a great day. Thank you for having me. Talk to you soon. Bye. Up next we have Alex Jacobson from 137 Ventures.

We've had Christian Garrett on the show many times for 137 Ventures and we're excited to catch up with Alex of ours. We have a great conversation with him a couple months ago excited to catch up and chat with him about everything in the 137 Ventures profile. How are you doing Alex? Good to see you. Oh good. You too. Sorry we're running late. We were yeah we ran a little late we got a bunch of very yappy people this guy Mark Andreessen came on he was talking a lot we were asking a lot of questions so we put us behind but good to catch up with you. How are How are you doing?

Good. I was just watching, but I was three minutes behind, so I thought I had a bit more time. Sorry about that. It's amazing. Yeah, we're going to figure out how to feed you the show so that as soon as you join the room, you're seeing it live instead of with the delay. But anyway, thank you so much for joining. What's new in your world?

What's the biggest news in the 137 Ventures portfolio? I'd love to just get the general update. Oh. The best start. Well, I'll start. I'll start. Christian had like a screen. screenshot. He was on Squatbox about Palantir.

It was, I know, but it was it basically had the 137 portfolio up and it basically looked like the top 10 most in demand private companies. And I was like, oh, it looks like a pretty good portfolio to have. So you guys have been everywhere. Yeah. I mean, we have, we're building this general concept of mag seven for private. And so we have there, there's this idea of there's these companies that grow indefinitely in the public markets and they've built them themselves to do that in the public markets. The strategy for doing that in the private markets is different because you need to do different things in different ways.

And so SpaceX has led the way on this because they give employees equity and they then also run tenders regularly so that equity can become liquidity. And the important bit about that strategy is that the price at which they're providing liquidity is materially lower than what the market would pay in a fully liquid public market. So you can say, oh, that's terrible. The employees aren't getting the right thing. But that's not fair. The actual thing that's happening is when employees are issued equity when they get hired,

they know it's going to go up because SpaceX keeps growing. SpaceX is one of all of these companies have the property of having a huge amount of power in the market and even larger TAM. And so you can understand this indefinite growth. And you can talk about it for the public companies or the private companies, but the general thing that you're looking at is this combination of power and large TAM. And so in a private market context, All these companies need to hire people to go capture that TAM.

And the value of these companies keeps going up. And the strategy is to give the employees certainty that the equity they've been issued will be worth a lot more in the future. That's harder to do in the public markets because everything's perfectly priced. And so you don't really know if the stock's going to go up. You think so. You hope so. But it's harder to tell. In private markets, the nice thing is you control your price.

and you can, if you run your process properly, you can have this very deterministic outcome for your employees, which I think is super powerful. We talked to Harley at Shopify about the benefits of being public. You have liquid currency for acquisitions. It sets him up. He referred to it as being in the major leagues now. Give me the pitch for staying private as long as possible.

I mean, I don't think. I think SpaceX isn't major league? Yes, I agree. So there's... What are the benefits of staying private? I think the big one is hiring. Okay. That's the thing that I think people underestimate,

which is that if you can issue people private company stock at your 409A and then you control the price, it keeps growing, then this is all a much more more deterministic thing and it's super powerful to be in these companies because the nice thing about these companies there's this funniness of they're always under price so there's always investor demand for them yeah so as long as they as long as you just have this positive feedback loop of there's always investor demand for them so the investors want to buy more and the stock price keeps going

up which means the employees can look at these companies and go oh this equity is really worth something because they know it's going to go up. If you're getting public company stock, you don't get that. Yeah, yeah. And there's not the volatility that comes with. I check my portfolio today. I'm down a little bit just because, oh, there's some weird tariff thing going on in some foreign country and how does that affect my business?

Whereas, yeah, if you're in the private market, you don't have to deal with that. What about these new, what's your take on some of these new initiatives to give retail investors access to private company shares or put private company shares on chain and create tokens and SPVs and all these different financial engineering efforts to bring, to give retail traders access to private market company stocks. What's your take on all that? We've been seeing people trying to do this forever. It's an ongoing thing of, oh, we find it inconvenient that these companies are private,

so we're going to try to do things to make them seem more like public companies. But one of the big values of being a private company is controlling who your shareholders are and controlling your pricing. And so if you're trying to stop that from happening, maybe that's an opportunity. And maybe you can force one of these companies to be de facto public. But that's not a service. If they want to be public, you know, there's this mechanism called the public markets to do that. Yeah. What about a higher level of abstraction? Like, I believe, I believe Golden Sacks is public. I mean, the other big, big thing is, you know, certain private companies don't want to be public because they don't want to be on this sort of regular reporting cadence. I just don't feel like they're ready for it.

But the idea of like taking private companies, putting them on chain, letting private company shares, putting them on chain, and then just letting anybody in the world buy access, you know, buy them, but not giving them any of the information that allows, that makes the public market so beautiful, which is like anybody can read this information, and you can come up with a decision on your own. And there's like rules and frameworks to make sure that people aren't insider trading

and things like that. And so it's like, who's benefiting here? It's like retail has a potential to get even more hosed. And then the companies are having to deal with shareholders that now have an opinion about how they're operating the company, but none of the real rights associated with owning the shares. Let me be charitable to these people. There's a real thing of the only people who get access to these private Mag7 companies are institutional LPs who invest in our fund or funds like ours. And, you know, we're not, we're not, we don't have retail investors in our fund.

There's there's a sort of in the manager class, there's something we refer to as retail, which is there's a longer tail of wealth that is now trying to play in and invest in these things. So there is something that is in some sense a longer tail. But there isn't the people who have, who are public markets players. There aren't non-institutional players in the game, really. And there's something real about giving people access to the growth that these companies are going to have. And it's unfair in some general sense that these retail investors don't have access to this growth. And so I think there is some amount of institutional design around how do we give them access to this growth. But turning these private companies into public companies is not necessarily the way to do that.

Yeah, it's almost like, what if the Yale endowment was publicly traded? It's like, then I'd have broad exposure to a bunch of venture capital firms, which have exposure to a bunch of private companies. And, you know, it still just rolls up to just a few line items on the cap table. I'm not, you know, this random retail trader doesn't, isn't actually on the cap table. And it's also, it's also disingenuous to say that like everyday people don't benefit from the private Mag 7, as you've described it, because a lot of venture capitalists raised from pension funds. Sure, sure, sure.

Teachers and firefighters and things like that who are actually, you know, benefiting from the performance. I mean, I think there's a pension. We're definitely in the pension funds are definitely beneficiaries and that's very real. But I don't know how much of the, like, there's definitely these, you know, those types of pension funds that are out there. But there's lots of people who aren't in these sorts of pension funds. pension funds. I think you're right. There's definitely that that sort of universe is taken care of. But I think there's plenty of people who aren't in that type of pension fund either. Yeah. Totally.

And generically, they're in the pension, it's in a pension fund. They're not able to say, oh, I would like more exposure to privates. Yeah. Right. So at a, at the class, there's there, I, I understand the feeling. of that these people are underserved in some way. They're not my customers, but I get the feeling. It is funny to think we need to design institutions for it. It is funny to think about if a bunch of Stripe shares were dropped on Solana,

like what Stripe would actually trade at? Because I would guess that it would look like it would be like the most insane pop. And then suddenly it's like, okay, if you want to buy Stripe on chain, you got to pay like $400 billion or something. Or a trillion dollars, right? And for some of the even bigger, like, potentially like me-me-y private companies, it could be even crazier. Yeah.

I mean, there's the whole Elon universe of companies. Yeah. We have a sample of one of them that's public. Yep. We know what happens there. Speaking of that, I mean, he was talking loosely about taking Tesla private at one point. That, you know, obviously didn't materialize. But is there a world where the private market?

it evolved to such a point and it becomes so elusive that you see a take private of a company specifically to get back on that track or once you IPO has the ship sailed and it's gone forever? I mean, I live at the private markets. I like the private markets. I think there's a huge amount of value of being here in the private markets. And so I'm definitely talking my book. But the big thing that has changed over the past some number of years, is how big the private markets now are.

So we, so the, how much we can invest in SpaceX or Anneral or Gusto or any of these companies has gotten to be large numbers. And so, you know, there's this argument of when you're public, you have this currency that you didn't have because the public markets have more money. And I don't think that's as true anymore. I think that there's this real chance that you could decide, I just. want more control, I don't want to deal with this regulatory nonsense, and be private. There is this, they're both, it's a regulatory ritual, but it's not actually different company, and you might want to just have a different regulatory environment. Yeah, I'm thinking specifically about SNAP had earnings.

They're down 17 percent today. They live and die by the earnings call. Meanwhile, Apple's up 5 percent, meta platforms that have 1 percent. Amazon's up 4 percent. And it's only a $13 billion company. The capital exists and there's always been this narrative like it's easier to turn the cruise ship when you're in the private markets. You go to just a few investors. You say, hey, we're going to miss earnings really bad for a couple quarters, but then we're going to build something new. I don't know. It's a wild idea.

You could have three snaps or one perplexity. Wow. I think it's there at 20. I mean, the part of, I mean, the hard part is these things are now priced. Yeah. And so you're going to take it private and you're now to go resell this to the. the private capital markets who have looked at this in the public markets and gone, well, that didn't look right. And so I think there's a hard sell of going to the private markets and selling, you know,

selling snap in the private markets. I think it's entirely possible that it could do better. But, you know, they have to go. They have to sell it. There's a founder-led story. I mean, I think we can like sit here and script it. and maybe we can make it work. It would be a super exciting story.

I think the best example is probably Dell, which went through Take Private, and then later went back public at a much higher price. But I mean, these stories are extremely few and far between. Well, switching gears. I wanted, we had Dan on from Armada earlier this week. Super exciting company. I know you guys have been involved from probably before the company was created.

wanted to get your view on the opportunity and what made you so bullish. Now or back then? I mean, both. I mean, we got it, last time we talked, I think we got a good, good overview, but it was kind of hearing the kind of landscape and the initial catalyst was interesting. I mean, the original version of this is if you think about what a data center is, A data center is a point source, is trying to maximize the value of a point source of connectivity and power. That's like structurally what it is.

And so, you know, power is available at different prices, different places, fiber is available at different levels of reliability. How many fiber points do you have? Are they really one or two? It's doing the underwriting of what, it's actually a lot of work to make a data center. But the whole premise is that you have sufficiently reliable and large power and sufficiently reliable and large connectivity. SpaceX changes that. So SpaceX says, actually, there is no longer this concept of a point source of connectivity.

It's available everywhere. And in any one point, it's not as good as fiber, but the world's pretty big. And so the aggregate bandwidth of Starlink is huge, even if the availability in the square foot you're in is not as big as holding a fiber in your hand. And so the insight is, okay, well, SpaceX gives us this opportunity to rethink how we do connectivity. So does that mean we can rethink how we think about power? And the interesting thing is, if you're looser about connectivity, now you can. look at power in a bunch of different places and you can say, hey, there's solar everywhere. So now you just simply put a solar panel and a dish and you're live.

And is the solar panel enough for the amount of compute you want to do? Maybe, maybe not. But that's the beginning of the conversation. And then, you know, so in the original design for this, I sketched out a shipping container that unfolds solar panels on a roof. And I figured out that you need a lot of solar. panels to power a rat but you can do it but and so that's but you know the math on that actually worked so it was like a 3% IRR when you did the model and the 3% IRR isn't glorious but it was

just a model and then you have the testing so it's say oh this is it's efficiently interesting that it's worth testing and then when we then we got dan and dan is this phenomenal has phenomenal network is phenomenal sales and his solution wasn't halo let's build it and sell cloud. His solution is calling people he knows and asking them, what do you think? Because that's how he works. And so we get on the phone with the CT over Ramco, somewhat of a surprise to me. So I'm suddenly on the phone with this guy who is...

San Francisco Company, right? Yeah, founded in San Francisco. Right. West Coast Tech wins again. Yes. And so I'm suddenly pitching this to this guy. And the insight is, well, there's this other problem. The underlying story that I just told you is about stranded energy.

There's stranded energy all over the place. Let's use stranded energy. But the thing that's happening on an oil platform is there's stranded data. And so these oil platforms produce huge amounts of data that get effectively dropped on the floor. And they have some amount of compute that is in a closet somewhere. that some guy on the platform manages. And Halliburton support and Haliburton sells in the software. It's like the actual architecture of that market's crazy.

And so it's like, oh, wait, we could do better. And so if we provide a cloud at this locus of stranded data and stranded energy and in global communications, we can create a lot of value. That's great. Very exciting. Well, next time you join, we'll have you on for an hour. Yeah, we can go so much deeper. Thank you so much for stopping by.

We'll talk to you soon. Great to see you. All right. Good to see you. You hang out soon. Bye. Cheers. Of next, we have Nick from Rillett coming into the studio.

I think we got to get the mallet ready. We got to get the wallet ready. Warm it up. the gong ready. Jordy, you want to take this one? I'll take this one happily. I think I hit the last one. But let's bring him in. Let's play some music. Let's play some soundboard. Let's bring in Rillet. How you doing? Welcome the stream. Thanks for joining.

Sorry, we're a little bit late. How you doing? Hey, check, check. Do you hear me okay? Oh, you're great. Is it Nick, Nicholas, is it Rillette or relay? Rillette. Relat. So the American spelling. Yeah. There we go. Okay. I wasn't sure if it was French. Anyway, give me the update, give me the news, give me the overview of what you're building.

Awesome, yeah. Excited to be here, guys. Thanks for having you on. So today we're announcing our $70 million series B, co-laid by Andreessen Harowitz and Iconic. Congratulations. Oh, that was a good timing, good timing. Congratulations. Boom.

Thank you. Give us a little prehistory on the company. When did it start? How's the growth been? And key customers, how you're building and how you're thinking about, solving AI native ERP. Yeah. So we've, so quick, quick backstory also myself, so deeply accounting and finance versed, sort of

full background there. Started this company roughly three and a half or so years ago, very much in stealth for a very long time. As you can imagine, a full-on accounting system or ERP is not a small feat to build. It's a lot of surface area. These are very entrenched systems traditionally dominated by Netswee, Oracle, SAP, these type of names. So we, yeah, we're very fortunate to have had a stellar team of accountants and engineers,

building and building. And then we came out of stealth last summer, roughly a year ago, and things have gone vertical since then. So it's been an awesome journey. Race our Series A here just a couple months ago, three months ago, 12 weeks back. And, yeah, back to back here with a B. What does it take to get a company to rip out their existing ERP? and bet on a new company, right?

Like, that's the real challenge. I was about to ask, are people ripping out Oracle and SAP and NetSuite, or is it I'm upgrading from a spreadsheet? Yeah, great question. So we are getting 70% of our customers today coming from softers called QuickBooks and Zero. Sure.

And then 30% of our customers coming from NetSuite and Sage Intact. Okay. So it's very much a mid-market-focused ERP software. That's good. Key reasons why people use Rilet or want to buy Rilet, despite us being newer to the market over a legacy player, is number one, we have super strong native integrations that really suck in all that key upstream information that you need for AI automation really seamlessly into our platform. And then our automations on top produce reporting that's just unmatched in the market.

And that would, yeah, leaves us eventually to getting customers like Windsor, one of the fastest growing AI companies in recent history, mind and others to trust Rillett over their current systems. Rillet, now used by Google and cognition potentially. That's the best part about a zombie acquisition. You get potentially two new customers. Maybe. Anyway, I want to know about, do you want to put a chat box in your product?

Or do you want to use AI behind the scenes and not surface that to the customer? Whenever I see tools like this, I think I love the idea that you're using. AI, but I don't want to open it up every day and prompt, hey, clean up my ERP. I want you just to do that. And then when I show up, I want it to look like a nice groomed garden. Yeah, love it. Hopefully, a bit of both, actually. So for very important for our controllers, accountants, and CFOs, having a human in the loop on some of their AI processes is actually really important and a positive. So yes, these are background processes, but always with a human in the loop for the most critical ones. there. And then for you as a business owner specifically or a strategic CFO or investor,

you want none of that detail, right? You want to go in in a clean box, ask a question, get the information you need, move on. So we're building both to both ends of these spectrums. And you really need both to be successful here. I guess my question is a lot of people want to go, a lot of CFOs, a lot of investors want to go into the ERP, the accounting suite, and get information. But I feel like 99% of the time they're asking for the same thing. They want a really clean balance sheet, a really clean income statement, a really clean cash flow statement. And they don't want to think of a prompt like, oh, I want, let me describe to you what a balance sheet. It's like, I want a balance sheet. I just want it to be accurate and I want everything to be

tagged correctly. My prompt is don't make mistakes. So talking about the tradeoff there. Yeah, great question. So definitely there are these pre-can reports. You don't want to be teaching and AI to rebuild your income statement. Every single fucking, sorry, that's for it. Every single time, every single time the same way. And so 100% agree there. There are a lot of more custom analyses, though, that you want. If you're a professional, like specific for your business, maybe track also some non-gap metrics and non-accounting metrics on a financial side. That is really handy to just describe it in natural language and get what you need.

Yeah. And then the other question is, like, I feel like one shot prompting in this context is even less relevant because if I, if I'm, if I'm, if I'm, if I'm, if I'm designing a new non-gap metric, I probably want to say, like, set up a workflow so we produce this every month and that it's just on. Like vibe adjusted diva-dive. Exactly. I want to vibe code and adjusted divitaph. But then I want it to be the same every month because when I've reported non-gap metrics in the past, like churn, I mean, even Dow Mao, that's a non-gap metric. But like you want it to be accurate and you want it to be classified the same way because there's often these times where you change one underlying thing in your product and then that flows to a complete change in the non-gap metric and then you're going, hey, it was a non-gap the whole time. Don't worry, don't get mad at me, investors. I know you thought that that was never going to change, but the metric change. And it's actually not that different for the, we're just looking at the business in a different

way now. But talk to me about like how AI can improve the development and ongoing maintenance of non-gap metrics. Yeah. Great question. So there is definitely an element of like repetitive adjustments, certain things that can even be codified outside of AI in terms of certain customizations of your non-gap metrics, ARR, is a very prominent one for software AI business and recurring revenue businesses, for example. And so there, yes, it is definitely something you can do basically in the UI, customize your metrics the way you want it in a repeatable fashion. But there is definitely also an element of certain workflows and I almost call it openness to

human interpretation where you need these more like L11 driven workflows that help interpret maybe look at each chart of accounts, look at historical transactions, what happened there and try and make sense of things. And that's where AI today is very powerful. But again, you do need usually a human in a loop to do these repeatedly and reliably, just given them mission criticality of the workflows. What's next? What's more important for you? You're in the mid-market. Do you want to go downward, upward, upward, do you want to expand and offer go multi-product? What's most interesting to you? Yeah.

Yeah. So for us, more of the same. The opportunity that we have in the mid-market is insane. The amount of customer love we're getting and product quality that we have with the reviews that we have combined with like really entrenched and frankly universally disliked ERP systems, you'll be hard pressed to find someone that truly enjoys using their accounting system or ERP creates a huge influx of just demand on the product and the company. So what we're going to do with the money that we're raising is double down on product,

build more of the same additional features, go up market from here with the money that we've raised, we have a stronger balance sheet to capture even bigger names and logos that can work with us. And then the other piece is just investing heavily into our customer success and onboarding motions. We're just like we work with very talented accountants in these teams to help on board, help on board their platform into their systems. And we're massively expanding these teams to just absorb the demand that we have. It's awesome. What an opportunity. It's great.

Yep. Congratulations and good luck. Many people have tried and failed to rebuild the, to build the new ERP. There was a gap in the pre-AI era where there were a lot of people that took a run of it. And it was a little rough. finally. The mind share is there, the opportunity is here. It's really exciting.

Amazing. Congratulations. We'll talk to you soon. Have a good day. Cheers. Thanks for having us. Bye. What else should we cover?

ARP. I didn't like the way he was talking about NetSuite. I would die for NetSuite. I would die for SAP. I love. They can all win. I love all. They can all win.

I love all in. Equally like my children. I like them all equally. Anyway, perplexity. You mentioned it briefly. It's now worth almost 15 times JetBlue. They raised another $200 million.

One and a half snaps. A little under, yeah, I think snaps at 13. But perplexity is now at $20 billion. It's unreported, but it's been leaked by RFA Rock. Who knows if it's true, but he usually gets it right. So we're chatting about it here. 138,000 people saw the news. I'm sure Arvin will comment if it's fake news.

But in other news, what else do we have? We wanted to send our regards to the Doge staffer who was assaulted or gotten a dust-up with someone who was attacked. He stepped in. And fortunately, we saw another picture that he has been cleaned up and healed and seems like he is on the mend. And so sending our best to Mr. Balls. rough, but sounds like he acted with, as this poster would say, bravery and courage. Yes. And that's pretty much everything, Jordy.

I don't know if there's anything as you want to cover. It's a good place send it. Maybe leave us five stars on Apple Podcasts and Spotify. Maybe thank you if you've been in the chat today. I like this post from Aidan Belt Skies says timeline, not in turmoil. John and Jordy pulled off some S-tier handshakes during. the Figma IPO live stream. This is what it's all about, boys.

Keep it up. Thank you from the chat. That's on the X chat. And Mark is live. Live handshakes. High stakes. Nothing. High stakes. We've seen some viral botched handshake slash dapp ups.

You got to have a shared language. Stick to the default handshake. It's Linde is not going anywhere. Do you think it's better to look at the person's eyes or the hand coming in? Because I think it might be more about like a like a, like kind of you don't actually you want to keep the I think it's kind of coming in it's kind of like a game of rock paper scissors and so high stakes game but but but the optimal the optimal strategy is to just stick with the handshake and never deviate never never never never never doubt yourself

with one of these make it very clear from early on you're coming in with the handshake and you're not blinking it's a game of chicken so if you come in with it doesn't matter if you come in with knucks or you come in with with the I'm going to hit you here doesn't matter As long as you stay here forever, they will see that you're not back and down. They'll conform. And you'll assert your dominance. Yeah. So go out into the world and practice. Practice live.

Yeah. I mean, it is tough because I hit people with all sorts of different things. Every time I go to the bathroom, Ben hits me with the knuckles on the way back. It pumps me up, gets me back on the stream. Back in the game. Anyway, thank you to John Exley for moderating the chat. Fantastic work today. Thank you to Gabe for chiming in with hilarious comments all stream. Thank you for Mark for watching and enjoying the stream.

And we will see you tomorrow. Tomorrow is going to be an absolutely insane day. We can't say why, but it's going to be insane. Something's going to happen tomorrow. And I cannot wait. I'm pumped. It's a stacked lineup. We will announce it soon. See you tomorrow.

Bye. Cheers.