Weekly Recap: Sam vs. Elon, Cluely's Crazy Ad, UFC x Paramount, Apple's AI Robot

You're watching tvpn.

Jordy, what is going on in the timeline?

We got a red flag.

We got a red flag on the play. The timeline is in turmoil. Elon Musk and Sam Altman are fighting. Mom and dad are fighting. Mom and dad are fighting. Shots fired. This all started because of Apple, actually. Not OpenAI, not X, not the Everything app, but. But Apple and the. And the iOS charts. So this started with Elon Musk posting Apple is behaving in a manner that makes it impossible for any AI company besides OpenAI to reach number one in the app Store, which is an unequivocal antitrust violation. XAI will take immediate legal action. Well, the. The community notes section took immediate action and added a number of extra context that they thought people might want to know. First, In January of 2025, DeepSeek reached number one overall in the App Store. Similar to how Ramp is our number one sponsor on this show, Time is Money Save. Both easy, your corporate cards, bill payments, accounting, and a whole lot more all in one place, folks. Go to ramp.com and just one month ago, on July 18, 2025, perplexity reached number one overall in India. App Store. That's interesting. I didn't.

Grok has hit number one in Japan.

Oh, that's right. Yeah. But probably contributing to that. So most of the AI apps live in the. In the productivity section. So that must have been in that. But it's odd because it's like we need a whole new set of. Of of categories for the new AI apps. I feel like they should have their own category immediately. What are all the categories? I mean, productivity. I feel like that's everything from like spreadsheets to ChatGPT. Like it's a pretty broad. It's a pretty broad CA. And so I mean, Apple probably needs to redo the categories. Where even are the categories? I feel like you can't. I feel like they're hard to find these days. They have editor choice apps, which we'll get into because, okay, so browse categories. They have Apple Vision Pro apps at the top. That's an odd choice for the iPhone version of the App Store. Then Apple Watch apps, then AR apps, then books, Business developer tools, education, Entertainment, finance, Food and drink. Food and drink gets its own category.

Well, I'm looking at the global top charts right now or global for the US so across all the different categories and right now ChatGPT is number one. Number two is tea on her dating advice, which.

There's a new Tea app which is.

Helping men date Safe.

Wait, wait. So it's the men version?

The male version.

The male version. It's men's cereal.

Very poorly the T appraisal. But it is number one in lifestyle and number two on the overall charts. Number three is T dating. Four is Threads, which apparently just crossed 400 million monthly active users.

Adam Mosseri is not messing around over.

At Meta hq and Grok is number five.

Okay, so it's not that bad.

This is despite Apple. I do believe that Elon is correct and that Apple's not including Grok in features, which are editorial.

That's a separate thing.

Editorial.

Yeah. So yeah, I mean Doge designer says Grok is the number two app in productivity, number six overall. So why is it so hard to find? And what he's pointing out is that Elon Musk shared this. Grok is the smartest AI in the world and on the toughest test and came first by far. Encoding but is not mentioned all at all under AI but by Apple. So Apple did an editorial view on the AI powered apps and they listed out a bunch. ChatGPT, Google Gemini, Microsoft Copilot, Lightroom, photo and video editor, which, I mean, I use it every once in a while, but it barely has AI features. It does not compare to a foundation model. It's not a chat interface. Then you have Canva. Then you have Duolingo, which of course Duolingo uses artificial intelligence in some ways, but you don't think of it as like a competitor to ChatGPT. So this is clearly just like here are some fun apps and we're using the AI buzzword over at Apple.

And I think it's given. Given the series of PR crises that Grok has experienced in the last call it six weeks, I think it is fairly fair for the App Store to say, hey, let's let this breathe a little bit. We're not removing them from the App Store, we're not removing them from the rankings. They're still ranked, but maybe let's not feature it given that it's been saying.

Yeah, the other interesting thing is Grok usage has to be split across X and Grok the actual app. Like I bet I wouldn't be surprised if the majority of actual prompts or tokens flow through X the everything app. Like that's certainly how I use Grok. When I use it, I usually start with the post, click the Grok button, then I'm going back and forth a little bit there. And then also they actually have a pretty awesome flow where now on any image you see On X, you can press and hold and say, edit this in Grok. And then you can just type Studio Ghibli and it just Ghibli's it right there in Grok.

And they launched a feature to turn it into a video.

Video, exactly. So all of those Grok tokens, all those Grok usages, that's going to feed into the X app, which again, they've always put X in news and not in social networking because I think it couldn't beat Facebook and Instagram, but it can beat like the, like the traditional news apps. So all of the different rankings are all very, it's all game to different ways. Some of them are momentum based, some of them are different categories. And I mean, we heard this from Nikita Beer where he was saying that his last app that went super viral, he hid in the games section. Those were his words. He hid because the game section is so crowded that people wouldn't see him climbing the rankings.

But social charts are based on acceleration. Broadly.

Exactly.

Total usage.

And there are only like 10 dominant social platforms. So if he had a ton of momentum, he'd probably be in the top 10. People would be screenshotting and being like, what is this? And it would be on Mark Zuckerberg's desk. And Mark would have to be like, what's our plan to kill this? Right. As opposed to if you're hanging out in games next to 75 Candy Crush clones, you could be doing well, but realistically you're not really gaining that much. You're not drawing that much attention. Then at the last second, he swaps over to social networking where he actually wanted to be, and then he's at the top of the charts anyway.

So Elon says Apple's behaving in a manner that makes it impossible for any AI company besides OpenAI to reach number one in the app Store. He starts getting community noted. Sam Altman fires back, says, this is a remarkable claim. Claims that Elon is using X to benefit himself and it's his own companies and harm his competitors. Sam starts getting Community noted. Elon fires back and says, you got 3 million views on your BS post, you liar. Far more than I've received on many of mine. Despite me having 50 times your follower count. Elon's getting community noted. But Sam fires back, Skill issue or bots?

Skill issue or bots.

Anyways, timeline's in turmoil. Timeline turmoil. Will you sign an affidavit that you never directed changes to the X algorithm in a way that hurt your computer competitors or helped Your own companies.

Wait, who said that?

This is Sam.

Sam said that. An affidavit.

I apologize if so. One thing is that OpenAI has consistently dominated the narrative on X organically. Many of their.

To some degree. I feel like OpenAI has been more dominant just in the real world and actually Anthropic's done very well on X. And there's been this whole theme of like, Claude has better vibes and big model smell and oh, I would just.

Say if Sam Altman puts a post up, he can put up a picture of a Death star and get 21.

You're 100% correct.

And so I don't believe that he's shadow banned. I don't believe that anyone in OpenAI is shadow banned or being held back in the algorithm. But somebody responded to this really quickly.

If Sam's worried about Reach, he should be using Restream, putting his content all over the Internet. Restream, one live destination, one livestream, 30 plus destinations, multiple multistream and reach your audience wherever they are. And thank you to everyone in the chat today. Mark Siraj, we got Taylor, David, we got a whole crew here. Techno Chief, I believe, said good morning to me. Hello, Techno Chief, Good morning. Good to see you. And it's fun having all the crew assemble in the chat rooms. We actually can see the chat wherever it's happening, whether it's on x or on YouTube because of Restream. So thank you to Restream. Anyway, continue.

A friend just messaged and said, somebody tell Elon to unshadow ban me.

Okay.

And I think that, oh, I think.

I know who this is.

I think it's much more likely that it's a skill issue.

I think it is a skill issue.

For this particular person. Sam is saying here, you never know. You never know. But somebody responded to Sam and Elon's exchange and they said, grok, who's right? Don't be biased. And Grok fires back and says based on verified evidence, Sam Altman is right. Musk's Apple antitrust claim is undermined by apps like Deep Sea can perplexity reaching number one in 2025. Conversely, Musk has a history of directing X algorithm changes to boost his post and favor his interests per 2023 reports and ongoing probes. Hypocrisy noted. So we'll see how long.

Yeah, but this is 4D chess, because in general, people don't believe Grok because they believe it to be insane. Well, it said some wild things, crazy things. And so if Grok is siding With Sam, people will be like, well, Grok's wrong. Elon's actually correct.

Yeah.

That's the beauty of these things.

Elon is thinking, what, have I created a monster? This is like the child that. What's that mythology where the child kills the.

Oedipus Rex?

Of course. Yes. Oedipus Rex moment. The child coming for the father.

Yeah.

Well, we'll see how long Rock lasts.

What would you want to play on that on that analogy gets pretty weird.

It gets pretty weird.

I mean, Oedipus marries his mother.

Yes. But we'll see how long Grok lasts in the timeline today.

We'll see, we'll see, we'll see. So the, the. The Wall Street Journal wrote it up. Apple is behaving in a manner that makes it impossible for any AI company besides OpenAI to reach number one. The App Store, which features lists of top and trending apps, is a key way for apps to access new customers. Apple doesn't to seem say much about how or why it chooses particular apps to feature. Among the factors it considers is usability and positive reviews and ratings. Yeah, it's definitely getting more and more complex.

It's just important to break out editorial decisions from the App Store team and true rankings.

I mean, I think a lot of people have not taken a tour of the modern app store on iOS because it used to be that you just show up and there were. Then there was just a list of like, the charts were like a few clicks away. I'm like seven clicks deep. There's an ad for just Google in here. There's essentials that are just top games right now. Like you have to scroll and scroll and scroll and then you go over to apps, then you can browse the categories. But even when you browse the categories, you go to productivity, it's going to give you like the. When I go to productivity, you think I'd be browsing to hit like, okay, show me the top productivity apps. No, it's. It's showing me an app called Sofa, a downtime organizer. Then it's showing me essential productivity apps that are selected by the App Store editors. Then the best calendar apps, best to do apps, best email apps, best focus timer, habit tracking. It's going, going, going.

And the big question is how much, how much discovery is actually happening in the App Store where people are actually.

Finding the top charts. But it's seven. It's seven clicks deep. And so, yeah, if I go to free apps, it's ChatGPT number one, then Grok number two then Microsoft authenticator number three. AGI cheat. Microsoft authenticator doesn't get enough love.

No, but the question here is, are people discovering various LLMs through TikTok and X and Instagram, various platforms? Is that where discovery happens is happening? And I'm sure there's some discovery happening in the App Store, but it has to be a very small fraction of the downloads being driven are actually driven by the editorial decisions from the App Store.

Of course. Of course. Yeah. And games don't even show up in the top. Like it's not. Games are not. It's a completely separate tab. You cannot select games as one of the subcategories of apps because games are not considered apps anymore in the App Store. To answer the question about like hiding in games. Yeah. The other, the other question is just like how much, how fair is it to give yourself a little boost in your app that you paid $70 billion for? Yeah, like I was thinking about this and I was saying like 44 billion. It was 44. Yeah, yeah, it was 44. It wasn't 69. I thought it was 69. For some reason that was the price per share was he paid. So ridiculous. Anyway, I was thinking about like before the launch of Instagram stories, I didn't have a way to share Snapchat stories on Instagram. Like you could screenshot them, but there's no sort of like native way to share a link to a story that if someone on my Instagram feed sees it, it clicks and it automatically opens in Snapchat. Like it was always a walled garden. And then when Instagram stories came out, there was really no way to tie those two platforms together. Like in the, in the pre meta era, there was a way when you shared a post on Instagram to check a box and automatically share it to your Twitter. And it would not just share a link, it would share the actual photo with the comment and it would work as a native post on X. And it could, you know, not really go viral, but it could, it, it existed on Twitter as like a first class citizen in the sense that the image was there, there was no link. It was just like you had just actually opened the app, posted the same photo with proper cross post. Right. And then eventually once Facebook bought Instagram, Twitter had to respond by breaking that integration. So early on, Twitter had this very, this was like the Adam Bain era, like the early back the Dick Costolo era. Twitter had this idea of like, we're gonna be the command line for your life. And there was this idea of like I'm gonna be able to tweet at you to send money to you. Like, the everything app idea is something that's over 13 years old at this point, maybe 15 years old at this point. Like, the Twitter executives were thinking about this a long time ago. Like, being able to tweet money at you, being able to, like, use. Use Twitter as kind of this, like, substrate or this, like API for all sorts of things. Other apps would be built on top of Twitter. There was tweetdeck and Tweety Bird and like all these different. Like the mobile app was by an independent developer and they just didn't. And they were like, why would we need one? We just developed the protocol. Well, yeah, well, you need one because you want to stuff ads in it, obviously, and you want to monetize it. It makes a ton of sense. And so Elon King killed all that.

But the other challenge here is Grok has a variety of NSFW features. And when Apple is making editorial decisions about which apps to feature in the App Store, that is certainly going to be on their mind.

We're having Eugenia from Replica on the show later today. I'm wondering if Replica, I think is currently in the app store with 220, 26,000 reviews, hasn't gotten kicked out. I'm trying to think if any other. If any other apps. So like, only Fans, for example, does not have an app. I'm wondering if any of the. What character AI does. Right. So Apple's clearly drawn the line at like, like AI chatbots, even if they're somewhat erotic, are okay, but we.

But that doesn't mean they're gonna. That doesn't mean they're gonna feature.

They're not gonna promote it to the.

Average person in the App Store.

Yeah, no, that's absolutely correct.

So, and then again, I think it's very hard to say how Apple makes these editorial decisions, but it's very possible that they are talking with people on the Apple team and being like, what apps do you like? What apps are you using making decisions that way? Right. I think, I think part of Grok over the last month or so is like figuring out like, how do we get strong consumer adoption? Which categories are we going to get strong?

I'll tell you how they get strong consumer adoption. SOC 2 compliance. They need to get on Vanta Automate Compliance, Manage risk, prove trust continuously Vantage Trust Management platform takes the manual work out of your security compliance process and replaces it with continuous automation. Whether you're putting. Pursuing your first framework or managing a Complex program? No, that is a good question. How can Grok actually get consumer adoption? The companion space seems like the. Maybe the right move, even if it is a little weird. But is there any other path? I mean, forking it into X and Twitter and like having it show up on every post seems like a reasonable way. I wonder. It is weird that they have two apps. I thought X was supposed to be the everything app. Like, like why are they trying to get me to install a second app when they should be? Maybe just say, hey, we're actually going to be the everything app. You can use Grok fully and you're all the upgrade and you can get on Grok Heavy and you can be paying $200 a month for a top tier, super insane level LLM. But it all exists and it's all managed within the X app that already has a lot of, a lot of installations.

Why your Grok is top or near the top of a bunch of different benchmarks and yet that doesn't guarantee consumer adoption. Right. People don't switch products because something is 20% better in this technical way. Right. They develop habits. The other thing, it was only a month ago Grok or XAI announced a partnership with the Department of Defense, $200 million contract. So that's serious. But again, you know, figuring out where they're going to ramp revenue with consumers that when it's not just bundled into X, the social media app.

Yeah. I think it's interesting to debate like when can a new technology be bolted on and become additive to an existing platform versus you need a new app to actually get it to work. So Facebook actually rolled out a search engine in Facebook, the blue app, like years ago. And when you would go to search, you could search for your friends, but then you could also go and it would search the web too and it would say, hey, you're searching for this thing. There's no posts about it out there. So let's just pull from the web. And they basically built a search engine team.

Pull up this post from ChatGPT. I want to see John's live reaction to it.

Okay, let me see this post. Whereas like Stories was something that was added.

So ChatGPT quoted Grok saying that Sam Altman is right, said, good bot, good bot.

Wow. Timelines and turmoil. You love to see it. It's a great day on the Internet, but there's always. Yeah. So I think there is something about like the pixel screen, the screen pixel real estate that is somewhat like screen pixels are scarce and you can only do so much. Instagram was able to add reels as a tab, integrate those into the main feed. They were able to add stories up at the top as a little bar and it got crowded. But Instagram did not see massive churn. But I think that if they, I really think if they try and stuff like, okay, yes, this is also like Claude code. And you're going to do your, you're going to do your homework in Instagram now. You're going to do your, your, your software engineering in Instagram right now.

Here.

You're going to have a chat below.

I think people actually know. I think it's. I think it's too, too big of a step. Whereas going from photos to videos to reels to stories, like, all of that works within the same kind of UX UI space. But if you try and bolt on, like a productivity tool into a social network, it winds up being a little rough. And so maybe that's the reason why Elon, like, realizes that, like, Grok has to be a separate app. Because if I'm using Grok heavy to solve Olympic level math, I might not want that to happen on the Chaos Timeline feed. Right?

Yeah. I mean, if.

But it's an uphill battle.

If you're a software engineer working at a company and every time your boss walks up to talk to you about something, you're in X, the Everything app, just scroll on the timeline. While coding, that might be tough, right? It might be better to be doing that coding work with Grok. Right?

Put a spreadsheet in there, put a database in there. Let me build an entire business within the Everything app. Tyler, what's your reaction today?

I think so. One thing I've been thinking about is like, I forget who it was, but someone said, like, as a company, you want to, like, do things that your competitor that, like, go against your competitor's morals.

Right.

Do you remember this was Paul Graham?

Yeah.

Okay. It wasn't, it wasn't morals, but it was just like the idea of counter positioning this idea that, like Avi Schiffman at Friend, like, he's making a product that is a wearable AI device. And if he makes it look like an Apple product and he makes it like, oh, it's instead of an AirPod and, you know, an iPhone, it's a necklace. Like, and it has Apple design language, like, Apple will totally be able to just steamroll that by launching something very similar. Whereas if he comes out with something that has such a bizarre design that Apple would be like, wow, if we try and copy and he's not trying.

To make it useful for anything other than being your friend.

Right.

He's, he's also counter positioning against ChatGPT, which is a tool, its default state is not to be your friend. People use it that way and in some ways it's designed to function like that. But ultimately, yeah, I think he's found he's carved out his own. And Grok did the same thing by coming out and like leaning really heavy into companionship. Right. That is a, I think down that path.

If they're going to go like in that like kind of counter positioning like idea, I think they could also. I could see them doing almost like a cluly type thing where it's like, okay, you can use this for your homework. Like you make it very good at like, you don't have to use like cheating per se. But I think that's like an interesting way that like if they want to go further down that like consumer kind of positioning.

Yeah.

It's just, it's odd that it feels like they're trying to do a lot right now. They're still in the exploration phase. Because I have, I mean maybe this gets to like the value of transfer learning in reinforcement learning, but I have it. I have a hard time believing that many people who are in the market for like a companion are like, I definitely want it to be able to ace humanities last exam. Like I in fact, in fact super important. In fact, I think that you might want a companion not to be better than you at math. I think you might want a companion that just kind of like is like oh yeah, I agree man.

Like that math problem is a toxic friendship dynamic. People are like, I want my friend to be great but not better than me.

Exactly. That's real. So they need to dumb it down. They need to be the inverse of bench hacking.

Yeah.

The truth is a new meta is.

Going to be teaching your AI how to do things. Teaching it history.

Yeah.

Teaching it how to do your homework.

Yeah, yeah, yeah, yeah. I mean the Feynman method, you teach and then you learn, you learn. Like I would probably learn more from teaching an LLM the answer to a problem than it trying to teach me. What were you about to say?

I was going to say like the truth seeking like idea of Grok seems very much like, opposed to the companion aspect. Seem like two different ends of the.

Hero of our group chat right now is one of the greatest fabulous in history. He's a, he's a serial liar, but. But he's the most entertaining character. He's the court jester of our group chat because he hallucinates more than GPT2 and it's fantastic and it genuinely provides a lot of value than just someone who's like, oh, I will only give you things that are 100%. I will only say things that are 100% factual. I will only say things that are. That are definitively provable. It's like, no. He throws out a lot of wild pitches, but every once in a while you just slam it into the outfield. Roy Lee is in the news again. He bought a backward news in Manhattan and he said, alex Cohen says, fine. I'm starting to like the Clulee guys because Roy put up a billboard that says, hi, I'm Roy. I'm 21. Missing an apostrophe on the second. I'm clearly deliberate, absolute wordsmith. This was very expensive. Pls buy my thing. Please buy my thing. Clearly.com and John Chu Coastal Ventures says, this is why I invested in Roy Lee once in a generation. Talent tapping into culture and zeitgeist. This was a spicy post. People kind of going back and forth on whether or not this would actually deliver. Roy also put out a post saying that like everything he does is for short form. So he bought this billboard. He expects it to convert on the short form that will be created around the billboard. Certainly feels like people will create. Will take pictures of this.

I mean, when you. When you look. When you look at. I think what I'm interested to see from Clulee is where they really get. Where they really get, you know, retention. Right. Because they have the top of funnel. They kind of. Yeah, they got plenty extremely full top of funnel marketing. But when you look at like it doesn't. It seems like if they just focused on students. Right. They seem to be in this kind of place where they're. They're thinking about exploring a lot of different markets. For sure. Different markets. But if you look at the way that ChatGPT is usage, basically tokens generated dropped as the summer started. That's a good. It's a good argument to say Clulid could just focus on products for. For students that will eventually graduate into the workforce and things like that.

Sure. But anyway, they didn't make the top AI agents by revenue chart that a non sunwall shared said, who would you put your money today to win in the long term? The top was cursor at 500 million in annual revenue and they're making $3.2 million in revenue per employee. Glean's at 100 million. Mercour's at 100 million.

Replit, how are you defining an agent?

I think it's just AI startups. Anyone who's used the AI agent like meme basically stolen.

The agents are gonna be pissed about this stolen valor. I don't know, it's hard to call Merkhor.

I agree with that one. That one feels particularly odd but.

And glean even glean is enterprise search.

Yeah, search. But I would say Replit is an agent company. It's kind of like every AI company is now now a now an agent company. Harvey, you know, hasn't really branded themselves.

In agent but it's interesting like you're not putting Claude, you're going to put cursor on here which generates revenue from Claude code.

Yeah.

And is a gentic ide.

Yeah. Is crazy. And also not including ChatGPT which has deep research, which is probably the most used AI agent in the world right now. I have to imagine just by number of queries there. It's missing some stuff. It's missing some stuff, but it's okay. Also Devin's not on here which is kind of odd. But that's the nature of these lists. It's CB insights folks. They do the best they can. Let's pull up the videos of Sam Altman on AI ads and I think it will crystallize a little bit of like what mean when we mean like monetizing a free LLM, a free AI chat app. It doesn't necessarily mean stuffing display ads in there. Just like the answer to Facebook's monetization problem was not banner ads on the in the right bar. It was in feed ads that look if you're watching reels and you see a reels ad, it looks exactly like a reel. And in fact the best performing ads on Instagram reels feel just like user generated content. They don't look like super bowl ads.

They're additive.

Yeah, they're additive and people often enjoy them. And so that will be at least this is the semi analysis argument that I sort of agree with. That will be the, the, the like what, what we say about ads in AI it will be more like commissions for agentic checkout at least, at least to start. So let's pull up the first video example right now.

There's a lot of people that have websites that monetize with referrals to Amazon.

Yeah.

And they're frustrated because a lot of the, I mean traffic just like organic SEO is way down.

Yep.

And the general read here is that OpenAI will ultimately start to earn that same type of revenue that the publishers historically did?

Yep. So there was a fireside chat at Harvard Business School with Sam Altman.

Let's give it up for Harvard Business School.

Let's give it up for Harvard Business School.

It's the Harvard of business School.

That's what they've been saying. They've been saying. So he got a question from the audience about ad monetization, and we'll hear how Sam Altman responded to it. Although fair, it could be a barrier for early stage entrepreneurs or startups or even small businesses. Given this context, do you envision OpenAI exploring alternative monetization strategy that could include.

Like, free API access, perhaps supported by.

Advertising or other methods to foster innovation in the future?

I will disclose, just as like a personal bias, that I hate ads.

I think.

I think ads were important. We love ads to give the early.

Internet ads everywhere, but I think they do sort of somewhat fundamentally misalign a user's incentives with the company providing the service. I'm not totally against them.

I'm not sure you're ready for any.

Twitter ads, but I don't like them in general. And I think that ads +AI is sort of uniquely unsettling to me. You know, when I think of like GPT writing me a response if I had to go figure out, you know, exactly, exactly how much was HU paying here to influence what I'm being shown. I don't think I would like that. I think I would like that even less. So there's something I really like about the simplicity of our model, which is we make great AI and you pay us for it. And it's like we're just trying to do the best we can for you. And then, Senator, we run ads that has some inherent lack of access and inequality. We commit as a company to use a lot of what basically the rich people pay to give free access to the poor people or the poorer people. You see us do that today with the ChatGPT free tier. You'll see us do a lot more to make the free tier much better over time. And I'm interested in figuring out how we bring the equivalent concept to the API.

But.

But.

I kind of think of ads as like a last resort for us for a business model. I would do it if it meant that was the.

That's where he says. He says, I kind of think of ads as a last resort, as a business model. But recently he dropped a new podcast. This was from, I think a month ago. And it's from the OpenAI podcast. So you have to imagine that the run of show and the talking podcast points in here are very carefully selected to, you know, move the narrative forward and kind of educate the community on where the company is going. And so we'll pull up Sam Altman's interview on AGI GPT5 and what's next from the OpenAI podcast.

So that brings up the other question from people who are using this or skeptical is that OpenAI now has access to this data. And there's the concern. One was about training, which OpenAI has been very clear about. When or when not it's training, you have the options to turn that off. The other thing is like advertising things like that. What's OpenAI's approach towards that? How are you going to handle that responsibility?

We haven't done any advertising product yet. I kind of, I mean, I'm not totally against it.

Yay.

Areas where I like ads. I think ads on Instagram kind of cool.

Yes, Ads on Instagram, very cool. Let's go.

I am like, I think it'd be very hard to. It would take a lot of care to get right.

I, I have faith. I think you can do it.

People have a very high degree of trust in ChatGPT, which is interesting because like, AI hallucinates, it should be the tech that you don't advise.

My friends too, so I trust them.

People really do.

But I think part of that is if you compare us to social media or, you know, web search or something where you can kind of tell that you are being monetized in. The company is trying to, like, yeah, you can deliver good, no doubt.

But also, yeah, kind of like, this is the monetized block. This is the monetized block.

Whatever.

Like, you know, how much.

How much.

Do you believe that, like you're getting the thing that that company actually thinks is the best content for you versus something that's also trying to like, interact with the ads. I, I think there's like, there's a psychological thing there. So, for example, I think if we started modifying the output, like the stream that comes back from the LLM in exchange for who is paying us more, that would feel really bad.

This is a great solution, give you the actual answer you want. But, hey, these are the best headphones for you. But if you want me to buy them, I'm gonna have to go cook as an agent. I'm doing the work and I'm going to take a cut of that. That's. That's amazing. I'm so down for that.

Like the agent's. Like, I'm getting paid either way.

Yeah, exactly. And I could say, okay, which headphones do I want? Do I want the Sony's or do I want the. Or do I want the Apple AirPod Max's? And if I decide the Apple ones, it goes, checks out. It uses some coupon code to get some.

Yeah. I mean, comparing this to the other ways that people discover products and services.

Yeah.

If somebody searches best luxury hotel in Hawaii, they're going to get ads against that and then they're going to get organic rankings that aren't necessarily the truth. Right. Because the truth is for something like best luxury hotel in Hawaii is very subjective.

Yeah.

Then they might go and try to get recommendations from an influencer.

Yep.

In hopefully the influencer is disclosed.

Yes.

Whether or not they're being compensated by the advertiser.

Yeah. And so if I were to ask you as an influencer, like, what design software would you recommend? Like, what would you say? Just honestly. Figma.com Think bigger. Build faster, build faster. Figma helps design and development teams build great products together. This is a paid advertising.

Yeah. Disclosure is super important.

Exactly.

And if the influencer is saying, like, oh, yeah, I love this hotel, but that hotel is giving them like, you know, two weeks free a year, then like, that's not, that's not a super ethical.

Totally. Yeah. It needs to be disclosed. And then also the beauty of the LLM is that, is that like, like the, the recommendations are going to be able to be tailored. So best luxury hotel. Well, if you're, if you really want a certain type of pillow or you really want, you know, a pool in your unit, or you want it to be wheelchair accessible, or you want your high CE or you want, you know, beachfront access, like there's a million different parameters that could go, yeah, and the thing. And then it could just. And then it just saves you the time at the very last step.

Well, the thing that ChatGPT needs to navigate is maintaining that trust.

Yep.

Right. I trust that Instagram is going to serve me ads that, that I, I trust that they're going to try to serve me ads for things that I will want to buy. Right. Sometimes they serve me an ad. I'm like, this is, this looks garbage. I'm not going to buy it. Other times they serve me an ad and I'm like, this looks great. You're actually good. Good call. I am interested in this product. And the thing is, if ChatGPT has to maintain that trust because if they recommend you a hotel and they're like, you're going to love this. I know what you like, you're going to love this hotel and you go there and you spend all this money and you stay there and it's terrible. It's the same thing. If you go to a friend for a recommendation for a hotel, they recommend you a hotel. You show up there and it's like, this is terrible. Like, why did you recommend this? And if they go, oh yeah, I recommended it because I was getting like 7% referral fee, you're going to be like, what are you doing? Why are you monetizing me? Right. So I think, like, it's a very, it's an interesting challenge that they have where they're going to be directing, already directing so much economic activity. And how do you monetize that? In a sustainable ethical.

Exclusive from the Wall Street Journal, AI startup Perplexity made an unsolicited long shot offer to buy Google's chrome browser for 34.5 billion.

This inspired me. John. Everybody needs to be making long shot offers. Tvpn. Let's, let's submit an offer to buy the Golden Gate Bridge.

That's a good one. That's a good one.

You got to be thinking bigger, what.

Else should we buy? I don't know. It's tough.

Yeah, it's really. When I saw this, it's funny to think that Sundar has to go to the board or like, probably send them an email and say, like, I'm required to let you know that we've received this offer from Perplexity to buy one of the crown jewels of our product ecosystem.

I wonder how, how real this is. Like, it seems like we would like. The reporting is pretty loose. We can read through this Wall Street Journal article and then kind of like dig into it a little bit more. So Perplexity's offer is significantly more than its own valuation, which is estimated 18 billion. Has this ever been done before? Has any tech company been able to pull off an acquisition that's bigger than their own valuation? Like totally.

I mean, what was the slow ventures company that Metropolis and they buy a company that was much larger, you know.

Basically more kind of reverse merger almost.

Yeah, exactly.

Interesting.

These things happen.

So there's, there's also pressure. So estimates of Chrome's enterprise value vary widely, but recent ones have ranged between 20 billion and 50 billion. I have no idea how you got that. I feel like Chrome is pretty probably more valuable than that just on. I mean, I imagine that the default Search engine in Chrome has got to be as valuable as the default search engine in iOS Safari. Right. And how much is Apple paying Google every year?

20, 20 billion per year.

I mean, you'd pay like, you'd pay way more than $50 billion to get $20 billion of profit every year. It's crazy. So that feels low. I don't know where. I don't know where those estimates came from, but they are apparently under pressure. U.S. district Judge Amit Mehta is weighing whether to force Google to sell the browser as a means of weakening Google's stranglehold on web search. It's so, it's so interesting that, like, we're having this debate now, like, Google's had 99% of the web search market for two decades, and they finally have a crack in the, in the, in the. In the foundation. With ChatGPT actually eroding things AI LLMs kind of upending web search dominance. Like, yeah, we're having.

We're having James from Profound on later today, which. Which is built a big business very quickly because consumers are now using something besides Google search to. To do web searches.

Yeah. This is like the last. This is. This is like completely the wrong time to be focused on this. Like, the ship has completely sailed. But that's kind of the. That's kind of the pace that the government operates at. So.

Yeah, the Journal says Perplexity offer could be an attempt to signal to the judge that there's an interested buyer should he force a sale. I think there would be a lot of interested buyers if there was a forced sale.

Yeah. Who else would buy it? Apple Meta. Microsoft. Get bingo in again.

Bing. It's Bing. Bing it.

I mean, Microsoft would have a wild strategy because they'd have all the code to ChatGPT, all the OpenAI intellectual property. Like they could do. They could do what Perplexity is thinking about. Right?

Yeah. I mean, what does this say? What does this say about. I mean. So they just launched the Perplexity Comet browser.

Yeah. Daniel says Chrome's worth at least 90 to 100. 120. Billy. I agree. I think 20 is. 20 is too low.

But, yeah, so they. So Perplexity just launched Comet. We had Arvind on the show. He said this is a big, important bet for the company. If you're. If you just launched a browser and now you're saying that you want to buy to compete with Chrome and now you're trying to buy Chrome, what does that say about the state of your new product?

Distribution is king. It's very clear that even if you built a better product, the best product is not going to win. And we see this with the LLMs, the leaderboards and the vibes shift constantly. But the consumer adoption is just around chat. Everyone just uses chat. That's the one that's broken through. That's the one that people are familiar with all over the United States and all over the world. And so it's, it's very hard to break through even if, even if Perplexity can like rocket to the top of the App Store in India for a little bit. Like, it's hard to hold on to when, when the chatgpt snowball is just growing bigger and bigger every single day. Yeah, much like the graphite Snowball code review for the age of AI Graphite helps teams on GitHub ship higher quality software faster. Get started for free graphite.dev Is there a. Is there a polymarket on Perplexity, by the way? Can you look that up? I will.

Perplexity acquire Chrome in 2025.

What's the market on it?

When I posted about it earlier, it was at 12% and I said 12 seems quite high. And it's now at 3%.

It went down. Even though they're in the Journal saying they're going to do it.

They're like, well, it came out, the market opened at 34%, like 9am when the news dropped, and it's now down at 3%.

Wait, just today it went from 34 down to 3. Wow. So this has basically been like debunked. Yeah, Makes sense.

Yeah. I mean there's, I would put the odds of this deal happening closer to 0% than 1%. That sounds fair, but great way for Perplexity to get in the Journal.

Call me. No, this is great for SEO. It definitely puts Perplexity and web browser in the same keywords.

Perplexity was also bidding on TikTok earlier this year.

Arvin loves a long shot bet. A long shot bid.

I invited him on the show this morning. I don't think they want it.

He's not ready for press, not ready to chat.

But we love a size Lord launch offer.

Keep throwing these out. More star should do this. Google hasn't indicated a willingness to sell Chrome. In testimony this year, Pichai told the judge that forcing the company to sell it or share data with rivals would harm Google's business, deter it from investing in new technology, and potentially create security risks. Chrome has roughly three and a half billion users worldwide. It's like it's Like Perplexity has probably like millions of users, but it's just going to take decades, even at reasonable growth rate to hit 3 billion. It's just so many people and accounts for more than 60% of the global browser market. That's lower than I thought it would be found in 2022. San Francisco San Francisco based Perplexity recently released its own web browser called Comet. Some of its users. Tyler, you've demoed these?

Yeah.

What are you daily driving these days in terms of the web browser?

I'm using old reliable Chrome.

You're using Chrome? Yeah. I don't know.

None of the agent browsers really seemed very helpful to me.

Have you churned from Clulee to.

Yeah, well I started using that like you know, right after the show ended. I was, I mean I'm not doing interviews.

I was thinking about like the clue.

What if Tyler was like been using Clulee every day, hours a day. We're like, wait, you've been interviewing with other companies?

Well, that's the thing is that I, when I was thinking about Cluli, I was actually wondering like if we ran the show through Cluli and we had like this just this idea of like streaming audio essentially and streaming pixels and then just somewhat random LLM queries firing all the time, like is that a good pattern versus ChatGPT where you have to decide okay, this is the time to query it. And so when I think about like the going to you and saying like the pull the poly market up on Perplexity or tell me the market share for Chrome, if we could have something like Cluli running and it's just a big screen there that I can see and it's just pulling up answers to questions that it thinks we're about to talk about or ask or just adding context like that actually might be kind of useful to us. But I don't know if a tool like that exists. It also just might be be so sloppy and so noisy and so inference heavy that it just doesn't math out. I don't know, are there any value or like what's actually seeing breakout for you in terms of AI adoption?

I mean most of my time using AI tools is just like coding stuff. So it's like claude code cursor a little bit. Sometimes I just go raw chatgpt.com oh, that's raw. Yeah, yeah, just straight prompt.

What about raw text file? What happened to good old fashioned just handwritten code farm to table code?

Apparently not has to raise code.

But yeah, I don't know I mean, like, clearly I'm not like doing. I'm not doing like sales calls or like anything that really requires that kind of like interaction.

Yeah, yeah.

So not super useful for me.

But yeah, I wonder how the product will change because, like the marketing is not going to stop. Like they are definitely going to keep filling that top funnel until they figure out the bottom of the funnel and the churn. But I wonder what retention's like. I mean, I imagine that you bring in such a broad top of funnel, like there's probably people that are sticking around. There's probably people that just sign up and forget their credit cards down and they just pay for a while. The nature of just getting someone to convert for 20 bucks a month. They don't even have a free tier, right? Like you have to pay to get up and running.

No, they do have a free tier.

They do. Okay.

Yeah.

Interesting. There was something. Oh yeah, Tyler, did you see Vibe Sort? Someone wrote a sorting algorithm. It uses an LLM you have to put in your OpenAI API key and then it just passes the random list or the array of numbers to chatgpt and asks chatgpt to sort it and it just comes back with whatever it can do. And I was thinking that. Have you ever, did you ever have to do Fizzbuzz back in the day?

Have you ever heard of this, the code of folder? Yeah, I've heard of it.

Yeah, we should definitely do Vibe Fizzbuzz, where for each number between 1 and 100, it queries the GPT5 API. To ask it should it respond with the number fizz or buzz? The most ineffective, efficient and most inference heavy fizzbuzz answer possible. Anyway, back to perplexity. Told Sundar Pichai as part of the proposed acquisition, it would maintain and support Chromium, the open source project that supports Chrome and other browsers. It also said it would continue placing Google as the default search engine within Chrome, though users could change settings. That's crazy because you would think the first thing you would do if you own Chrome you would would be change the browser to mine. I'm getting those ad dollars, but I don't know, maybe it's a nice olive branch in this somewhat, somewhat hypey offer. The Justice Department filed the antitrust case against Google in 2020. And just a few five years later, in addition to they're here. In addition to forcing a sale of Chrome, the judge is considering limiting Google's ability to pay to be the default search engine on devices and browsers and requiring it to Share data with rivals and among other things, in weighing potential remedies. Earlier this year, Meta questioned how much new AI chatbots might be whittling away at the traditional search business, of which Google has 90% market share, so 60% share in browser, but 90% in search. So even for there's 30% of people out there that are using Firefox or Brave or Edge. It's Edge, right? Is the Microsoft one. Yes, but then they use Edge, but then they still set their default to Google because Google has dominant. Or maybe just Google users are querying more.

I'm so fascinated by perplexity and I can't wait to see what ultimately it's perplexing and I can't wait to see what ultimately becomes of the business. They're offering to buy TikTok, they're offering to buy Chrome, they're rumored. You know, there's rumor mills saying that Apple's kicking the tires on it. It's hard to tell where that's coming from. Right. Is that just more marketing complexity they have? I think the last reported number or the alleged number is something like 30 million monthly actives, which is a lot, but they have such a tiny share of overall queries that it becomes. And it's just clear that search feels like it'll be a winner take all market. And they're not going two years from now if they're sitting at 30 million monthly actives or even a multiple of that. It's hard to see it being worth significantly more than Snapchat, for example.

Yeah, and I think that's why you're seeing Arvind talk about potentially maybe pivoting to B2B context with the Bloomberg thing. It feels like if you're, if you're in a winner take all market and you're in the writings on the wall that you're not going to be the winner in that market. A B2B pivot feels like a decent move. You could even put kind of anthropic in that bucket where there was a time when it was like, oh, is everyone going to switch to Claude like the app on their phone? And then it didn't play out that way. Now they have a dominant B2B business. So you can imagine the perplexity niches down like the perplexity interface, the perplexity ability to, to, you know, like answer questions, act as a search engine that feels valuable in enterprises, it feels value in finance, it feels valuable in medicine if you, if you, you fine tune the service for that. But it's just a very different business. So they have to go through a little bit of a rebuild or kind of readjustment maybe. I don't know.

We'll see. Yeah. I mean, at the end of the day, when Elon bought X, it took someone like Elon to be able to put together the amount of equity and debt to buy that business. And he had to put a lot of his own capital on the line and in this situation. And that was buying a business that generated a lot of advertising revenue. Right. It wasn't. It wasn't known as. It was a bit of a dog in some ways. Right. Maybe. Maybe never was living up to its full potential. Where in this case, if you offer to buy Chrome and then in theory, you have to create an entire new economic model for the business because you don't have the Google Ads engine built into it, it's hard to see. It feels like basically impossible to pull that amount of cap to find people.

Yeah, call me when the Morgan Stanley associates are losing sleep.

Good point.

Unless Morgan Stanley is burning the midnight oil, this deal probably isn't happening.

I'm remembering that exchange between Ellison and Musk basically saying, how much? Ellison's like, how much should I do a billion? And Elon goes, I don't know. I think you should do like two.

Yeah, it's great. Yeah. So, I mean, Arvin's gotta find some elves, gotta get Larry, get a few of those guys on board for sure.

34 Larry's.

Well, we have some big news in the TVPN world. We are officially partnered with. With Waymo via Spotify because Spotify has a partnership with Waymo. So you can listen to tvpn. Waymo's now.

No way. Yeah, Big.

This just dropped. There's a new upgraded Spotify in Waymo launching today. Waymo put out the post two hours ago. You connect your accounts and you can vibe. You can listen to TVPN in your Waymo, which is probably the most tech thing you could possibly do.

A little. Bertie just told me that they spoke to a number of bankers that have not heard of the Perplexity deal, which I imagine as it. As it firms up. If it firms up, they'll certainly hear about it.

UFC landed a $7.7 billion deal with Paramount after a whirlwind 48 hours, according to TKO executives. But zoom out for me. Explain UFC. This is. This is the punching one or the punching one and the kicking one. How does.

This is one of your best bets? So Senra and I love ufc.

Yep.

And we'll text John and say, hey, John, do you want to watch? You want to watch? Come over and watch, like the pay per view tonight, John. Awesome.

I love watching my boys.

Who's playing tonight?

Yeah, I mean, I'm there if the teams are good. Yeah.

If there's a good matchup.

Yeah. But it depends on who's starting.

I want to know the starting. It'll depend on the starting line.

I want to know who's playing.

So ufc, the kicking and punching one. The kicking and punching one.

They can do both.

And wrestling.

They can do wrestling too.

And choking.

Okay.

Rear naked chokes.

Got it. Got it. Kind of a mixture of experts model in fighting sports.

Got it. Yes.

Multimodal.

Yes, exactly. So long relied on the pay per view model. Worked well before the Internet and illegal streaming. And the pay per view model has been in decline. UFC has also struggled to surface new American superstars. A lot of the biggest stars are international, but UFC's biggest market is in the U.S. international stars have a lot of fans in the U.S. but they're not driving pay per views.

Right.

People in Brazil are not paying $100 to watch a single card. Right. And so the pay per view model has been in decline. The UFC relationship with ESPN has been a little bit rocky.

Pay per view model in decline. Is that. Is that just because of the trends in, like, how. How yappy and how popular the fighters are, or is there something else going on at, like, a technological level where, like, I don't have to click any buttons to open Instagram Reels or TikTok or YouTube. And with pay per View, it's, like, technically difficult. It's kind of janky tech systems. Is that a real headwind or is that, like, fake news?

So there's definitely friction to buying pay per view. You got to, like, be on a traditional television.

It's not like in app purchase, like double tap and face id, like that level.

It's not that easy. Okay. I think you can go in the ESPN app and, like, buy the pay per view, but a lot of people aren't buying the pay per view to watch it on their phone.

Sure.

I think the bigger challenge and something that.

And Apple would take 30% if you paid on your phone. So even then, I don't know.

I think they would be real world.

It's digital.

Oh. Cause it's streaming.

It's for sure. It's for sure digital.

But the funny thing is Dana's been in this challenge because he gets very frustrated about illegal streams. And every now and then he'll be like, if you're watching a stream tonight, a legal stream tonight, we're coming for you. And everyone's like, okay, I don't know how you're planning to do that, but they try to take. I think that they try to DDoS the various streaming sites.

Oh, interesting.

Like, you can just send a ton.

Of text messages, and they probably have someone talking to Twitch and being like, we are going live right at this time. You need to have someone searching and taking down streams actively.

Yeah, but there's plenty of. Plenty of other, like, more decentralized, illegal.

Streams that are happening. All these different ones. Yeah.

Random URLs, even just somebody spinning up, like, their own thing. And so Dana was in a tough position because the more he talks about illegal streaming affecting the business, the more people are like, wait, I can stream it illegally.

Streisand effect. Yeah.

And so he had been looking for a new home for ufc. People had debated, you know, would Netflix be a potential buyer there. There's a lot of potential candidates.

Meta quest. You have to buy the headset. You got to watch it in VR, I'm sure.

I mean, Dana's obviously on the meta board, and I'm sure that that idea.

I mean, $7.7 billion. That's one researcher. Just. Just cut one. One headcount from MSL and you're good. And Zuck can just own all. All of UFC and be like, look, put on the headset or you don't get to watch.

Yep. The deal that was reported. We're in Yahoo Sports, how the UFC landed a $7.7 billion deal with Paramount after whirlwind 48 hours. According to TKO execs, even in 2026, all UFC events will stream on Paramount, plus with select events also airing on CBS. And so anyways, Ben, folks, is reporting. Just how big of a surprise was the UFC's bombshell $7.7 billion deal with Paramount and CBS? So big that even UFC CEO Dana White didn't see it coming. No, this is a quote. No, I didn't think this is where we'd end up. White said Monday during an appearance on the Pat McAfee Show. Shout out to Pat. Let's discuss the new broadcasting rights deal. But this sort of. This. This is sort of how it played out. I love it. These guys are obviously smart guys, very aggressive. He's talking about the team, the new team over at Paramount. So in an interview with cnbc, TKO President and CEO Mark Shapiro said he initially expected to make a deal that would Bring only the UFC Fight Night events to Paramount. But after Skydance and Fight Night events have always aired on espn, just fully ad supported. Yeah, so got it. But after Skydance Media completed its deal to purchase control of Paramount last week, Shapiro said the deal for the entirety of the UFC's U.S. president broadcast rights came together in just 48 hours. Now, instead of just the 30 UFC Fight Night events per year, Paramount will feature. And the Fight Night events were events that wouldn't draw big pay per view buys.

Those were not the numbered UFC events. Correct. So it's like UFC 776. That's like pay per view fight nights, just random on the.

And there's only 13 of those events per year.

13 numbered and 1330 fight nights. So almost every week there's something if you're UFC fan, but you're really once a month is when you really want to tune in.

But for the, they sort of depend on more casual fans historically to drive pay per views. People that are like, I watch UFC when Conor McGregor's fighting, right?

That's what really spike the pay per view and have kind of a power law outcome.

People getting together, being like, hey, we're going to buy the pay per view, we're going to, we're to, going, going to watch it, that kind of thing. So possibly the biggest news in all of this for fight fans is the end of UFC's pay per view era. Ever since the first UFC event in 1993, pay per view has been a vital part of the UFC strategy. Remember, UFC when it started airing on tv was thought to be. It was so shocking and violent to people that even politicians wanted to get it banned.

Wow.

Like, they were like, this should not be on television. So it had to be in that bucket.

Oh, it had to be paper. Because it was like, yeah, it was like adult restricted. Do the fighters need to sleep well before the fights?

Of course.

They have to be on eight sleep pod five.

That's right.

Five year warranty, 30 night risk, free trial, free returns, free shipping.

We got to make sure Bo, Nicole, the show is on an eight sleep. I'm going to message him after this.

Okay, perfect.

Let's. Let's figure this out. Ever since the first UFC event in 1993, pay per view has been a vital part of the UFC strategy. Under the current deal with ESPN, each UFC pay per view costs $79.99 in the US plus the cost of the ESPN plus subscription on top of that. So you got to subscribe and then you got to pay per event. With the pay per view revenues reportedly in decline, it makes sense for the UFC to finally ditch that model. The fact that it's doing so as part of a deal that will essentially double the roughly 550 million per year the UFC currently receives from ESPN likely only made that calculation easier.

This is crazy. This is crazy. When they find out, wait, if I, if I just sign up for Paramount plus for 12.99amonth, I'm going to automatically get UFC's numbered fights and the rest of the portfolio. That's a message we want to amplify. That is a crazy, crazy value change. Because if you were doing, if you were doing all 13, if you were the hardcore UFC pay per view buyer, you're up at $1,000 a year on UFC pay per view. Basically, like 700, $800. Like, that's a lot of money. And you're going from there down to 12. 99amonth. That's 150 bucks. Basically. That, that is a steep, steep savings.

Yeah.

The hardcore fans are gonna be excited.

Yeah. And a lot, A lot of, I mean, a lot of fans would, you know, somebody's hosting the pay per view one night, they split it or something.

Yeah, it makes sense. But still, I mean, that's a big, big savings. I do wonder what the ad load will be like, because when we watched UFC that a couple months ago with David Senra, I noticed that it felt extremely, extremely content, rich. And like, the flow of the show was fantastic because you'd just be watching a fight and then it would cut to the commentary. The post show, the in ring Joe Rogan's doing an interview, and then pretty quickly they start telling you who's coming up next and the actual fight to, like, you're only watching maybe like a half an hour of fighting, over an hour or two of content. But it didn't feel like, oh, I'm not crazy. Commercial break, and I'm totally taken out of the stadium. Like, even, even the. They did a really good job with the production to show, like, when they show you the stats, it's virtually laid over in the stadium. Like you're watching a camera pan around the stadium. So it actually kind of feels like you're there the whole time. It was really well produced. I wonder if they'll lose any of that by stuffing ads all over the place.

Yeah, we'll see. I mean, and you compare that to NFL. NFL's, like, roughly, I think 25% of the time allotted to NFL is advertising.

Yeah.

And it certainly hasn't been the case for pay per View.

Yeah. And it certainly feels like it's. Yeah, it's like it's a shift in the, in the type of content. Yeah, they just got to work on making seamless transitions to advertisements like we do here. With Fin AI, the number one AI agent for customer service, number one in performance benchmarks, number one in competitive Bake Off's number one ranking on G2. You can get started at Fin AI.

And anyway, so Paramount is or Skydance Young Allison is just getting aggressive across the board. Right. There's been rumors of this Free Press acquisition that I'm sure is still in the works, but it had been reported it was quarter billion dollars for the Free Press. Yeah, from our sense, they have about a million people subscribed to their newsletter, most of which vast majority of which are non paid. And so that just says that Paramount sees cbs Paramount, you know, the combined and Skydance sort of this combined entity just see the Free Press and as highly. And Barry Weiss specifically is just highly strategic.

And then CBS also cut the Late show, which is losing 40 million a year. So they're really like kind of shuffling the chips around like really restructuring the organization. Like what does CBS actually do? Well now it's potentially UFC and Barry Weiss as opposed to Stephen Colbert, I guess in terms of like headline stuff. I love this. I love this article. Back in back in 2009, Dana White swore that if UFC 100 did a million pay per view buys, he'd bungee jump off the Mandalay Bay. It did. He didn't. When UFC 151 was canceled after Jon Jones refused an appointment an opponent switch White called Jones's coach, Greg Jackson, a sport killer. He left a crater in the UFC schedule that only MMA fans could fully appreciate. A few years later, when Conor McGregor and Nate Diaz shattered the pay per view world record at UFC 196, it was a testament to how big the sport had become. We cared about those numbers as much as we did the outcome since UFC won when people paid. Out of morbid curiosity, this is what you were talking about earlier. Pay per views have been a vital a part of the identity of the sport. It's hard to get nostalgic over being gouged, but what follows here should shouldn't be mistaken as such, but Monday's news of UFC coming $7.7 billion partnership with Paramount came with a small pang of sadness upon realizing the pay per view model will soon belong to a Biden.

Yeah, I'm interested to see how this impacts the fighters because early in their UFC careers, they don't get any exposure to pay per views points, which is like a percentage of pay per view sales that top fighters would earn. So if you're a rising fighter and you got on a pay per view card, you weren't necessarily getting cut into that revenue, but if you were Conor McGregor or a true superstar, you were actually getting some percentage of the sales. Right. They call these pay per view points. And so now that that model is going away, I mean, I'm sure the number one thing thing on the fighter's mind with this new deal is how are we going to actually basically get cut into this new structure? Because historically a lot of fighters have critiqued the way that UFC fighters are comped.

But this is insane nostalgic article. This writer clearly really loves ufc. Back in the mid aughts, the UFC combined the Tuxedo affairs of D 1990s boxing with the vibes of an underground temptation. From there, it slowly stockpiled its greatest passions behind the paywall. Remember how Red Dana's face would turn as he tried to sell the pay per view at the end of the televised portion of the card? Remember the names he goes through? BJ Penn, Matt Hughes, Chuck Liddell, Tito Ortiz, Randy Couture, Conor McGregor, Ronda Rousey. Go through the posters of the past, and they were special. They were the special attractions. The names on the marquee for the numbered events. Those were some good parties we shelled out for as MMA fans. They were ours.

Well, end of an era.

Yeah. And probably a lot for us.

MKO holdings, the company that of course owns WWWE and the UFC, is a $36.8 billion company. As of today, it's up 11. Let's take it over to Mark Gurman himself.

Mark Gurman.

He has some new reporting. Apple plots expansion into AI robots, home security and smart displays. Apple Inc. Is plotting its artificial intelligence comeback. Don't call it a comeback. Actually, you can call it a comeback. Give you.

I mean, it makes sense. They've been. They've had. They've gotten a lot of hate over the last few years for the role, particularly the rollout of Apple Intelligence. Very hyped. A lot of features on display at WWDC that didn't ship when the new iPhone launched. They sold the new iPhone. Yeah, exactly. And so I think that there were. There were high expectations because of Apple's position with on device inference being very.

I gave you Genmoji. Are you not entertained?

They shipped a few things. I think that I would love to See the Genmoji actual usage data because I feel like even though it's not popular in T pop and on X and in the tech, the technorati, it might actually be seeing like decent adoption and joy. I mean, I've seen people share generative imagery from Microsoft Copilot, which I had no idea was a popular image generation platform.

Facebook crowd.

Exactly, exactly. So I wouldn't be so sure that Genmoji is producing zero value, but at the same time, it's clearly been a tumultuous time for the company. They've been kind of on defense in terms of public relations, but Mark Gurman is still scooping every single day.

Apple's plotting its AI come back with an ambitious slate of new devices, including robots, a lifelike version of Siri, a smart speaker with a display and home security cameras. A tabletop robot that serves as a virtual companion targeted for 2027 is the centerpiece of the AI strategy, according to people with knowledge of the matter. The smart speaker with a display, meanwhile, slated to arrive next year, part of a push into entry level smart home products. Tabletop robot. That's a virtual companion. Yep, that's big.

Lot of, lot of software they gotta write to make that work. They gotta get on graphite.dev code review for the age of AI. Graphite helps teams on GitHub ship higher quality software faster. Tim Cook, if you're listening, get on graphite.

Get on graphite. So this is big too. Home security is seen as another big growth opportunity. New cameras will anchor an Apple security system that can automate household functions. The approach should help make Apple's product ecosystem stickier with consumers. Said the people who asked not to be identified because the initiatives haven't been announced.

I mean, this seems like something that they could knock out of the park, like just the smart thermostat like these. These patterns have been defined for a decade and that's like the sweet spot for Apple to come in. And like the Sonos home speaker, where we were saying like it. They're just like the speak. Everyone knows what they want out of a connected speaker system. They just want it at the level of Apple execution. It doesn't require going 0 to 1 in some new territory. And so I feel like they could be very successful in the home video monitoring, home thermostat, home speaker. So I think I like this overall. The robot is where it gets funny because this feels crazy.

But yeah, going into companionship feels interesting.

I don't think it's companionship. I don't I don't think of it like that.

Tabletop robot that serves as a virtual companion is the centerpiece of the AI strategy.

I mean, so let's read what the actual report is on this robot. The tabletop robot resembles an iPad mounted on a movable limb that can swivel and repo position itself to follow users in a room. Like a human head, it can turn toward a person who is speaking or summoning it and even seek to draw the attention of someone not facing it. The hope is to bring AI into life. I don't think of this as like companion necessarily. This could just be Jarvis Assistant. Yeah, Jarvis feels like jarvis. I mean, I guess JARVIS isn't a companion, but companion has a bad flavor to it right now. A bad, a bad vibe because people are saying like I married my companion. And that feels very black mirror. I do think that there is a serious black mirror risk, but I'll go into that in a second. But what were you about to say?

Yeah, I'm not going out and saying this. This feels like a competitor to Friend.com or any other players that are actually focused on companionship. But this does feel like giving Siri a presence in the home. Like an always on presence in the home.

It will be interesting to see how they solve it. A partnership with OpenAI or Anthropic seems logical, but Apple Siri partnership with OpenAI Anthropic before September on Polymarket is a 6% chance. By December it's at a 44% chance though, so. But this is a very low volume market. But still it feels like we haven't seen a lot of messaging from Tim Cook that they're going to be building a. A huge data center, going to be training foundation model. It feels like they're going to need to partner on some of the underlying infrastructure for that because I would not want Siri hanging out in my kitchen talking to me like Siri is just not reliable enough. But I would welcome Claude or ChatGPT into my house, but I would treat it specifically as like an Alexa. What's the weather like? What's on my calendar? Look up the history of the Roman Empire. I would ask it questions like that and I would use it as a knowledge retrieval tool. I might start to trust it in some agentic contexts.

The question is, do you want, do we want a robot iPad with an arm swinging around your kitchen? And so. And hopefully when you want to know the weather, it's you're within earshot.

I think this could be really cool. I think it could be really cool. I mean, they say FaceTime calls and.

It'S just funny because they're still competing with their own.

Yeah, but if you're ever in the kitchen, like this is an example. So FaceTime calls will be a key function of the device. During video conferencing, the display will be able to shift to lock onto people around a room. And so as you're walking around, the FaceTime is just following you. That actually feels more social. I remember during COVID I bought Facebook's camera. They had some sort of device that you would mount on your TV and it would act as a video conferencing system. And I used it a few times. I ultimately churned. But it was, but it was, it had a really wide angle lens and it would like zoom in on the people. It had really good face detection and so it would center you no matter where you were in the room. But as soon as you walked to the other side of the room, it wouldn't be able to follow you because it didn't have any motors in it. And I feel like, can we pull.

Up the link in the chat? This all reminds me of the Meta portal.

Yeah, that's the one I'm talking about.

The portal basically didn't market it as a robot, but it would sit on a table and it would sort of follow you around the room. And Meta is no longer selling metaportal. They wound up it down. So anyways, Tim Cook told employees I.

Gifted some of these and they were pretty well received. Most of them wound up being used as like digital, digital photo frames though, and they weren't actually used that often for calls. I don't know if putting it on a robotic arm is enough to make it useful. I think you do have to nail the underlying speech recognition and in the conversational nature, like it has to be powered by a Frontier LLM.

You know what's funny? What is I feel like the product. I could imagine them having a mood board internally with the Pixar lamp hopping around.

That's literally listed here as like one of their inspirations.

Apparently so. Tim Cook told employees in an all hands meeting this month that Apple must win in AI and hinted at the upcoming devices. The quote, the product pipeline, which I can't talk about. Thankfully Mark Gurman has his malls in Apple that are talking. Tim Cook says it's amazing, guys, it's amazing. Some of it you'll see soon, some of it will come later, but there's a lot to see. Yeah, so Apple obviously always is cooking On Moonshots, very little of it surfaces to Mark Gurman and even less actually services, you know, publicly via Apple's own channels.

Mark Gurman. If you have a little extra time, you should leak Julius's user numbers because Julius has over 2 million users now. They're trusted by folks at Princeton, BCG Zapier. What analysis do you want to run? It's the AI data analyst that works for you. Connect your data, ask questions in plain English and get insights in seconds. No coding required.

Need to get our logo up there. Yeah, the fourth logo for sure.

There's a lot more. So Apple is planning to put Siri at the center of the device operating system and and give it a visual personality to make it feel lifelike. The approach, dubbed bubbles, is vaguely reminiscent of Clippy, which we are super bullish on. And we think Microsoft should totally bring back Clippy, an animated paperclip from the 90s that served as a virtual assistant in Microsoft Office. Apple has tested making Siri look like an animated version of the Finder logo, the iconic smiley face representing the Mac's file management system. The final decision on its appearance has been made with designers considering ideas that veer closer to Memoji the player, the playful characters that represent Apple user accounts. Device prototypes use a roughly 7 inch horizontal display approaching the size of an iPad mini. The motorized arm can extend the display away from the base roughly half a foot in any direction. So that's not really hoping.

I was having five across the room.

It reaches across and pins you against the. Against the fridge. Yeah. I think Apple should be able to nail delivering this in a way that doesn't feel black mirror and dystopian.

Oh, that's funny. So I actually didn't see this, but there's a quote in here. Some people familiar with the product call it the Pixar lamp.

Yep, exactly. I think it could work really, really well. Now they have to be cognizant of the black mirror.

That would be very cool if it could hop off.

That's not Nat Friedman core. This is doable with modern technology, but the risk is that they botch it because Apple's had a rough go with the recent marketing and they've had a couple real backlashes. They had that one where they smashed all the pianos in the press. Do you remember this? So they had a huge hydraulic press and they pressed an easel with paints and a bunch of paint buckets and pianos and violins and all this stuff.

They get a bunch of blowback.

They got a ton of blowback from this, we can, we can try and pull it up. Do you remember this?

People were saying that they're like destroying these timeless pieces. Yeah.

And it's just like the things that I.

Instruments and stuff like that.

Yeah. The thing that I love. You're literally, you're destroying it for this particular. Shots fired. Bill Bishop from the Substack Stream. You need better guests to talk about china and chips, not just Nvidia partners who are regurgitating Nvidia's talking points. Bill Bishop, get on the stream. Call in. We'd love to have you, Bill. Bill, you're welcome. Hop on.

Of course Aaron is riding with Jensen.

Yeah. He is the Jensen surrogate. He's. He's a surrogate for Jensen the juggler. The other ad that didn't go as viral and result in major backlash was the launch of the Apple Vision Pro. So when the Apple Vision Pro launched, one of the, the videos that they showed was of a middle aged man looking at VR video of his kids. But he was alone in a dark room and it. And Ben Thompson was like, it feels like he's divorced or something. Or maybe like the kids died. It felt like Minority Report, which is a movie I know you haven't seen, but in Minority Report, Tom Cruise's son goes missing from the pool one day and he has a, he has a virtual 3D video that he plays back and gets very emotional about. And, and the, the imagery was very similar in the Apple Vision Pro, but it wasn't an optimistic scenario. It was like, yeah, like I guess if my son passed away or got kidnapped, like I would want to relive that. But that's not like going to inspire me to buy an Apple Vision Pro. And so with, with this like Pixar lamp, this, this robot in your kitchen or in your house, I think they have to lean into light mode, not dark mode. They have to really be aware and reality check themselves on the black mirror effect. Because the idea of someone hanging out in their kitchen, being lonely and having this be the only companion they ever talk to, that's a lot less fun than, hey, we're all cooking together in the kitchen. And this is helping us keep track of the ingredients. Or my idea was teach it to. Teach it to pour some beers and get it at the fraternity. This thing needs to be tending bar at a frat house and inspiring dancing silhouettes like the original ipod commercial in order to be successful. Get the iRobot a stack of red Solo cups ASAP.

Apple's canceled for fraternity coded marketing.

I think I think that would be amazing if this thing could. If this thing could pour a beer without foam, I think they got a winner because it needs to inject itself into social. Even the ipod, like, yes, you could put it in the headphones and you could tune out the world and be that dancing silhouette during the ipod commercial. But also you could say, pass me the aux cord. I want to play a song for my ipod that everyone can enjoy. And it's a social experience. And so it's important to show both of those sides more now more than ever. And in. In ChatGPT demos, in any AI product demo, there's always going to be the user who. Who uses the product to be more isolating. But the marketing should always be aspirational and pro social. Just like social media. Social media, it's at its best when we're highlighting the fact that, like, I have so many friends that just send me a random Instagram reel every single day and like, that's our interaction and it's. And for me, it's a way to, like, check in with a friend and like, share a laugh. And share a laugh. And it's a heart and it's a laugh.

I can imagine the ad creative for the Apple Pixar lamp. You know, it's sitting on the center of a kitchen island, family's hanging out and they're FaceTiming somebody.

Exactly, yeah. One of the family members.

I can actually imagine us having a Pixar lamp here on the table.

Really? Really. It's like. Yeah, it's like you want to be connecting two families or two sides of the family. The east coast side of the family, the west coast side of the family. They're both having Thanksgiving and they're able to connect, but everyone's being extremely social. Do not film this thing with the lonely person. Having the Pixar lamp be the only thing to keep it company. Anyway, that's my advice for Adam. What else is there?

Projects coated Lynwood and Glenwood core to the new home devices and current products like iPhones and iPads is an overhaul to the underpinnings of Siri. Engineers are working on a version code named Linwood with an entirely new brain built around LLMs. The foundation of gen AI. The goal is to tap into personal data to fulfill queries and ability that was delayed due to hiccups with the current version.

It should be able to do it.

Hopefully they should have a. A moonshot division just trying to solve imessage search. They could probably start there.

Every tech company is that Amazon search is rough. Google Gmail search is rough. Their search has gotten so like dominated.

Billion dollar business in figuring out search.

I mean that would be the killer. That might be a killer app for Comet and some of the browsers is just like actually being able to search my Gmail engine box better. That would be pretty fantastic.

I feel like it should be at.

The app layer anyway. Whatever they're building, they're building products they got to get on Linear. Linear is a purpose built tool for planning and building products. Meet the system for modern software development, streamline issues, projects and product roadmaps.

OpenAI's on linear. Let's get Apple on Linear. Let's get the Siri team on Linear. They very well might be already so somebody Craig Federighi who you all know says the work we've done on this end to end revamp of Syria has given us the result we needed. He added that this has put us in a position to not just deliver what we announced but to deliver a much bigger upgrade than we envisioned. He said there is no project people are taking more seriously. So that is very positive. A final decision hasn't been made on which models will be used, but Apple has been testing and anthropic PBC's Claude for this purpose. Mike Rockwell, the former Vision Pro chief who was put in charge of Siri earlier this year, is overseeing both the Linwood and Glenwood efforts. Apparently so during the development of the tabletop robot, Apple engineers have made heavy use of ChatGPT and Google Gemini to build and test features within Apple AI and Siri teams. As the whole software developers are increasingly using third party systems as part of their development process. Getting into a ring competitor here Apple is working on A camera codenamed J450 is designed for home security, detecting people and automating tasks. The device will be battery powered and could last for several months to a year on a single charge. That's interesting. If they can do this for they can. They can do this for a random home security camera. They can't do it for so they have. You're telling me they have the capability to have an always on camera.

Put the tinfoil hat on.

We need a tinfoil hat. The device has facial recognition and infrared sensors to determine who is in a room. Apple believes users will replace will place cameras throughout their home to help with automation. That could mean turning lights off when someone leaves a room or automatically playing music. So Apple, you're telling me that you can make a home security camera of which I will purchase Multiple and put them around my house and it will be able to have a single battery that could last for several months to a year on a single charge. And you can't make an iPhone.

It's a little bit. The screen's pretty bright.

The screen's bright, yeah, but it's always on camera. How long do you think your phone battery would last if you just constantly had the camera?

I actually think running the camera sensor is a lot less power intensive than running a screen for sure.

And obviously a bunch of different applications but still pretty powerful. Pretty exciting stuff.

I don't think people want an always on iPhone. They want the thrill of charging the iPhone every night. If you're an adrenaline junkie, there's nothing like being on 1% trying to fire off that last imessage, trying to search it.

You know, I actually had a moment yesterday. My one year old is starting to walk and and I had this moment yesterday where I filmed a video of her walking for about 10 seconds. Everyone's in the house is like laughing, enjoying the moment and then my battery died and completely lost the video because I didn't hit like end of video. So the battery died right in the middle of this memory gone forever.

There is actually that has something to do with the encoder. There is a certain file type that you can use when you're filming on one of these cameras that if the battery dies it will actually save the last of the file. But there's certain codecs where if you don't hit finish recording the whole thing will be corrupted. It's very annoying. We have the debate of the century, the debate of the year. Showdown between former founder fund founders, fund colleagues, friends turned foes, friends turned rivals. Everett Randall, he's been on the show before. Delian Asperuhov, he's also been on the show.

They haven't been mincing words, John.

They haven't.

They have been throwing shots.

Yes.

Back and forth every TVPN appearance.

Yes.

They're calling the other one out.

Yes. And so they will.

And right now we're going to settle this strategies.

We're going to settle it today on the stream. The slop versus steel debate. Which is better, high margin software or capex intensive re industrialization efforts?

That's right.

We'll bring in Delian and Everett into the studio. Welcome to the stream. How you guys doing?

I like that background. Very good.

Energized.

Here we go. We're breaking it down live here.

Yeah, we're going to be breaking it.

Down live every time One of you gets a point. I'll put a little point.

If things get out of hand, we'll be banging the gong and bringing order like it's a gavel. But I'm sure, I'm sure this will be, I'm sure everyone will be civil.

Oh, I'm sure.

Keep the name calling to a minimum. Good to have you both. Thanks so much for being here. Let's, let's kick it off.

Let's start off. What's your least favorite thing about the other person?

Kick it off with the original story. Like, how did this all start?

Yeah.

Yeah.

Give us some backstory.

Yeah.

You want to, you want to give it.

Yep. I'm happy to. So we were back at Founders Fund. We were starting to beef up our, like, CRM and data science efforts. And so we were, we were integrating some external data into our CRM, figuring out how we could filter opportunities better to each of the, each of the investment team professionals. And we were looking, and we were looking at the different data. I was like, I'd be really nice if we could filter this by gross margin so that all of the negative growth margin companies that come into our CRM, we could give them all the Delian because it seems like those are the types of companies that he loves to invest in. The rivalry between the low gross margin side of the house and the high gross margin side of the house was born then.

Okay, Delian, justify why do you like the businesses? Is that even a fair characterization?

Fair characterization? I think, you know, my sort of one liner would be I'm not sure that you said gross margin is actually the right thing to sort of focus on in a business especially, you know, sort of early on, what you want to be thinking about is obviously EBITDA margin in particular terminal EBITDA margin. And so when I think about the like, at least founders feeling ethos to you investing, we think that that terminal EBITDA margin mostly is determined by ultimately how much of a monopoly your company can be in the long term. And so if you look at sort of mag 7 today, obviously there's a decent chunk of them that have some phenomenal sort of gross margins. And those tend to be the ones that are a little more software oriented. If you look at the one that is at the biggest scale and has the best EBITDA margins, it's the one that is the most basically hardware oriented for sure. Some of it propped up by like Cuda and they're like, you know, sort of software side of the house. But like Nvidia is the one that is performing the best of all those. And then even if you study with within those, which of those companies on the hardware side have monopolies versus not you see, it's the one that with the monopoly clearly outperform the ones that don't. Tesla obviously in that sort of mag 7. But a part of why they sort of suffer much worse margins than an Apple or an Nvidia is because they actually do have competition. And so my General characterization of SaaS is people always study their original gross margin but weren't burdening in the cost of sales, marketing, et cetera. And because you just have much less of monopoly typically in SaaS that ends up totally hurting your EBITDA margin profile. So take the like the favorite terminal scale thought of as a monopoly sort of SaaS company that I'm sure loves Salesforce, their market share and all things CRM is 25%. And so that's why you end up seeing like yeah, gross margin profile is only burdened by like you know, cloud sort of cost but their EBITDA margin profile that is like 40%. And so you know, the reason that I like these negative gross margin businesses is yes, they're like tougher to start. They may be more equity intensive at the beginning but end up with way better, you know, sort of terminal margin profiles versus you know, envelopes to you know, invest in the, you know, AI slop codes that might have early gross margins in revenues.

But so what's the bull case for for software? What's the bull case for SaaS? What's the bull case for AI slopkos?

Look, so to quote the Godfather Neil Mehta himself, the laws of great businesses are the laws of great businesses. The job of a business in a capital society is to maximize and find the efficiency frontier for three roic, AKA return on invested capital, the amount of capital you can actually deploy and how long you can deploy that amount of capital at and above market roic. There's a lot of different framings for the path to do this and how companies can actually do this. The one that people like in tech circles is Hamilton Hemler's seven powers. A company accumulates power in the form of scale economies, network effects whatever power you want to take and then uses that power to produce above market roix for as long as possible and with as much capital invested in the business as possible. There are great Adams based businesses that do this. There are terrible atoms based businesses, businesses that don't do this. There are great digital businesses that do this. There's great or there's terrible digital businesses that don't do this. I mean, you want to hear about great Adams based business that does this. Listen to the acquired pod on Costco. Like, it's certainly not like a atoms versus SAS thing necessarily. The advantage that digital businesses have is that in this process of producing above market ROIC for a long time is that their product form factor and the way that they distribute their product lends itself more to the process of creating power, I'd argue, than most atoms based businesses. If you think about network effects, the best place to create network effects is in a digital marketplace like an Uber and Airbnb or DoorDash. There's a lot of these forms of power that naturally lend themselves to digital products. And the scalability of digital products tends to be a lot greater than physical products. And so you can see, see these, these rapid growth trajectories like we're seeing from OpenAI and Anthropic and many others.

When did you guys find common ground? Was it in the E Scooter era, the sort of 15 minute delivery era? Were you ever able to kind of come together and say like, yeah, this is, you know, we can both agree that this is, this is not it.

We'Re good or good. I mean, to be fair, Kleiner Perkins Founders Fund have both invested in figma, Stripe, airbnb. There is some portfolio overlap.

Rippling, right?

Rippling as well. There's a Modern Health, I believe as well, there's a few others. But yeah, to Jordy's point, where else is the common ground and where, where else is the divide or the consensus in the disagreement?

Yeah, you say everyone were texting before this of like, you know, what are sort of two companies that I think both of us were enthusiastic about in 2021 that actually both, you know, sort of trended well. But are, you know, sort of counterpoints are two arguments. The ones that we kind of came up with were, you know, 2021, I was really sort of, you know, high conviction on Hadrian in 2021 was super high conviction on rippling. Both those investments have performed quite well over the last couple of years. But look, you know, sort of wildly different in terms of profile, you know, rippling, like many other sort of SaaS companies, does end up having, you know, an initial very high gross margin, but does still have to spend a lot on sales and marketing to bring in sort of net new customers. Hadrian on the flip side, deeply negative gross margin to start. But now as they've gotten to Scale, they actually have super limited sales and marketing spend because there's only 10, 15 customers that matter and the moment that you're delivering for them, they just proactively start throwing revenue at you. I think there are times where both of our stories obviously can play out. The thing that I'd be curious to hear from Everett is to actually compare contrast. You're bringing up some of these digital businesses, businesses that end up having these network effects. I would kind of argue that 2010s negative gross margin businesses like the Uber DoorDash types I think of as more as Adams businesses. But there was a whole set of investors in the mid 2010s that were generally unwilling to approach both Adams based businesses that started with negative gross margin. But even some of these local marketplaces that started with negative gross margin that swore off of the Uber ubers, the DoorDashes, etc. It's very clear that Uber DoorDash, through lots of investment, through building out these local sort of networks of both supply and demand, were able to, and drivers were able to eventually get to a point where now they actually have very attractive financial profiles. Today the equivalent of that is like there's all these investors that back in the 2010s would have refused to invest into any company that had negative gross margin and are all now pouring cash into to both the like AI application layer companies and like you know, foundation models that all have like ridiculously I mean I forget, I think it's girly is non stop, you know, not my favorite person in the world. But girly is nonstop talking about like, you know, what is going on here? They're selling a buck for 90 cents.

And so I guess so I think, I think it's an important example because you had that, you know, plenty of examples of these chained losses during that, that era where a restaurant was selling something below cost to a platform that was selling something below cost to a logistics provider, an individual contractor that like maybe wasn't actually making money if you factored in depreciation and fuel costs of their vehicle. And that ultimately worked out right. Doordash is a, is a massive fantastic business based on the power of the American consumer. But when you compare that to today, where a lot of the conversation on the timeline this week has been the margin profile of this new generation of software companies that has to pay a lot for sales and marketing, but also inference. And so I think like the debate should really be, you know, continue to be around just how quickly will the cost per token fall. And I think a lot of People have a lot of confidence around that. But I think that that is the key thing that Everett's sort of like broad investment these as right now is dependent on.

Yeah. Do you think there's going to be that same path of like Uber for a while had a bunch of negative gross margin people going into it? Like, do you actually think I want.

To pull this post up? Everett actually posted this in January 31st of 2024. So over 18 months ago he said, I'm making a real effort to not take for granted the $3 Uber across town era of AI. And I hope you are too. And so I, I guess the question is, because then, then a bunch of people, I thought it was a good point. I thought it was a hot take then. And I think then, you know, a bunch of people kind of parroted that take all over the timeline. Stole your whole flow, as you like to say. But, but, but, but I guess the question is like, are we in some sort of different regime right now where the traditional gravity and like fundamentals of software investing have changed because we are out of the zero marginal cost era? And does that impose risks to the strategy that, you know, you've sort of employed or like, we're kind of putting you in this whole, in this box. But if the, if the fundamental structure of zero marginal cost era is going away, that, that presumably forces like a rewrite of your logic around investing, I would imagine.

Yeah, I think that, I think that the biggest variable that that's changed from the 2000s SaaS era to today is that in the 2000s, and you basically made this point without making it delane, though, is that the thing that was missing from your talk track is that the competitive intensity of SaaS during the 2010s was much, much, much lower than it is today. During the 2010s, there was an entire crop of companies in the 2000s, but then especially in the 2010s, you could basically pick either a vertical segment, you know, like H Vac or car dealerships, or you could do a horizontal function like the CRM or, you know, some very niche workflow for like the finance team. You could build a software product around that workflow around that vertical, and you really only had to deal with typically like two to three competitors. Like there really wasn't that much competition relative to what there is today. And there was less just like general pricing pressure, competitive pressure, just the general pressure that you actually had a lot with some of the digital marketplaces early on. I think there was A whole crop of investors then, and the SaaS investors then were like, well, we don't need a bunch of cash burn. And it's a really unhealthy indicator if these SaaS companies are producing a bunch of burn because they're not competing with anybody. So if they can't sell their product for good unit economics on day one, when the competitive intensity isn't very high, then they're probably not a very good business. I think the thing that's changed now is one, you have the change from zero marginal cost to actual meaningful marginal costs in the form of inference. And it's also just a hell of a lot more competitive than it used to be. And so you are. And by the way, there's an immense, probably 10x more capital than there was 15 years ago to go into these companies. And so every single category now has.

Become.

Mini rideshare or mini uber market, where it's like, hey, there's probably a really big pot of gold at the end of the tunnel and we need to be the ones that get first to scale. And in a lot of these categories, the ones that have gotten first to scale have gotten a lot of brand equity out of it and have gotten a pretty resounding lead. I think the, the only other piece, I would say I lost my train of thought. So.

Yeah, yeah, but it's going to be basically, it's going to be like a capital fight now on the, on the SAS side, I wonder if, if the contrarian trade around hard tech is entering a similar era where it's become consensus and so we're going to see more capital fights. And when a founder goes out and says, yeah, I'm going to do something crazy, but I need to spend a billion dollars of capex, people are just like, yeah, this could be the next basic.

It made sense to have a capital war in rideshare, but now we have a capital war in this niche agentic workflow in some industry that, that most people have never heard of.

And then also a capital.

Here's $200 million funding for military boats.

And UAS and UAP. Like all these different subsegments are going to wind up if capital wars start popping up there. That could potentially be a headwind to Delian's model. Is that roughly correct? How would you fight back against that?

Look, I think it's always sort of important to talk about sort of specifics here, right. One of EVs sort of major investments in the last year is this company called Captions that basically does AI captioning of various sort of videos on social media. When I think about, you know, handing, you know, sort of two Stanford grads and $100 million to go try and, you know, sort of replicate that. Yeah, feels like, you know, they could go do something like that. There's like, you know, clear, you know, voice recognition models. They can go, you know, sort of pay on ads on TikTok etc. And you could probably go and replicate that. And so, you know, you know, our one liner at Founders Fund is competition is for losers. And so, you know, I think I was loser for investing.

Now the chat was saying spicy enough and you just delivered deli. So thank you.

Now, you know, if you take, you know, sort of two Stanford grads and $200 million and tell them, hey, I need you to go replicate this manufacturing facility and go start building a bunch of, you know, sort of satellites, reentry vehicles, you know, bioreactors that can actually survive the environment of space. Most, you know, sort of Stanford grads, you know, can't go, you know, ask chat GPS how to go do that and yet hasn't really faced significant additional competition irrespective of the fact that, you know, all things space factories are thought to be, you know, sort of the hot new thing.

To be clear, we use captions here on clips. We enjoy the Captions app. We thank EV for making it possible and subsidizing our and there is a.

Yc, there is a Varda esque YC company so coming for you Indian Varda.

I think will be a little bit less competitive than Indian Captions. Also, if you're the caption CEO and Founders Fund is trying to invest in your next round. Please don't let us do that. Helpful counterpoint for me.

Delian, you were correct that it was getting too friendly of a debate. I did want to make sure I could pin this one on you. If you can recite the equation for a return on invested capital, I will victory to you and I will donate $5,000 to a charity of your choice.

Hopefully he's got Cluly running exactly my.

Like, you know, equivalent for Everett will be if you can explain, you know, basically why you can't create microgravity down here on Earth. I will also donate $5,000 to a charity of your choice. But I don't think, you know, I may not have the, you know, basic understanding of business physics, but you don't have the basic understanding of physics and one's more important about understanding the universe around you.

Okay.

I mean I'm pretty fixated on the 20, 20, 35 Midas list. That's really kind of the final, the bigger. That's this.

Have you been on Midas brink yet or. I forget whether or not you've made it up there. Not even on the brink yet, you know, rejoins, you know, KP after you and she beats you.

Laughing.

Me. It's okay. It's okay. Eventually we're gonna, we're gonna bring back the extra names and Kleiner Perkins. It used to be Kleiner Perkins Caulfield Byers. It's gonna be Kleiner Perkins Randall Braswell eventually. Once we're working on it, we're working on it, we're pitching it. Where, where should we go next?

Jordy, I guess. Everett, how are you? How quickly, like how much should people be fixated on the cost per token with these Frontier models over the next six months? Like how, how long can. Can venture capital sort of like backstop these chained losses?

Yeah, I think that the way to delineate the whole. So obviously, like, I think there was this kind of consensus narrative that like every, you know, 12, 18 months token costs were going down in order of magnitude. And I think that did hold for a while. I think what you've seen now is like, actually for Frontier models, that started to peter out a bit and like pricing has actually started. It's still going down. It's not going down nearly as much as it, as it used to when, like when we were kind of in, in the, in the, like the meat of the curve of capability improvements on Frontier LLMs in terms of pricing curve. So I think the way that you want to delineate it is like there is a certain, like what I always tell everyone is that like there hasn't been a chatbot query since gpt4 that like my mom hasn't been able to ask and have it answered by the model. So there's like the mom test of models where like there's a growing subset of tasks, like economic or knowledge tasks that the models are tasked to do that no longer need frontier intelligence. And when you're not on the frontier, either through open source or just the, like the cheapening and distilling of, of older models, like the price still falls off a cliff. Sure there's going to be a very, very large set of tasks that models do that are not on the frontier and those are going to continue to get dirt cheap. Actually think that at the Frontier you're probably going to see continued price decreases on a per token basis. But Nowhere near what you saw before, which was like this, this order of magnitude decrease on a very regular cadence. And so I think, I think for, like, depending on the company, it's going to depend on one, if you've actually built a company that has enough power where you have pricing power, where you can price above the kind of marginal token price from the actual model providers, and then to, like, how much of your inference actually needs to be at the frontier? Like, how much of your inference can be an older model that's much, much cheaper versus how much do you need to do on the actual frontier? I think that's what you're seeing. Like, you know, everyone loves to talk about Cursor. And Chris Paket over at Pace Capital had this really great kind of like mini essay, I think only like last night or a couple nights ago, and he talked about, like, no one knows if Cursor has power yet because, you know, coders and developers, they're very, very like, they're tastemakers. They're very good at understanding the quality of the models and how much inference they're getting. And there's a lot of price sensitivity for them because they have a really good understanding of how much inference they're getting. And so no one really knows. I think no one can definitively say whether a lot of those types of companies have actual power with their users or if they're just drawn to an interface for frontier models or not. And so I think that's what everyone needs to be looking out for, is those two things, like, do you actually have power? Like, will people give you margin above the marginal cost of tokens? And then two, like, do you even need the frontier inference for the vast majority of your product? Or is that. Or is there a lot that you can offload to cheaper models?

Yeah, I mean, I guess your counter said there, Everett, is that, you know, a majority of what the foundation models are providing in terms of value to their end users is starting to be, you know, sort of obviated by the like, you know, historical generation. Even some of the ones that are sort of open source, so seem to imply that where value is accruing and where you'd expect like the highest revenue growth wouldn't necessarily be at the foundation layer, but you'd see it more at the application layer since those folks can swap models out. But in reality, that's literally just not what actually is happening. If you look at which companies are sort of fastest on revenue growth, user growth, etc. It is the foundation model companies, it seems like A part of it is that they also have the most pricing power where yes, Your mom uses GPT4 but she's not the one that's necessarily paying $101,000 $10,000 per month versus the true frontier capabilities on AI coding, the Pro users, the one that actually do care about. Maybe your mom is fine with 115 IQ model and that's fine for the rest of her life because she's just not asking it that difficult of questions versus the people that actually are willing to pay are the ones that actually do care about the 140, 160, 180 IQ. Again maybe at some point that gets commoditized as well. But my counter to you would be you've made this argument that seems to imply hey things will accrue to the AI application layer which if I understand your guys portfolio is largely where you guys invest invested. But in reality that's not what's played out. They like places that have captured the most revenue growth, the most market share have been the ones that are actually pushing the true frontier of the technology forward. And so so far, at least in the last 18 months your thesis is not playing out at all well.

To be clear, isn't it somewhat widely understood that Anthropic has negative gross margins as well? So it's not like they're doing like.

Point was that you want to invest in these companies that have the seven powers and like you know, in the know days of like uber, you know, DoorDash, etc. That did end up using translating it.

Seems like Nvidia has the most labs maybe then the application layer we'll see how much power develops in the application layer. But EV we'll let you respond.

Oh yeah, I was going to say that basically what Delian said was just wrong because even though it is, even though like if you, if you think about okay like let's take like whatever OpenAI and Anthropic's recently reported revenue run rate is the majority of all of that or at least the plurality of all of that is ChatGPT and ChatGPT, even though it is served by a foundation model company, is an application. It is a consumer subscription that has an immense amount of power. It has an immense amount of branding. It is the only it is the first billion plus user consumer application that's been developed by a new company in a really long time. I think that you could put whatever models you wanted through ChatGPT at this point and it would not knock it off of its I think that is power. You could run Cloud 3 Sonnet through ChatGPT and I guarantee people like the average user wouldn't actually know the difference. And that to me is power. And just because the foundation model companies are producing apps themselves doesn't mean that it's not the application layer that is accruing the value.

Okay then my question is, you've got OpenAI with the best possible consumer application layer. You've got anthropic that shifted over to positive gross margins and those margins are expanding. And yet Kleiner is not investing into either of those foundation model companies.

Why I cannot comment on our current investment activities.

Okay, switching gears.

Twitch. Come on, man.

I mean, do you like making money or do you like, you know.

Can you comment on Donald Boat? Have either of you bought anything for Donald Boat, the notorious e beggar on x.com the everything app.

Like my little brother played the Uno reverse card and tried to get down a boat to buy him something smart contrarian.

Let's talk about revenue quality because I think that you guys run into this in your respective domains every single day. Just like in AI, you can have low quality revenue. That might be the explosion of consumer prompt to apply activity might not be the highest quality revenue. Meanwhile, on the hard tech side, if somebody gets like a random siber or experimental gets like experimental budget from some branch of the military and it's like a fixed length contract, it's not necessarily the right strategy to slap like a 50x revenue multiple on it. So what's your view on both of those? And then I want to talk about if we should get into if accounting rules even matter at this point.

Yeah, yeah, for sure.

Yeah.

I mean in hardware land, we think about this all the time of like there's clear differences in quality of revenue. Everything from like defense program of record. You have to value that very differently than even like a $50 million SBIR. And so it has been interesting to see a bunch of investors coming into this field where I think there's a lot of preexisting 10 years of work rules around software of what healthy revenue looks like. Rule 40. There's all these things that even if you're somewhat unsophisticated, infinite blog posts, when you look at that in the world of hardware and defense investing or aerospace, there aren't infinite blog posts for people to study. And so I admit that I'm sometimes amazed when I watch people come in even for I should never trash my own portfolio. But sometimes even my own portfolio companies, I watch people invest into them and I'm like, wow, you just have a deep underappreciation for just how long this company has until gross margin flips to positive, how long it's going to be until they're actually ready to go scale revenue. Even if it on the back end it might be attractive, it may be years and years for them to get there. And so yeah, I see huge variation on that. And then mostly what I end up sort of seeing is people just come in and slap a 10 to. I even saw 100x rev rate multiple on this hardware company recently and I was like, holy shit, wow. People like not having IRR for a long time.

Yeah, I think so. Delian's hero and close mentor Bill Gurley had an essay a long time ago called the 10X Revenue Club. And I think it's like a good abstraction for kind of like tech revenue quality and what makes up revenue quality. And it's things like how durable is the revenue. If you sign a customer, are they going to stay for a year, 20 years? How much contribution profit is going to come off of that revenue stream over time? All the basics. And I think you can like take those same building blocks and apply it to AI. I think there's several things that are worse for AI than at least than relative to SaaS for now. So generally gross, like gross margins are lower, which means contribution profit coming off is lower. I actually think that like depending on the category, you could have customers that are more sticky or less sticky. Like I know the meme is that everything's experimental run rate and none of these customers are actually sticky. I think, I think we see something very, very different among our group of portfolio companies. I think the biggest lever that didn't exist in SaaS that exists in AI, that could be a huge call option boon for the revenue quality of AI is the actual contract sizes as people start to eat into potential labor budgets. I know this is still kind of like inning one and ending two, and it's also a little bit of a meme where everyone's like, oh, it's going to replace labor. And labor is 10 times sass. And it hasn't really happened yet. But I think if you look at some of these coding tools and you look at something like cloud code, that is the first place where you can really actually say like, no, this is replacing the labor that a developer would do and it is paid for on like a metered consumption basis. And the monetization numbers we're hearing around developers using cloud code are pretty crazy in terms of like wow, that's like you're paying like 110 of like a developer's full in cost to a company on an annualized basis for this product. And so I think that the like the thing to watch is like durability of revenue plus the amount of actual revenue that a customer can give you. And I think that you're going to end up the amount of gross profit that a customer can contribute over time. And I do think as some customers crack these agent products that look and monetize More like labor, AI revenue could actually exceed the quality of SaaS revenue just because you're getting so much more gross profit per customer or like customer relationship than you would on the SaaS side. Even though there are clearly things that are worse about AI revenue at this current point in time than there are about SaaS revenue.

Dillian, how do you think about the moral imperative of a venture capitalist to invest in positive sum versus zero sum markets? This idea that you're re industrializing America, you're saving the west versus moving chips around you personally versus moving chips around the poker table, taking from some legacy, you know, Web 1.0 company and putting it into an AI company. What's your thinking and argument? There is a market beating roic. All that you need.

Yeah, I think Peter always reminds us like our number one job is deliver returns for user LPs. And so I actually tend to not try to overly moralize when analyzing the things that I want to invest into. For sure, when it comes into like policy and I'm in D.C. and I like need to sort of report to the Security Council that Bill Gurley is that Chinese spy, like investments that he's making should probably be banned from the United States. Yeah, for sure there I have sort of moral imperatives and things that influence that may end up shifting roic. Right. So you know, but when it comes to you know, like which literal investments are we making, I think of it as just like yeah, you just have to you know, sort of make the sort of best possible investments irrespective of sort of moral imperatives. But in some ways I tend to think it turns out actually if you go to immoral then that ends up affecting roic. So the last thing that I at least close on for my question for Everett is one of the upsides of Founders Fund is we're very sort of, let's say non centralized distributed, not many sort of rules which you know, ever for some reason you sort of chose to leave. And so I know nowadays everything that he says publicly, you know, probably, you know, five comms people and five compliance people that need to, you know, sort of approve it. And so I have. My only request to you is, you know, so blink twice if somebody's, you know, got a gun behind the camera saying that you. That's all you got to tell us, brother. Let us know.

Hey.

Our wonderful marketing partner Ali is behind the camera with a green and red paddle, and she hasn't raised the red paddle yet, so that's.

Well, thank you both.

Last question. Are you worried about Uncle Sam potentially having sharp elbows now that we're hearing about Intel? Yeah. The federal government taking a stake in Intel. Any. Any concerns about him going down the stack into the. Into the early stage game, competing for those seed and series A allocations?

Look, if Trump Capital wants you, you know, sort of mark up some of the, you know, re industrialization companies, I'm all for it, baby. Cheap cost of capital.

You're all for it?

I'll say two things. I would say. One, I think that the EV of, like, the Enterprise Value of Founders Fund probably 3x the night that Trump got elected. So I don't think Dalian would complain about that. And then two, as a parting gift. Dalian, I think this conversation's been great and it's made me realize why you want to build factories in space because your math on earth doesn't make any sense.

Well, thank you both for joining this.

You're both good sports. We'll have to do this again.

I think it might be a draw. We'll have to have you both back soon. Thanks so much for hopping on.

Great stuff.

We'll see you guys later.

Cheers.