Siri AI, Fable 5 Launch, Rivian CEO Joins | RJ Scaringe, Chris Miller, Evan Beard, Nick Fleisher, Chris Matarese, Alex Heath, Rob Schroder Jr.

You're watching TVPN. Today is Tuesday, June 9th, 2026, and we are live from the TVPN UltraDome, the temple of technology, the fortress of finance, the capital of capital. Let me tell you about ramp.com. Time is money, save both. Easy to use corporate cards, bill pay, accounting, and a whole lot more all in one place. We have a special guest today. RJ from Rivian is coming on. It's the launch of the R2 today. It's also just WWDC. It's also Anthropic launched Mythos or Fable. Fable is the main consumer model, but Mythos, there's some details about that as well. More for cybersecurity.

And China is waking up.

China is waking up right on schedule. They do.

$295 billion infrastructure fund of sorts that they're working on.

That is very competitive.

Yes, we have Chris Miller joining today.

That feels like right in scope with the capital that's been deployed into the big labs. You know, maybe if you total up NVIDIA and Google and Microsoft, you get— You got to bigger number, but, you know, it's certainly in the game. It's not France, right? Yeah, it's, it's a bit— it's a big number and it's definitely competing. France. Uh, yeah, it's serious money, so we're going to dive into that. Uh, Tim Cook took his last bow at Apple's showcase event. This is history on the front page of the Wall Street Journal. Uh, yeah, he needs a 21-gun salute, a 21-iAirPod salute where you throw them into the wall, or a 21-point nuke. 21 point near. I don't think so. I think the stock's going to do great. So this is actually fascinating. So the focus of Tim Cook's last WWDC was Siri. And reviews are good. They launched a bunch of new features. I think people are going to be very satisfied with it and all the decisions that they made. Did you know Tim Cook's first WWDC was when they announced Siri the first time? 2011.. So we can play this clip from the first Apple Siri announcement. Listen to what they actually say, how they talk about Siri, and then look at where we are today just 16 years later, 15 years later. That you're going to be able to talk to technology and it'll do things for us. Haven't we seen this before over and over? But it never comes true. They have very limited capability. It's crazy. You just learn a syntax. Call a name, dial a number, play a song. It is such a letdown. What we really want to do is just talk to our device. Ask a simple question. What's the weather gonna be like today? And get a response. In fact, we don't want to be told how to talk to it. We want to talk to it any way we'd like. Someone else might ask, "Will it rain in Cupertino?" Or, "Is the weather gonna get worse today?" "Or do I need an umbrella today?" And your device, in this case your phone, will figure out what you mean and help you get what you want done. That's a feature of the iPhone 4S we call Siri. Siri is your intelligent assistant that helps you get things done just by asking. Pretty crazy. Called it. What happened? But it just took 15 years for the technology to actually catch up. And so I think they launched a functionality. When Siri launched, it was the best voice assistant. It was a great experience. You didn't have to think in syntax.

I don't think it's—

I mean, it took 15 years for their technology to catch up.

Yes, but it took 13 years or 12 years for the technology industry to catch up. And so they're only 10% behind if you look at it that way. Like, it's been a 15-year project from that announcement. It's been 12 years for the leading labs to get there. And so they're 3 years behind. Feels like an eternity in AI world, but it's really only 10%, 10% slower. And they wanted to do it their way and they got there. It is very interesting to me that they're calling it Siri AI because Siri, the whole pitch was AI and like Siri, the name Siri comes from SRI International, which was the Stanford Research Institute.. And it's from SRI International Artificial Intelligence Center. Like these were Stanford AI researchers who started this company in like 2008 or something like that and grew it and then eventually sold it to Apple as Siri Inc. Acquired October, initially released October 4th, 2011, 14 years ago. Hmm, maybe 15 years ago. They might need, or I guess 14, it's gonna be 15, it's gonna be the 15-year anniversary in a couple months.. But do you know what happened the day after October 4th, 2011, the Let's Talk iPhone WWDC where they announced Siri? Steve— Tim Cook was the first— this was his first WWDC as CEO. He obviously hands it off to his colleague to introduce Siri. Steve Jobs passed away the next day. Pretty crazy. And so you have this bookending of Siri on both sides of Tim Cook's career where sort of nothing happened in an interesting way, you know? But the stock did fantastically well. Tim Cook created an immense amount of value, but it was this like AI winter.

Lighter, cheaper, faster, stronger.

Look at the stock price. Look at the earnings.

I understand what you're saying. I'm not saying that in a negative way. I'm just—

So you have like Tim Cook operationalizing the company, making so much money from the App Store, building the business, but at the same time, he just happened to be the CEO during the greatest AI winter ever basically.

Yeah.

From 2011 to 2022, 2023 when we started getting LLMs, started getting chatbots. And it did take them a few years because they'd lived through a decade of like, Okay, so the Siri team, have you guys got it better? And they're like, not yet. The technology is not moving fast enough. Like we haven't invented anything new. So we are the best of a very mediocre category. And Siri sort of slowly fell behind. But even when people say, even if you go back to 2019, pre-ChatGPT, pre-AI, pre-LLMs, and you ask the question of like, Is Siri good? People would say, "Ah, nah." It's not really delivering in the promise of like, talk to your phone, it does whatever. Like, people have the idea of the C-3PO, the agent, the computer. It's not delivering there. But it's not like people were switching in droves to Android because Google Assistant was so much better. Or Alexa was so much better. Alexa was the same system. I'm sorry, I'm triggering a whole bunch of smart home devices.

You're triggering every—

Oh, this is interesting. So during the keynote for, uh, at WWDC, they, whenever the keyword SIRI was mentioned, they took out the 4 kHz, 6 kHz, a few different spaces so it wouldn't trigger the at-home devices. So they basically in the recording, uh, like nerfed it so it wouldn't fire off your, your smart device, which I thought was cool, cool, like technical, uh, solution to that problem. Anyway, um, So they were like sort of behind, but it didn't really matter because they wouldn't— they weren't getting their lunch eaten until the dawn of the LLM, the dawn of the chat app, ChatGPT, and other apps that came out and became like superpowered relative to Siri's capabilities. Today they're catching up. And it's a very interesting bookend on this, on the, the Tim Cook era at Apple where he was incredibly effective. But the technology was just not in a— it was not at an inflection point at any time during his career basically until the very, very end. And yeah, they got, you know, they were a little bit behind the ball, but ultimately they did catch up. So, there are a bunch of interesting other things in here. The facts first we should go through. So, Apple spent roughly 12 minutes yesterday detailing the expansion of its child safety and parental control features. This was interesting because it was a notably short keynote. It wasn't a particularly long keynote. I think they spent like 15 minutes talking about products and then 12 minutes on parental features. Sort of gives you an idea of where the energy, where the focus is, where they want to spend time.

Yeah, potentially what you were saying yesterday. I think people, more and more people are waking up to maybe phones or causing systemic issues in society.

Yes, yes. Brain rot.

And I don't wanna get ahead of that.

Utility stuff. And so that's what I wrote about today in the TPN newsletter. So quick on the facts, child accounts with built-in age protections like limiting adult websites and showing age-appropriate media, good. Ask to browse, which asks for the parent's approval before the visit, before the child can visit a new website. So any, so the child can say, hey, I gotta go to this website for my homework. Click a button and it sends a notification to the parent's phone and they approve and they say, "Yeah, that's fine." So you can really lock it down. Communication controls allow parents to manage who their child talks to and will ask for confirmation before their child can add a new contact. Apple also unveiled a new photo editing tool. This is separate, but spatial reframing. So if you don't like the angle of the photo you were taking, you took it over this way, you can, with AI, reframe it so that it's more straight on. This is a very cool feature. I think this is awesome for a few reasons. One is that it doesn't feel like full gen AI to me where it's like going to slop it up. It's more just like a nice feature that's in the actual camera app, photo roll, camera roll. I thought that was cool. And also this doesn't feel like, oh yeah, this is something that every— that there was already a startup doing and it was already baked into Instagram. And so Apple's just rolling it out and they're like behind the ball. This feels like the first time I've seen this. Seems really obvious. You could probably one-shot this in Images or Nanobanana or any of the image models. But it was cool and it wasn't like, oh yeah, this is a startup and they're just like rolling it in. This feels like Apple's DNA of like understanding the technologies and then doing something cool with it and unique was on display here. So I thought it was cool. What was interesting— Yeah, sorry.

I think some of the pushback is that a lot of this kind of functionality had been available in other apps, like post-production apps, and now bringing this, basically bringing it into effectively the realm of the camera, where the computer is now the camera. It does feel notable to me.

But so notable how? Like pushback? Because there was pushback, which was interesting. People were saying that like, this is too much AI. Like, I just view this as as a camera.

It's a camera. This is a way to capture reality. And now in the reality capture device, you can use AI to—

So Apple has long prided themselves on like what you see is what you get in the camera app. But that started to change with the artificial depth of field, the tone mapping that happens. There's a lot of things that you can do.

Yeah, there's a lot of like upscaling. And sometimes you see images, people are like, oh, that's AI. But it's actually just upscaled Apple image.

Yeah, yeah, especially if it's low light, noisy image, it'll clean all that up. I think it's fine. I— you could just not do it if you want to not have your photos reframed. But it was interesting seeing that, that there's a lot of folks that do land in that camp of Apple is the— is not going too hard into AI, they're not stuffing AI everywhere, and then here's a feature that feels very, uh, on the nose in terms of AI. We'll see. We'll see. And there's mixed feelings on the timeline about spatial reframing, with some calling it one of Apple's most exciting AI features, while others are expressing concern about the detection direction AI-generated imaging is headed. Because you can— I mean, Apple's been mediocre at like the remove trash in the background feature, like Gemini or I guess Android AI, like the AI camera on Android devices has been smoking them there. They're probably going to get better at that. But pretty soon, it's like you take the photo in super low light, it up-reses it, camera filter, color grade, spatial reframing, remove everyone else and you get a totally different photo. And I think that's fine. I don't have a problem with that. But like, I understand why people would be like—

I mean, it's a little bit strange because you're like—

Just take the normal photo. Just turn off all the features.

At least— yeah.

No, and I think a lot of people— It's an option. You can always pull it into Photoshop and just do whatever you want with it.

Yeah, it'll just be interesting.

Put a dinosaur in it.

We're going to enter an era where people's memories of their lived experience are different than—

The Joe Weisenthal view of the future. Joe Weisenthal got in some hot water because he said that, why do you need to store any photos in the cloud when you could just embed them all and then on demand—

We're not that—

Show me a photo of my kid. Riding a dinosaur when he was 5. And it just shows you a photo of that. And people were like, that is not the world I want to live in, Joe.

Yeah, we're just not that far off from, hey Siri, your phone's in your pocket.

Yeah.

And you just have your AirPods in. You say, hey Siri, make sure to generate some images of my time at Disneyland today.

Hmm. Okay.

Yeah. It's going to happen.

For certain people, that's their option. Let them do whatever they want, you know? You can go to Disneyland and you can say, "I'm going to make a— I'm going to print out a photo of a stock photo from Disneyland." You can do that. It's weird. I don't know why you'd want to do that. Take a photo of yourself or your family. But you could. You can go get a professional photo taken at Disneyland, pay for it. You can take one with your phone. You can bring a professional camera. You can have somebody draw it. You could have a sketch artist. Yeah, exactly. Like, these are all just options.

I don't know.

I don't have a problem with it. But people do have a problem with it in the context of their kids because they want control and they want to be the decider of those options. They don't want the kids making the decisions 'cause the kids are gonna say, "Max brain rot, please." And so that's what's going on with the parental security thing and parental privacy.

So—

I am interested to see what's gonna happen on Instagram.

Yeah.

And if there ends up being a divide between accounts that are constantly getting the tag with like the AI tag on them.

Oh yeah.

And people that just say like, nope, no AI ever. Yeah. And I think there'll be a pretty clear split.

Yeah, and when do you apply that tag? Little spatial reframing?

No, I think Instagram tag, I think they do the tagging.

But would they for a color grade? Would they cut?

Yeah, they'll have to figure out what the line is.

But to me— If you take a photo and you pass through a neural network to say, make it black and white, does that count?

Spatial reframing is slightly— because it's capturing a picture that never existed, whereas a filter is just sort of like enhancing color.

It's kind of just separating the layers, rearranging them. I don't know. It's clearly like a blurry line. Yeah, but it's the blurriest we've ever seen.

But we could use— no one thinks Midjourney is not AI, you know, make it less blurry.

Yeah. And that's literally what they do because you can see that the rest of the space is sort of blurry and then it puts together. Tyler, what do you think about this?

Yeah, I was just saying I was kind of surprised that they didn't release more like AI detection kind of features around this. Like I would imagine that, like I assume there will be some metadata if you do the reframing that says that it used the reframing.

Maybe everything you just take a screenshot and you're good.

AI true. But it's like, yeah, that's true.

Yeah, yeah, easy. But yeah, I would be interested. I mean, they could have leaned into that more. They could do like when someone sends you a photo in iMessage, we will add that tag. That tag's being added on X and Instagram like Apple could easily roll that out and that seems like a consumer-friendly thing. It seems like there's demand for the AI tag, even though I think it's kind of silly 'cause you can get around it. And what is AI? But interesting. Anyway, so this was my wildcard topic yesterday. Maybe they'll talk about something related to the fertility crisis or something. And this feels like 3 steps away from it, but I like the direction that they went for a few reasons. So we got a bunch of new AI features, especially the new Siri. Technology and UI patterns are mature. We talked about this. So it's easy to, for Apple to nail. Also, as I predicted, they're giving Siri an app and the models are strong enough that I expect random hallucinations to be basically completely acceptable from a daily use perspective. But my wildcard was that something in the presentation loosely would link to the topic of the fertility crisis that Derek Thompson has been covering recently. There was that article in the Financial Times, our phones causing brain rot or problems like, How do you address this? The correlation isn't perfect for sure, but there's enough evidence that reasonable people are starting to put phones in like the probably linked bucket of reasons why there's a modern fertility decline. It's just one of them. I think Derek Thompson says like, "These may be 30% of the reason," something like that. But what's interesting here is that Apple is not a fear-based marketing company. So their execs don't go on podcasts and make freewheeling proclamations about doom scenarios, they identified environmental concerns about fossil fuels and energy use early, like early, early, early. And you cannot find a clip of Tim Cook talking about like climate apocalypse, we're all gonna die because the sea levels are gonna rise. That's not, he wasn't saying that for years and then figured out how to do clean energy for Apple's data centers. That's not what they did. They just worked privately. And then started talking about the solutions in their marketing materials. They didn't put the cart before the horse. And it seems like they're doing the same thing here with parental controls, phone addiction. Instead of giving a long speech about how bad endless scrolling can be and how it might lead to a less flourishing life down the road, they only talk about solutions, parental controls. 12 minutes of content in a notably short keynote. I think many parents are very hesitant to give kids smartphones these days, so it's also a good business reason to say, "Hey, you can buy your kid an iPhone because There are so many controls that you are fully in control. You know what's going to happen. You're safe with us. And there's already a whole niche market in these dumb phone devices for kids. There's this watch that has GPS and the ability to call home but not much else. There's no screens, there's no social media. And so being proactive about building solutions and then only talking about the solutions is just extremely refreshing to me and I think it's a good move. Anyway. The timeline was in turmoil and we'll take you through it after I tell you about Cisco critical infrastructure for the AI era. Unlock seamless real-time experiences and new value with Cisco. So we got Gruber versus Gurman, two of our best friends from the show. We love both of these guys.

The Gurmanator versus the Gruberator.

The Gruberator. It's rough, but we'll take you through it. What happened? So John Gruber and Mark Gurman threw the timeline in turmoil over the rollout of Apple's new Siri integration strategy. Here's the play-by-play.

Citizen says, again, this all sounds complicated. Just generate your kids.

Generate your kids. That's the real doom scenario. But you won't hear Apple talking about that, thankfully. That's the domain of, you know, journalists and podcasters. Anyway, so back in March of 20, March 26th, 2026. So 4 months ago, 3 months ago, Mark Gurman reported in Bloomberg that Apple plans to open up Siri to rival AI assistants in iOS 27 update. And so that was the headline. He got some scoops and he says the company's working on developing new tools to allow AI chatbot apps installed via the App Store to integrate with the Siri AI assistant. The chatbots also work with upcoming Siri app and other features in the Apple Intelligence Platform. That means, for instance, if you have Gemini or Claude or ChatGPT installed, they'll be able to send those queries to those services within the Siri voice assistant. Assistant, just like they have been able to do with ChatGPT since Apple Intelligence launched in 2024. And so, but as of yesterday, according to John Gruber, Apple has not fully announced exactly that, at least according to John Gruber, who says, quote, maybe Apple ran out of time today and will announce this tomorrow. That was yesterday. Maybe they forgot to announce it. Maybe they scrapped the next generation Siri that existed 2 months ago and in the last month rebuilt an entirely new Next Generation Siri. I'll bet something like that is what happened. So he's taking shots, he's putting Mark Gurman in the truth zone, but Gurman fired back and said, here's some screenshots of integrations that basically match his original reporting. And so Gurman's saying, no, no, you can actually go, there's a model picker in the Siri app, you can configure Siri to work with other apps. But there's a deeper question here. I think they're sort of quibbling over like, minor leaks and rumors and like the interpretation of these things. And I actually think there's like enough of a gap here that, that it's not totally unreasonable for them, but for both of their interpretations to be roughly correct.

But someone in the X chat, I feel like I'm watching this in 2x speed.

0.5x buttons right there, buddy. 0.5x buttons right there. I actually don't know if it's on livestream.

This is great. This is, this is a feature, not a bug. Yeah, I'm sorry, Humble.

You got to warm up for DALL-E every once in a while. Okay.

Anyway, so John's on a tear. Let him cook.

I think the core question is about expectations versus reality. There was some hope from hardcore AI users that the Siri button— me— the Siri button would be able to fully remap to their AI model of choice. So you're deeply integrated in ChatGPT and it knows a bunch of things about you. You're confident with that model. You know what keywords to use when you have the integration set up, you press the Siri button and you wind up getting that model by default. Currently you have to say, hey Siri, go ask ChatGPT to do this for me. And then it gives you this little pop-up with this button. You have to click okay. Yeah, you're playing telephone. It's really— you're playing telephone. It's not good. And, and so, and then also there is a demand for the AI app that they have installed to get access to everything across iOS with roughly the same privileges as Siri. And is that announced? Not quite. And so this would be a big departure from the current flow where you specifically have to ask ChatGPT GPT. Now, there are screenshots of a model picker dropdown in the Siri app, but it's unclear if that resets on some regular cadence, like maybe every query, or if it's a permanent change. And I'm sure we'll learn more. Marques Brownlee had an interesting thing that they're calling— what was it? Maybe like the new operating system is called Golden Gate. And he was saying that like the nominative determinism of calling something Golden Gate is like it's a golden gate in the sense that they are maintaining their gated wall walls, their walled garden, and producing endless gold from mining what's inside the gates. It could just be that they like San Francisco. But—

Well, it's just one or the other.

As someone who loves nominative determinism, I thought it was a fun take. Anyway, let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. We talked about— reframing. And so the Brian MacDuff shares exactly what's going on with the cleanup feature, drastically improved in iOS 27. This feels like it might be nano banana under the hood or something fine-tuned on that under the hood. It was very rough in iOS 26. There's also this interesting detail that I got more information on the private cloud. So the private cloud has been extended into Google Cloud. So for AFM, which is Apple Foundation Models, AFM 3, this is the third iteration, AFM 3 Cloud Pro, which is their reasoning model that's fine-tuned of Gemini or some train on Gemini. They say they worked with Google and NVIDIA to extend private cloud compute, which they've put a bunch of resources into securing and branding, into— onto NVIDIA GPUs in Google Cloud while maintaining the same guarantees to protect our user privacy. And so Apple's been very privacy forward. They have a ton of trust. People put everything on their iPhones. They're very, very confident in that. And Apple's bringing that, but they're vending it into Google, which is like the advertiser, the dangerous one, the one that is listening to you. Not really. It's actually Facebook that's more of like the listening one, but they're sort of extending their brand in. But what's interesting here is that Apple has made a ton of environmental claims like, "Our data centers run on renewable energy," but Google just did a deal with Colossus that does not run on renewable energy. And so, if the Apple, like, Google Cloud or Apple Cloud Pro model is running on Google Cloud, Nvidia chips that are actually a colossus. You have an environmental concern there, but Apple's never claimed that they don't rent servers from non-renewable sources. They just say our data centers. And so it's probably something for them to work through over the long term. I would imagine if they're getting in this game long term, we would see some CapEx and some clean data centers, some environmentally neutral, zero emissions data centers, because that has always been a very key a key marketing point, a key strategy in Apple's rollout. So, insane that this is the main image Apple is using to show off the new AI Siri. Aga sent you a message about Calathea. She described it as a patterned tropical houseplant. Have you heard of Calathea?

Because it's not summarizing anything.

Yeah.

—like the AI—

Sometimes you don't need a summary. It should answer the question.

When you see this, you immediately understand that she sent you a message about Calathea.

That's very funny. 17,000 letters. Tropical houseplant. Yeah, the demo should be Aga asked you about this tropical houseplant. Here's a picture of it. Have you seen this? That would be helpful, like to go a step further and actually go search the web. Find the image, show me the image, and say, hey, your friend asked you if you ever seen this houseplant. Anyway, it's early days. They're going to continue to iterate here. I've been—

you could ask Siri. I've been— I don't remember if I've heard of Calathea. I've been—

have I? I've been a Siri hater for too long. I'm turning the tide. I'm a Siri supporter now. But back in 2024, I said— I just asked Siri, what is an LLM? And it gave me this and it said, LLM stands for Master of Laws, which is an advanced certification for lawyers. So in 2024, 2 years after ChatGPT launched, or a year and a half after ChatGPT launched. Tyler. Yeah.

Have you applied for the beta yet?

Oh yeah. You gotta run it on your phone.

We got you this phone. Right? Because you bricked, we asked you to apply on your phone.

Do you have the latest, latest phone? Or are you one revision behind? Because only, only the latest phones that have 12 gigs of RAM can run the latest local foundation models. And we got to test those. We got to see if they're funny. I made a prediction back in 2023. I said, with all these incredible advances in conversational AI chatbots, I'm willing to put down a firm prediction by the year 2043, Siri will be usable. And I think it came true. I was correct. I think Siri's available now.

You haven't been able to verify this.

Ahead of schedule. Oh yeah, it doesn't release yet. It's not live for a couple more months.

Okay, I'm downloading the beta. Let's go.

Okay. At one point, actually, I replaced Siri. You're your archive of posts. I know, I know. You got to know this lore. At one point, I replaced Siri with ChatGPT via an iOS shortcut.

January 8th, 2023. Yeah, early. The most complicated shortcut.

Yeah, so you go in., and it would take an API key, and you would trigger this with a shortcut. You'd say, "Hey Siri, trigger Siri Pro Mode," or something, or you'd ask Siri Pro Mode a message, and then it would route that to ChatGPT. Before they had the integration, you had to do this manually, and you'd have to set all this up, but it was pretty easy to copy and paste that. But important Apple history, Siri was an acquisition. M&A for AI is in Apple's DNA. This was a reminder from last year when Apple met with Mira Murati, the former CTO of OpenAI, to discuss a potential deal for her new AI startup Thinking Machines Lab. The talks never progressed to an advanced stage. What a different fork in the road moment. They went with Gemini, but they could have had Thinking Machines Lab inside Apple. All of those incredible engineers and that, that talented team, but they would have had to pay through the nose and Apple You know, Siri was, I think, a couple hundred million dollar acquisition. It was not in the billions. Yeah, I think it was $200 million. Yeah, and Thinking Machines would have been $10 million or something. It would have been a lot to actually get that deal across based on where they were. Gaussian splatting is coming to Apple Maps. This is very cool. Wow, it looks so much better. Look at this. Wow. Normal Apple Maps— no, that is what looks good right there. Once it gets the color grade, I like that. I don't know when you would use this. Maybe if you're looking around a town, I don't know.

I feel like this would be very cool if you were in a new city and you were trying to kind of like learn orientation, right? Yeah. Like let's say you check into a hotel in a city you've never been to before and you can navigate around and kind of get a feel.

Or like you could do this at Disneyland with your AI-generated kids and you could pick the particular region of Disneyland that you want to generate a fake image of your fake kids and then you could post that on your fake social network and you could just slop it up from start to finish. No, it's beautiful. I've seen some people use Google Maps and Google Earth for motion graphics, you know, in a YouTube video you want to show where the action happened. Let's zoom in from space and this looks great. Apple did do a really, really good job with the geometry of some of these cities. They just went a lot further than Google has on the actual mapping and texturing. Whatever magic went in there, it's good and it's getting better. So hats off to the Apple Maps team over at Apple. Anyway, we have our next guest joining in just a minute. Let me tell you about Figma agents. Meet the Canvas. Your AI agents can now create and modify your Figma files with design system context. There are some other stories that we're going to be going through throughout the day. This one's crazy. The North Korean economy becomes the world's most unlikely success. We've been talking up Pyongyang for a while. We've been saying a lot of people raise from PIF. That's usually the Public Investment Fund from Saudi Arabia. But if you just tell people I raised from PIF, and it turns out it's the Pyongyang Investment Fund, you might go to jail. 'cause they're sanctioned, but you might be able to sneak it across the finish line. But they are booming. Tyler, you read this story. You had some takeaways. What'd they do? How many? Oh, they built 10,000 homes.

Yeah, which I believe is more than Los Angeles. Los Angeles, yeah.

How are they beating us?

So they're kind of, I think they read "Abundance," right?

Yeah. Derek Thompson's book. Yeah, that makes sense. They got abundance-pilled and then they said, yeah, we gotta build housing. I mean, it's like, if they're doing it over there, it's like, what are we doing here? Like, the most sanctioned and locked-up economy, but they are making money. They're selling arms to Russia and China. They have lots of nuclear weapons. They're building more by the day. And so, they're on track. Leopold also made it to the front page of the Wall Street Journal. Congrats. There we go. He's doing well.

There we go. Got his new picture in there. Claude Fable 5 is out. We mentioned it earlier in the show. A bunch of feedback coming in. People on X have had access to the model for a while now. Dan Shipper over at M3.

What does the information have against it? They called it a neutered version of Mythos. That's like, it's sort of— I think they said— The dropping is not framing it as neutered. They're framing it as more safe. It's a safer version. Safer and more reliable for certain things and probably cheaper.

I don't know. Yeah, absolutely crushes the benchmarks. Dan Shipper is doing a live vibe check right now. He had access to the model. We will summarize all the posts from today, get them into Twitter.

The pricing is $10 per input millions of tokens, $50 of output per millions of tokens. So if you're token maxing, get ready to pay. But the capabilities—

But still, yeah, quite a bit more compute.

People are talking about running it for 9 hours straight, running it over the weekend for just long, long time without getting confused. Very exciting model. Noam Brown had an interesting piece on the implications of large-scale test-time compute that I thought tied to this a little bit. And there was a very interesting quote in here where he said, so frequently when I discuss this, people, ask why we don't just evaluate a harness that pushes test time compute until performance plateaus. The problem is that empirically, the plateau is very far out. Sometimes we may not observe a plateau at all within practical budgets. You can just spend, spend, spend. So Karpathy, Andrej Karpathy, saw this in the AutoResearch experiment where the performance continues to improve even after hundreds of experiments. And he goes on to talk about benchmarks a little bit, but at one point, At one point he says that a model might run for longer than it takes to train the next run, which I thought was such an interesting, such an interesting concept. Just this idea that you launch a product, you're running it, you say go do a job, and then before it completes that job, you got the next model ready and released, which is a crazy, crazy, world to be living in and something I don't think people predicted. But more and more compute, more and more inference, more and more reasoning all across the board. Meta launched a workforce academy to train workers to build data centers with 5-week program, which is free of charge, guarantees a job, follows the recent layoffs of 8,000 employees. This is the learn to weld meme in real life. Forget learning to code. Meta Platform says it's time to pick up a wrench. Company is starting a workforce academy to train Americans to build its data centers as skilled trade workers become sought-after commodity. 5-week training program in partnership with CBRE and the Associated Builders and Contractors.

And Tyler has already been accepted into the program. Huge news. With a disguise and a fake name, he will be learning to weld. Huge news. Yeah, that would be fun.

We should consider it. Um, RJ is running just a few minutes behind. He'll be joining in just a minute.

So we got to talk about, um, what else? We got to talk about Flock safety. Okay, can we pull up this video? This is a crazy endorsement for Flock. The CEO's been— Garrett's great. Admitted criminal discussing the impact of Flock. Okay, let's play it. On just crime in San Francisco, period.

That shit over with, brother.

Oh my mama, nigga, they got drones. They're called TPP Unsafe.

We need to bleep this before we play it.

Okay. We're going to summarize it for you because I didn't realize that there were so many swear words. But basically, they say you can't do crime in San Francisco anymore.

And these two gentlemen are admitting to have previously participated in illegal activity. Exactly.

And so they are actually—

they're sort of thought leaders.

If you steal a car, a drone will start following you immediately. Yes. And it will just trail you from thousands of feet up. Yes. You won't even necessarily know that it's following you. Yeah. And the gentleman hosting the podcast—

So you won't have a chance to run away from a cop that finds you, because they'll just wait until you're actually parked or stopped, and then they'll box you in.

And then the host of the podcast says, so can you even steal a car, run up on your ops, and ditch the car after? And they say no. And so—

He doesn't just say that. He describes it as the classic steal a car, run up on your ops, and then ditch the car. Like it's a thing that's generally accepted. Not to lay out a bear case for Flock, but has anyone considered listening to podcasts where criminals admit to crimes and going and arresting them? It seems like there's a— Talk about their methods? It seems like there's a dearth of these podcast clips, but who knows?

Yeah, I mean, these aren't even just like random petty criminals. They're listed as— the gentleman on the left is one of the most prolific criminals in all of San Francisco. Wow. So presumably, like, yeah, someone in the police force can just be like, oh, that's an interesting tweet. Just Google it. This guy's name is—

He claims that he can't even do drive-bys anymore.

Wow. That is wild. That is a wild time. Anyway, let me tell you about Codex. Codex is a powerful workforce for getting work done with AI agents, whether you're writing code, analyzing data, creating content, or automating business workflows. Codex helps you move projects forward from start to finish. Moving on to the next story while we wait for our next guest. There are a few more.

Pat McAfee is on a tear. We can wait for that one because Someone over on X is telling Google's AI that I'm eating and drinking increasing amounts of salami and Monster Energy until it starts shouting at me to get help. The person says, I just ate another kilo. This is a good response.

This is a great response. This is actually a great breakthrough in lack of sycophancy and guardrails. Like, this is exactly what it should be doing. And to give you the phone number there, I mean, we're only one step away from it just calls on your behalf. Yeah. Right? Like, it should be able to make a tool call and know where you are, and the emergency rescue just shows up on your doorstep. Certainly possible. But this is good. Yeah, good response. The bad response would be like, wow, you really are a GOAT. You are the legend who ate the most salami. You're not just— Consuming, you're not just consuming salami, you're setting records. You should be the next world record eater of salami or whatever. This is good. Anyway, we have our next guest, RJ from Rivian with us in the TD Pin UltraDome. RJ, how are you doing? Welcome back. Welcome back.

Good to see you guys. Thanks for having me on.

Thank you so much for hopping on.

Congratulations on the launch. Look at these pictures. This is a great setup.

It's in your blood. Are these all you? Were you driving go-karts as a kid? It's actually a mix.

There's one of me and then they're mostly of my kids.

Oh, that's amazing. Nice.

Yeah, but it's cool 'cause it's like, and some of them it's the same go-kart that I was driving and then there's a few other go-karts.

Okay, but I'm seeing something very controversial behind you. One of those pictures looks like it's gas-powered. What's going on? Yeah, well, you know, I was— That was the spark. You had to fix it.

'83, so there weren't electric go-karts back then.

Oh, okay. Have you been tempted to make a Rivian go-kart?

I might be sad to do it. We have, we actually have an electric go-kart that we bought now. Okay, cool.

But too busy. I know, but think of the Rivian.

I assembled it.

It can't possibly compare to the Rivian equivalent.

I agree. But far too busy to work on go-karts because you're launching the R2 today. Take us through it. How is it going? What does this launch signify for the business? And what was like the key the key idea value prop? How long has this been in the tank?

Well, I mean, the idea of R2 has been sort of part of the business since the very, very early days. And essentially the product strategy that we have is launched with the flagship set of products. That's our R1S and our R1T. Yeah, we launched those in late 2021, early 2022, and that was really— think of it like the handshake with the world that established the brand. But these are, these are expensive products. The average selling price of an R1 is over $90,000. Yeah. So R2 is really our mass market product and what like the whole business has been designed around this because it's what brings us scale. And so, you know, that's not just getting the vehicle ready that you just showed images of, but also getting the rest of the business. So service infrastructure, distribution infrastructure, obviously manufacturing footprint suppliers. But today we start deliveries to external customers. We've been delivering to internal customers, employees for a while. But these are the first external deliveries starting today.

The ultimate employee perk. I love it. Why is 330 miles of range the right number? Like psychologically, you could see sort of like the inverse of the 999 and go 301 or 305. But 330, is there some rationale for that? Is it bounded by the batteries or is it bounded by the lifestyle of your customers? How'd you land there? It's a good question.

I have not been asked that question yet. Why 330 versus 301? Yeah, well, it's actually a— it's an output rather than an input. The input was we needed to be greater than 300. Yeah. On all configurations. And so we have an off-road tire that takes the range down to 307. Sure. And then the on-road tire gets us to 330. And at the beginning of the program, you're sizing, you're making a bunch of decisions around like battery size. Target for vehicle weight, target for vehicle efficiency. And in the end, we ended up exceeding a number of targets such that the least efficient version is 307 miles and the most efficient version, which is actually the base version with rear-wheel drive, is 345. And then the all-wheel version, which we're launching with, is 330. Do the door handles actually matter?

I've become completely normalized to the flat door handles. I prefer them now. They were controversial early on, and the excuse that was, you know, given by anyone who was doing flat door handles was, well, it'll give you more range. But is it like 15 feet of range? It just doesn't feel like it can be 20 miles of range, right?

A little more than 15 feet. It is.

Like, I like range. It's good. Save the energy. But at the same time, it just feels like of all the things that you could change, that can't be that big, but is it? Is it actually material or is it just like the flat door handles?

It's not gonna make or break the vehicle. It also is a design decision. It allows, you know, the surface language of the vehicle to be carried through the handle. So it's, as you said, it's a lot of folks have adopted this type of a handle, even in non-EVs you're now seeing them.

Yeah. I was so fascinated, like one of the most impressive things about Rivian is the fact that the full-size SUV and the truck launched basically in the same year, like very, very quickly. I don't know the exact shipping dates, but like that is— shipping one car is crazy, but shipping two vehicles the same time is— obviously, there are redundancies and similarities. But walk me through the decision not to do a small truck as fast. Is that market demand? I'm not really sure where— the market for small trucks is, but take me through the thinking.

Yeah, I mean, to your point, we actually launched the R1T, the R1S, and we have a commercial van as well that we produce for Amazon. Amazon's a big customer. And those all launched within 3 months of each other. That's crazy. And, you know, the classic question of like, what would you do differently if you were to build the company again? And, you know, and often you try to give a non-answer to those kinds of questions, but the honest real answer is don't launch 3 cars at the same time. That is like the singular biggest thing I would change. It was enormously complex to ramp up the supply chain and get the, you know, especially this is during COVID for one vehicle, but to do that times 3, like we were like choking on it. And so learning from that with R2, we have a launch edition that we're launching with. We have a narrower set of colors we're launching with. There's actually the whole R2 program only has has less than 200 build combinations. Wow. And just to put that into perspective, on R1 when we launched, this is, I say, this is a learning. We had hundreds of thousands of different possible build combinations. And so in theory, you don't build everything, but like if you click every box in different ways, like there's enough combinations that we had to support that level of complexity. And so R2 is this extreme exercise in, being very aggressive on driving cost out of the vehicle in every possible way and learning from some of those challenges on R1. So to your question, there's lots of cool variants we can imagine of R2. Yeah. And even that platform, including things we've shown, like we have an R3. Yeah. But as tempting as it was, and I guarantee you we had lots of debates and I've been tempted. Yeah. We decided not to pull R3 into launch at the same time as R2, but to let R2 launch Ramp, have some breathing room, and then bring R3 in.

But I can still do that aftermarket. If I wanna wrap it in corporate SaaS logos, you're not gonna sue me, I can do whatever I want aftermarket? Yeah, if you wanna wrap it in— Make it look like a NASCAR for Silicon Valley? Maybe that's actually—

I don't know if I'd love it, but you can do it.

But you'll be able to, that's great.

Last time, yeah, last time you were on, you talked about like just how different a Rivian is, the way that you guys have built it from the ground up to be a singular, singular, basically computer, singular software system, all this stuff, and the advantages that you get from that versus maybe taking a legacy car design and then trying to electrify it. With the recent, with a number of high-profile EV launches, in the last, let's say this year. I would love for you to kind of like revisit that. And just because I thought it was a fascinating explanation and particularly interesting now as we've seen these other manufacturers just try to enter the category and come in with products that I don't believe are incredibly compelling.

Yeah, well, there's a lot embedded in that. I think at the core, there's a number of things that more by coincidence than by necessity, when for a new EV, let's say a Rivian, and I'd say Tesla falls in this category, when you architect a vehicle from a clean sheet, you end up approaching a number of things differently. Obviously there's the battery and electric drivetrain that's different than an engine. But I'd say the biggest other difference that is probably not appreciated or talked about as much, but it manifests certainly in terms of customer features and things like over-the-air updates, is around software and electronics. And the software and electronics that exist in really every vehicle from a legacy existing manufacturer, so it's every manufacturer outside of a Tesla or Rivian, what they look like is, is really a reflection of a long multi-decade evolutionary track, which started ironically with the first computer that made its way into a car, which was to power or to drive the fuel injection system, which is somewhat ironic. But prior to early 1960s, cars were completely analog. There was no microcontrollers, there was no computers in the car. The first computer came in and it was specific to the domain of fuel injection. So that was designed to support just putting the electronics in for the fuel injector. Subsequent to that, starting in the 1960s and all the way through today, as different functions started to have software capabilities embedded in them, let's say seats that have memory or power windows that have intelligence to not like grip your fingers or HVAC systems that had settings and things that were not just mechanically controlled. Each of those functional domains had a little ECU that was created for it by the supplier. And over time, this has grown. I describe it sort of like a field of weeds. So it was not a planned garden, it was not architected. You ended up with what you have today in most cars, which is sometimes upwards of 100 to 150 little computers, what we call little electronic control units that correspond to a very specific function or domain. And those little computers have their own little island of software written by a separate team. Often it's not even written by the supplier, it's written by a third party, often overseas, often in India. And so you have this like mess, like absolutely a mess of software where you have multiple different, you know, little islands of code, none of which are coordinated, all of which now to deliver complex features that have things working together require a lot of coordination between suppliers. And it's for precisely that reason that you see very few significant over-the-air updates on traditional cars. Because to like, let's say, update the sequence of events that occur when you walk up to the car, where the horn makes an audible noise, the door unlocks, the HVAC comes on, the seat adjusts, the steering wheel adjusts, Let's say there's a— the music adjusts inside the car, the battery system comes to life. All those are different controllers. So just to change the order of events, it's very hard. You've got to coordinate like between 10 to 15 companies in a classical sense. And so what we built is something where you consolidate all that compute onto a very small number of computers. We now call those zonal controllers because they work in a physical zone of the car. And it runs one OS, which we built. And that single OS allows us to very easily change things. And that, that architecture is what underpins not only Rivian's, but we did a large $5.8 billion software licensing deal with Volkswagen Group to take that technology and help them on their journey with their future electric products.

That's very cool. Thank you for re-explaining. So the R2, boasts more AI infotainment compute power than any vehicle currently available in North America. And I'm interested with the backdrop of WWDC, what does the future of AI infotainment in a vehicle look like? Because you're starting to bump up against the ability to run on-device models, I'm sure. And there's this trade-off. We saw it with Apple where there's some stuff you can do on any phones, basically. There's other things you can do with 12 gigs of RAM just on the latest phones, other things that's going to go to their cloud. And I'm sure you have a series of abilities at varying levels. But when I get to the examples, I don't necessarily know that— like, people were sort of laughing at you get a text message that asks you a question, and it summarizes it, and it just asks you the same question. You probably don't want to be the CEO that's going around to the employees saying like, "Stuff AI everywhere. Like people are going to want to, you know, have AI decide how far the window goes up or the windshield wipers go on." But there probably are interesting ideas. So how do you see that evolving? What's the role of the car as a compute source versus cloud? How do you think about that?

Yeah, well, I'd separate the features that we deliver from the platform. Okay. And the way we've architected R2 is thinking of it really as a platform. So we want to embed in it enough headroom for features to grow dramatically and for things to evolve in ways that even today we can't imagine. Yeah. And so on everything that's infotainment related, putting a really powerful tuner tops, you know, platform into the vehicle allows us to run much larger models in the vehicle than like, let's say what we can run in R1. So R1, we just launched a Rivian Assistant and we can run a small model on board, but for anything that's more complex, we have to go up to the cloud because we just don't have enough compute in the vehicle to be able to run these larger models. R2 is quite different. So R2 has a much higher headroom, so we can run a lot more, a much larger model in the car. And it is helpful because it reduces latency. So suddenly you can have much more, you know, natural back and forth conversations without having to wait for things to send up to the cloud and back. And so we built that in knowing that we're going to add more over time. What we have today is a better version of our Rivian Assistant that'll be going into the car. We have— that'll be actually continually getting updated as well and continually getting better. And you'll start to see, as you just described it in the case of iPhones, you're going to see differences in like our Gen 2 R1s and our Gen 1 R1s aren't gonna have the same hardware capability that we have with R2.

How are you thinking about self-driving and what the next few years looks like there? I've been surprised that we've seen this like insane capital war in the foundation model companies and hyperscalers spending hundreds of billions of dollars. And we saw that a little bit in, automotive years ago with the Cruise acquisition by GM. And there's been initiatives and the technology's on a good track, but it feels— no one has self-driving psychosis. Like, no one is saying that like, oh, every automaker is going to go bankrupt because they're investing too much in this. Everyone's like, yeah, the self-driving is getting better and we're on a smooth path. And in a couple of years, the cars are— you're going to be really not touching the steering wheel at all. But is that conscious? Is that based on the research trajectory? Do you expect that to accelerate? Like, how are you thinking about self-driving these days?

It's a great question. And it's important just to talk about like the history of self-driving and how much the technology has changed. And so self-driving's been a, I mean, it's been a topic of science fiction for many decades, but it's been a topic of, of real applied work since like early 2010. So call it, you know, 2013, 2014, it started to emerge as something that people were working on and the way those early models were being built. And this was true up until 2022, 2023, where you would essentially take the knowledge of driving and try to codify it. You try to put that into code that would then have— you'd have cameras and a perception stack, maybe cameras plus radar plus LiDAR depending on the vehicle. But that would perceive the world, it would hand all that information to a planner. The planner's code that's representing human knowledge of driving, that's coded by humans.

Yeah, it was almost all C++ and it was, uh, other than, other than processing the images.

I think it's like highly rules-based environment. Yeah. And then output, uh, controls. Yeah, control directives to the vehicle to then steer, brake, accelerate, what have you. That approach has been completely superseded by a foundation model-based approach. So instead of trying to write code to represent driving, we're training these multi-billion parameter models. And when I say we, I mean the people who are doing this correctly, I would say, to be honest. And not everyone falls in this category, but the right way to do this is clearly to build a large neural net, multi-billion parameter model where you're training the model with the benefit of millions and millions of miles of, from vehicles being driven. That model is getting better and better over time through the learning of the installed car park. And so the way we designed our 2 was we wanted to put a lot of compute into the vehicle. So a lot of inference. We want to have a really robust sensor set that's not just there to drive the car, but importantly allows us to collect a lot of data and to allow us, because we want to have really high data quality, to allow us to catch up rapidly to where Tesla is and ultimately be a leader in the path to Level 4. So the path to the idea of the vehicle driving empty, you know, are you sitting in the backseat or picking up groceries for you or dropping you at the airport? And so the timeline for us on that is getting to Level 4 in 2028. And backing up from that, we have point-to-point capabilities which will launch later this year. So that's You're in the car, you're in the driver's seat, you type the address in, the car will completely navigate to that address. Your hands are not on the wheel. You're still, you know, alert and aware of what's happening in the driver's seat, but you're not driving the car. That's what we call that is hands off, eyes on. Next year we'll go level 3, which is hands off, eyes off. So you do all that, but you no longer have to look at the road at all. You can be on your phone, you can be reading a book.

But you have to be there in case the car asks you to take over.

—right? Exactly. And that's level 3. And then level 4 is where the car won't ask you to take over in any situation. It's fully on its own.

Are there any—

But one thing I really want to make clear, if everything I just said was ignored, the thing to keep in mind is when you think about like technical progress on a topic, it's very natural. It's like human nature to look at previous progress and use that as a way to extrapolate what future progress is going to look like. And the reason we so naturally do this is it's generally true. So if you look at like how we're building homes today, we can look at the last 5 years and say, well, that's probably how we're going to build homes for the next 5 years. That ceases to be true when you have these big technical inflection points. And those don't happen often. Like, so the, like the invention of the semiconductor, a big tech inflection point. In the case of autonomy, the invention of what we think of today in terms of modern AI, so transformer-based encoding, The idea of these large-scale foundation models has completely shifted how autonomy is being developed. And so I would characterize it as the progress that was made over the last 5 years, let's say from 2021 to 2026, is not at all representative, not even a little bit representative of what I think the progress is going to be over the next 5 years. So from 2026 through 2031. And I actually think that the progress is going to be faster than society will be able to ingest. And so today, you know, if, you know, we're in San Francisco, it's easy to go like hop on a Waymo and experience this. That's a, you know, it's a tourist attraction for people that don't live in San Francisco. That is just gonna be cars. Like, I mean, in my view, I think like a car that you buy in 2030, 2031, you will expect it to have capabilities similar to that. And it's sort of hard to like believe it 'cause you're like, oh, we've been saying things like this for a while. But there was a big shift in how the models were developed. And it is— we're seeing it firsthand, like the rate of progress we're making on our own platform is materially different than what we've been doing for the last few years.

That's super exciting. I had an interesting conversation with someone who owns like, I think, 5 to 10 car dealerships. None, none of the, none of the brands that he sells has any type of like meaningful autonomy. Maybe they have a team working on it, but not really tangible progress. And I asked him, how much do buyers really care about autonomy right now? Do they come in and say, well, I love the look of this car, but it doesn't drive me, so I'm not super interested? And his dealerships are in the Midwest. And he gave me an interesting answer. He said, I don't— it's not really a big concern. It's not a top priority. For our buyers right now. And I think it was because a lot of, a lot of the people in that area have like a 15-minute commute, 20-minute commute, something like that. And I compared that to friends of mine. And I live about 40 minutes outside of like LA area. And I have two friends who just could not fathom driving anything but like their Model Y because they're just so addicted to autonomy now. They're just like, I don't care what any other car looks like. And so I think one of the big wake-up calls is going to be for the big manufacturers over this next 5-year period, where even if they're lagging 1 to 2 years behind as companies like Rivian achieve these new levels of autonomy, it will be just so hard to compete purely on aesthetics and what does the internal leather look and feel like. Feel, things like that that historically have mattered. So—

I couldn't agree more with everything you just said, including today it's not a huge— it doesn't drive a huge percentage of purchases. And I think part of the challenge is even with Tesla's FSD, it's a very, very good level 2, but you're still required to be in the driver's seat, still required to be alert. I think the big inflection point is the moment you start to— without the car dinging at you or asking you to look at the road, like where you truly get your time back, where you can, like, be on an iPad the whole ride, not look at the road once. And I think that you experience that once or twice and you don't think you needed it. You didn't think it was part of your purchase criteria. And then you're like, wow, I like that. Must have that in my next car, to your point. And I think you see with early adopters within the Tesla ecosystem, like, that's definitely true. If someone's been using FSD for a while, it's very hard for them to imagine a car that doesn't have at least those capabilities. Yeah.

Final question. I'm sure you have to jump, but the car enthusiast in me would love for the R2 to be such a massive success for you to sell so many units that you guys have the ability to fund maybe something like the 911 equivalent of an EV, you know, a somewhat accessible— a Taycan Mokk? You know, something like an accessible sporty EV for enthusiasts. What's on the bucket list? Hearing everything that you've said, I just feel like you're never going to actually be able to fully justify that. But I can dream. What about a track-only—

Well, Mark II is accelerating on 11 right now.

So you've got that. Yeah, it is. Well, what about a track-only version of the Amazon delivery van? That's what I'm in the market for. Very small market. Very small market. I want a wing on the back, some active aero.

Let's do it. No, but I think—

I think there's probably like 5 people.

No, but I think the like $80,000 enthusiast car for the EV market will eventually—

I mean, I would actually take the other side of this just to, you know, steel man, is just like the focus on the categories that you've chosen have clearly been remarkably prescient and accurate in the sense that like more and more Americans are buying full-size SUVs for their families. Trucks are still incredibly popular. Midsize SUVs are popular. And as much as the car nerds love the old sedans and the sports cars and the two-doors and the convertibles, like those are just a different tool, their recreational vehicle. It's like a special thing. And like, and like Jay Leno loves them, but like there aren't that many Jay Lenos if you're trying to build a big business. And maybe this comes down the pipe, but we're not gonna hold your feet to the fire today.

I'm gonna be praying for millions and millions and millions of R2s on the road. Me too. And so RJ can justify the go-kart.

Yes, of course.

So you can justify the go-kart, so you can justify the 911 equivalent. So we'll be praying for—

I think if I were to stack rank them though, and things we would do, the 911 equivalent would be higher than the go-kart and higher than the super van that you just—

track only delivery van. Okay, so, so we, we got only upwards to go. So that's good. Yeah.

So, so it's more likely than the other options. Yeah.

Well, well, I'm getting one. Thank you so much for coming on. We really appreciate it.

Yeah, great, great to get the update and congrats to the whole team. It's, it's a beautiful car.

Yeah, we're very excited. Have a great rest of your day. Congratulations. Goodbye. Let me tell you about Consul. Consul builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. And our next guest is already in the waiting room. We have Chris Miller, the author of Chip War, fantastic book if you haven't read it already. He's been on the show before. Welcome back. To the show.

Chris, how are you doing? What's going on?

Doing well. Thanks for having me.

Thanks for hopping on. Um, I guess the first place to start, since you've been on the show before, we did the whole deep dive on Chip War. I've read the book, it's great. Hopefully most of the audience has read it. If they haven't, they should go pick it up. But, uh, what do you think has changed the most since you wrote Chip War? What, what is— what have been your biggest updates to your sort of world model, your way of thinking about the chip war broadly?

I think the big change is that for the last 2 decades, if you look at semiconductor spend as a share of GDP, it was roughly flat. Yeah. And then in the last 4 years, it's roughly doubled. And that has all been driven by AI. And I think, you know, the question is doubling's fine. Can it triple? How much more can it go relative to trendline? That's the big question I think hanging over the industry right now.

And is that gated by TSMC capacity, energy capacity, AI demand? The AI demand seems to be there. Like everyone's inferencing models all day long. The models require more and more chips. A single model these days needs an NVL72 or a whole rack. But what do you see as sort of like the key inflection points that people should be watching for as the semiconductor industry continues to boom or, you know, tops out at some point?

Yeah, I think right now it is gated by TSMC and manufacturing capacity more broadly. But if you put yourself in the shoes of those companies, they're looking at AI demand, not sure how strong it's gonna be. Obviously strong, but is it strong enough to double or tripling of semiconductor spend as a share of GDP? That's a big jump. And of course they're on the hooks for the CapEx. They gotta build the factories, they gotta pay for them. And so I think companies in the manufacturing layer are excited but nervous about just how far this is going to run. And they have an incentive as a result to be a bit more cautious on building chip plants than AI companies would like.

In historical infrastructure buildouts, let's say like electricity and railroads, was there ever a key supplier that just didn't have any sort of like insane FOMO or fear of missing out? Like I'm thinking like railroads, was there a critical supplier that was like, you know what, we're excited about this trend, we're investing in it, but we don't wanna get ahead of it.

We're building enough railroad ties. We're not gonna build any more railroad ties.

Yeah, yeah, yeah. Like, you know, if there's any historical—

The curmudgeon steam engine.

It feels like on one hand, obviously, there's frustration from other players in the supply chain. But at the same time, it feels somewhat potentially healthy, even though I think you can see that bubbles have produced great results throughout history.

Yeah, I think we are in a somewhat unique moment in that there are a series of choke points down the supply chain where you just rely on a single company for most capacity. And that didn't happen with railway ties. Hasn't happened with electricity. Uh, you have it at the TSMC layer for sure, but then you look at their suppliers for key materials for advanced packaging, and there's often choke points there as well. And the trick is that the deeper you get in the supply chain, the further you are from the actual end AI demand, and the more skepticism builds up as to how real is this actually. And that's why there's been underinvestment in the materials and manufacturing relative to what Silicon Valley needs. Yeah, it's the bullwhip effect, right?

A small change in demand amplifies and amplifies and all of a sudden they don't want to get caught at the tail end. And also what I know, I know Nvidia traded all over the place during the crypto boom when cryptocurrency miners were using GPUs and then Ethereum went proof of stake instead of proof of work and, and there was a drop off in demand there. But I'm wondering, do you remember if TSMC had the same level of skepticism or was even affected by the previous rise and fall of NVIDIA?

You know, I think the key is that before the last couple of years, NVIDIA was one of many customers that TSMC had. Sure. And Apple was the most important. Whereas now it's kind of flipped. NVIDIA and AI in general is the key. Growth driver. So in the past, TSMC would have more closely tracked smartphone sales than they would have tracked data center buildout.

And smartphone sales were on a much more steady pace since Apple has their whole machinery of refreshing and planned obsolescence potentially.

And you look at individuals are not like, I need— like the iPhone's gotten so good, I need 5 phones.

Yeah, yeah, yeah. There was no exponential takeoff, but now there's like, okay, I need 10 more, 10 times more tokens, 1000 times 5 more tokens, et cetera, et cetera. Staying on TSMC, I'm curious about the culture there because TSMC is known as this like champion of Taiwan, the most elite institution, the crown jewel of Taiwan. And like why— we explained the bullwhip effect a little bit about the reticence to deploy more CapEx, but is there also some sort of cultural element or capital markets dynamic? Because when America has a crown jewel, we're like, send it to the moon. Like, oh, there's any excuse to 10x capacity or 10x the scale of our crown jewel, like, let's do it. And it seems interesting that you would have a company that has this opportunity, and there's like a glimmer of a hope for it being even bigger. And yet, is it, Or is it more of like this protective mindset of like, we have this good thing going, let's not risk it? Or is it just like they can raise prices so there's less downside there?

I think it is a conservative culture, but I think they believe their conservatism has served them well in the past. There have been cycles in the past, and being cautious going into the cycle gives you the opportunity to invest at the bottom of the cycle. And TSMC has done that really well historically. Now, I think right now, I think they realize that we're not at the top. They're building new factories as we speak, but I think they're also looking at the history and saying, we don't wanna over-invest relative to what our customers are gonna need.

Yeah, and then what does the domestic chip supply look like from your perspective right now? There's a ton of energy around Intel and it feels like Intel checked a lot of the boxes for the comeback. Like the comeback hasn't really arrived yet. But in terms of, you know, the backstop from the US government, additional investment, the stock price is up and a bunch of, a bunch of buyers and potential customers are falling in line for their, their most advanced node. It feels like things are on track there. But what else is going on and is that actually a pressure on TSMC at some point?

I think to some degree that that does create a bit of pressure. Agree that Intel looks like in a it's in a much stronger position. I think the customer traction is the key thing. And as you say, we haven't— we've seen news about customer traction, but not actually volumes yet. And Intel really needs substantial volumes of customer orders. So I think that's one positive sign. I think everyone's watching carefully what Elon is going to do with TerraFab, how that's going to grow. Could be transformative, could not be. It's obviously a hard industry to, to break into. And then, you You know, there's Samsung with a big plant in Texas they're bringing online, and also Tesla is reportedly one of their customers. So you look across the ecosystem and there's certainly a lot more optimism about leading-edge manufacturing than there would have been just a couple of years ago.

Bull and bear for TerraFab.

I would love your take. You know, I think, A, never bet against Elon, but B, you know, it seems like he's lining up a coalition of partners on the IP side, on the manufacturing side. It'll be interesting to watch how it develops.

I'm excited. Yeah, I guess the question is like, how much of the fab is him versus he's the buyer and it's actually Intel doing everything? Because I don't think anyone doubts that Intel can like run a fab since that's what they've done for decades. And if you bring on the right partners, it seems like it could be possible.

What do you know about ASML's current thinking around AI progress and the buildout and how they're thinking about investing against this progress?

Well, I think ASML is in a similar position to TSMC. They want to meet customer demand, but they don't want to overbuild. And they're also on the wrong side of the bullwhip. And so they have an incentive to be a bit more conservative than their customers do. But it's been a great year for ASML as obviously demand has just been through the roof and that's I think that's going to continue for some time. I think the other challenge that ASML is working through is that TSMC, their most important customer, has said they're not going to buy the latest high-NA lithography machines, at least for a couple of years. And that's also a medium-term challenge for ASML because they're trying to sell the next generation machines.

It sort of doesn't matter how AGI-pilled you are if your customer doesn't show up to buy the thing that you're making in the supply chain. How have you been processing the export controls back and forth? Because it feels like it's been every other day. It's like, okay, we can send anything and then we can sell the Nerf stuff and then we can now sell nothing and then everything. And it's been going back and forth. And that's sort of the nature of this administration. But is there a clearer through line that you've detected? Is there something else going on there or is it as chaotic as I've reading into it?

Yeah, I think, you know, step one is don't read the tweets, uh, and step two is look at Congress, um, because Congress, that's where you get the bipartisan support for these export controls. And I think Congress has been playing an important role in influencing the administration. You've seen a bunch of influential Republicans, for example, publicly support export controls, trying to put pressure on the, the White House. I think the third thing that's important is, uh, Methos really did change thinking in the White House about how powerful AI is going to be and how important it is to have the leading capabilities to yourself. And so I think there's been a shift around the AI debate in the White House just over the last couple of weeks.

Yeah. So related to that, there's reporting in Reuters today, China prepares $295 billion plan to fund nationwide AI buildout. How did you interpret this? Do you see this coming. And then I'm interested to know, like, do they have the supply chain to actually deploy $300 billion? Because the numbers that I've seen back of the envelope for if the chip controls get weakened and Nvidia can sell the, the TAM is like $10 billion, which would be great, but not like huge and certainly a small chunk of this. So are they advancing enough with SMIC and SME to to indigenize their supply chain and actually spend this money? Do they have other plans? Like, how do you see this playing out?

I think first, those numbers are over 5 years. So annually, that's less money than Google or any of the hyperscalers. So just to put into context.

Play the bald eagle sound, baby. We still got it. We still got it.

What do they know? What do they know? What do they know? What do they know, Chris? No, seriously. Seriously.

Yeah, how do you read into it?

They're— they're— this is the puzzle. Yeah. The puzzle is, is why isn't Xi Jinping more AGI-pilled? Sure. You know, if you thought AI was important, you thought chips were an important ingredient, China's been underspending for the last 4 years on AI. And I think it's because the Chinese government just doesn't really believe that AI is going to be nearly as important as we do. And if they did believe it, they'd be, A, buying H200 chips. And B, spending more on data centers.

They're not doing it. Is it possible that they just, like, they believe in AGI or something, but they just have longer timelines? Like, it feels like a lot of the underwriting that's happening in the United States is based on, again, 18 months, coincidentally exactly the VC timeline for a round. It's like, what will we do in the next 18 months? We're AGI in 2027, 2028. Really short timelines here. And then that's also backed up by Dwarkesh Patel had a post saying like, if AGI comes soon, America wins. If it comes longer, China wins. And it feels like that's almost like— I like his analysis, but it's also like a reflection of the, of the revealed preferences of the two countries where America is sort of building towards a fast takeoff AGI soon scenario and China might actually not be like necessarily behind the ball. They might just be optimizing for a 2035 AGI. And if they ramp up slowly, there's going to be less gyrations, less of a bullwhip effect, and they can take a longer view.

How do you process that? Yeah, I think one explanation is just that the US is a service sector heavy economy and AI is going to be more impactful in services first, manufacturing second. Yeah. And so that could be part of the story. I think, you know, second, US firms just could be more tech-forward. You see this in cloud computing. China has a tiny cloud computing market relative to GDP because its large enterprises just don't want to move to the cloud. And so that could be another analogy. And then I think third, we should be open to the prospect that the Chinese government just hasn't gotten the memo yet. Yeah, you know, for better or for worse, we have no deficit of venture capitalists and tech entrepreneurs in government. Yeah. And of course, China is not the center of government. And that could be another factor. Yeah.

You would think they would be excited about it purely from an authoritarian lens because like if you have like the all-powerful model that, that can basically theoretically— they already have— micromanage every single— micromanage every single citizen's— the content that they get, the services they get access to, you know, how much, how much the retirement benefits they get, all these different things. Like, in the US, we think of this authoritarian AI scenario as like a, you know, doomsday scenario.

It sounds terrible. But if you're an authoritarian.

But if you're an authoritarian, you think like, I gotta get one of these authoritarian models.

Yeah. I don't know. Yeah, I think that dynamic is there. I think the risk if you're an authoritarian leader is you don't know what the model's gonna output. They're not nearly deterministic enough for your preferences. And that's the downside for any authoritarian leader.

Yeah, there was recently— Or better training data.

There was recently reporting that Vladimir Putin instructed the Kremlin to disconnect or firewall or air gap all of the CCTV cameras that are, I guess, inside inside the Kremlin or like could potentially triangulate him because they were worried about a hack where his location is disclosed to a foreign adversary. And so, yeah, you can see that these tools cut both ways. Have you actually followed anything on what's going on in China? Or in Russia? In Russia.

But before we go there, I wanted to ask, how closely have you followed the whole Manis saga and debacle? Do you know the latest? It feels very unclear how something like that gets, you know, unwound. But yeah.

Are they still in China? Yeah, I don't know. Well, this is, I think, the really interesting part is that, you know, if you have a broader campaign of pulling passports from leading entrepreneurs, that is an extraordinarily chilling sign for the rest of the industry. And I think there's a real, you know, if I was sitting in Beijing, I'd be really nervous about being too tough on my entrepreneurs. 'Cause those are the people you need to invest and build companies. And that's a pretty risky move, I would think, for Beijing in terms of building their own domestic ecosystem.

I mean, Liang Wenfeng, the founder of Deepseek, has the most control, I think even higher level of control than Elon at SpaceX in terms of founder ownership and concentration, because he had Highflyer, the hedge fund, and then he funded DeepSeq. So he did the first couple of rounds and provided his own capital. I think it's like 80% ownership. And I mean, it's already a multibillion-dollar company, could be one of the biggest companies in China if AI labs are any indication in America, if there's any correlation there. And so he's going to— he should get a second passport or something. I don't know. Good luck to him.

But anyway. But Russia. Russia. Yes. We've been trying to— we've been They're trying to figure out how many GPUs they have.

Well, like 10 years ago, Vladimir Putin was giving a talk and he said, "Whoever wins AI wins the future." And he was AGI-pilled before, way before it was cool. Then he got into a crazy war and it seems like he lost all of his best talent to Nebius and what's the other one? ClickHouse and a bunch of other Yandex spinouts and stuff. So I'm just wondering, like, has Russia come up at all in your research of like where chips are going, where data centers are getting built? I just don't hear them in the conversation at all. And it's surprising to me because they have nuclear weapons.

Yeah, no, I think, I think you're absolutely right. They're not a big player. And the reason is because Nebbius is outside of Russia. That's an example of a team that was one of the leading groups of tech entrepreneurs in Russia, at least until 2022. And now they spun outside of Russia and are building a great company in the West. Uh, and that's a microcosm of the entire tech ecosystem. The best estimates that hundreds of thousands of Russians left the country after 2022, disproportionately well-educated, disproportionately in the tech sector. And so yeah, there, there have been some reports of GPU smuggling into Russia, but it's, I think, small scale because there's just not the demand for, uh, building big data centers. They're gaming.

They're, they're getting those GPUs gaming. Gaming. It's like they, they only smuggled "1080 Ti's? You can't run advanced LLMs on those." It's like, "Yeah, they're gaming." Talk to me about the reception in China of American GPUs, Western GPUs. There's like interesting smuggling dynamics, interesting debates over NVIDIA export controls, but then we see these articles pop up every once in a while saying like Beijing doesn't want 'em or Beijing is gonna pressure local companies to use the Huawei, a send chip, or go more indigenous. And I'm just interested because we label China as like authoritarian, but obviously no system is perfect one way or another, obviously. And so do they actually have the ability to restrict this? How much do they actually want? Do they really care? Because it feels like the pool of compute is pretty fungible and I don't know how Are they worried about like security risks or dependency? Like how would you rank their worries?

Yeah, yeah, I mean, what we know is there is smuggling at a reasonably large scale. There's also offshore compute access. So if you're a Chinese hyperscaler, you can get access to compute in Malaysia or wherever else. But it seems like the Chinese government is really fixated on pushing firms to buy Huawei. And it's surprising, I think, that they haven't let Chinese companies buy the H200s. Despite that, it's now been half a year in which the Trump administration has been allowing sales. And that goes back to what's the Chinese government trying to do. If you were AGI-pilled, you would buy the H200s. Yeah. Um, but if you're not, then you're focused on domestic manufacturing and supporting the Huawei ecosystem. If you're worried about data security and think there are backdoors, which I think the Chinese might be worried about, then maybe say ban the foreign chips. Yeah, there'll be a bit of smuggling, but better to build up your domestic ecosystem first. And that does seem to be the revealed preference of the Chinese leadership.

Why do you think— what are the keys to China's broad support of AI locally? AI polls so terribly in the United States. It polls terribly in the EU as well. But China is almost flipped. It's like 80% support artificial intelligence. Only in America, it's like 20%. What's going on?

You know, I think part of it might just be the media landscape. Uh, you know, in, in China, if there are topics that the government doesn't want discussed, they don't get discussed. That's part A. Um, you know, B is that the, the entire tech sector is much less controversial among the Chinese populace than in the U.S. We've had a decade of debates about Mark Zuckerberg and social media. That's been present in China to some degree, but much less toxic than in the US. And then, you know, I think third, US firms have been uniquely bad with publicity, promising to destroy jobs, promising to threaten the future existence of humanity. That's been the marketing of AI over the last couple of years. And, you know, when you hear that, it's not surprising people don't like it.

Yeah, that makes a lot of sense. Is anything happening in Japan that's relevant right now?

I think the big story in Japan is the the number of companies that have been discovered to have extraordinary materials capabilities that sell into the chip supply chain. So my favorite example is one of the toilet makers in Japan has an advanced ceramic capability that's used in the machines that make chips.

I know, and it's so funny because when the headlines started breaking around that, this was probably 2, 3 months ago, a lot of people were like, wow, top signal, like, The toilet companies getting into— AI. It was like an all-purpose thing. Yeah, it looked like just like a pump, but no, they actually have an industrial process that's very valuable and very real. Yeah.

I mean, yeah, great lineage there in— I mean, they were doing humanoid robots. The Honda ASIMO or something was in like 2002. And it feels like the Japanese economy just sort of lost a step and just didn't keep doubling down. They were just too early. And so they ceded a lot of that territory to China and America. But I'm rooting for a comeback because I like that Honda robot. It was cute. Didn't feel like Terminator at all. Felt like C-3PO. I don't know. That's right. That's right.

I think the Japanese robotics firms, I think they're so focused on factory automation and that's where their expertise really lies, that they're struggling to get their heads around what does AI plus robotics actually me. Sure, sure.

Interesting. Uh, last question from me, uh, is NVIDIA a car? What's going on with their moat? How are you thinking about NVIDIA in the supply chain? How important is their role? How defensible is their position? The debate was, uh, Jensen saying NVIDIA is not a car. It is not something you can just step into and get where you're going. It's a special product from a special company?

So I think we're into year 2 at least of the ASIC debate. It's clear ASICs will have some market share, but it also seems clear that NVIDIA is going to be the dominant player by far. I thought it was interesting to watch NVIDIA acquire Groq Technology 6 months ago, showing that they're going to be able to diversify into inference-focused hardware, building their empire of sorts. And CUDA still remains a very powerful part of the moat. So, you know, is the market going to decline from 90% to something lower than 90%? You know, that's what's happening. But does that mean NVIDIA is not going to be the dominant player? It certainly seems like it's well on track to be the key provider of GPUs for a long time to come.

When's the next book coming?

I was going to ask and tell us the title. Is he— but obviously you can't announce anything, but If you were to create the next chip war, would it be called Chip War II, or would it be called Battery War, or Robotic Supply Chain War, or, you know, Energy War? Like, is there a new war that will supersede the chip war, become the most relevant topic, or are we just doubling down on chips?

You're kind of using a little bit of fear-based marketing yourself, Chris.

Yeah, war, eh! You're like, oh, it's a war?

Chip War.

Why not just beautiful trading and collaboration? And yeah, there'll be prices. Chip trade, I don't think. Yeah, chip trade doesn't sell any books. Chip industry. Chip industry dynamics. Yeah, chip industry deep dive, the book. That's what, yeah, fear-based marketing. We caught him red-handed.

It's working on me. It's working on me. I want the next one.

The next book, the next hot topic, the next thing that you'll be digging into. What are you thinking?

You know, I think I'm spending all my time on chips right now. I don't know if there's gonna be a sequel, Chip War II.

Chip War II: Revenge of the Chips. I know that sounds so good.

Revenge of the Chips, that's where we are. But sequels are never as good as the original. So I think probably not a sequel on that topic.

Maybe Chip Peace.

You've done Chip War. Chip Peace.

I hope so. What is the path to Chip Peace? That's what I want to hear. A lot of fearmongering, a lot of worries about war. We need a lot of ways to get out of quagmires, I think. That's generally what I'm rooting for. Anyway, thank you so much for taking the time to come chat with us. This is a lot of fun.

Yeah, I really appreciate it.

Love to have you back on soon. And go get the book, Chip War. Available everywhere books are sold. We'll talk to you soon. Goodbye. Let me tell you about CrowdStrike. Your business is AI, their business is securing it. CrowdStrike secures AI and stops breaches. Our next guest is in the waiting room already. We have Evan Beard from Standard Bots. He's the co-founder, CEO, chief engineer, and he has a massive Series C to announce.

How are you doing, Evan? What's going on? Good to meet you.

How's it going? You're looking extremely sharp. What's the occasion? Introduce yourself. Tell us what's going on.

Thanks so much. Evan Beard, CEO of StandardBots. StandardBots is America's largest maker of AI-native robots. Announcing today our Series C. So we raised $200 million at a $1 billion valuation. And we're announcing also— Congratulations. Yes, they're waiting for the gong. Yes. Now the fundraise is official. Yeah, fantastic. And so we're also announcing we're quadrupling our manufacturing footprint. Footprint. We're growing so quickly. We expect next year to be 10% of all US robotics, industrial robot deployments.

Wow. So how small is the US industrial robotics? They all come from Germany and Japan or China. Like, who are the major players internationally? I mean, obviously it's important to, to reindustrialize, but it feels like for such a young company to get so much ground, like we really have to be behind the behind the ball here, right?

Yeah, last year we grew 10x year on year. We are behind the ball. When I first started the company, so I did Y Combinator. Cool. I, uh, I looked around and realized I wanted to start a robotics company. I realized that almost all the robots, it's like 99.9% of the robots used in the United States, are not made here. Yeah. So, uh, so that's the kind of starting point. And then you try the products and you have these menus that have like a single letter in them. It's like crazy stuff, like you would expect from I don't know, like mainframe computer type. Like it's beyond any other software I've ever seen. And so we realized we could make something 10x better. And then with AI now, obviously you can train a robot by demonstrating a task, whole new way to train it, can be more capable, handle more variability.

So that's what we do. Yeah, and it's probably not sold to like a technical audience. Not like Stripe for humanoid robots or industrial robots exists. The buyer's probably some firm that owns a whole, footprint of manufacturing plants and they just say like, stick it there and figure it out later.

Yeah, so has the journey so far been about finding product-market fit or has it been about just delivering an American equivalent of products that are already popular?

Yeah, it's a great question. When we first started, so I talked to 100 different manufacturers and I asked them, do you like your robots? Who programmed them? Can I play with them? How long did it take? What are they doing? It was really from the beginning, we knew exactly what we needed to build. And from the beginning, the problem was actually building it. It wasn't like, what do we build? Or how are we going to reach the people? Or do they want it? And what's been consistent from the beginning of the company, we did our seed round around 2020, has been we build the next thing that we know people need and we see even more takeoff and even more accelerated growth. So that's been, it's mostly been a technical problem, which is what I wanted.

When I got into this whole thing. Yeah, yeah, it's great if demand is all there.

How vertically integrated do you want to be? I've heard a lot of FUD about actuators and small drone motors and all these different pieces deeper in the supply chain. You hear about the American company that is the brand, and it's all made in China, or then the American company that's making the product for the for the American brand, but then all their parts are made in China. And we're going through this process of working through the entire supply chain.

Let's just say this, Jordi, he made that tie. That tie is a standard Botz tie. There we go. He's gotta go all the way.

You gotta go all the way. I imagine before—

Nike's, you know, I try to do, I rep the American brands, but we're trying to make every single component of the robot with this fundraise in the United States. And we've already designed most of it. So we designed our own motor, we designed our own motor controller. Like, we wrote the original firmware. I wrote the original firmware to do all the motor control, 60,000 times a second switching electricity to, you know, make the magnet and the electric— electrical— the electrical electromagnet, uh, you know. So, so we vertically integrated on a really low level. Yeah. And now it's about vertically integrating the manufacturer as well. So that's the next step here.

What is demand like for American-made industrial robots? Obviously, people online, commentators, investors have gotten the memo about the importance of reindustrialization. But if you're just some American manufacturer, you might not have the money, you might not really care, your customers might not care.

How resonant is this idea? Yeah, is part of it people just like knowing that you guys will be stopping by the factory and things like that? Yeah, basically, how are you differentiating besides like the country of origin? Yeah, yeah, I know.

So we— our whole focus is training through demonstration. So we have made, made it so that an average assembly worker or manufacturing worker can train a robot, and with our next software release, they can train an AI model too. And so this is, uh, it'd be the equivalent of if you could fine-tune like ChatGPT without any kind of software engineering degree. And we make it so simple and easy to show a robot what it is you want to do and then have the robot go and do it on its own. And I don't— none of the other incumbent robots, they're so far in the past that I don't think they're even beginning to work on this. And I think that there is a real potency to vertically integrating, to be able to make the hardware and the AI stack and the software all together, because you can make it work better. And so that's what we do, and that's what we've been focused on. And I think that it's, it's that combination that allows you to go address all these tasks that otherwise you just can't automate today.

Are there many robotics-led manufacturing robots? I'm thinking like, you know, a PE firm says like, hey, there's like a very specific type of manufacturer. They haven't adopted technology as fast as they could. We're gonna buy these companies, combine them, and, you know, let's say roll out standard bots across their whole fleet.

We are working with a couple of those types of private equity firms. So we have, you know, many hundreds of customers now. So there are some of those. It's a lot of just, You know, so we've got small and medium businesses in almost every state. We have iconic Fortune 500s, oil and gas, data centers, aerospace, automotive. So really spanning the gamut there. But it's people who are making the things that make America, the manufacturing components, also the end products that we use. And that, that gets us up out of bed because what robots are today are really they're the tool for us to make our economy and our country. Manufacturing affects 1/3 of GDP, 1/3 of all jobs. And so if we can move the needle on those things, it moves the needle for the country. And that's why I've been, I've been to testify to Congress twice in the last few months. And this is, uh, it's really across the aisle, there's agreement that we need to dominate robotics as a country in this next. And so I think to your other question, it's very important, I think, that all the components are made here because we don't have this supply chain here today. And you don't want to have something that is so fundamental and foundational to the future of this country not made here. It's just not a good, good play for our kids.

Last question for me. How do you see humanoid robots fitting into our robotic future? How do you think they fit in, break down? I'm always shocked by this, this stat that I think Amazon employs like a million humans, but also a million robots, or they've deployed a million robots while scaling their workforce into like the tens and hundreds of thousands. Hundreds of thousands, and there seems to be some synergy there. I imagine that there's a lot of energy around humanoid robots, but there's so many situations where you want a robot that just bolts to the floor with a lot of power that can just move a steel beam that a human can't perfectly, repeatedly via demonstration learning. But how are you thinking about that adoption curve? Because I imagine there'll be some takeoff off in exponential growth, but give me some nuance of what you're seeing on the ground.

Yeah, our view is for a controlled environment, factory, field warehouse, that a legged humanoid is— it's going to be more cost, more unique parts, but it's also going to be less efficient. So you just have a worse ROI. Yeah. So we think maybe humanoids are right for the home, but then you have the safety issue. Like when it's going up the stairs, you need two 5-year-olds to be able to jump on the back of it and pull it, try to trip it. It's the Home Alone Home Alone test is what I call it.

The Home Alone test. Yeah, yeah.

But, but I think we're years away from, 'cause like you see RL robots flipping with RL and it looks so impressive and I think the average person thinks like, okay, it's all solved, robotics is solved. But I think to, to solve the, to, to beat the Home Alone test, it's, we don't, you know, RL's not quite there. Yeah. And it's hard to simulate all the things that a, an 8-year-old could do to, to, so you can't even do the RL right now. Yeah., right? Because, okay, we don't simulate all the soft bodies and the meshes that are cutting. And so it's, it's an unsolved problem. I think it's going to be years away from being solved in the home. And in the factory, we make it— as you said, there's a lot of applications where you just need a static base, or maybe you, you, you, you— maybe some you need as a mobile base, and maybe you need one arm, maybe you need four arms. So we have a flexible platform where you can kind of mix and match depending on your needs, and we think that's the best ROI, the best performance. But humanoids are better at one use case, do you know what it is? Raising money.

So, I mean, if we're doing the Home Alone test at scale, we're gonna need to massively revamp child labor laws because the 5-year-old, the 8-year-old, they're gonna be working 40 hours a week paying for college, just beating up humanoids in some— sitting in some designed house. This is important work. Now, I'm sure they'll solve it some other way. Synthetic data. That's a good— anyway, that's a really good line. That's a good zinger. Thank you so much for coming on the show. Thanks for having me on. Congratulations. Yeah, congratulations to the whole team.

Appreciate what you're doing. Glad you guys are doing what you're doing.

And looking sharp. We'll talk to you soon. Have a good rest of your day. That's fantastic. Let me tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in-store, on mobile, on on social, on marketplaces, and now with AI agents. And our next guest is already in the waiting room. We have Nick Fleischer returning to the TVPL program from Sandstone. He's the co-founder and CEO. How you doing, Nick? Good to see you.

Great to be back. Great to be back. How are you guys?

Great to see you.

We're doing great. Overnight success. What's going on? You raised some more money. Tell us about it.

We, we, we raised some more money today. We announced a $30 million Series A led by Lightspeed. All right.

There we go. You're moving at light speed.

Tell us what's going on. What's the traction? What's the unlock? What's the actual product doing for people day to day?

Yeah, the product is legal relationship management, which means we centralize intake and knowledge for in-house legal teams so that they can actually deploy it in AI workflows that are useful. Since we launched about 150 days ago, we've deployed to, you know, dozens of Fortune 500s and mid-market companies across industries like manufacturing, commerce, tech. And we've grown the business by 40x. The team has tripled since we last spoke end of January. And we're excited about bringing this, this tech to legal teams who are really in need. Yeah. We, we like to say that we are playing against a legacy market, which is called contract lifecycle management. Very boring tools where in-house teams historically store contracts and that's about it. And they are burdened and gated by that piece of tech. And we want to help unlock them so they can actually use, use the great AI that's out there now.

Your investors haven't exactly sat out the foundation model race. There's been a lot of FUD around wrappers and application layer companies, all sorts of derogatory terms thrown around. At the same time, we just saw data from Cursor, Cognition, a couple other firms that appear to be in the direct path of the AI labs. They're all growing, they're accelerating, they're doing great. How did you frame that question during this round of pitches?

Yeah, and I mean, Claude is, you know, Claude has announced Claude for legal. Yeah. OpenAI made a big hire. Yeah. Last week in the legal tech space. And so they definitely want to play. And I think most of where we see them playing is on quick legal questions, legal research, and redlining of documents. Yeah. And we explicitly said from the start, we think, redlining and marking up documents is going to be commoditized by the labs. And so we said we want to actually build a workflow tool that's helpful for in-house teams. And so, you know, even the most forward-thinking tech companies are paying us, one, for opinions on what those workflows should look like and how to implement them. Two, the context layer to actually think about, you know, how does each legal decision and relationship map, you know, to everything else and how do we retrieve that at the right moment so that we're pulling the proper context when I'm answering a question or looking at a document. And then I think the final one is building a tool that learns from your past legal decisions and can work for not just the stakeholders within your legal team, but also the folks in the business who are requesting work from legal. That's a lot more complex than any chatbot or self-built tool is going to be able to do.

Has token maxing hit the legal community in any way? I would assume no, because there's maybe more finite amount of— like companies aren't— a software company wakes up in the morning, and there's so much software to build. We want to build as many individual features as we can. There's stuff to build internally and externally and all these things. Whereas like an in-house legal team is not necessarily waking up and being like, we want to 100x the amount of money.

A tech company might go back and say, review every line of code in our code base for bugs. A law firm probably isn't going back and saying like, review every document we've ever sent and do another redline on it when our client's not paying us. But how have you been interpreting it, the token maxing question?

I think generally we don't see we don't see it happening at scale, but you could see a world where people say, hey, can you redline this or review this agreement or answer this question in the 50 different ways that we potentially could and see how it plays out in a negotiation, right? Can we simulate what the outcome could be here? And so I think we see some people playing with the idea of using AI to do things that are maybe outside of the day-to-day. But I had a GC the other day on a demo tell me we are definitely not AI maxing, today and we won't be doing it soon. And I think they are much more cautious. And historically, legal is seen as a cost center and a bottleneck. I think that's going to change a lot over the next couple of months. But they are the last ones to get the ridiculous token limits and ability to go do those experiments.

Yeah, that makes sense.

How much effort are you putting into like harness development specifically these days? It feels like that was the unexpected win of the last like 6 months across CogCode, Codex, Cognition, what they're doing. Like people at least a year ago or 2 years ago thought like the model will be everything. Everything's gonna be in the model. You're not gonna, like this whole prompt engineering thing sort of evolved into harness engineering and it's, born fruit for so many firms. Are you focused on that?

Yeah, I mean, one of the things that we'll do with clients is we'll say, hey, take all the integrations that we have that you have connected through Sandstone and connect them with MCP into Claude and OpenAI and just see what, one, the quality is of the ability for it to retrieve context and answer your questions, and two, the cost, because you're not necessarily routing to the right model for the given task. And so, you know, If you do that experiment, you're going to see that, you know, 50x the results on the Sandstone side. And that happens with nearly every legal request that hits a legal team within a company. And so a lot of it is around how do we map data and context from other systems and actually go retrieve that, you know, when these questions are being asked or when we're reviewing documents. And then a lot of it is also just the, you know, underlying decisions around what models to be using when.

Typical path in law, at least from my understanding, is usually like law school takes forever, a couple years at a corporate law firm, then you can make the jump to in-house counsel somewhere. Is AI going to lower the age requirement, the experience requirement for in-house counsels? Like, should we expect to see 10-person startup, 20-person startup say, yeah, let's get someone who just graduated, they passed the bar, they're a lawyer, but instead of them spending 5 years at a white shoe law firm, let's have them jump over because we know they're going to be superpowered, or how does the dynamic play out on that side?

Potentially, I think we will see some of that. The two pushbacks I would have are one, you want to hire a GC or an in-house lawyer who has experience dealing with different types of counterparties so they can actually go up and have the negotiation and have a good backing and background when they come to those conversations. And I think, too, you still gain a lot from sort of the hustle and the grind of all of the work that you do in big law. And that will decrease a little bit, but I don't think it'll go away completely in the next 5 years. I do think we're going to see more non-lawyers on in-house legal teams. We have some clients who have hired engineers fully to their legal teams. And historically, legal has had this role called legal ops, which is essentially, if you think about RevOps on the The revenue side, it is sort of the person who pulls everything together, builds the systems, manages the processes. These people are becoming more and more technical, and I think we're just going to see a lot more of them. Like, you know, you look at Mercury's legal team, I think it's in the 20 to 30 person legal team range, and they have several people doing legal ops. Historically, it would have been maybe one or at most two for a 30 person legal team. But we see people focusing on it more and more.

That's very cool. What are you tracking on the consumer legal side? Like LegalZoom? This is sort of out of your wheelhouse, but I'm just sort of interested.

Yeah, well, it's more interesting to ask you than somebody that's actually building that space.

Yeah, I mean, LegalZoom was like so magical. Initially, it was form filling, and then we got Stripe Atlas. But is there more there? I mean, people are people are gonna go to the models and demand legal advice one way or another. They're gonna jailbreak them if they have to. But how does that market change? How do you think, what do you think's gonna happen in that market? Is there anything interesting you see as changing?

There's so much opportunity, right? Like those tools are historically very much just mapping you to the right person who can do the services. And it should all get disrupted. And there's very little investment and focus there. And so like if I was starting another company today, That's a really great market and I think it's very specialized, right? You see some popping up on the consumer divorce and family law side. Sure. And we'll see that in, in all the different areas. The, the other thing is just law is getting more complex and there's a lot more of it because of AI, right? All of our clients are telling us they're busier than ever because, you know, the people in the business are negotiating more. They're doing things with AI that they shouldn't be. They're creating more text and more words. The business users are token maxing. And so as a result, there's just more legal work, right? And there's more regulation. And so you're going to see people wanting cheaper options for getting the answers to their questions and getting through cases.

Yeah, Tyler Cowen had a funny take on this. There was like, even in the most aggressive ASI scenario, you're going to have tons and tons of government lawyers because we're going to want the humans to write the laws that govern in the AI and whatnot. Anyway, thank you so much for taking the time to come chat with us.

Fantastic progress. Congrats on the 40x to the whole team. We're gonna hold you to that the next time you come on. Better be a 40x. Ideally accelerate.

Yeah, yeah, maybe, maybe.

400x in 150 days.

There we go, he's calling the shot. Calling the shot.

Love it. Have a great rest of your day.

Thanks, guys. We'll talk to you soon. And I'll tell everyone about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI, own the data platform that powers it.

Yes. Did you know that Dwarkesh Patel's assistant is the brother of Leopold's fiancée's boss?

I heard that Dwarkesh Patel's hairstylist's uncle went to school with Demis at DeepMind in preschool. And so I thought that was something that needed to be disclosed.

I think you're thinking of the preschool teacher.

The preschool teacher. That's right. Okay. Well, yeah. Newsflash, Silicon Valley is a small place. And in fact, this is so funny because it's like, actually all the people in AI live in the same house. That's the real story here. It's not that there's some crazy— It's one Justin is the brother of the boss.

It's a small house and you're not in it.

Anyway, uh, it's very funny that somebody tried to drop this as a bombshell and wound up writing one of the most confusing lines on— actually, no, it's not AI.

We're still creating new sentence structures.

We are, we are. Never before in the English language is something so confusing. Before we bring in our next guest, Chris Ninja, we have Paul Graham said—

What'd he say? I strive to make my writing unsummarizable in the sense that it has so little fluff left in it that if you take any words out, as summaries by definition do, you lose a lot of interesting ideas. And somebody just says, well.

And they put, why is this block 3D?

They just blocked out the rest of the text.

But why is it 3D with a beautiful light gradient rip, like flourishing off of it. Like, I don't understand the— the— like, you could have just gone into iOS, put a little black box over this, or blur it or something. They chose to put a 3D block there that's with a blue light and a white light on the right. It looks very aesthetic. This feels—

but Paul Graham fires back and says, are you trying to prove my point or just doing it accidentally? If I'd written just that, no one would have understood what I meant. The only reason people can understand your version is that they can also see the original.

Owned. I wonder, I wonder, is "I strive to make my writing unsummarizable" a banger? Do you think that one gets more posts, more likes? The original got 5.2K likes. Do you think "I strive to make my writing unsummarizable" gets— it got 66,000 with this, but that's because it's funny. Anyway, we can debate that more in the future. Let's bring in our next guest. We have Chris from New York. Angel One. He's the president and co-founder. He's in the waiting room and now he's in the TVP and Ultra Room. How are you doing, Chris? Good to meet you. Great, how are you? Congratulations. Massive news, but it is your first time on the show, so I'd love a little introduction on yourself and the company first, and then we can go into the news.

Great. First, thanks for having me. I'm a big fan of the show.

Thanks so much.

Uh, and I'm a local in, uh, La Cañada, so pretty close to, uh, where you guys are.

Yeah, that's amazing.

Yeah, I knew that. Uh, but I'm Chris Matarisi, founder of NinjaOne. Um, at Ninja, we believe the hardest problem in IT today is complexity and fragmentation at scale, as organizations are stitching together literally dozens of point solutions just to manage, protect, back up, and support their endpoints, you know, and that fragmentation costs— it creates cost, risk, poor service, etc. And Ninja solves this challenge by bringing all of these operations into one unified, modern, cloud-native platform. You know, in short, we really just help our customers reduce spend, unify IT, and simplify work.

That's, that's our thesis. Okay. That makes a ton of sense. And I think anyone who's been inside of a big organization has felt that, uh, at the same time, point solutions, point solutions are great sometimes. And I'm wondering about how you do the dance between, well, we're so baked into this, we're so locked into this, or only this point solution does this particular strategy or technique for us or solves this problem for us. We just can't rip it out. How do you integrate different point solutions without having people actually need to pull back from the software or the solutions that are providing real value?

Yeah, I mean, well, first of all, we try to create our own solutions that are all unified together so we can replace all of these point solutions. Wow. The average Ninja customer replaces 7 point solutions. Wow. And I think our whole key— we're a product-led company. Yeah. You know, our thesis is to build the best product, best possible product. So we think we've built some of the best possible point solutions and unifying them together not only saves, you know, IT teams, you know, hassle, complexity, having to manage all these solutions, But the nature of having multiple point solutions is there are gaps between the solutions. Yeah. And those gaps, you know, not only are inefficient, but they create vectors of attack and risk. So by unifying those, we eliminate those gaps. Now, if there are great point solutions that we feel like we don't have, you know, an answer to, we integrate them into our platform. So our IT teams that use Ninja have one pane of glass you know, fully integrated. Hopefully they've removed, you know, 10-point solutions. They're using Ninja to take care of all of their problems. And if there are one or two-point solutions they love that, you know, they can't replace in Ninja, you know, we have integrations for a lot of them.

You have some huge customers: Deloitte, Porsche, Hyundai, Carnival Cruise Line. They're all companies that people know. Kawasaki, Birkenstock. Um, but at the same time, you have 40,000 organizations on board. So what does the long tail look like? What's the smallest customer you're working with? Like, how are you positioning yourself in the market and winning deals?

Yeah, I mean, people ask me, who is Ninja's customer? And I'd answer, it's every business in the world, literally, is a potential customer. Wow. So for small companies, we sell through MSPs, which are managed service providers. They may manage,, you know, 6 local accounting firms and 2 law firms and a pizza shop. And, you know, all the way up to the largest enterprise customers. You know, we've got some very large, you know, Fortune 10 customers we can't name. But so we can literally service everybody.

Yeah. What was the question? Talk about the sword.

Oh yeah, the sword.

Game of Thrones. Thrones fans. So for those of you who watch Game of Thrones, uh, that's a— do they have flash—

do they have flashbangs in Game of Thrones?

I don't think so. Rolling flashbangs. We have flashbangs here on the show. I'm a big video gamer too, so they've got flashbangs. What games? What games? We play everything.

Uh, I actually started, uh, a couple companies that were video game companies. No way. My co-founder is really big into Call Call of Duty, so we've got—

they got flashbangs in Call of Duty. They also got smoke grenades in Call of Duty. UAV. I want to— are you a—

more of the strategy games?

Are you, are you a Counter-Strike? Oh, you're more strategy games. Okay, wait, wait, 4X or, or, uh, like Age of Empires, or are we doing Hearts of Iron?

Uh, you know, I've done Age of Empires, you know, but even games like, um, Hearthstone. Hearthstone's great. Yeah, even my partner and I actually started playing Axis and Allies when we were kids. That's a great game. We always say Ninja is one big strategy game. You know, it's just a high-stakes strategy game, but it's still fun.

I love it. My business is so addicting.

Is it getting more high stakes in the age of AI? Is it because of security risk? Or is it a pure accelerant to your business? Like, how have you been processing the last few years? It feels like it's a great place to be with 40,000 organizations, 140 countries, lots of opportunity, but what is the shape of the impact actually been?

Yeah, one thing I'd say is the 40,000, that's 40,000 customers. I mentioned some of those customers are MSPs that service hundreds of customers. Wow. So we might have close to a million business customers. Businesses using Ninja. Whoa, that's crazy. But, um, as far as AI, you know, we see AI as a real tailwind, you know, that's been accelerating our growth and it's going to continue to accelerate our growth, you know, for 3 reasons. First, we monitor devices, uh, and we've seen over the last several years device proliferation, and I think we're going to continue to see that, you know, laptops, computers, but it's smartwatches, Yes.

Microsoft launched some like smart badge and like Microsoft launched a desktop device and like it's clearly leaking into everything.

Even fast food restaurants, you know, it's not only the computers in the restaurants, it's not only the point of service equipment, the order boards, but now the fryers are endpoints, the grillers are endpoints and we monitor all of those. You have IoT fryers and brooms. IoT and even, even more devices coming in. So the more devices, the better it is for us because that's more devices for us to manage. So, you know, that's one real tailwind. Secondly, we're incorporating AI into all of our products. So AI is making our product more efficient and more powerful and thus more valuable to our customers. And we're also coming up with a number of AI-native products that we just released, the vulnerability product. We have a patch intelligence product. And then finally, and perhaps more importantly, and we've seen in the news even today, that AI is creating more vectors of attack. So more vulnerabilities are exposed, more things are being— need to be monitored and managed because of those vulnerabilities, and they need to be patched quickly. And that makes our product even more valuable. Very cool.

Well, what, uh, where's the company based primarily? You said, you said you're, you're in LA, but is the company distributed?

I'm in LA, but we're all over the world. Our largest offices, you know, are in Austin, Texas, uh, Florida, and, uh, Berlin, Germany. But we've got offices in Australia, England, Singapore, Spain, Philippines, you know, uh, I don't know, Indonesia, a bunch more.

Global. I love it.

I love it. Well, thanks so much for taking the time to come talk to us.

Yeah, it's great to meet you, Chris. Thanks for coming on, and congrats on a massive milestone. I'm sure you'll be back on very soon.

Yeah, I don't think we hit the gong. Tell us about the funding raise.

Yeah, we got it, we got it.

How much did you raise?

So we raised $400 million in Series C extensions at a $12.3 billion company valuation. Fantastic.

Here we go. Doubling the valuation since February 2025. Thank you so much for taking the time to come. Not bad at all. Have a good rest of your day. We'll talk to you soon.

Great hanging out, Chris. Remarkable ability from Chris to be completely unfazed by the flashbang. And you just didn't even— and the smoke grenade. Just like, I guess it's just another day. Another day.

Another day at work. Yeah. Where should we go next? Pat McAfee is negotiating a new deal that can pay him over $60 million a year from ESPN. This is in The Athletic. Big news for one of our biggest role models. ESPN and representatives for Pat McAfee are discussing an extension to his contract that would pay him more than $60 million per year. Deal's not completed yet, and if an agreement can be reached, it could be a sliding scale based on his new responsibilities. McAfee, already omnipresent, could be on the air even more with a bigger role in NFL coverage. I'd love that.

And last night he had a like watch-along, like on actual TV for the Knicks game.

That's super cool. Yeah. And he's been brought into the college football ecosystem and been a really important voice there. He's 39, 2 years remaining on his current contract. And he's a college game day panelist and has various appearances on other, on other programs. The makeup of the new deal would be similar in structure. ESPN's view of the arrangement as a production contract and a separate talent agreement, differentiating it from the deals with most of the on-air personalities. Hosts a 3-hour daily show with his crew, The Pat McAfee Show. The first 2 hours are on ESPN, all 3 are on YouTube. Very interesting split there.

Let's head over to Verizon CEO Dan Shulman to talk about some of the advantages he's getting. He and Verizon are getting out of AI. Okay.

Let's see. In the last 3 months, we've been experimenting with agents that are replacing some of our customer service reps, and those agents their customer satisfaction rate is 1,280 basis points better than, than what we had before. 1,280 basis point increase. That's a, that's a, that's a 1,000 basis point lift. Think about that. 1,200 basis point lift. That's 12% if you're putting it in normal terms.

It really does sound way more than 12%.

It sounds like 100 times as much because the number is 100 times bigger. But as a reminder, if you— Dan Schulman, the CEO, was in the Wall Street Journal recently for preaching straight talk about AI and job losses. This is April 19th. Uh, and he claimed that within the next 2 to 5 years, unemployment in America would be 20 to 30%.

So what is that, 2,000 or 3,000 basis points?

Yeah, 3,000 basis point unemployment rate. Uh, he warns that advancements in humanoid robots could upend the manual labor jobs still seen as safe today.

It is notable that, that people outside of AI actually have the most aggressive near-term predictions around employment?

Yeah. Yeah. Well, I think a lot of that is that the actual idea of job losses and whatnot, that came out of like the most insider-y tech, like the LessWrong, the Dwarkesh Patel podcast, where people were having very very mature conversations with a lot of caveats, a lot of like, oh, here's my probability distribution for if this happens, then this happens. And if we don't do anything about it, then that will happen. And in reality, there's a lot of things— well, oh, OK, well, if the unemployment rate spikes, we'd cut interest rates, right? Or like, we'd do like more federal jobs. Or, you know, there's a lot of things that could happen. But those quotes were taken out of context. Where it was like, in isolation, if only the bad stuff happens and none of the good stuff that would mitigate it happens, you could land at— and then the original thing was Dario saying 50% of early-stage white-collar work, which still puts you at less than 10% unemployment rate. And then he revised that down to 10% unemployment more broadly, which is still lower than this. And then people took it and were like, oh, well, if the smart guy who just put up a 10x annual run rate rain year said it, I should go a little bit further.

I got to 10x that again.

I got to 10x that again. And then you wind up with— we saw that Ice Coffee Hour clip of someone saying, what was it, 80% of jobs? I think he said 80%. I've heard 50% multiple times. And then Luke Belmar, that influencer, said, oh, 100% of jobs. And so, like, you just get to see—

It's notable, though, it's never their job. No, no. It's never their job. It's always— everyone else.

Because you need to be scared and you need to come work for me.

And get—

buy the course. Quit your job now, come work for me for half what you're making.

One more post before we bring in our next guest. Meredith says it's so easy to spot AI writing because it always sounds like it was written by a fedora. Silly.

Let me tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web app servers, databases, and more while Railway automatically automatically takes care of scaling, monitoring, and security.

See if the Ethanator has.

And without further ado, let's bring in Alex Heath. He's the author of Sources: The Ethanator. Welcome to the show. How you doing, Alex?

Always scooping, not afraid to use AI. Do you ever throw on the fedora when you're gonna—

Oh, he has that baked into the prompt. Don't talk like a fedora. It sounds great.

Don't talk like a fedora, guys. Although I am excited to deliver the first Fable newsletter later today or tomorrow.

But you had the leak or the scoop yesterday. You were the first person I saw in the media confirm.

In sources we trust, gentlemen. In sources.

But yeah, and I honestly use sources as an example where like you could use all the AI in the world. You could say like, I don't even touch my keyboard. I just sort of dictate to the AI the things that I know. And then I, and it publishes and it wouldn't make the product worse, right? 'Cause like I'm coming to you for like facts, like new information plus your takes on like how you think about something and however it gets into Substack doesn't really matter to me. So anyway.

You're saying you don't like my beautiful prose? I love your prose.

No, I think it's great. I go through the Verge all the time. I think it's great, but I'm just saying like, But that's not why I subscribe because of like who you are and the access that you have and the insight, but I don't care like if it was you clacking the keys. You know, it's funny you bring that up.

I've been having some conversation with Open Claw pilled folks recently, and they've been telling me to build something like it. I shouldn't maybe even be saying this on air, maybe I'm giving away some alpha, but build a kind of real-time MCP hook into my brain essentially. To just deliver infinitely personalized, agentically delivered media and reporting. What do you guys think about that?

Wait, actually personalized?

So like— Well, it would be personalized by the person's claw agent, what have you, right? Because the idea is the way that I decide to present a story is one way, but there may be things in like an hour interview that I don't even put in the story that may be hyper relevant.

Yeah, especially if it's delivered agentically. And also that actually does, that does feel like the more modern instantiation of the sub-newsletter list. Like you go to Bloomberg and you're like, I want the markets newsletter, the tech newsletter, but not the politics newsletter. And with an MCP server or some sort of AI intermediation, I could say like, give me the WWDC scoop, pass on whatever other scoop is less interesting to me. Like you'd know all my preferences, and then you can pump out a lot of different stuff and find very high delivery rates based on what I'm actually interested in at a particular time.

Yeah, I mean, I have a throughput problem of getting reporting into the world, right? Sure. And so it's a good problem to have, but what it means is like WWDC yesterday is a great example. I went to that tech talk that Craig Federighi and all the Apple execs did. I broke out some stuff that seemed interesting to the widest possible audience, but You know, there's a lot there. It was like an hour of stuff. And like, what if you just had direct, you know, IV drip access, you know, for your agent to my reporting and the rooms I'm in?

Yeah, I don't know. We'll see. Yeah, yeah. I mean, I'm sure you'll be the one building.

Good time to be experimenting.

Anyway, uh, on WWC, what, like, what was your reaction? What was your expectation going into this? And then, like, what were the biggest announcements that stuck out to you specifically? Because I, yeah, I mean, that's really what I'm coming for, is I want, I want your interpretation, even if you give me, even if you run an LLM to clean up the final language, I want your taste applied to this question.

I think they finally have a Siri that works. It's arguably what they should've shipped 2 years ago when they had that big trailer with, I think it was like people on an airplane and it was doing all this agentic stuff on the phone that was like, "Wow, it's crazy Siri can do this." It didn't do that right. They had to basically apologize for it. They rebuilt the Siri stack to be, I think their word was expandable. So more modern, more AI native. And now, you know, actually this morning I was in Cupertino for a hands-on with the new Siri. So I got to see it in depth and ask questions about it. And you know, it is a step forward, but there's basic things about it that if you're, if you're Clawpilled or you're using Codex or in chat all day, you're still gonna be like, eh. For example, it doesn't have persistent memory. So it kind of forgets everything in each query. They haven't built that yet. I wouldn't have expected that.

Hold on. I wouldn't have expected that because it feels like that's a source of, like, memory is not like the number one feature that people benchmark the current models around. They're like computer use, MCP integrations, and like, can I set Codex or Claude code to run for like 9 hours and it comes back with something really polished and just does, writes amazing code. The memory thing, it feels like personal superintelligence, feels very Apple and they have all of this. That's very surprising to hear. Anyway, what else, Tucker? It is.

I think they're moving very slow and they're very worried about the privacy aspect of all this, which I don't know if people really care about this with regards to AI. They seem to think they do. I think we'll see, maybe there'll be there will be a Cambridge Analytica moment for AI and then that will wake everyone up. I don't know, they're seeming to think that will happen. Yeah. There was a demo you brought up, Computer Use, where they were showing the new Siri and Apple Intelligence powered by, you know, Gemini and all that stuff on a Mac renaming folders and renaming files inside folders. And you know, if you use Computer Use and Codex or Claude, Cowork, like you can do this, it's very easy. You can abstract it away from a prompt, You can't actually even do it via a prompt in Siri. You have to do it via the menu, right-click, have Siri do something, or do it through Shortcuts, which like, I don't know if you guys use Shortcuts. I don't use Shortcuts.

The last shortcut I used was to remap Siri to ChatGPT actually, like 2 years ago. In 2023, in January of 2023. Yeah, it would take whatever I said and fire it off into the API and spill it back to me. It was sort of cool.

Yeah, I mean, I feel like one of the challenges for Apple is like software is moving more quickly than ever. And the Apple kind of ethos and approach is this annual release and like, let's make everything perfect and all this stuff. And it just feels like a really, really tough challenge to be able to— and it's not to say that they need to be at the frontier of everything. Right? Like, all anyone's been asking for is just like a usable version of Siri. Like, I haven't prompted Siri— I haven't gone to Siri for anything in—

I use it to set timers. Set timers.

Yeah, that's the weather. That's a good— that's a good use case, weather. But even then, I'm so— but, but anyways, um, but getting better.

So the other, the other big thing about this new Siri is, uh, they really nerfed the ChatGPT partnership. Yeah. It's, you know, that was a big thing a couple years ago, right? Yeah. And you have to prompt it each time to send something to chat and you have to do it via text. So there's not even like a software UI element of this anymore. So I was actually thinking about like meta narratives for WWDC this year. And one is that, you know, the Google partnership obviously, which we could talk about. But the other is I really think there's this, OpenAI is kind of like the new Facebook to Apple where they, they work together closely. If you guys remember when Steve Jobs and Mark Zuckerberg integrated Facebook and iOS, Facebook had, you know, a bunch of its, you know, press challenges, privacy challenges, et cetera. And then they just kind of drifted apart and then it became this like messaging war. And you started to see that bubble up yesterday with Craig saying stuff about like, we're not gonna have ads or just these subtle hints that they have, like ideologically drift away from, from OpenAI and ChatGPT and that partnership is not gone well. That was another big thing that, that stood out.

So on the, on the actual integration, that's somewhat, I mean, somewhat to be expected, right? Like you can't like hire— maybe you can't, you can't hire, you can't hire like Johnny and say like, we're going to, we're going to create a bunch of devices and poach a bunch of people and then expect like it to all be Good. And in so many ways it makes sense for a Google and an Apple.

'Cause Google and Apple have been working together perfectly for years at the search partnership. And this feels like a search partnership.

Enemy of my enemy is my friend kind of thing. Yeah. I think so.

They know Google well. And a bunch of Google people have come into Apple to reboot the AI strategy. Yeah. And the whole point of that very unique, which Apple's never done this by the way, an on-the-record like post-keynote, thing with their execs. It was very unique. We got in this tiny theater, got to ask Craig and the execs questions. He did this, like, I called it AI— in the newsletter I called it AI 101. He had this like almost whiteboard-like graphic where he was explaining AI stacks, how they work normally, how Apple's works, and really doing a lot to try to show that yes, they're distilling on Google, but they really control everything. It's insanely post-trained on everything Apple wants. It's not gonna give the same kind of responses that Jim and I would in a Google Surface. And I think that was the point of that entire thing was to just say like, yes, we're working with Google on AI, but we are not actually using like the Gemini product.

I wonder if it's gonna have its own stylistic flourish.

We gotta do a salami bench where you tell Siri, people, if you tell Siri, or you tell Gemini that you've had a kilo of salami, it will tell you to dial emergency services, so. 999. You've tried this personally, Jordi? No, no, but there's been a— some chatter about it.

Yeah, I wonder if it'll have its own turn of phrase, like, "It's not this, it's that," or, "You're absolutely right." There's like all these different phrases that each different model sort of like leaves its touches on.

What about there's only a certain set of devices that the new Siri is gonna work for? I think there's—

It's anything that Apple Intelligence runs on right now.

And then the latest models can be do, or the latest phones with 12 gigs of RAM can do the, like the highest tier. And then beyond that, you go into the Apple private cloud, which is NVIDIA GPUs in Google Cloud Platform, which is like their own thing. Have you dug into like where in Google Cloud it's going? Because they just did that deal with SpaceX for Colossus. Oh yeah. And that feels like not perfectly in line with like Apple's commitment to environmental considerations necessarily?

I think that's gonna be a way slower ramp than people realize. Okay. So to start, the new Siri is only in English. It's not in China and the EU. Okay. And they're doing a beta invite waitlist approach for—

Oh, this makes so much sense. I think through the year. Because I was looking at like a billion users. If a billion users are even just asking for set a timer and it's doing inference, like that's a ton of tokens. You've seen the ramp at Google as they've stuffed everything everywhere. Like, they're just gonna wind up with a big token budget at some point, but it sounds like it's a wall crawl run strategy.

I think it's a slow ramp. They were showing us some of the new photo stuff, photo retouching, removing things in photos, and it would take it like 25 seconds to change the backdrop of an image. So they're doing cloud inference, but they're, again, it was like, even in the demo in Apple Park yesterday on launch day, it was like slow. When there were bugs because they were like, "Yeah, our servers are melting. We're getting way more interested in this than we thought." So I think they're gonna take a very slow approach, but that was actually the number one question that people were asking was, "What are the inference costs of this? How is Apple doing this?" And like, "How big are they actually going with GCP?" But I think it's gonna be a slow ramp.

Yeah, because I mean, if you look at the like leaked financials from the AI labs, like we're not in this paradigm of like, training is the only cost. Like inference does have cost, it does CapEx associated with it. And Apple's been telling a very beautiful story of like, hey, we're sort of an AI winner without any of the CapEx. But if you start offering inference, you're gonna have a Google Cloud bill, you're gonna need data centers, you're gonna need clean data centers that take longer to build. And all of that is gonna show up at some point. But I think the hope is that they're just not jumping the gun, they're not, stretching themselves. They have the cash, they have access to the debt market.

What's the latest on Liquid Glass?

Oh yeah, Liquid Glass.

You know, they barely talked about it, and, and, uh, they didn't talk about the Vision Pro really either, which is where you see Liquid Glass kind of at its fullest. I love that product set certainly on the backdrop.

I know all that out. I'm the biggest user. Oh, you are? I love the Vision Pro. I watch movies in it.

You use it on a plane? Uh, yeah.

But mostly like— You're that guy. I'm that guy, yeah. I know, I love it. It is the anti-brain rot device for me. I put it on and I'm not on my phone and I just watch a movie. I watch Citizen Kane. It's a movie that would be like, you don't wanna watch Citizen Kane on your phone. You don't wanna watch it. And even if you watch it on the TV, you're gonna get up, go get popcorn or like check your phone. Somebody's gonna text you. You're gonna—

He's laughing at you. 'Cause it's a slow—

You seriously watch Citizen Kane on your phone? Finish on my Apple Vision Pro. Lawrence of Arabia. Lawrence of Arabia. It's a 3-hour movie. It's a long movie. It's hard just to make the time to sit down.

Do you mind if we pull up an image of you on a plane using a Vision Pro? Do you actually have one? Ben has one.

Wait, really? No way. No way. Um, yes. Yeah, let's pull it up.

Yeah, you can pull this up.

This is hilarious. When did this happen?

This must have been a year ago. Yeah, yeah.

Were you in the ad? The Apple ad? It should have been.

I'm a believer. Put me in the ad.

I'm going to be in it. Pull it up, boys. There we go.

There we go. Oh my God. Locked in. I'm that guy. I'm that guy.

Boys, are we even in business in this? Like, what is this seat?

I think that could very well be— This was economy premium. I think this was years ago.

Economy premium. Okay.

Oh, was that my last dollar? I spent my last dollar on the Vision Pro, okay? Okay. I don't know what's going on in this photo.

This is a long time ago.

This is fun though. It's incredible. Anyway, they did launch a feature for Liquid Glass. You can turn it off now. There's a slider that's basically an off switch, which is very funny. But you know, listen to your users, I suppose.

Yeah, it took a backseat. I mean, everything was all Siri. I've actually never seen a keynote so focused on one product. Yep. Apple usually does this State of the Union across all the platforms and like you get 15 minutes on CarPlay, you get 15 minutes on Vision Pro, all that stuff. It was basically an hour of Siri.

Have you been tracking any backlash to spatial reframing? Spatial, that's what it's called, right? Like, because—

I saw that up close in my briefing. Yeah. Oh, people like were booing there or something?

No, no, no, I didn't.

There was no boos. I'm saying I was in a briefing where like a product manager was showing it to us and letting us play with it. Yeah, yeah. It's pretty cool. I think it's cool. Yeah, I mean, Apple is like, they said like, you know, we're not gonna let you put a dragon in a photo. Like we wanna try to stay real to the essence of a photo, but we know that there are ways that a photo could look better with AI, and whether that's reframing or whatever. But yeah, they're not gonna do, you know, Midjourney level, Gemini level, Nanobanana level stuff. Yeah. But it is this whole, you know, idea of like, what is a photo?

It's devastating for the dragon community.

Yeah. Yeah. Yeah, if you're into live action role playing, LARPing, or the Renaissance Faire, like, You're an Android user. Yeah, you're an Android user, probably. But yeah, I mean, it is interesting that like there's a backlash to AI. Apple was like, they missed AI, but like the silver lining there is that everyone sort of sees them as like they're holding the ground, they're saying no to putting AI everywhere. And it's like, actually they maybe just didn't have the people in the right places at the time. Like they totally would've loved to do it. They actually acquired an AI company in 2011 called Siri for hundreds of millions of dollars. They were ready to go and they just sort of like missed on the rollout. But now I did see like a small bit of backlash to the spatial reframing saying that like, I know Apple as the camera company, what you see is what you get. There's not, the images aren't overprocessed. And so this is like, you know, they're leaning into AI too much, but I don't think it'll go anywhere. I think it's pretty moderate.

You can already remove things in the Photos app with the retouch tool or whatever. Can you?

Like this is— Can you? Have you tried to remove something from the image with the iOS 26?

Clearly you have and have not been able to. —so I guess maybe I'm—

It looks like a content-aware fill attempt from like a few years ago. It is not at the level of any of the Frontier AI image models. The new version is flawless though. It is good. So they fixed it.

It's not flawless. They showed us doing it to a boy who was covered, like his leg was partially covered, and they took something away and it made his leg like super long, like really awkwardly long. Maybe they're promoting hypermaxing.

Leg extension surgery is very popular. Together these days. You never know. Subtle messaging. Well, speaking of that, I wouldn't want kids getting body dysmorphia from editing their photos on an iPhone inappropriately. What did you think of all the privacy parental control talk? It felt like that had a remarkable amount of airtime in a relatively short keynote. Why do you think they're doing that? What was the outcome? How was it received? What do you think comes from it?

Yeah, there was this rumor in the industry, a lot of the bigger apps in the App Store, I was hearing it from their executive last year that Apple was going around and suggesting that it was gonna charge, like basically create a new service line to verify it. So basically like have these developers somehow pay to be a part of this system. I don't think that ever materialized or at least they haven't announced it, but I mean, you see the regulation on this topic and how it's heating up everywhere in every, you know, big country. And it was kind of funny, like the day of the keynote, you know, every year there's protestors outside of Apple Park around the keynote. One year it was about like, it was around the actor's strike and there were, it was like a services-related thing.

Tim Cook's comp too, I think.

Yeah, one year it was around that.

This year, this year was around kids' safety. So there was like these child safety advocates that were out there with their bullhorns, like saying this stuff. And then Apple went ahead and did what I would consider like the fullest, biggest suite of kids' safety features of any platform on that very day, which I think kind of shows that maybe it's just an issue they see really bubbling up and becoming really huge and trickling down from Meta and the social platforms and those huge trials that we've seen in recent months. Totally. To the platform layer. Yeah. And you know, all the apps will say, oh, we want Apple to verify, you know, the age because it's a lot easier than all of us doing it on our own and not having a shared consensus. So I think the industry is actually loving this because it, and it depends on how much, if Apple's gonna charge or not, but I don't think they are, at least in what they announced, but it seems like a win-win. I mean, the device should be verifying age. It's an obvious thing they can do. They already have the secure enclave. They can store it safely. Why would they not? So I don't know. I actually really thought that was a good move on their part, and maybe they are just trying to get ahead of things too.

Kids, get ready to learn Linux if you want to browse the web without your parents' approval. That's always been the case. I think this is good. I think it's good to have the parental controls and then the kids will hack their way around.

Did they talk about ads at all besides that small shot around that they took? Because they're bringing ads to Apple Maps. They have ads.

Yeah, you got to imagine that they'll put ads in this thing eventually.

Yeah, why not? The idea that, that they really can't take the stance that ads are universally bad or they would have to self-identify as a bad actor. Ooh.

Mm, I mean, when you look at the new Siri app, right? Which I looked at this morning and played with, it is a very, very bare bones, I would say almost knockoff of ChatGPT. So a lot of surface area for ads. A lot of surface area for things. I mean, there's barely anything in there.

I don't think they'll do it anytime soon. Like the idea, like they're putting ads in Maps, so you will search something and it will pop up, like the first result will be for somebody that paid to be there. And then to be like, oh, well, we would never put ads in, you know, in a web search experience. It just, it doesn't really, it doesn't really make sense.

Well, also working with Google, the largest advertising company on earth, to help supply your AI strategy. Yeah, there's, yeah.

I mean, one of my initial guesses that for like the longer term is that eventually the dollar flow would flip. So right now, Google pays Apple for the default search in Safari to be Google because they monetize Google searches with ads. And I—

They give them a percentage of revenue. Yes. Very, very important distinction, but yes.

Yes, and it's huge. And so my prediction was that eventually Siri would be so integrated with Gemini that there would be ads in the Gemini stream and Gemini tokens would be monetized because the cost would drop and the monetization would increase. And eventually Google would be paying Apple for the right to be in those query streams that then are monetizable with links or affiliate links or ads or whatever they want what they're gonna do, but I think all of the big tech companies are being, you know, walk, crawl, run on this, but so maybe it's interesting.

I thought that too, John, but I don't think that's the strategy at all. I think Apple sees, I think what the smartest thing I heard, I think it was Dave Morin, he told me I could quote him, shout out Dave Morin, I saw him after the keynote, was that, yeah, shout out Dave, was that Siri is the harness. So Apple's OS and Siri are the harness, and they've built this flexible new infrastructure underneath that could be a future model. They were hinting but not saying explicitly in all my briefings that there will be future models that you can choose from or plug into this. And Google right now is their best partner for a myriad of reasons. We've talked about some of them, but they wanna do all the post-training, they wanna do the reinforcement learning, They wanna make it Apple-y. And they have their foundational model strategy and they're not at the frontier, but I think they want to be eventually and they're getting that flywheel going. And I think this is potentially just a short-term thing to them owning their destiny on the AI side more. I think they feel like they finally have the pieces in place to get there. It's still gonna take probably a couple years. As we see with Meta, like rebooting a lab is no small feat. You have to get the flywheel going. I think they're on that path now and they feel good about it. But I thought what you said was correct, right? That it was gonna be just Gemini tokens that they monetize like search. And I don't think it's going that direction at all, actually.

Interesting. Would you expect Apple to make an AI-related acquisition at north of $1 billion in the next year?

You were looking at thinking machines a year ago?

You know, I thought they would've already done that by now. They looked at Perplexity. Yeah, they looked at Thinking Machines.

If they got this far, what, what are they?

Yeah, so they have this web knowledge graph now that I think was what the Perplexity acquisition would have unlocked for them. So they already went and did that. So I don't know what value Perplexity brings them now. I don't know what Thinking Machines' strategy, frankly, even is, uh, for it to make sense for Apple, much less like what Thinking Machines' strategy is. So yeah, I guess, um, I just don't get it. So I, I think I think they could make an acquisition, but it would be talent-based. I don't think there's anything that's gonna be fully resetting the strategy or like foundational, like an Oculus or something like that. I mean, I think that moment's kind of passed. It feels like, it feels like all these companies are huge and mature and you're not gonna buy Anthropic, you know, even if you're Apple, you can't afford it and they wouldn't sell. So you have to go out on your own.

Yeah, yeah, makes sense.

Yeah, and if you look at the history, it's like, Beats was like, I think, their biggest acquisition ever, but it was like pretty conservative from a revenue.

What did that do? It got them some like Dr. Dre ads.

Well, yeah, but it's just like they just— their style is like, let's buy, you know, let's do like acqui— like effectively acquihires, get really great people on board. Yeah, but the issue is that all the, all the great teams get marked up so quickly in the billions now that they get kind of they're no longer that great of candidates for acquihires, 'cause the investors on board are gonna say like, "Yeah, I gave you guys $200 million to take a big swing, just take a big swing." So how open do you think, to close it out, how open do you think the Siri ecosystem will be over the next year?

Because I'm optimistic that the base Siri model will be good on launch, But even if I put aside the ChatGPT relationship, I imagine that Gemini is gonna launch 3 new models that are better in Q1, Q2, Q3. You know, they're gonna be launching much more rapidly. And if I just wanna switch over the default model to something that's more frontier, I saw in the Siri app I could select from the dropdown dropdown. It was unclear if that was a default or something that I have to do per query. And then there was a big question in my mind about will I be able to effectively remap the side button to just say, hey Siri AI fine-tuned on Gemini 3.1 or 3.5 or whatever, that's great, but 4.0 just came out from Google and I want to use that when I hit the button. Do you think that'll be an option?

I think they wanna get there to some degree, but the question is how much this harness needs to be custom fitted for a specific model and how plug and play these things are. I mean, we just saw with Fable today, right? Like this is kind of a new paradigm in a model. And would that work in a harness like Siri as it exists today? Yeah. You know, maybe the future of these models is actually being built in, you know, sync with the harness that they're meant to exist in. Yeah, yeah. So I don't know if models will exist like search engines in the sense of like plug and play like that. The products are just so integrated and models are changing the way we think about operating systems with agents, right? So could you, maybe to some degree, I think they will do something like a half measure there, but I don't think it's gonna be quite as full-featured as you think. But you did kind of hit the nail on the head, which we can end on this, like the new Siri, it's up to the developer to opt into this new Siri in the sense of it can see everything on the screen if it's like basic text and photos, but it can't understand a UI unless the developer opts in, which means you're basically letting Siri take all your data. If you're Canva, it's like you're giving it up to Siri. And it lives on device and they say it's private and Apple won't train on it, but you're still kind of fundamentally doing that, right? And so that's gonna be the big question is will the Metas of the world, will the OpenAIs, the Anthropics open their apps up to Siri of their own accord because the users demand it? Or they think it's a good business strategy, but then you're giving that data, that valuable UI to Siri. So I don't know, that's gonna be the next thing to watch, I think, in the coming months.

Very fun. Well, good time to be a subscriber of Sources.

So go subscribe if you haven't already.

It's always a great time.

It's always a great time, but it's the best time ever.

Every day is the best time. Love catching up with you, Alex. Always great.

Great to see you a couple weeks ago. Enjoy the West. Safe travels. Wills, we'll talk to you soon. Have fun. All right, talk soon. Goodbye. Well, whether you're bullish on Apple or bearish about Apple, go express it on public.com. Investing for those who take it seriously. They got stocks, options, bonds, crypto, treasuries, and more with great customer service. And our next guest is already in the waiting room. We have Rob Schroeder from Vinyl Equity. He's the co-founder and CEO. Rob, how are you doing?

Welcome to the show. Hey guys, uh, thanks for having me. I appreciate it.

Thanks for hopping on the show. Appreciate that. Uh, first time on the show. Why don't you kick us off with an introduction on yourself and the company?

Yeah, so my name is Rob Skoder. Uh, I am the founder and CEO of Vinyl Equity. Vinyl Equity is a transfer agent for public companies or soon-to-be public companies. Um, we've been around since 2022 and we just raised or just announced our Series A raise today. How much?

Get that gong. How much did you raise? $20 million.

We're now backed by JumpCap. Fantastic. Okay, explain transfer agents like I'm 5 years old.

Yeah, it's a good question. Uh, most people have no idea. So, um, you could think of, uh, transfer agent as, uh doing for the public markets what Carta has done for the private markets. And so, uh, when a company goes public, they have to have what's called a master security file, and it is the basis of their equity ownership. And so transfer agents for a long time, uh, have maintained these records. You could think all the way back to the beginning of Wall Street, uh, used to have somebody who would take a paper stock certificate, trade on Wall Street, and they'd walk all the way down, um, to the transfer agent to record that transaction on their ledger.

It's interesting, so I just bought a, I put my whole life savings in SpaceX, and the guy said we didn't need to use a transfer agent, and then I saw him on Instagram, he's in Miami, and then I think he was in South America or something.

Yeah, and the stock certificates look like Monopoly paper dollars.

I'm realizing now I might have needed a transfer agent. You should have called us.

How have they modernized to date? We're no longer walking paper stock certificates around the trading floor. Historically, they haven't.

Yeah, long and short of it. So I actually started the business after running portfolio operations at AngelList, and we did, I think, 200+ IPOs in the couple of years that I was there out of the portfolio. And every time we worked with a transfer agent, it just felt like we were sending paper around, uh, all around the country just trying to get our shares into a brokerage account. So tons of market exposure post-lockup. And, you know, as you look at the market today, it has changed drastically in the last 6 months. I mean, even just in the last 6 months, you've got tokenization, you've got these massive IPOs, that have cap tables that are substantially more complex than they've ever been before. And user expectations have shifted. And you guys have talked about it. You're buying SpaceX shares down in Miami from people on the street. But you're also expecting to get those shares instantaneously. And, you know, we live in a world today where you can get packages delivered in less than a day. And yet from the transfer agent to your brokerage account, by and large, you can't get your shares delivered in 24 hours. And so Vynyl was built to solve that problem effectively.

Quick question, the name, why Vynyl? Transfer agents keep records. Oh, okay, yeah.

We spent a lot. Vynyl's booming. We spent a lot of time trying to figure that out. Yeah, that's right, yeah, we're back, baby.

Yeah, so with, yeah, The SpaceX IPO is insane also, just because of all the layers.

There actually are ironically a lot of unwinding.

And part of why solving this is so important is like, you know, the way markets feel like they're, at least in this, whatever part of the cycle we're in right now, such violent movement, like 24-hour delay could meaningfully impact, meaningfully impact your returns. How are you guys, like, what does your go-to-market look like at this point? Like, where do you guys kind of like take the baton? I don't know if that's the right way to think about it.

Yeah, we generally come on board with clients as they're getting ready to go public. So they're thinking about their IPO, in the context of an IPO. And so if they're making a switch from, cap table administrator to a plans administrator, we think that's a really good time to align. But there are companies, depending on their readiness, that will bring on a transfer agent, you know, 2 months before an IPO. We help facilitate the transfer of that information seamlessly. We do use APIs. We are an API-native business. And so it is fairly seamless and straightforward for us to ingest those records. When we get our company stood up, we actually give them a staging environment so they can go and test all of the data and make sure it's correct. And that's an anomaly in the world that we live in. And so, you know, for a very long time, this has been a just-in-time industry that's operated on paper, that's operated very reactively. And we want to pull all of these motions forward so that companies can go into the public market markets with certainty. I think what you guys are referring to is the liquidity process and the transfer agents involved in the liquidity process. I talked to a family office the other day who told me that on average they see, they see shares show up post-lockup 9 days after the lockup is lifted, and those shares will often see 10% delta. And when then the pricing, when from when the lockup was lifted. And so our ambition is to make a more fluid capital markets from the point of issuance to delivery. And by being proactive and having more nimble systems, we're able to do that.

A little bit of a somewhat related question. Bill Gurley hit the timeline earlier last night, last night, 8 p.m. He said if someone is telling you you are subject to a lockup agreement and your firm signed nothing with the lead underwriter. They are either lying or misinformed. Lockups are a silly contract with the underwriter, primarily to help engineer a secondary. Prove me wrong. All correct. He turned off reply, so you can't prove him wrong necessarily. But he said also it's never been enforced. The underwriting bank would have to sue for breach of contract. He asked AI the best example of lockup breach, couldn't find one, try it yourself, find a counterpoint, please share. I want— can you explain what he's talking about? Do you have any idea what he's getting at? Is there any truth to this, or do you want to push back and prove him wrong?

So, so lockups are, um, are, are agreed on with underwriters for the purposes of smoothing out market entry and smoothing out volatility, for all intents and purposes. And so, um, as he refers to it as sort of an abstract or created construct, it is, right? It is agreed on with the company as a way to manage the inflows and outflows of securities into the markets. The vast majority of companies don't see the kind of demands that we're gonna see and probably the kind of volatility that we're gonna see over the course of the next several months. And so you see some creative constructs that have been proposed in, you know, the IPO that we're going to have on— we're going to see on Friday, and then presumably some of the other IPOs coming down the pipe.

So— And then also the company is signing a contract with the underwriter, the bank, and the company probably does have pretty solid legal recourse over the employee stock option pool, maybe the founders' or current executives, but what he's talking about is being an early-stage venture capital investor who maybe didn't sign many contracts that relate anything to lockups, and so he doesn't necessarily need to sign on. Of course, a good underwriter might go around and say, "Wait a minute, there's a VC who wants— who's heading for the exit, who owns 20%? Like, let's bake that in, get them to sign a contract." But he's saying, If I didn't sign anything, I'm selling on day one, I guess, something like that.

Yeah, by and large, you're gonna find that liquidity is pretty tightly held or gated by the transfer agents and the brokers. Sure, sure. And so, and when those lockups are agreed to, they tend to get managed pretty tightly.

Well, thanks for helping break it down. I know it's a hairy debate, but it helps contextualize these things. I really appreciate it. Jordi, anything else?

Thank you for modernizing investment capital flows. Thank you for everything you do. It's really important work. Thank you. Yeah, yeah, for sure.

It's an area that I think nobody's really thought about for the last, I don't know, forever. I'm glad you have. And we're excited to be thinking about it here.

So hopefully everyone will get their shares more effectively. Yeah, good problem for the AngelList Mafia. I'm sure those 200 IPOs that you went through were, Uh, you know, uh, stressful.

But well, thank you so much for coming on the show. We'll talk to you soon, and congratulations.

Congrats to the team.

Appreciate you guys having us. Goodbye. Um, what else is going on in the timeline today? Uh, the system prompt already leaked for Siri. Is this real?

Hard to tell if it's real, Siri. But you know what is real? What is real? A Colorado man has been arrested alongside his passenger for allegedly driving more than 130 miles an hour through Northern Colorado. He was doing a cannonball run, John. And I wanted, I'm just gonna assume that this is someone in the audience. They're probably not watching right now because they are in prison. But yeah, it is, cannonball is not for the faint of heart.

It is crazy. That people are still going for the Cannonball record post-COVID because all the records were smashed. So if you're not familiar with the Cannonball, it's a coast-to-coast race. You race from the Red Ball Garage in New York City to the Portofino Hotel in La Jolla, I think, in Southern California. And it takes something like originally like days and days and days. People did it on motorcycles. People have done it in all sorts of different things, electric cars, self-driving cars. But the true record was, you know, sort of no limits. People go as fast as they can, very dangerous, very illegal. You have to wait for the statute of limitations to lift before you share your run or share the video or else you'll be held accountable for what you did. Very controversial in the car community too. Some people love it, some people don't think it's safe and for good reason, it's very dangerous. A lot of people bring iPads and Waze and different maps and different radar detectors and all sorts of things. Some Cannonball record holders actually had friends who piloted planes that flew in front of their car. They're going 200 miles an hour, planes going 250, and looking for cops ahead of them. So that was the level of detail. There's multiple stopping points. They plan out the gas, they have a fuel tank. In the back, a fuel cell that holds an extra 50 gallons of gas or something like that. So you can stop less, a 2-stop, a 3-stop. These are the techniques. I'm roughly correct here. But during COVID no one was driving. No one was on the roads. And this was before people were like, oh, well, I can't go to a crowded building, but I can go camping. So I will get on the road. But even the trucks were down. Or off the road. And the trucks are the slowest part, because you get stuck behind a truck, you might not be able to get around. And so all the Cannonball records were smashed. Ed Bolian from VinWiki participated and did very well. I think he held the record for a while. Maybe he still does. But to go after it today, in the modern era where technology is even better, the cops have better radar than ever, there's cameras and all sorts of things, truly one of the boldest things you can do. So good luck to him on his legal fight. Maybe he'll be using an illegal AI tool to get himself out of jail.

I don't know. Harvey, don't make mistakes. Maybe, maybe. Anything else? Well, folks, we will be back tomorrow, tomorrow for a massive show.

We got to talk about the Trojan horse for the Odyssey, the popcorn bucket, the TBPN effect. We have a giant horse which might be a Trojan horse. We never opened that thing up. We don't know what's inside.

That is true. Can you imagine?

But the Odyssey has released a limited edition Trojan horse popcorn bucket to commemorate the Odyssey. You can get it. People were fans of the sandworm popcorn bucket. This doesn't feel like it came from the mind of Christopher Nolan. This feels like some producer, some money man came up with this. I like it. I think it actually looks cool. I don't know how practical it is to eat your popcorn out of this Trojan horse, but it is very fun. I like the popcorn's hidden inside the horse. I'll be picking one of these up. Production team, how is the process of obtaining tickets to The Odyssey? I've been seeing that it's very stressful. Hasn't been going well? We got them. We got some. You got some? Where are you going? Where are we going? You going to the Chinese Theater? Chinese Theater IMAX. That's a good one.

And then you guys got tickets for for like the whole team, right?

Because like, no, just production teams. Wow. Okay, I see. Oh, little division here. Okay, we'll solve it, we'll solve it out. No, we gotta, we gotta get everyone out to the Odyssey. You should get your friends out to the Odyssey. Uh, call everyone, buy a bunch of tickets. They're refundable on Fandango if you do it that way. And then the Coogan method. Get, get 20 tickets. Text everyone you know who's in town. Tell them, be here at this time. You'll get some, some churn. Not everyone will show up. You refund those tickets, then you charge everyone after the fact. It's a good strategy for having fun with some friends, touching grass, remember, reminding yourself of what's real. It's important. Stay sane out there. Leave us 5 stars on that podcast, Spotify. Sign up for our newsletter, tbfan.com. See you tomorrow! Goodbye!