Today is Wednesday, September 9th, 2026. You want to tell us how many days we got till Christmas?
Cuz we are live from the TVP Ultradome, the Temple of Technology, Ports of Finance, the Capital Capital. How many days until Christmas?
106 days until Christmas.
Countdown really makes it feel like it's going faster. I feel like just the other day we were at—
It's flying by. Start shopping. Just do it.
Yeah. And you actually have time to, again, like we talked about yesterday, remove your flooring, floorboards, all that stuff,
and just dig straight down and install a permanent Christmas tree. Don't just get ready for Christmas 2026. Be ready for next year.
You also need to start thinking about gift giving. It's harder than ever to find good gifts. The best gift you can give someone is helping them save time and money. Get them on Ramp.
Time is money. Save both. He used to use corporate cards, bill payments, accounting, and a whole lot more. Mom, I got you a corporate card that you're going to love.
Anyway, pull this up. Pull this up.
iPhone Duo is going to be under a lot of Christmas trees. But what's more important than the new iPhone?
Pull up this graphic. Foldable.
Pull up this graphic. I was hoping that Apple would announce this today. If you zoom in,
The tree is actually—the house is actually built around the tree.
Stop building a home and adding a tree to it.
Yeah, and find a Christmas tree and build a house around it. Yes. That's the Christmas spirit.
Our mutual friend John sent this to me this morning, so he's on top of it.
Cool. Let's do it. Well, Apple has announced the iPhone Duo, their foldable phone. We'll pull up some video and react to a little bit of it. I—
How is this going to impact screen time? Does your screen time go twice as fast because you're actually locked in on two screens?
Should you do pixel-adjusted screen time, too? Like, if you're watching a laptop versus an Apple Watch, if I spend an hour on Apple Watch, is that like 15 minutes on the laptop? On the triple monitor setup, getting triple the screen time? We have the AirPods with the cameras.
They haven't announced those. I don't think they're going to be reporting,
but when that comes, that could potentially help you add to your screen time by tracking screens out in the real world that you're seeing that aren't getting picked up.
True. Yeah, digital billboards, that should go in your screen time. Eventually, it's going to be 24/7 screen time. Everyone's looking at a screen everywhere. There's a screen everywhere. You're looking at the windshield of your car; that's the screen. Can we pull up some real pictures of the Duo? Here it is.
Here it is, the Duo. In this video, there is no crease. I'm sure the YouTube tech reviewers will be hunting in all different lighting for creases, but it feels like Apple waited long enough to avoid a crease whatsoever. So, you will get no crease in your—
That is crazy. Sitting there. Boom.
I think this is going to sell like hotcakes. I think people are ready to spend. It's going to be like two grand, maybe three grand when you max it out. I think the Apple buyer, the Apple fanboy, is ready for something that actually is a flex beyond a slight upgrade. You paid 30% more for the Pro. This says
you pulled out the big bucks. You dropped two grand.
Well, and that's been the thing that—
relative to how much value you get out of a phone, phones have always been incredibly cheap, even the highest-priced.
You spend more time on your phone than you do in your car, and your car is 10 times the price of your phone, if not more. And so I think this will sell very well. It'll be interesting to see if any new apps come from this, if people need to redesign stuff. Like, we're having Mikey from Suno on the show later today, and Suno has an app that's really full-featured for iPhone. You can make songs, you can scroll a feed of other people's songs, but you do have to pop over to desktop if you want to get into their composer and actually deconstruct the song and shift around different pieces of the song, and then you get a sort of GarageBand interface. What's his plan? Is he thinking, okay, this is going to sell enough that I need to re-engineer my app for a different flow? Even using a lot of normal iPhone apps on iPads, it usually just doesn't feel quite right scrolling. So I think there'll be a couple UI/UX opportunities, challenges both. Interesting little bezel there where you can see the screen is curved, but it's on a blocked-out, squared-off edge there in the center. Two cameras on the back, one on the front. Yeah.
Leaving room for the iPhone Duo Pro with five cameras or something like that.
Yeah, it really shows that the phone seems to be—at least the iPhone is certainly in its final form, right? This feels like the biggest step forward in a long time, and yet it does absolutely nothing new for you.
A rave review. Until the new paper that folds out. It's moderately exciting because everyone's addicted to their phone and it's like, cool new screen, but it does nothing new for you. It doesn't move humanity forward in any way at all. Jordy, you're going to be able to watch two Reels at once and they're both going to be vertical.
From the demo that I saw, it looked like you put the video up top and then you iMessage.
Well, that's because that was a horizontal video. I don't think you'll be able to open two copies of Instagram at the same time.
No, I think it'll be more like the iPadOS where you can have different windows.
So, you think you'll be able to do YouTube Shorts and Instagram Reels side by side?
Ideally, brain rot final boss over here wants to do it,
and then you can have the third one on the back.
Well, it's definitely going to be better for looking at charts on Public.com investing for those who take it seriously. They got stocks, options, bonds, crypto, treasuries, and more with great customer service. What is Apple actually doing today in the markets?
Brian says this could literally 2x GDP.
Does the market like this? I mean, all this has been, you know, priced in, but let's check on Apple: down 1.27% today, down 5% over the last five days, but sort of flat on the month. So it's exciting to see John Ternus get a big win early, first thing as the CEO, first major Apple event, and it's an entirely new product, the foldable phone, something that's been rumored for years. They didn't tee him up just to do a minor refresh.
Exactly. Yeah, it's exciting that he got to come out strong with this. And it's also less of a risk than the Apple Vision Pro. If they'd thrown him there, then he'd be remembered like, 'Oh, his first thing was sort of a flop,' even though I don't—
Yeah, I think this is going to be a hit.
I think it's going to sell very well. I think a lot of people are going to buy them for a variety of reasons. But I think it's just a good way to kick off a really strong decade-long run, hopefully four decades. Cook couldn't make it to 95 in the seat, but Ternus, I think you've got what it takes. I want to see Ternus running this place.
And so you're 100% getting the Duo.
Yeah, I think I'll get it. I'll need to feel how it feels in my hand. But at one point I was considering carrying an iPad mini as my daily driver phone because they come with Wi-Fi.
6'8". So I have big pockets, and so it would actually fit in my pocket. And I don't like the phone feature of the phone; I don't like getting phone calls. So you lose that because it doesn't have a cellular connection. So you can never call me,
and it's not like you need a phone number for anything else besides calls, like logging into pretty much any service.
Yeah, you can do two-factor on an actual authentication app. I think you could get by. You might need to have a Mac Mini hooked up somewhere that has a phone number attached to it that can route you two-factor codes or something, but it did seem doable. Or, you know, you just check your actual phone with the phone number at home later, but every day you just have this iPad mini that you pull out, and it has a cellular connection. I don't know that it can actually get phone calls, but it has a cellular connection so you can get data while you're out and about, so you can be out in the world and scroll.
Real data. What else do you need? You don't need to make big data.
Yeah. I mean, it really comes down to the fact that the phone is lighter,
It's not lighter. It's not thinner. It is probably faster. It's bigger. This is the opposite of your meme. Typically, they come out and they say it's lighter, thinner, faster, thinner, thinner, faster, thinner, faster 12 times. It's the best iPhone ever. This is actually a different direction. This is a response to your joke.
Would you rather search work in iMessage or be able to get the Duo?
Search work in iMessage. What do you think?
Oh, you want that to work? Yeah.
Personally, I would be excited about functional features of the core software that I use with the iPhone.
Among many other things. But Kate over on X says, 'Mentally ready for my iPhone to start being slow, laggy, and having a bad battery.'
You have that to look forward to.
So they also updated the normal iPhone, is that right? Did you watch a little bit of the earlier keynote?
That one is thinner, lighter, faster, better, stronger, all the different things.
They updated Apple Intelligence a bunch. I thought they haven't launched it yet. I thought this was the launch of the Siri, like it's actually going to go out now.
Correct. Yes. Yeah, this was in the launch video.
And also the new normal iPhone 18, they were like, 'Oh, the battery is much better.' So all those things that Jordy said about the new iPhone.
Which one do I have? Because I got mine last year, I believe. I have a 17. So they launched the 18, which is the same form factor, just lighter, thinner, faster, better, stronger.
Okay, but looks exactly the same.
From what I—I'm sure some people can tell the difference.
But yeah, better chip or something. So, I don't know, maybe you'll notice it with Apple Intelligence. I wonder if there's any gated features, because I know that there were initially some gated features for the Pro models. You need more memory for the on-device stuff. I don't know. It'll be interesting. I'm excited to get it in hand, see how it works. It might be too much to open that thing up. I'm thinking it's pretty easy just—
For you, it's great because you're such a newspaper guy.
Put your feet up on the desk and fold your phone open.
I'm scrolling. Yeah, something like that might make sense. I was reflecting on Steve Jobs. I saw a viral video about Steve Jobs predicting ChatGPT in 1985. And the video is incredible because he arrives to this talk—he's giving a talk at a university on the future. It's just Steve Jobs' view of the future in 1985. And he arrives on a helicopter with an entourage, and they all get out of this helicopter and they go into this. It's amazing.
King of the castle, king of the castle.
Because the video kind of takes place in a castle. But we can probably play this video here. Let me pull this up. I'll send it in, because you've got to watch Steve Jobs in 1985 predict modern AI pretty well. What is that sound? Is that us or is that the soundboard? You've got a new vibe chatter in the background. Anyway, while we pull that up, let me tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web app servers, databases, and more, while Railway automatically takes care of scaling, monitoring, and security. So yeah, let's play Steve Jobs predicts ChatGPT in 1985, because I had a bunch of thoughts based on—
Aristotle. And when I read this, I became immensely jealous. And I think I would have enjoyed that a great deal. And—
So, pause. What is the second video that's playing? I think you guys have two videos playing. He's saying—I'll just read the transcript. He says, 'Did you know that Alexander the Great's tutor for 14 years was Aristotle? When Alexander the Great had a question, he would just go to Aristotle.' I think we can play the clip from the beginning now. Let's see.
who Alexander the Great's tutor was for about 14 years: Aristotle. And when I read this, I became immensely jealous, and I think I would have enjoyed that a great deal. Through the miracle of the printed page, I can at least read what Aristotle wrote without an intermediary. And maybe if there's a professor, they can add to that, but at least I can go directly to the source material, and that is, of course, the foundation upon which our Western civilization is—
blazing the printing press—
But I can't ask Aristotle a question. I mean, I can, but I won't get an answer. And so,
is that in our lifetimes, we can make a tool of a new kind, of an interactive kind. And so my hope is, someday when the next Aristotle is alive, we can capture the underlying worldview of that Aristotle in a computer. And someday some student will be able to not only read the words Aristotle wrote, but ask Aristotle a question and get an answer. And that's what I hope that we can do.
It's so crazy. He's just like, we've got to compress all the data down and be able to give it a conversational interface.
Large language models with a simple interface that anyone can use.
It's a crazy prediction 40-some years ago. There are so many interesting details about that video. Steve Jobs is probably the most famous techno-optimist, the most successful techno-optimist. The 'bicycle for the mind' quote, that one continues to hit constantly. You can see his excitement every time he gives a keynote; he'll just be on stage giddily testing GarageBand. But his delivery there, too, is something really special. You notice that, right? Where he asks a hypothetical question, then, 'Well, I could ask, but I wouldn't get an answer.' And he lets the laugh line land and lets the audience—he has really—
Basically the Mark Gurman of CEOs.
And is the greatest founder CEO ever.
Yeah. It's just really cool timing.
Surprised it took you that long to—
Because I thought you were talking about Mark Gurman's comedic timing, which is fantastic, but—
Mark Gurman's delivery of Apple—
Truly, he truly has Apple Intelligence in his delivery.
That's true, that's true.
Ben Thompson has called this out a few times, but when Steve Jobs introduces GarageBand, he is just so excited about this product. You can see him in this one video introducing GarageBand 1.0 with John Mayer, which is really cool. He has John Mayer come out and just play some stuff on the guitar.
This is where I want to show you loops, and I'm going to show you what a—
And he's just chilling there on this computer, on this stool,
It's going to be me using this tool, something using loops. So we come—
and just doing a live demo. And if you scroll forward about halfway, you'll see John Mayer comes in and plays some songs. It's toward the end.
There you go. It's a British invasion.
He's just having fun. And with the power of modern LLMs, you can ask an AI version of Steve Jobs, 'What would he think about AI?' I did, and I think the answer is somewhat interesting. He said probably yes, with very strong opinions about what makes AI worth using. That's speculation, but his own words give us a surprisingly good basis for it. In 1985, Jobs imagined computers capturing a thinker's underlying ideas so that a student could interact with them. You could actually ask Aristotle a question. He was already enthusiastic about something resembling a central promise of conversational AI. But his enthusiasm would come with three demands in this hypothetical Steve Jobs analysis:
One, you've got to give people new abilities. Learning, designing, composing, building would probably excite him most. Two, you have to make it effortless. I'd expect him to push for AI that understands what you want with minimal instruction. And three, you have to make something excellent. I suspect he'd value AI's ability to help someone realize an original vision much more than its ability to produce enormous quantities of generic content. He would be anti-slop, but pro-AI.
Aristotle also had an interesting AI take in this imagined LLM version of Aristotle. He'd care more about what AI does to our character than how intelligent it becomes. The conclusion of this is Aristotle would favor AI as an instrument of human flourishing while resisting the idea that a life requiring less human effort is automatically a better life. You must continue to create. You must continue to work. You will be unsatisfied if the—
Create more than you consume.
Exactly. AI that strengthens our ability is what's important. And so what I was going back and forth with you about was it seems pretty clear that Steve Jobs would be excited about AI or some form of AI, but what would the reaction to that be in the modern era? Because I tend to think that there would have been the natural cycle of reaction because there's a whole host of things: political polarization, economic stagnation, wealth inequality, negativity bias on social media. There are so many different incentives that the impetus to tear someone down is unavoidable. And so I think we would be in this weird era where a lot of people would be excited about that particular vision for AI, and yet there would still be viral videos complaining about the latest Steve Jobs keynote. What do you think? Because you've called for the Steve Jobs of AI, someone who's an optimist, someone who talks about empowering humanity, focused on human flourishing, agrees with the Aristotelian philosophy around AI. But what would the actual reaction be? Is it actually a panacea, or would it fall by the wayside like all other attempts at positivity seem to have fallen by the wayside?
Yeah, I've come around to the technology itself being fundamentally scary, even though it's very cool.
And then you factor that in with certain actors that are trying to position it as even more scary than maybe it is inherently. And I think even if we still had Steve with us, it would still be a mess.
And he might be nicely counter-positioned, but it doesn't stop other people from poisoning—
Maybe not poisoning the discourse.
What do you think his p(doom) would be? Over/under 10%?
1%. He's the 1%. What did Joanna Stern have to say about the Apple launch? There's a video here you wanted to—
New feature called Live Rewind.
What is this feature? Let's play it.
Hi, welcome. Today's specials are an heirloom tomato BLT, Caesar salad, caldo verde lentil soup, chicken alfredo, the smoked salmon, and the avocado tart.
Just double-press the Digital Crown, and Live Rewind transcribes what was just said. You can ask Siri about the content of that transcript later. This is great for when you're trying to recall a book recommendation, a colleague's brilliant idea, or something you didn't hear the first time. The next audio intelligence feature will help you remember key moments from daily conversations so you can be more present.
So, is it just recording everything? Because that seems genuinely useful. Probably not in the—I mean, Joanna Stern, her take here is, 'I don't know, maybe just ask the waitress to repeat herself.' That's a good point. But if you're actually talking to someone, like your lawyer is breaking down a contract and you forget which part of the contract that the lawyer flagged as important for you to go dig into and get to the bottom of, having a transcript there makes sense. And you could ask the lawyer to repeat themselves, but it costs a thousand bucks an hour, so maybe you'd want to just transcribe it. I don't know. What do you think about this?
There's got to be some people out there that are thrilled and incredibly excited about this new feature, and I would be surprised if there's more than a hundred of them globally. It just feels like this whole event, and it's hard to be excited about anything at all.
A lot of these demos, they get harder and harder to feel natural and human. Like the GarageBand examples, it's so much easier to make it look cool because you're there with John Mayer, and you're imagining, 'Okay, I got a buddy who can play drums, but I can't play bass, so I'm going to get GarageBand to fill in the bass digitally, and I need to compose everything together.' I don't really have an alternative that's super logical, like just asking someone to repeat themselves. Anyway, it's got to be a tough job working through how do we showcase this in a non-Black Mirror—
@SemiAnalysis asks, 'How we feeling on the crease?'
Yeah. Are we feeling—I haven't been able to get my eyes on the crease.
No, it feels like whatever they did in the photography or—
That's the one thing you would expect for them entering the foldable market. You would expect them to absolutely—
There's not going to be a crease on this thing. I will be shocked if there's a visible crease whatsoever. Arthur from Vibe just resigned today.
It's obviously a joke. Read it out since you brought it up.
'I spent the last three years doing TV ad performance at both Vibe and Walmart. Neither company is acting responsibly. They are racing straight to self-improving TV advertising performance and gambling with our ROI. More thoughts below.' I like it. Very good. I only read the first line. Stupid. Apple foldable crease—let's see if there's anything here. 'This is how Apple dealt with the crease issue on the iPhone. This is definitely going to be the best folding on the market. I think passport size of it.' There are some photos. 'This is a canyon. It's not a crease; it's a canyon.' 'The crease in the middle isn't very noticeable. It's been handled well.' I don't know, we'll see. It definitely seems like they were waiting for the technology to become post-crease before they launched this, because foldable phones have been out for five years, seven years or something like that, and they've always been a little bit rough. The hinge gets dirty. This feels like it will hold up for at least one year to get you to the next one. What do you think? You think that's the plan?
Hinge lifespan of 11 months and 27 days.
I'm trying to find an image of the new Apple Watch, which I think they're calling the Apple Watch—what is that for?
It is just crazy to look at. It's got the Jubilee bracelet, right? It's got a really chunky metal bracelet on it. Anyway, let's see what the Gruberator says about the iPhone Air and iPhone 17 Pro. He said, 'Last year, an unplanned personal hiatus coincided with the time I should have been reviewing the then-new iPhone 17 and 17 Air. 354 days is arguably more than a few, but given I missed the window for reviewing them when they were new, I thought it might be interesting to consider the iPhone 17 Pro and Air on the cusp of there being obviated by the next generation of iPhones. So, the iPhone Air is the nicest phone Apple has ever made. Probably the nicest device Apple has ever made, period. Just so nice and loves it.' iPhone 17 Pro: 'I'm eagerly awaiting the Daring Fireball review of the new foldable phone, but he's got to get his hands on it. I've got to know what he's thinking about.' Yes, the Hermès Paris Apple Watch. Is this the Ultra? That is a chunky, chunky device.
Yeah, this is the Ultra 4.
It's horrific. It's horrific. Is that the dial or the bracelet?
It does have a very odd interface to the actual watch with that bar that comes out of the clasp. Like, the clasp doesn't actually integrate fully into the watch; it sort of has these two points that bolt on there. Pretty chunky. But honestly, the Apple Watch Ultra, this device had gotten so chunky that I think the chunkier band might fit better with the chunky watch, because if you put a really thin strap on that, then you just have this massive bulge on there.
Yeah. I almost think they should have launched it as an ankle bracelet. Imagine checking your super chunky watch by just pulling up your foot on an ankle monitor.
And you could be in a meeting just looking down.
Yeah, I've seen some guys doing that with their WHOOPs; they put it on their ankles.
Yeah. People are saying, Bobby Cosmic says, 'No pendant.' Imagine you have this skunky—
John Ternus was vagueposting before this at 6:00 a.m. He posted just a light, and it was maybe a little bit too vague. Was that it? It was just a light. It didn't really hint at anything. Was this a—
What's wrong, Ternus? Afraid you might say something?
Was this a teaser? Is there a folding thing here that I'm not picking up on? Is this an homage? Samsung Mobile US hit X and said,
'So far, so same. #It'sTimeToSwitch.' Throwing shots.
Using the hashtag in 2026. They're going to be so dedicated to the hashtags on that brand account that it's going to become ironic and cool again to use hashtags, and they're going to be in the perfect spot. Let me tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web apps, servers, databases, and more, while Railway automatically takes care of scaling, monitoring, and security. In the Wall Street Journal, they didn't talk about math. It made it to the digital version; didn't make it into the paper. Navier-Stokes did not make the paper edition of the Wall Street Journal, but Meta and their new personal AI agent did. It's right here: 'Meta Launches—'
The absolute state of traditional media.
Probably it's more profitable as a business perspective, right?
It's like, what are you going to do, sell math theories? Navier-Stokes was a $1 million prize. You don't think Alex Wang can pull a million bucks out of this?
This could be a trillion-dollar prize.
They haven't monetized it yet, but they've got to put ads in this thing. And will they do agentic commerce? What will they do? We were very excited talking to Sourenator yesterday about AGI for small businesses for folks who run ads on Meta. I still love that idea, but they're going a different direction: they're doing personal superintelligence. Zuck's been talking about this for a year at least. He's put out a lengthy blog post about how he wants AI to diffuse, how he wants everyone to have AI. And now everyone does with the Meta Muse app. That's a separate app that you get that will book restaurants for you. The Wall Street Journal has a little bit of a story here: 'Meta Platforms is betting that making AI agents easy to use will get mainstream consumers to start adopting them in their daily lives. The company on Tuesday announced a new personal AI agent that can buy goods online.' I think they partnered with Stripe on Link. So, I don't know if that just means that when Link is enabled on a particular checkout, it'll be a little bit smoother, or if it's going to try and steer you towards things that are linked, right? Because some websites do it; it's really nice when they do. But certainly getting closer to this: 'can respond to emails'—so do you have to link it to your Gmail, I guess, to get that working—'and complete other tasks users authorize it to do for them. The agent, called Muse, is accessible via a dedicated app. Users operate the agent by conversing with it like a typical chatbot and can choose their own names for it.'
I saw some really funny ones. Tyler called his Mark. I saw someone else gave it a really funny name, like they said, 'You can pick one,' and then the name of the agent was just 'You can pick one' or something like that. A bunch of funny hallucinations always happens with these products; everyone's going to have fun with them. 'The new tool is Meta's latest attempt to monetize its massive AI investments, which are expected to top 130 billion this year. The AI agent is free to access for most of what people need and has two tiers of subscription plans for power users: 100 bucks a month and $20 a month. Anthropic and OpenAI have quickly built multibillion-dollar businesses around agents capable of performing complex computer tasks, but most of that usage is driven by software coding or other business applications. Consumer adoption of technology lags behind, offering Meta, Apple, and other companies potential for catching up. "The fact that we've given such a powerful technology in a package that is going to be really digestible for the product is one of the things that will set us apart from lots of other agentic products," Meta's AI chief Alex Wang said in an interview.'
Ben Thompson was raving about the fact that it gives you a pretty powerful VM: 8 gigs of memory and 8 gigs of storage that is yours, sequestered. And they're also implementing a lot of the security features that they have done with WhatsApp where there's end-to-end encryption.
Yeah, they got Moxie pulled into it, apparently.
That's really cool. Moxie is a legend, so great on the recruiting front. Everyone loves Signal, and great team there.
Yeah, the question is how—it sounds like they're actually giving every user a computer.
A VM, it's virtualized, but yes, serious computing resources.
Yeah, but the cost is going to probably be pretty significant.
So the VM most likely spins up when you use it and then gets taken down when you're not using it. So the actual compute hours that are associated to anyone but a power user—and of course very quickly you can wind up spending 20 bucks or 100 bucks, and that's enough to just pay for a virtual machine. You can go and rent a true virtual machine to host your website on DigitalOcean for like 20 bucks.
So Kaz, the CEO of Opendoor, posted a screenshot of him asking Meta AI in WhatsApp, and he said, 'Please make WhatsApp the main interface for these things, Alex Wang, please.' And he just says, 'How do I use Muse?' And Meta responds, 'Muse is a free AI assistant from Anthropic. You can use it on the web, on mobile, or through their API. Go to claude.ai, create an account, and log in.' So, I'm looking at this post, trying to find why is nobody talking about how Meta AI thinks that Muse is a product from Anthropic and is telling people to go log in there?
That is weird that it would hallucinate that Muse is from Anthropic, because there's not even any connection between those. It's not like Anthropic ever had a product called Muse that I know of. Very odd, but Meta AI, of course, is running an older LLM, and then I believe you still get thrown into a Llama model when you're on Instagram if you wind up in an LLM chat. And so there's going to need to be unification around all of these different products, and they're going to need to eventually talk to all the same models. But of course, there's scaling and all sorts of different issues, because some of these products that they've gotten to scale are pretty significant. Even if they're not the most talked about, when you are a hyperscaler with billions of users and you vend an LLM into something, you've deployed that at scale; it's doing a lot. It's like the new Gemini 3.7 Flash: a bunch of great reviews, great benchmarks on things, but they can't just stuff that in Google search on day one. They have to actually scale it up.
Yeah. The question I have is, does this seem like this was something they were working on prior to Instinct gaining traction, or is this an Instinct clone? I saw one post that was saying the rumor that they tried to buy it—
Ruroc says, 'There's a rumor going around that Meta and others made a 10-figure offer for Instinct. That's a crazy valuation for a one-year-old company, something like that. Instinct turned it down.'
Now Meta has launched something like that.
And that's literally what the post is: 'Meta has launched Muse and now the timeline is reacting. It'll be a big deal. Sounds familiar.' Yeah, I saw some people saying now the people with distribution are coming in. It'll be interesting to see what distribution looks like in the Meta ecosystem, because Meta was able to win Stories. How did they do that? They didn't win by launching a Snapchat clone as a separate app; they won by putting it in Instagram, which already had the install base. And so if Muse works its way into Instagram, which a lot of people do use, it'll probably get a lot of adoption. If it's a separate app, it's a little bit trickier. Now, the Threads playbook worked: they routed a lot of people to there, Threads shows up on Instagram, it has an advertising surface area, it's pulled over. What are they at? A couple hundred million MAUs?
Hundreds of millions MAUs, and again, it's hard to fully understand the value of that because—
They they're counting somebody who clicked through clickbait on on in—
Yeah, there's that and then also—
—but there's a really robust, there's a robust, like real community of just pure Threads users. The other, the other thing that I think about is like Meta has the advantage of being able to, when you open all of their apps, just put a huge billboard. They make ad inventory just for themselves with the specs where you open around Christmas. Sorry, sorry. Just a little little slip there. But when you opened Instagram in Q4, it's like buy some of our, you know, classes, right? And so I would expect—
I don't think they're going to do that until Muse has like—
Yeah, they'll get the early adopters on and figure out all the kinks and stuff.
Says Meta has been working on Muse for almost a year now. You have to believe that that's true. The Manus acquisition ties into that. The Zuck messaging around personal super intelligence. I mean, when we interviewed him a year ago, we were asking about, oh, am I going to be able to use your AI to actually buy, if I see a pair of shoes, will I be able to just buy it? And he was kind of like, I'm not ready to announce that yet, but it was very clear that that's where this was going. So, I don't think it's an Instinct clone. I think it will compete with Instinct. And I think it's tricky because Instinct is pretty early and they are going up against a big company that's been investing and has a lot of resources to go after this very, very hard. At the same time, consumers come to Meta to be entertained. Muse feels like a product for productivity, and I'm not sure that people will put them in the same buckets like that. It's very unclear. At the same time, people do a lot of shopping on Instagram, basically just scrolling and—
Oh, for sure. Not not actually checking. Are you joking? What? Like—
It's like it is like the most dominant shopping app of all time, that positioned as—
—but particularly people will like go there and not just search things, but just like casually scroll and like linger over things so that they get retargeted, visit websites, and then have the have the ads come to them. And so you could imagine that that on the shopping side that there's a lot of traction there.
Yeah, a lot of excitement around this climbing the chart. Muse is currently number three on the charts in the App Store after ESPN Fantasy Sports and more and ChatGPT.
Growing Daniel came in with an absolutely brutal ratio. He quoted Zuck's announcement with a screenshot of the exchange that Zuck had with somebody during the early days of Facebook where he said, "Yeah, so if you ever need info about anyone at Harvard, just ask. I have over 4,000—"
—emails, pictures, addresses, and student numbers. "What? How'd you manage that one?" "People just submitted it. I don't know why. They quote-unquote trust me, dumb f***s." Anyway, Growing Daniel never forgets.
Obviously, Zuck has grown up a lot since then.
Who punched him in the nose? That's the question Ryan Petersen is asking.
Let me find the—you sent me the image. Is this true?
Yeah. So here's the thing: his nose is starting to look like this guy, Merab, who's a UFC champion. And Merab has one of the most busted-up noses in the UFC. And Merab was the UFC fighter that Zuck was training with on that deck. That's a lot of people are asking. So pull up this picture of Merab, and you can see their noses are starting to look—
That's a really that's a really good sign that—zoom in there.
He's pushing. Look, their noses.
Zuck was like, "Give me one of them right now."
"Give me a nose like that right now." I like it. Do you think so?
And I would expect, who knows, Merab could go and get probably surgery to correct this one. It's a very cool battle scar—
—and two, this tells me that Zuck is going way harder in training than he lets on, which I love to see. He's like, he's ready for that Elon fight. He's like, "There's a chance I might catch Elon on a good day. He opts in, I got to be ready."
He's also going extremely hard on the Irish countryside, buying castles. We call him Lord Zuck. We got a video now breaking down his latest real estate purchase, Strancally Castle in Ireland. Let's play a little bit of this because there's some good voiceover, some good context.
—castle, and it can be found on the banks of the River Blackwater in County Waterford. This Gothic Revival mansion was designed by the famous brothers for John Kiely, a local landlord, former Tory MP for Clonmel and High Sheriff of County Waterford. The house was complete in 1830 and its scale—
He's got a bunker under there.
According to records, it took a brisk four and a half minute walk—
—just to get from the kitchen to the dining room. Within the grounds there was a folly—
Chinese pagoda and an aviary.
The folly, which is known as Strancally Tower, was built in 1826—
We're having David Perell on the show later to talk about architectural revival.
—medieval, and it served as—
He's going to have the resources to actually restore this and do this. Some of these castles are cheap relative to how epic they are, but the carrying cost of actually renovating or living in, even visiting any of these places—just cleaning out the cobwebs is probably six figures a year, and the landscaping's got to be extremely expensive. But I think this is the answer to where's the tech wealth going. At some point, he's either going to get bored of this or donate it or pass it on, or it's going to fall out of the family because seven generations later the family's like, "Yeah, we don't really want this castle anymore." And boom, you got a beautiful museum that was properly renovated and brought up to modern standards in 2026 and is prepared to actually house a bunch of—
This other video. We're going to watch this quick video, and then we're going to talk about absolute doom and despair.
But first, this little palate cleanser.
What are we doing? Emperor, dude. Does that feel like that ever?
No. I wish that people did what I wanted them to do. It would be fun. But in the meantime, you can have some fun.
Yeah. Oh, yeah. I would have a chopper, I'd have a—I don't even know what I would have. I would have an underground tunnel.
I do have an underground tunnel.
Yeah. It's more of an underground storage type situation, but yeah. No, it's a—
Oh, wow. Like a damn emperor, dude.
I love that the Theo Von arc happened where for some reason all the tech CEOs were like, "It's a rite of passage now. We got to go." I'm disappointed we didn't get Satya on Theo, and I'm really disappointed we didn't get CC.
Zuck got a bunker. All right, let's talk about doom.
Doom. An Anthropic researcher is quitting the artificial intelligence industry. Whoa. He didn't just quit the company, quit the industry over fears that the lab and its competitors are racing to build systems that could spiral out of control and destroy humanity. This was not received well. A bunch of back and forth with another researcher chiming in saying that they have a p(doom) of 10%, which is something that's come up years ago and we kind of went back and forth on. This all started with Jacob Coxin who said, "I resigned from Anthropic today. I spent the last three years doing pre-training research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives." He is not happy with where things go. And it got a 100 million views and 600,000 likes. It's going straight on a Bernie Sanders billboard for sure. "I worked on an actual thermonuclear weapons project," Martin Casado from a16z says, "and the dissonance was less bizarre." And so he's comparing it to what it was like working on very scary technology, very different than what's going on today. And Evan Hubinger, who works on alignment science at Anthropic, previously MIRI, OpenAI, Google, Ripple, Yelp—I guess Ripple was the cryptocurrency, is that right? Isn't Ripple XRP? Yeah. He says, "Jacob is correct here. We really do earnestly believe AI could kill all humans. I personally think it's above 10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are clearly not on track to." So everyone's sort of pushing for a slowdown now, but then you go into the regulatory capture thing. I've been seeing what—
Jess Buckingham says, "Bro got that PIP."
Yeah. A lot of people are like, "Are you giving back your equity? Blah blah blah." Odd. Nick Carter says, "If you actually believe this, you'd be committing acts of violent sabotage and terrorism against your employer. You clearly do not, so stop it with this self-aggrandizing twaddle." Eliezer on the other side says, "Thank you for saying plainly what you believe. I do appreciate it, and the world could be even grimmer than it is otherwise." We have Eliezer Yudkowsky's co-author coming on the show tomorrow, Nate Soares, to talk about 'If Anyone Builds It, Everyone Dies.' We'll continue the doom debate. It seems like it's in the news. Lots of people were just going back and forth on everything. Ryan Petersen sees it as a bull case. He says, "I need to buy more stock in this company. I've never seen such a commitment to making money regardless of the consequences." Dario says, "If you publicly assign a greater than 10% probability to human extinction within a decade, you owe people a clear explanation of how you reach that number, plus what would change your mind?"
Yes, yes. A lot of people were upset that this felt like a vibe prediction, very unresearched. Evan did link to the Anthropic risk report, which you can read in full. They have a reasonable scaling policy and they're talking about pacing the frontier, collaborating with other labs. And I'm oddly optimistic that this will happen. Obviously, you know, Sam and Dario can't even hold hands on stage in that India conference, but AI has made strange bedfellows out of people. Like if you had said, "Oh yeah, in six months Elon Musk, the guy who's calling Anthropic 'Misanthropic,' will be doing a deal and bring Dario to watch rockets take off at Starbase," I would have been like, "No way. No way can Elon Musk and Dario become friends." That just seems like impossible. But the funniest outcome is the most likely, to quote Elon again. And yeah, maybe there's this weird—people wind up working together. There are other consequences of that. There are some very serious—
Just thinking, believing in your whole heart that things are that serious, and then just rage quitting and not—
—and trying. Like, do you really think that the rage quit method—
I mean, some of these organizations are really big, and if you're at a certain level, it can be very hard to actually get things done in a large organization. So you can be there and say like, "Yeah, I've been sending a Slack message and emailing the whole executive team every day that this is the thing that I want to have happen, and it's not happening, and I don't have the authority within this particular role to actually have the change." People sort of tend to think, oh, if you work for the company, you can exact your will, but that's not true. Like you can work in the marketing organization, think you want to run billboards in Tallahassee, Florida, and not be able to get that done because someone higher up than you said, "No, we're not doing billboards in Florida," right? Things get vetoed for a variety of reasons. So it's not like you just take a job at a 5,000-person company and then you can just do whatever you want. That's not the way companies work. It's a weird, weird time. We'll see. But there are serious regulatory frameworks for not allowing collusion between big companies. And so we might wind up in this other weird scenario where if all the big labs are like, "Yes, okay, the government's not stepping in regulating. We're going to self-regulate, we're going to work together," then that could actually be a legal problem because you're not allowed to price fix, you're not allowed to collude with your—that is like textbook antitrust, right? And then the other thing is that there's this weird dynamic I've seen playing out where the AI leaders make a prediction and then the outsiders criticize that prediction as like they're only saying that to raise the next round, or this was the same thing with all the labs saying like, "We want regulation," and then, "They're only saying that for regulatory capture." Regulatory capture might be a side effect of regulation, but the labs can still want it. And now I'm seeing people online look at what's happening and say we do need regulation, and I'm willing to put up with that regulatory capture that may happen under that scenario. And so there's all these weird predictions that happen way too early for everyone to digest because the people inside the lab see things much earlier and understand, oh yeah, okay, the next leg up is going to work, the agentic era is going to arrive, swarms are going to be a thing, we're on track to solve this. Anyway, Jordy, one last thing, and then we'll bring in—
Friend of the show Bowdo Capital quoted Evan with his 10% prediction and said, "Hey Dario, congrats again on the great quarter. Nice to see enterprise reacceleration. Hoping you could provide a little color on that 10% extinction number and some of the puts and takes that might have on further enterprise adoption, and then just a quick follow-up after."
Yeah. What are Anthropic earnings calls going to be like? It's just going to be the CFO and Dario, or are Daniela going to be on? I don't know. That's not that crazy to imagine that happens, that that's a thing that happens on the call. Very, very odd. I mean, I guess you do have defense tech companies that have earnings calls, or defense primes; they might get asked about geopolitical things that are pretty heady. I don't know. Anyway, we'll figure it out.
In the meantime, let me tell you about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI, own the data platform that powers it. Up next, we have David Cahn from Sequoia Capital to talk about the state of AI. Good news, there's lots happening, so plenty to talk about. Welcome back. Thanks so much for hopping on.
Good to see you guys again. You guys got a lot of range, from Zuck's castle to p(doom).
There's a lot of range. There's a lot of range. You think he's got a bunker in the castle, or do you think he has enough? It seems likely.
I think he's got a doomsday dungeon.
Dungeon. Okay, the doom dungeon to go with Zuck's bunker. Bunker's bunker. Anyway, what is on the top of your mind? Because there's so many different things happening in the application layer. We had Cognition on, Harvey just raised a bunch of money. Things are going really well with diffusion and the actual application of the technology. There's value, jobs apocalypse is delayed—
—and should p(doom) be factored into financial models for Series A investments?
Yeah. Broadly in AI right now, the thing that seems really interesting to me is just the concentration of capital going into these big companies. The thing top of mind for me is just the way the venture landscape has changed.
When you think about doing this job five, six years ago, you had a hundred interesting companies that everybody was looking at and you're trying to find the next one. That's still happening, and yet so much of the conversation today is about the top 10 companies.
Yeah, on that point, is the concentration of capital into OpenAI, Anthropic, maybe xAI, SpaceX, too—is that the biggest driver, or when you're talking about reshaping venture, I've been much more keyed in on the fact that you get Hugging Face and OpenRouter, and there are now many decacorn liquidity events that I would think changes the underwriting if you're a VC and you're like, "Yeah, it's an early round, it's a billion-dollar valuation, but there are plenty of examples of 10x returns at this scale."
Well, to me that's always been the case. We were early investors in Hugging Face, and I think there's always been this thesis on you have a big community and you have a lot of bottom-up adoption, and that there's a lot of option value in those type of companies. I think you'll see that get reflected back into the market where there's always these companies where investors ask a hundred questions about what does this become, and yet you just have so much sucking sound from the market that those are going to become interesting companies. It's so hard to be relevant and build something important, and I think Hugging Face did that and OpenRouter did that. And so I think there'll continue to be examples of that. The thing that's interesting to me just in this moment—and it's the fall of 2026, we're about to see these IPOs—is this precipice of moving from so much capital, and honestly this is not even like a tech market VC market thing, like the entire universe of capital has concentrated on this one sliver of the market, which is kind of pre-IPO late-stage AI companies, and we're about to see that transition into the public market. I don't think anybody knows how exactly that's going to play out, but it does feel like we're on the precipice of something pretty interesting.
It is wrong to think that it will play out as a binary one-day liquidity wave, correct? Because there will be lockups, and then many LPs will receive their stakes and want to hold. Some venture capital firms, registered investment advisors will hold as well. So we're looking at like how exactly does that change the market, because there will be the optionality of liquidity, but not necessarily actual cash on balance sheets ready to go into the next generation of startups.
Yeah. I mean, one thing I've been thinking about—you guys remember 2021, you had CrowdStrike and Datadog were public, and then you were investing in private CrowdStrike and Datadog, right? And there was Databricks, right? And you could go invest in Databricks and it was the same business model. You sort of understood how this thing was going to trade when it was liquid. And I think that's what's interesting about AI right now: there's no real public comps for any of these private companies, but as soon as one of these companies—let's say Anthropic is the first to get out—as soon as one of these companies gets out, then I think you actually have a better valuation framework for the rest of the market. And so I think it is a healthy thing for the market when you kind of can look at the liquid, you can look at the illiquid, you see how these things trade, and it will provide a lot of clarity, I think, to the private market.
Why is SpaceX not a good comp for Anthropic, OpenAI, and then even the smaller firms?
I think SpaceX was the beginning of this. Obviously it's held up really well, and for all the volatility—you know, every day you see a news headline that's like, "Stock was up 2%," and it's like stocks go up and down 2% every day, nothing happened, and there's always like some reason why it happened, by the way. It's like, oh, went up and down. SpaceX has held up well. There have been liquidity unlocks. Everyone was super worried going into the liquidity unlocks in the media, and nothing, it was like a non-event.
I remember the day of the big unlock, somehow traded up like 7%, if I remember correctly.
Like a short squeeze, where all these guys were short and then they got squeezed out. But so I do think SpaceX is the bellwether, and it's not, to your point, it's not going to be a one-day event, but I think it's an exciting time that these companies are growing so fast. I think there's a lot of demand for these companies, and it's going to be interesting to see that play out. I think the second thing that we're paying attention to, and I think a lot of investors are focused on, is just like the full stack of the data center buildout that's happening.
I wrote this blog post maybe a year and a half ago: servers, steel, and power are kind of the three pillars. Obviously, everyone's paying attention to the server side, which is chips. The steel side is the industrial side. You know, you have Elon now making his own gas turbines. So there's a lot of emphasis on the industrial revolution component of this, where we can't make more of the components that go into data centers. And then, of course, the power piece, where everyone's looking at, hey, there's no more power for the next 24 months, and what are the timelines? With these big capex projects, you have to plan two, three, four years out. Who even knows what the world is going to look like two, three, four years from now? So I think deeper below the cusp of frontier models is all the things that make them possible. And I think the market's pretty well educated on the compute market now, and investors are familiar with all parts of the stack, but I think there's going to be a lot of innovation all across that.
Do you think that the opportunity in the grid, opportunity in power plays out more like the neocloud market than the rocket launch market? Like I've always said, it'd be amazing if we had an Elon of solar or even an Elon of nuclear. And there's some nuclear companies that are doing really well. Some of those founders could go on to build the trillion-dollar energy company, but it feels like everyone's focused on it. There's so much capital flowing into energy. Default would be very oligopolistic, a bunch of great outcomes, but not this winner-take-all behemoth. But what's your model for where energy and grid buildout goes?
My mental model is probably closer to the Elon case than the neocloud case, and I can explain why. On the neocloud side, the primary innovation was financing. That's sort of how I look at it.
Real estate, private equity.
Yeah, exactly. Like, we're buying a bunch of chips from Jensen. And Jensen, to his credit, was an amazing player in the ecosystem and has helped really build the ecosystem. He made CoreWeave what CoreWeave is today. There was sort of this moment where CoreWeave realized we can buy these chips, you can finance them with Blackstone, somebody's going to finance the GPUs. There's a lot of financing innovation. Now there's a lot of companies that have followed suit. Obviously, you need to be a good dealmaker, but in some ways you could start a neocloud tomorrow. You know how to do it. Getting access to power is hard, it's not easy, but there's 15 neoclouds; there's going to be more. I think, by contrast—and my partner Shaun Maguire always has a great framework around this on the hardware side. He wrote this hardware manifesto long before hardware was cool, and he had this concept of, in hardware, there's potential energy. And these hardware companies spend five, six, seven, eight years building this potential energy, and it's because they're doing real, very difficult R&D work. At some point, that potential energy converts into kinetic energy, and that's when stock prices move. You see that with SpaceX, which took 24 years to bake, and you see that with some of these hardware companies. And so at Sequoia, as we're looking at the energy ecosystem, we've sort of chosen a few places to make investments. Shaun was on the show talking about Valar, which is our nuclear investment. Last week we announced our investment in a company called Form Energy, which is doing grid-scale batteries. That's a company founded by the guy who ran the energy business—he actually started the energy business at Tesla—a guy named Yet-Ming Chiang, who is probably one of the top three battery experts alive today. So just a very, very strong team. And they're nine years in. They spent nine years building these billion-dollar batteries. They just signed up a big project with Google earlier this year. And then you ask yourself, okay, well, we're going to have a lot of solar and wind, are we going to have a lot of batteries on the grid? I think the answer is yes. And then you have a company that spent nine years building IP to go put those batteries on the grid, and then this unique moment to go scale that up for AI. So I think there's going to be, to your question, three, four, five interesting kind of companies with deep, deep IP that you can go invest in. But I think it's different from the neocloud space, where what you really need is capital and access to GPUs, whereas on the grid side, you ideally want some real fundamental IP that is difficult, and then of course you need access to capital. On the nuclear stuff and a lot of these companies, what's interesting is you need to raise billions of dollars before you get to the promised land. And so that alone—one thing I say to founders when I meet a lot of seed-stage founders in hardware is, even if I think your IP is amazing, I can't invest if I don't think you can raise a billion dollars, because you just won't get there and you'll end up getting consolidated. Access to capital and consolidating capital obviously is an important part of building these companies as well.
So what are you looking for? What are the tells of a founder who can raise a billion dollars? Is it great storyteller, is it just actual—
Isaiah is a great example even at the seed stage. Honestly, I look at the El Segundo crew and I look at Augustus and Isaiah, and their ability to sell you their vision was just on another level even before they had raised $10 million. When they were raising their first round, they were just different. They had a natural talent and then they've refined that, and you have to be in a category that allows you to sell this vision of a trillion-dollar opportunity, a hundred-billion-dollar opportunity. But I think you have that natural talent and selling ability that is somehow inherent to some people. There are people that are just way less compelling naturally, and they get more compelling over time because they get more confident in their business, and round after round and hundreds of pitch meetings and meetings with candidates, they just get better and better and better. But then some people just start at the 10-yard line, and they just refine it a little bit, and then they can go from raising a $5 million round to raising a billion dollars.
Yeah. In our partner meeting, we actually talk about this a lot. Naturally, this is probably the most important question that you have to get right in early-stage investing. And the framework we often use is: would you go work for this person? Because I think everything is sort of subsumed in that. Other people have to be willing to go work for this person. The decision of whether to go work for a person is effectively putting 100% of your equity in this one company, right? So would you put 100% of your equity in this company? And often we had Mateo Jaramillo, the Form Energy CEO, come in, and you have people come and say, "Wow, this person seems really serious. This person seems like they have an incredible track record." In his case, he had built the energy business at Tesla. They've done something that's really impressive early on in life. And there's a clarity of thought component. Sometimes pitches are coherent and sometimes they're not coherent. There's a coherent story, which is like: Form Energy is building an iron-air battery; iron's really cheap, air's really cheap, that is why they're building an iron-air battery. It's so simple and yet kind of makes sense. And then it's like, "Well, why do we need batteries?" Well, there's no sun out at night. It's really simple things. And I think something we look for in a lot of these conversations with founders is clarity of thought. If you can articulate to an investor who's not the most sophisticated—the investors are never the most sophisticated people around the table—if you can articulate a very clear vision of why this vision makes sense, why your plan makes sense, then you're going to be able to articulate that to an employee. And then we have people like Shaun who have PhDs in physics and can go incredibly deep on the science and technology behind it, and you sort of marry those two things. I think that's kind of how we make these decisions. And we spend a lot of time getting to know the founders as well. Mateo I had known for six years before we invested; Shaun knew Isaiah for a long time before we invested.
You're not making the decision in a single pitch.
How are you thinking about trying to time the acceleration of the real world? Right now it feels like it's much faster to do things on your computer, and it's creating efficiencies for a bunch of people and companies and all these things. But we're not seeing the kind of GDP growth that has at least been predicted. It feels like in order to have that, we're going to need to see, one, general advancements of AI models that can actually be a drop-in addition to your team or replacement or whatever that ends up being. But it feels like we're going to need to do a lot of things in the real world to feel that acceleration. And even with the launch of Astra—we haven't actually played this video on the show yet today, but somebody had basically hooked up Astra and it can just drive your car now, right? It wasn't something that the model was designed to do, but the models are getting sufficiently advanced at understanding the world. And it feels like we're on the cusp of this acceleration of the real world, which I think is going to be—it's a very magical experience to say, "Generate me an image of this scene in my head" and be able to see it on your screen. It's going to be 100 times more magical to be like, "Make this thing in the world," and then it just makes it, right? And that can be applied to real estate, to building widgets, the list goes on and on. But robots building other robots and building things, I feel like will be the greatest investment cycle ever and much more exciting for normal people than just, "Hey, we made this software feature in two weeks versus 20 weeks."
Two thoughts on that. One is I think it's so important to synchronize the timeline of your investment with the timeline of the market. To your point, you need to understand when the market opportunity is going to convert, and ideally you're synchronizing the timeline of the company with the timeline of the market. So for example, if you're trying to solve AI power and you're starting today, it's kind of late. We need more power literally in a year. So hopefully you've been working on that problem already for a few years. If you're trying to solve nuclear and that's a 2040s problem, great, you can synchronize the timeline of being an early-stage company chasing an opportunity that is really going to hit its inflection in the 2040s. So I think that synchronization matters. And then I think the second thing—it is interesting that VCs have really shifted toward hardware investing, right? There's this broader trend. Andrew came on the show and talked about Send. We've been doing a lot of stuff in the hardware reindustrialization space. Part of that is because of this feeling that the timelines are so fast in the digital world that it's hard to create moats and it's hard to create value. So I think there's a reason why in hardware there's a feeling that you can create moats over many, many years. And eventually, obviously, the ultimate moat, to your point, is going to be the machine that makes the machine, right? It's the reindustrialization of this country, and I hope we're going to get there. I think we're going to get there. I do think sci-fi kind of preempts reality, and Isaac Asimov obviously imagined in 'I, Robot' the most valuable company in the world is a company that makes all the robots. There's a reason I think a lot of investors own Tesla for that reason, because they think that he's going to win on Optimus. So anyways, I think there's a lot of companies that should benefit from acceleration in the physical world, and there's a lot of investment going in there.
That's awesome. Well, thank you so much for coming on the show, breaking it down for us. Excited to have you back soon. Always a good time. Have a great rest of your day. Appreciate the takes.
Let me tell you about Console. Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. We have Mikey Shulman, co-founder, CEO of Suno, here with us in the waiting room. Let's bring him in.
A GOAT. A GOAT. Since you came on the show last, we have all been completely one-shotted by Suno. Every time we come into the studio, we're blasting Suno songs. We have a lot of fun with them, make a song for our friends, whatever the news is of the day. We're having a great time. But how are things going with you?
Much better now that I know that you guys are enjoying the product.
Yeah. I'm going to give you a compliment sandwich.
You don't need to compliment sandwich me. That's fine.
No, no, no. One, the product is probably the AI product that has brought me the most belly laughs, period. It's just a joy to use. Daily user. We have so much fun with it on the team. I've pulled it up in situations in my own home that I think are hilarious to me, less funny to members of my family. But anyways, the product is just amazing. It's wonderful. I get so much joy out of it. I think the software quality is bad.
In the app. Like, I think that it—
Not on the model side. The model is amazing. The things that matter are so good. It's so unbelievably fast, and I haven't even tried the new model yet.
But the actual software, I think you've been—
No, no, no. It just feels like it's kind of buggy, and it feels like you guys have been vibe coding too hard. I want you guys to basically do some cycles on just cleaning it up a little bit.
Okay. More optimal compliment sandwich.
What do you think? That's a good compliment sandwich. Let me say actually—
The final bread, and it's amazing.
No, no, no. The actual quality of the output is the real level of listenable, whereas like video generators, you're looking at it and no matter how much effort... It passes the test of listenable so you can have fun with it. Anyway, continue.
I appreciate that. We should probably stop releasing models while you guys are on air so you can actually have some time to test them out. But yeah, I'll just take the opportunity to talk about what we're doing today, which—
is releasing a new model. Okay.
And so way better than the last one.
It's actually way faster. And so you'll see the the way that you can use it when the songs come back so faster. It's just very different, much more fun, way more expressive.
Faster? I don't know why it needs to be faster. It was already fast enough because I would type the prompt and I would click the button and then it'd be available immediately. And I think that's because you stream it in, right? Like it's still rendering in the background. Is that right?
No, we're still streaming it.
No, when I feel like it's slow is when I'm one-shot by a song, but I want to change a certain word in the lyrics and I'm like, this has you covered. That was the thing when we first started using—when we first got one-shot by Suno, we would fire off a prompt and put your phone down and be like, "I'll come back to it in like 5 minutes, see where it's at." But then it'd be back in like 20 seconds.
Yeah. I think the thing that people don't realize is that when you're building things in entertainment, actually the loop speed—how quickly you give something to your users—ends up mattering a lot. We can actually measure that pretty directly and we know how much it matters. But just going back to the model, so V6 is better, it's faster, the songs are way better, it's more expressive. We have V6 Wild, which is a model that makes just some of the most interesting stuff, even though the songs themselves may not be quite as good. I'm having a ton of fun with that one. And it's actually like a big day for Suno and actually for the whole music industry where this is done in concert, in partnership with some of the best music companies in the world. It's really kind of laying out a blueprint for how we think AI can really grow the entire music ecosystem. So we're super, super excited.
Yeah, we were talking about that. Yeah, we can go there. Sure.
Yeah, the partnership. Everyone's waiting for a partnership. Everyone's been interested in how the economics will flow, what happens to songs that are used in training. Obviously, there's a fair use argument. There's obviously the Spotify "I want a royalty on everything forever." And it feels like this is the start of moving the conversation forward at least one click.
This is definitely an amazing start, yeah. The way I think about this is yes, these are now not only new business lines and new revenue streams for rights holders, for artists, for all kinds of creatives. Most importantly, these are really long-term partnerships that let us build products together with the greatest music companies, with the best creatives. And the thing that's so exciting is that for the first time in a long time, there's actually product differentiation. There's the ability to build things that are just different. And I think we've been so ingrained in the last 20 years just to think about digital music as one thing.
And so we'll be releasing more products in the near future that let you play with the content from your favorite artists. When those artists opt in, when users play with their content, the revenues that are attributable to those users, a portion of that will flow to those artists. And then again, just building more from that in the future of what should the future of music look like? What are the most engaging fan experiences? How do we pull people deeper into music and change the place that music has in their lives, to change the place that music has in culture? And the thing that's so amazing is that there's kind of acknowledgement from all sides here, like we have a very good start. We don't know what all of the products should look like, and we're excited to build them together.
How does AI video play into this? I notice that when I open, I often get served a vertical video feed of others' creations that are often AI video animated in a pretty convincing way, like you get a full music video basically. Is that something that you're planning to—is it already available for everyone? Is there just a button that can make these, or are there sort of like more pro creators that are using outside models bringing a music video to the platform?
It's a great question. Right now, that is mostly creators bringing outside technologies to bear. I would just back up though and think the purpose of Suno is actually to inspire you to do things that you haven't done and not just to do the old thing of scrolling again. And so what you may have noticed is those videos are actually really tailored to get you to lean in, to take a piece of content, you can actually remix it right off of that video. I think we're not used to thinking of ourselves as creators, us mere mortals. Maybe you guys are on TV every day, but it feels amazing to be creative and just getting people over the hurdle of actually I am a creator and I will find it really enjoyable to be creative. That's kind of the magical spark that we're trying to bring to people.
Yeah. And I have that all the time. John will send me a Suno song and I'll be thinking to myself, even though all I know that John just wrote like a handful of sentences and he put the words and the letters in the right orders and it got that output, I'm still thinking there's a spark of like John is a genius for putting this set of ideas together. And I know that he was thinking about the lyrics and how to arrange it and things like that. And so that joy of creation and compacting it into something that—and again, I've always seen this as like net new, right? Like the idea of like a gag song, like a joke, has been around forever, but the friction to do that—
I thought we were enjoying it specifically because it allowed us to go after music that was basically low TAM. So we could make a song about the most feared lawyer, about John Quinn, corporate lawyer, that would never exist in any timeline other than when you drop the cost of creation to effectively zero, then you can actually get this. And I thought that was just us, but when I was scrolling, I saw similar experiences happening in other pockets. Like someone made a song about the Revolutionary War and they took a bunch of history and it's like George Washington. Have you seen this video? I don't know, but it's all animated with AI and I was like, "Oh, this sounds like a Fourth of July banger, but it's historical, but no one is really going to sit there and arrange that." And all of a sudden, you have this like very niche overlap of you want a July 4th hard-hitting song, but you also want to learn history, and it puts it together in this really weird niche way, which I think is happening all over the place.
Yeah. Look, let me say two things to that. One is I'll plug you can go all the way down to a TAM of one or two people. You guys, I'm sure, have iMessages with one another. You can actually access Suno right from iMessage if you have the latest version of the app. And so it'll just be easier to interact there.
But the other thing I would say is I think maybe the best example is, I don't know if you guys saw this so-called Oreo Gate.
This is a lady who took the text messages from her mom's chat about her kids eating Oreos or sending Oreos to school and just using music as a platform to express I don't know what is special and absurd and banal about everyday life. And when I think about what was the alternative for her to be expressive like that, to use her vantage point, she basically would have had to become a stand-up comic to bring this to life. And now you can use something like music to do this in an entirely new way. And it's something that I think tons and tons of people really can relate to on a fundamental level.
How do you think about all the ways other than speed and overall feeling of music or musical quality going from model generation to model generation? One thing I noticed with the last generation was that it would just constantly be referencing glass in different ways. In a bunch of different songs, it would come back to the glass shattering.
Yeah, I was talking to John. To me, it was like a tell that Suno developed its own terms of phrases.
Yeah, "It's not this, it's that." Or you're absolutely right, like this version of that was different, but it seemed to exist. How much are you hunting for those tells?
You know, it's a really tricky thing in music. Unlike, let's say, the big labs LLM intelligence, there are no benchmarks. And so these questions of taste end up mattering a lot more, and even just measuring these things is really hard. So we are always on the hunt for kind of that telltale sign. And then, yeah, there's a good amount of judgment that goes into how do we fine-tune, how do we post-train these models? Actually what we're releasing today, having both a V6 and a V6 Wild—V6 Wild being very differently post-trained to be way more interesting and basically a little less good at making full songs—is kind of a start at bridging that of how do we make sure that people actually have the most creative experience that they can.
Back on the deal with Warner Music Group, BMG, Believe, I'm interested to understand how you're talking to investors, how you see the market evolving over the long term, because the most extreme economic share might be something like what we see with Spotify. I think they give maybe 30% of top line to artists across everything. There's a huge stake of revenue that goes to creators. Still a hundred billion dollar company, fantastic business. We love Spotify. And I'm interested to think about as the industry outside of Suno, but just AI-generated tools, what type of economic structure would you see landing in equilibrium?
I think about this in a couple of ways. There are things where we can think about how do we take the existing technology and really lower the cost and give it to way more people. That's what you said before, and you would never be savoring these moments if it weren't so easy to make music. And then there's the whole other side of things of how do we actually now apply this to markets that people couldn't even fathom before? And this is actually something that I kind of live and breathe every day.
If you rewind a couple years ago, people think music is really small, "I don't want to invest in it." And then you realize that it's not just that old TAM. Actually, what we're doing is fundamentally changing how people interact with media, how people entertain themselves, and realizing that actually is much bigger. And so to answer the question more directly, this is actually evolving. We have some economic models now. We have a lot of alignment to build products together and figure out the economic models that go with those products instead of having to figure everything out all at once. In the crazy time that we live in about AI progressing so quickly, we kind of don't have the luxury of trying to get it right in this very moment because things change so fast.
What is the demand side from consumers on editability, the pro workflow? Obviously, the model is outputting basically a complete waveform, but you can also use AI to deconstruct that into stems, into notes, into sheet music, and then a more experienced musician can go in and make fine-grained tweaks. Just like you can generate an image and bring it into Photoshop if you want. What is demand like? What are you building there? Where do you see that side of the business going?
We have a pretty broad product suite here, from you can type in a few words or access it from iMessage all the way down to Suno Studio, which is our digital audio workstation, which is really professional workflows. What we've seen is actually by making those workflows not only different and powerful, but actually much more fun, more of the professional tooling ends up just appealing to average people who fall in love with making music. I think this is a lot like I might buy the signature guitar of my favorite guitarist. I don't think I'm going to be a famous guitarist, but it actually makes me feel really good knowing that I use the same tools that the greats use, and it makes me feel better about my creativity. And I think that it's a failure of imagination to think I can't actually fall in love with digital tools in the same way. Like I know I am using the same tool set, at least in part, that makes some of the best music. I feel more creative as a result. I fall in love with making music, and there are virtuous cycles there.
What are you learning from the early sparks of progress in AI video? I was trying to think about the most viral AI video from last year, and I couldn't put my finger on it easily. I can think about the previous year because that was Harry Potter Balenciaga. That was the moment we were like, "Wow, this really worked." And I think for this year, music like Rubbers by Phoenix Flexing, it's now out that it was AI generated. I don't know if he used it or not, but it sparked the conversation. And then I'm wondering what next year looks like if we're trying to pattern match against the ebbs and flows of progress in AI video.
It's a great question. I almost think video is having a super interesting moment right now, as there's a lot of new models and they're getting more convincing, they're getting easier to use. And so we look at that very closely. I think the one maybe easy thing for me to say here is that actually speed ends up mattering a lot, and as it becomes more affordable to really iterate a lot on video—because if you rewind let's say six months, it's actually kind of prohibitive to make a full-length video—I would imagine that we're going to see some very interesting cultural moment. And I think of these less as technology per se. It's actually you gave the technology to somebody who has an interesting point of view who's able to express that point of view to the world with these tools. And so I think it's very hard to predict exactly which point of view that will be. It's like a very cultural zeitgeisty moment.
Yeah, I remember I talked to Emad Mostaque—
There is going to be at least one very big cultural moment. I will tell you that. That's all I can say right now. But there's going to be a big cultural moment involving music and technology.
I was talking to Emad Mostaque after Stable Diffusion launched years ago. This was maybe 2022, and he was predicting that by 2024, 2025, there would be full movies that were AI generated. And we kind of got there, but they didn't really break through. But AI video is actually really successful. It's just taken a very different approach and it's diffused in an interesting way where it's making its way into VFX workflows all over the place and it's become very useful. But there hasn't been the fully AI-generated movie that is so good it gets over the hill of like, "Oh, I don't want to like it, but I have to like it because it's good." And I'm wondering if that will be what next year in AI music looks like. It's a tool that's in every studio, every production, and it's something that certain artists take off the shelf occasionally, but it's not fully AI-created albums that are out, but it's also not fallen by the wayside. Is that a reasonable prediction, or do you think that's too conservative?
No, I think that's actually not too conservative for the reason that an album or even a song is a point of view, and it's very hard to just one-shot the point of view that is going to really resonate with the whole world. It does happen, but I think of this as a process of refining your own point of view over and over again. And you may use an AI tool set, you may not use an AI tool set. And so I think that a year from now, it's actually quite mundane that there's a lot of AI in music. It's used in every single part of the production chain. Some of it will be Suno, some of it will not be. But the idea that there's just going to be charts full of one-shotted AI bangers, to me, actually kind of misses the point. It doesn't express the point of view of the artist. And I think that this is kind of a faulty comparison to, let's say, we just saw OpenAI maybe prove a huge math theorem, and this is something that is long-running that is verifiable. Music is just not that way, and so I'm not sure we expect to see the same types of things.
Wait until I'm spinning up 1 million agents commands, 10 billion instances of Suno to create the giga banger chart topper that sits at Billboard Hot 100 number one for 100 years straight.
Yes, yes. The final song. I'm sure you can do it.
What was your takeaway from the Astro launch over the weekend? Because lots of Blender models, lots of video games created, and I feel like a million one-shotted web games were spawned this weekend. And I could not be more optimistic about GTA 6 blowing out the predictions. At the same time, I feel like as soon as I stop playing GTA 6, I'm going to get a text from a friend with a really hilarious micro game that they vibe coded that's GTA 6 but for TBPN or for tech or something, and I'm going to be like, "Oh, this is cool. This is the special thing that only makes sense to me and you." And so is that how you process that, or is there anything new on the overall AI capability side that's exciting to you these days?
Maybe I'm too dumb to predict the future, and so I would just look at, let's say, the last revolution there, which would be Roblox letting people really start to make their own games. And at the beginning, it starts really small, and people are really just playing their own games. And eventually that technology progresses where you get more of that power law, you get people playing other people's games, etc., and then you make an ecosystem there. It's not crazy that we'll see something similar there. But again, I think a game's purpose is to be played by people, and it is very different from the general intelligence model of things, and so I think it's hard to really apply logic from one into the other.
Yeah. How much should we read into the iPhone Duo launch? Do you think Apple's so Suno-pilled that they want every person to have more real estate to be able to make music? What's your read?
I really hope Apple's so Suno-pilled. Humbly, I don't think we're the reason that they're making that phone, but if you want to suggest that, I'm very happy to take it.
As somebody who makes mobile apps, are you guys already looking into the implications of this and how you're going to need to change your application to be able to adapt to this kind of new—
Might just be one prompt: make it work on the Duo.
For a lot of developers, that might be the answer.
Yeah, absolutely. I think about this as opportunity. If there is different interaction with your phone, we actually are able to now inspire you at different times, and you can take out your phone and do something without unlocking it. You can use one camera versus the other camera. I don't know what the plans are for the Vision Pro, but when I think about a final form of Suno, it's actually how do I get a Vision Pro app that I can put on and conduct an orchestra? And so I think when we think about interfaces, this is where in the creativity game we will just work on forever. There's no end to working on how we delight people by letting them interact with technology in new ways.
Last question. Have you guys ever run any billboards or wheat pasting campaigns?
Not that I know of. Why? If you suggest the billboards, we should—
Well, I just feel like you should be running a billboard that says, "Make a song with a sentence." And you should just go hard on that because people need to experience the magic of this product. It's just like—
We got to not get people to pick up their phones while they're driving. That would be bad. But we got to place those billboards in the right place.
Everybody's in Waymos now, Tesla Cybercabs.
Bull case for billboards. In Waymos, you can—
No, no, no. It's actually a bear case because everyone's going to be looking down on their phone.
It's going to be rough. Anyway, thank you so much.
The future's bright. You're going to be sitting in your Waymo, you're going to have a lot more time to be creative. Way better than driving, way safer than driving.
Yeah. You say, "Nothing from my end, thanks," and then you go back to making songs.
Great to see you, Mikey. Congrats on the launch. Excited to try the new model.
Have a good one. Bye-bye.
Let me tell you about Codex. Codex is a powerful workspace for getting work done with AI agents. Whether you're writing code, analyzing data, creating content, or automating business workflows, Codex helps you move projects forward from start to finish. Thieves raided the Renoir Museum in the latest French art heist. We'll get to that story after we chat with Keith Rabois, managing director at Khosla Ventures, returning to the show for the fourth, fifth, sixth time. It's always a great chat. Keith, how are you doing?
Great. It's great to be back.
Have you been following this French art heist? Is this bearish for France?
I have not been following it. I don't know how I missed it. I was just in France last week and somehow—
Oh, you were in France last week. Are you guilty? Let's figure it out.
Yeah, maybe. You never know. Maybe I was purloining some artworks.
Okay, well, you're back in America.
