I see a large IPM on the horizon. You're surrounded by journalists. Hold your position.
Overnight success. The place 3000 double right.
There's misinformation clearing order inbound.
Let's just roll. We are surrounded by general. Hold your position. Get up. Trust the experts here. We are experts.
Founder Mel 5. I see multiple journalists on the horizon. Stand by. UAV online.
Blaze. Double blaze. Triple blaze. Double kill. Fight is team deathmatch. We are experts. Triple blades. That's just wrong. Right.
Hardy clearing order inbound. We are surrounded by journalists. Hold your position.
Strike one. Strike two. Activate golden retriever mode.
Marky clearing order inbound. Five pudding. I see multiple journalists on the horizon. Stand by.
You're watching TBPN. Today is Thursday, September 24th, 2026. We are live from the TBPN 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. Very happy to be sponsored by Ramp. Thank you to the whole Ramp team.
Thank you to Chris Kringle because he's coming by in just 91 days.
That's right. The TBPN road to Christmas is continuing day by day. We're 91 days away. So that's right. Get ready. Get excited. Christmas is coming. It's around the corner.
Great weekend to pick up a tree if you can find one and struggle to keep it alive for three months.
Yeah, maybe. I mean, some operators I know are actually cycling through multiple trees on the road to Christmas.
Some operators, yeah, it's like that Chinese proverb: the best day to plant a tree was 10 years ago. The second best day is today. You know, get started. The best day to get a Christmas tree was 10 years ago. The second best day, today. No matter when it is, it's only 91 days away, so get ready. Anyway, Meta Connect yesterday, late. Zuck likes going late. 4 PM Pacific, 7 PM Eastern for the keynote.
Well, because he wanted to roll straight from the keynote into some beer pong.
Oh, that's true. We will get into that big hot debate on the timeline over his beer pong strategy. But really, really tight keynote, very well organized, very clear, didn't overstay its welcome. Did three live demos last year. There was a hiccup with a live demo Wi-Fi, and I believe Boz triggered everyone's agents by using the wake word. So this time he deliberately stayed away from using that wake word and used the name of his agent, so he didn't cause a lot of chaos in the dome or in the studio, wherever they were. But a bunch of great announcements, really good pacing, really tight one-hour keynote, and it was fun watching Zuck sort of do the live demos and be like, one down, like this is high stakes. It was clearly high stakes for him, but he got through it all. No mistakes.
Yeah. He had a funny line. He was like, and we love live demos here.
Yeah. It's funny because he totally could just not do live demos, but he wants to, so he pushes through it.
I think you need to.
Do you?
I think you absolutely need to. People have bought so many random hardware devices that haven't met expectations.
Yeah. And I guess there was hardware, if you want people to buy it, you have to do a live demo.
Yeah. There was a demo of a Gemini model running like two years ago that was sped up in post, and that faced a lot of criticism. And so I think, yeah, just demoing things as they are available in the real world is very valuable. And I think in general, the thesis for the products was meet people where they are, delivering tomorrow's technology with today's constraints. And that's just what I kept seeing as a theme was, what do people want? And of course, it basically created a huge dichotomy between the crazy superintelligence RSI ASI P(doom) X-risk crew and then just like Muse. It's a helpful agent. Neeraj Agrawal summed it up well. Anthropic marketing: we built a god only we can control and gave it a wet lab. Muse marketing: hey girl, I know you have dozens of unpaid parking tickets. Let's get that cleaned up. And it's funny because it's actually working. I think that if you went back to a couple months ago, the consensus was like, the popularity of AI is turning. It's deeply unpopular. You need to just be quiet or not lean in or just apologize, like what a lot of the lab leaders are doing, just being like, 'Yeah, it is bad. I'm sorry.' But Meta has taken a very different approach. They're not fear-mongering. They're just like, 'It's cute. It's lovable.' And they actually are successful. I mean, the New York Times yesterday, Jensen and Ezra are duking it out over X-risk and acceleration. Do we need to slow down? Meanwhile, the New York Times published a really positive review of Muse. They love it, and they're a tough cookie to win over, basically. So, success there. And it's interesting because there's been a few strikes for Meta over the past two years. The vertical AI slop app Vibes quickly disappeared. I don't even know if it's available anymore. It might have been quietly shuttered or tucked behind a couple different tabs. The core Meta chat app that was sort of just like a Gemini chat clone, I don't think that ever really took off.
Over the last year, you have massive spending, clearly massive focus for Meta, but the strategy was at most times completely incoherent.
It was scattered.
It was not legible to the markets. It was not legible to do everything. Yeah. You had nothing really, you know, Muse Code. You had them buying and then having to undo the Manus acquisition. You had... Oh, yeah.
The Meta Vibe. The first thing, and again, this was an unforced error. The first thing that they did post the whole talent raid was release the short-form slop video app, which wasn't even...
It was Midjourney. Yeah. It was basically a Midjourney wrapper.
Yeah. And we were, and everyone was like, 'Okay, this doesn't align to what you're putting out, which is this personal superintelligence narrative.' And so really yesterday, everything felt like...
It came together really nicely, right? Tying in all the hardware investments, tying in making personal superintelligence feel a bit more real, right? Totally. AI that hustles for you. I think that's smart. Yeah. Positioning, but it really felt like things came together for the first time.
And I wouldn't necessarily go so far as to give them full credit as if you were really going to give them full credit for like, this was the master plan all along, then there's a bunch of other stuff you wouldn't have done unless you were just trying to confuse the market. Yeah.
If you're trying to just confuse people, are we doing coding? Yeah. Are we doing a knowledge work harness? Are we doing short-form slop? Right. And so I think it was...
You're not giving him the benefit of the 4D chess.
No. Okay.
No, no, no. But going out, razor credit, 4D razor...
Total credit to them for figuring out how all the pieces connect together. Yeah.
And then putting that in a one-hour presentation that actually makes sense.
So what are the pieces that come together? Why does it make sense for Meta's business? Why are they seeing success here when the models haven't jumped to the frontier? The benchmarks have always been a little... there's been a lot of confusion on are they going to sell compute, sell tokens, sell, you know, go after the coding market, desktop agents, all these different things that they were testing. This seems to be working, and I think there's sort of four key points that come together to make this make sense. So first is like they have scaled and secured systems for years. Encryption is a great example. They have a lot of compute, and so they can offer a VM for millions of people that is secure and also well-resourced. And they won't be losing sleep over cost because they already have a lot of the compute built out. And the question was like, when will the demand come? Are those GPUs sitting idle? Is that infrastructure getting used efficiently? They're building a lot more of it. Do they need to? Well, they got to find a product to actually distribute all that through, and it seems like they have. Two, they have the distribution to continually educate people about features and benefits of Muse through big ad placements across the family of apps. I think that there is the world where they've been able to funnel what, like half a billion people over to Threads. It's not a particularly differentiated app. Obviously, it's a different community and it's done well, but it's a different challenge because with Threads you can just show someone a text post. People know what that looks like. It's more content. They go over there. Job's finished. Once you show someone, all you need to show is that their community exists over there. They're scrolling Instagram watching basketball highlights and you say, 'Hey, there's some basketball commentary going on in Threads.' You hop over there. With all the AI agent products, the capabilities change so rapidly that you have to re-educate the consumer on what it's capable of. Like, can it actually deal with parking tickets? If it can, you need to confidently tell someone that. And that capability might change. It might not work one week and then the model updates and then all of a sudden you can do that the next week. And so ChatGPT came out and it was like, it can answer Google searches, it can do your homework. Then it was like, it can write some code, and there's this constant re-education of the community. Obviously the early adopters, the enterprise, the business, the startup community tests everything, but consumers don't. Consumers just... they need to be told the new iPhone has an ultrawide camera. The new iPhone has Memoji. Even though Memoji was not a wildly popular feature, Apple told everyone what it was. We know what it is. Everyone sort of knows that concept because they ran billboard campaigns, and Meta has the ability to actually go and tell everyone what Muse is capable of and run an ad campaign across the whole platform to get people back up to speed on the latest and greatest feature. The third is that no one needs their...
Says deal, quote, deal with parking tickets, hack me out of it, then we are talking. Well, no, no. So the thing is that they are behind on the frontier. They're not dominating all the benchmarks, but it actually doesn't matter because no one needs their personal AI agent to solve the Millennium Prize math problem or secure their enterprise infrastructure from cyber attacks. Like, that's not a requirement. The model is definitely good enough for a lot of this. They obviously will increase the model and the capabilities, but it's not what consumers worry about. Consumers don't care about benchmarks. They don't understand them. The benchmarks are often hacked or not indicative of the feel of using a model. And so it leans into the capabilities. They have near-far-near frontier intelligence. They have good enough intelligence. And so if it's good enough to do this, you just roll out the capability and the harness is what matters. The integration, the distribution, that's what matters. So they actually didn't launch Watermelon. They've been teasing that for a while. There's been some leaks. It totally could have happened here, but they made the decision to either not rush it because maybe it does just need more time, but also even if it was ready, I don't think they should have launched it here because Meta Connect's not for that audience. They don't care. Yeah.
Watermelon might scare people.
Yeah. Yeah. Yeah. You probably don't want it to be too powerful. And then fourth, they of course already have a business model that's heavily aligned with this type of use case: advertising. We've talked about this idea that brands would love to sponsor agentic workflows. You open up a personal AI agent and it says, 'Hey, you know, the new Nathan Fielder movie is coming out in a couple months. I saw that you were scrolling reels. You seem to be a fan of this. You shared the movie trailer with your friend. Do you want me to get tickets, text some of your friends, share it on WhatsApp, share it in your Instagram DMs, and actually make that happen? Send a calendar invite, do everything necessary to take care of that.' And I think movie studios are just one example of a business that would happily pay for placement, for extra attention around that particular workflow that they can monetize through the traditional pathway. And then of course like small businesses will buy pixels anywhere. So, you know, if you open it up and there's a full-bleed ad, like that's fine and people will pay for that. So...
Yeah. The other kind of examples that get pretty crazy is your personal agent, for example, knows that you have home insurance that's set to renew in three months and it's like, do you want me to go get three competing bids before we renew?
And it's like, yes, right? And that's like an opportunity for some type of paid placement as well.
For sure. So the vibes, not the app, but the actual vibes at Meta internally, how are the employees feeling?
I feel like this is a turning point. I feel like if you're at Meta, you're like, 'Okay, finally there's a thing to align around. We have a strategy. As long as I do something that's in support of Muse and what's going on with the personal superintelligence mission that the CEO and founder has laid out, I'm good.' Even if I'm on the Facebook Marketplace team and I feel like I'm going to get steamrolled because of the Nikita Bier take that everyone's going to be negotiating and lowballing on Facebook Marketplace. Well, the Facebook Marketplace team, they can say, 'Okay, well, now we're going to build tools for sellers to make that not annoying.' And so there's more work to be done and it's more aligned with the overall mission. The vibe should be good. At the same time, The Verge ran a story that said the vibes are internally way off still at Meta, that people are upset about the scattershot nature of the approach, that they're still reeling from the metaverse misadventures. But I feel like that was the last hit piece you could write about Meta. And I think things...
Hard to be feeling that bad when the stock's up 38% in the last month.
Yeah. And so I think it's pretty easy to align all the efforts across what's happening with personal superintelligence. And even if you were on like the glasses team, there's a whole narrative of like, well, you were trying to build VR, you came, you were in the division that was building the cool Oculus stuff. It's evolved and now it's kind of just a GoPro on your face. There's been some negative, oh, these are just for, you know, people to spy on each other, blah, blah, blah. But now they even have a version without cameras and it's very aligned to what the users want. And so the topline hero initiative, even if you're like just on the glasses team and you're like, 'Oh, like they're building the data center and that's not really anything to do with me or they're working on a coding model and like that's not really relevant to like getting more people to wear the Oakleys or the Meta Ray-Ban bands.' Well, now you are aligned. Now you are on the same team. You're rowing in the same direction. And so I would expect the vibes to get better. I think, and yeah, this is all good news. Super reflected in the timeline. But I'm even more excited about the VR glasses. That was the highlight for me. It was the one more thing. I was hoping that we'd get a new VR headset last year. We didn't. And no one was speculating that this would drop this year, but it did. And it's very, very cool. So the headline is 1,300 bucks VR glasses. It's a hundred grams, which is the weight of a pack of deck of cards. And if you go back, John Carmack was the CTO of Oculus, working with Palmer Luckey. He came into Meta through the acquisition. He's the creator of Doom. Legendary computer scientist now running Keen, working on AI. He wound up leaving. Very, very disappointing to not have him realize his vision there at the company. And he had been pounding the table at Meta. And actually there were some leaks and there was like, oh, this is like a source of frustration because his goal and his mission was to get to 100 grams for the headset, very light at $100. He said it needs to be cheap and light and then it can go everywhere. Then you can see mass adoption and then you'll get the developer ecosystem, etc. Obviously $1,300 is not $100. There's been a lot of inflation, but not that much inflation. But it is half the price of the Apple Vision Pro. Basically, you're in a more accessible territory.
Half the price of specs.
Oh, yeah. It's half the price of specs. And it's, I think it's less than 1% the price of a Ferrari Lucho. So, yeah, there's that. If you're in the market, you can just put that on and imagine that you're driving an electric Ferrari.
The glasses and the car sound the same.
Oh, yeah. They're both very quiet. But he pitched it as three key things. It's a cinema, it's a workstation, and it's a gaming console. You can play VR games on it.
So I almost felt like they were listening or just like same line of thought as Ben Thompson because they're rolling out a bunch of like a hundred experiences. It sounds like a year.
Yeah, where you're going to be able to get some entertainment, live entertainment value from your headset.
And specifically sports.
Sports, sit anywhere in the stadium.
Yes. Yeah. Yes. Yes. And you can pick your seat. And I don't think they will be doing the crazy frenetic editing or the cut downs. They have a great deal with UFC. Obviously, Dana White's on the board. But it seems like they also rolled out partnerships with so many other leagues to get that there. They also put the compute and the battery pack on a puck that you wear on your belt. I love that design. It was very controversial, very unApple when Apple Vision Pro did it and everyone was like, Steve Jobs never would have stood for this. Well, like that's the reality of the technology. It doesn't make sense to put a battery on your face and so put it in your pocket. I've never had a problem with the way the Apple Vision Pro works with the tether. And they did a good job in the promotional videos of like hiding the puck, like you can't even see it there. So they're clearly aware that you want to forget about it. And I think in the best case, you actually can. So very good there.
The other, sorry, other dynamic tabletop typing. So if you have a flat surface in front of you... bad, bad, bad. No one wants to do that.
G says it's insane.
I don't think it'll... I trust him with my life. Okay. Okay. Maybe...
I think people will either connect this to a computer and just be using their computer or they'll just be using voice. I think that's like a half measure, like the writing, and I think it'll be 99% of the way there. So in a demo you'll be like, ah, this is pretty good. But if you're thinking about actually wearing this for work hours a day, you'll be like, yeah, let me get a $20 keyboard. It has tactile feedback. Like you can actually feel the keys move. I don't know. It's good that they did it. Like it's clearly cool. It's possible. And especially if you're in a situation where you can't bring a keyboard or a laptop, like that makes a lot of sense. That is a glimpse into the future. I would be surprised if that's the dominant input, the way people input stuff. They'll either just talk to it because the voice models are good or they'll be locked in at a computer with an actual keyboard.
This is an awesome step forward to, you know, humanity living inside a world owned and controlled and monitored by Mark Zuckerberg.
Yeah. The wireheading future. Yeah. No, no, no P(doom) around wireheading. That is one of the X-risk things. Remember, don't forget about that. It's not just bioweapons. It's also we all turn inward and lock into the VR worlds and then stop reproducing and that's the end of humanity.
The battery is only three hours though, so you can't even watch all of Lawrence of Arabia. Like, how can you be wireheading in?
It's very easy because that puck, you plug it in.
You plug it in. And that's what you do when you have the USB-C. The USB-C is also really important because Palmer Luckey was in the promotional video and he's getting his mind blown. He's saying you can do DisplayPort over USB-C. And what that means is that you'll be able to just take a USB-C cable from your laptop into the actual puck. It's charging your device and then also you're getting zero latency because the latency of having to go from screen sharing on your computer to your Wi-Fi router back to your glasses in Apple Vision Pro. I was trying to game on it. I did all these crazy things to try and wire up a PlayStation into it. It was a huge hassle. Never really worked. There was always a ton of latency. This cures that. I'm not exactly sure how the game consoles support USB-C, but I know all the Mac products do really, really easily DisplayPort over USB-C. So it looks like a USB-C cable, but it's basically an HDMI cable. So zero latency and so you just get a huge screen with whatever you want on it. So it doesn't matter if the developers don't show up. And they didn't bring GTA 6 to the Meta VR glasses, which are probably going to sell less than a million units in the short term and it doesn't come out for six months. But will you be able to play GTA 6 on this thing? Absolutely, right? Because it has DisplayPort. And so if you want that experience, it'll be available to you. It won't be a fully immersive VR experience. But if you're someone who only has $1,300 and can't afford a movie theater or doesn't even have space for a huge TV, you get this. You plug it into a laptop or a gaming console and you have a movie theater available to you. So very, very excited to try it. Very disappointed that they have to wait until spring of 2027. That's a long way away. But the road to Meta VR glasses begins today. So start counting down 180 days or so. When is the end of spring? When would that be? I guess the end of spring is what, May, right? Then summer starts. So we're up in like 230 days or something like that. But it'll be fun to see what people actually do with it. It's still the dominant platform for developers in VR even though everyone's pivoting.
I think that's interesting to me, the positioning of it being a computer.
Yeah. It's a small tweak, right? I don't remember them ever calling previous devices like, this is a computer. No, but it's easier for people to kind of imagine it as, hey, this is a device that I can do a bunch of stuff.
And this is the thing with the iPhone. Like the iPhone was a phone, an iPod, and an internet communicator, basically a web browser. And it was like, we know that at the very least people like phones, so if we can be as good as a phone as what they have right now, and so it's like, if we can be as good as a home theater, then people will buy it. And so instead of trying to purely position it as like this entirely new thing, metaverse, that you live in and VR and everything's immersive and it's full immersion, instead you're like, this can replace your triple monitor display and you're like, ah, you know what, I was in the market for a triple monitor display, was going to be two grand across three monitors, why don't I pick up this instead? And if they can deliver on that, I still think this might be V1 and it might be the second one that actually gets the breakout success and really broad adoption.
Yeah, he did say this is a glimpse of the future and anytime you hear that, you're kind of thinking, okay, like I'm going to get some cool demos and it's not going to be dailying it.
Yeah, I mean it'll take a while for the ecosystem, but it was the same thing with the iPhone. Like the first iPhone didn't have an app store, didn't have 3G, but it kicked off the beginning of a big revolution, and this feels like it could be that. Also, the pixel density, they haven't shared the exact specs. It feels like it might be 5K per eye, which is about 53%, about half as much as the Apple Vision Pro. So there's a world where in six months we're talking about, ah, it's still not that high fidelity to actually do work on for a full day. A lot of people are running 5K for a single monitor that's only, you know, in this much of your vision. You're talking about 5K spread over your full field of vision could be a little bit of a setback, but I think that is something that definitely gets resolved on the next iteration because the display technology...
What do we think about hologram calling?
Let's discuss it after I tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. I think it's cool. They did legs in VR. It's very difficult to do legs in VR. Boz took the victory lap on that and I think it blew people away. I think it's cool. I did an Apple Vision Pro call with someone who got one the same week. It didn't really, it never really took off because there was never a broad install base. It's all based on like, do they sell millions and millions of these things. It's always been hard for Facebook to develop like a new network that requires hardware distribution. It's much easier to go from a friend in real life to we're texting each other. We have each other's emails and then we're also sending Instagram reels to each other on DMs. Much harder to be like, 'Oh, did you get the device? Oh, no, I don't have it. Oh, I don't have it set up.' You know? But of course you can portal into someone with just WhatsApp or anything else. So it does seem good.
Well, there's one thing that is 100% sure, which is that you will be buying these and I will not.
That is 100% true. 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. Let's go through the timeline. React to some of the reactions. Mark Gurman loves it. He says they're incredible. Half the price of the Vision Pro, 16th the weight, external pack. He's stoked on it. What else is going on? They launch in spring 2027. The interface is basically the same as the Vision Pro, but it has an absolutely bonkers tabletop virtual keyboard that blows the Vision Pro keyboard out of the water. Yeah, because the Vision Pro keyboard like hovers there. You're not supposed to really use it. It's a terrible keyboard. But also, the tough thing with a keyboard is like, the moments when I was using VR, I wanted to be on a couch. I wanted to be on a chair that didn't necessarily have like a flat surface in front of me. So it requires a little bit of a different workflow to actually go to the coffee shop and sit down at the table and have a flat surface in front of you. So with the Apple Vision Pro, there was never a moment where I was like, 'Oh, I would love to be able to put this down onto a flat surface because I usually wouldn't use it on a flat surface.' So we'll see how that goes out. What else is in here? I underestimated Zuck's game. Ryan says, 'It's a hearing aid. You can't tell me I can't wear my Meta glasses anymore or else you'll be going against ADA, the Americans with Disabilities Act.' I don't know that that's actually what's going on here, but that is sort of crazy that there's all this backlash to smart glasses. Maybe there will be a next wave of like restaurants and bars that say, no, you can't wear those there. But if you say, no, it's my hearing aid, you can't. That's some 4D chess. Are you giving Mark Zuckerberg the benefit of Hanlon's razor on this one?
No, I just think it's cool.
I think it's just cool and independent. I don't think it's actually looking...
I was meeting with an older friend the other day and his hearing aid setup was like this, like, you know, it seemed kind of clunky. And if he could just throw on, he was wearing glasses, if he could have just thrown on a pair of glasses and be like, cool, I'm in business, that would have been awesome. So I think again, what many people forget about glasses is that smart glasses are way more appealing if you have a need for them outside of just wanting smart glasses. So if you just want, you know, if you need help with your vision or you need help with your hearing, I think these are going to have much stronger product-market fit.
Yeah. 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. Don't you think if you're wearing camera glasses, VR or AR, you know, Meta Ray-Bans and it has a camera there, it can see your hands, right? It can see your elbow. So if you were playing beer pong, it could tell you, 'Hey, back up. You're going to lose a point.'
It could go to your agent and kick off an agentic workflow to schedule you with a professional beer pong coach, have them come to your house and do hours and hours of drills so that you never make that mistake.
Is this vaporware if it didn't happen? Because we saw Mark Zuckerberg get caught with his elbow over the line at the beer pong table.
I just don't understand why smart glasses CEOs go out in public without wearing the smart glasses.
Oh, he's not wearing them.
And it's like Evan is doing interviews all the time without wearing them, speaking at places like Sun Valley, not wearing the glasses. Zuck is all the time out in public doing things, not using the glasses. I mean, they make clear versions of them, right? So if the product's so good and everyone should buy them, shouldn't you have them on at least half the time that you're out socializing?
I personally think if you see Jensen at a conference, he should be having an NVL72 in a backpack. He should have a rack and he should be carrying it around. No. I did notice he was wearing a very special watch and everyone's obsessed with Mark Zuckerberg's watch collection. He always debuts a new unique piece every time he's on stage. But I think the watch thing, it's getting a little old. I would like to see him get into cars a little more and I'd like to see him drive onto stage on like a Gordon Murray T.50.
And do a quick burnout.
Quick burnout. Yeah, exactly. Something to show that he's not one-dimensional. He's not just into watches. He's also into very special cars. Pull up in a Tourbillon. You know, you got the Tourbillon on your wrist. Why not the Bugatti Tourbillon? Drive onto the stage. Get out. No. My hot take about this beer pong elbow violation is I don't care about these. I think that people who play beer pong and worry about elbow violations are pedantic and annoying. And I think that it violates the spirit of the game. So I think that there's no problem with this at all. I think if you're the type of person that's like, 'Oh, you put your elbow over, you nerd,' and you're not having fun and you're ruining the fun for everyone. And it doesn't actually make that much of a difference whether you threw from two inches forward or two inches back.
I mean, it does make a difference, John. It makes a massive difference. Yes, but spirit, like the gentleman's game. You might address it if it's habitual, clearly.
Shane is bringing up the bunky in the chat. I think Zuck is going to have to get a proper setup. Get down in the bunky.
Spend two, three years in the bunk beer pong. Okay. Yeah. No, I think the gentleman's beer pong game, it doesn't need to be policed on a shot-by-shot basis. If there's a habitual stepping of the line and it's actually creating an unfair advantage because one person is following that particular rule and the other person is clearly violating it for increased performance, then yes, it's worth calling out. But one minor shot slightly over the line, I don't care about it.
I think cheating is wrong.
Oh, okay. You hold the line.
Law and order guy over here.
Look, you got to respect the spirit of the game. You can't break the rules.
Yeah, but then you can wind up in this thing of being, oh, you were 2mm over it. Turns it into...
I mean, this is blatant.
It's blatant.
It's not that blatant. It's still a long table. It's fine. He's not really leaning over that far. He could be doing so much worse. His elbow's over, but he just walked up to the table and threw the ball. And also, my man is down nine cups to one. Okay, give him a break. Do you see the score? When you're down, I guess. And the other team has five. He is losing so badly. Give him a little break. It's his event. He's throwing the party. He paid for the beer. Do something nice with the guy and let him get an extra on the shot.
Not a big deal. See, you miss the aesthetics and the dignity that comes with somebody who's hosting. They're hosting.
Let them in a little bit.
Are you doing the voices? I don't have the IM now, so I can't hear what you're doing. Has the whole thing been Chad or the...
Yeah, Chad. Tyler's Chad voice. Yours?
Okay. Okay. Okay. I got you. Well played. Well played. Well played.
Anyway, there's a bunch of fun things. Alex Wang on stage dripped out. People are speculating that it's high fashion. In fact, he corrected the record. It's not. He's thrifting. He's not buying Gucci and Prada pants. He's just wearing Crocs. Realtree Crocs. It is a big reverse because Alex Wang was on the cover of Fortune or Forbes in Louboutins and had his feet up with the red bottoms there and it was like a very... when I saw it, I was like, that's just a sick photo.
It's the barbell.
Yeah. But it does seem like he burnt out on the high fashion. It's not like he's... So, you know, he's like, yeah, I'm just gonna dress like this. I have my own style, which I think is very cool. So, good to see Alex Wang on a run and delivering, yeah, just like big announcements, big progress at, you know, one of the biggest companies in the world. Good stuff. 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.
Where should we go next? We got to talk...
Charm. Got to talk about the charm. The charm. The charm. The Tamagotchi.
So this was added. This was second. Mike Isaac says, 'Mark does a one more thing with a new hardware device AI agent called Muse Charm. Voice activated and palm-sized. Out in time for the holidays.' This feels super smart. People already think Muse is super cute. It's a tech-enabled Labubu.
Yeah. Basically, which is extremely popular, like the most dominant consumer product of the last couple years. And so it's distinctly sort of like outside of the tech bubble. The quirky Labubu is not like the refined edge of like a Cybertruck or even an Apple product. But Meta has been sort of not blinded by the clean aesthetics and instead they're like, 'Yeah, this is just a cute, fun thing.' And so people will actually like this. They know what people like.
Avi Schiffman over at Friend: the market has spent billions trying to compete with me and I'm still winning.
Let's go.
That is the right attitude. That being said, I don't see people really using the Muse Charm that much over time. Like, I think it's a fun kind of toy gimmick thing, but I think it's the kind of thing that people get. It's funny. It's worth doing for Meta because it should be...
Yeah.
But they'll still be on the phone 99% of the time, but...
Yeah. It'll be viral. You can talk to it.
I like these narrow one-off hardware devices for like one interaction. Also, it's just like...
It is a more wide-open space for them to sort of try and come into devices. Like you have the glasses, then you get this, and it's like slowly they're eating into more of the device footprint in people's lives, even if it's at the...
And again, I would expect them to sell this just at a loss. It's like good marketing for the core product, which...
Do we have a price? I don't think we do yet.
We don't have a price yet, but this feels like something that...
At most you price at like $99, but probably he tries like, again, he's selling a lot of hardware at negative.
People are going to wear this as a watch potentially. Like you're going to wind up with the AP Swatch thing, right, where people wrap this...
For your most swagless friend.
Brutal.
Give this as a gift with a watch strap to your most swagless homie this holiday season. It's only 91 days till Christmas, but the timing here is amazing. I have no real insight, but like clearly they were working on a personal agent for a long time. Unclear when this whole personal agent meets Labubu thing became part of the strategy, but it feels like, yeah, they're clearly like they have the hardware team set up. They can move super quickly on this kind of thing. And I think it's...
You mount them to glasses so you can look straight at it and then brain rot on your face. Let's talk about some of the downstream.
Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store, on mobile, on social marketplaces, and now with agents.
What do you want to talk about?
The chaos coming. Yeah. Well, coming away from this, a real challenge for Instinct.
Everyone building personal agents, right?
We talked about this a bit, but Meta has their own models.
Yep. That they can serve themselves at the cost of energy, right? And all this capex.
They design their own server racks. Like that's how vertically integrated they are. And also, yeah, they've done a ton of stuff talking about the data centers they're building like benefiting communities. They have this whole story about all the teachers in a school district got like a pretty large cash bonus when the data center came to town. Like they've done a lot to...
Yeah. And they were kind of born to navigate through this kind of era because of what they went through in the 2010s, right? Like the most veteran crisis comms. Like how do we navigate and position ourselves? They're already getting attacked by the social media addictions.
Can you imagine when a data center's coming to town and the Muse guy comes in like the Michelin man and he's just like, 'Don't worry about it. I don't even like water. What are you talking about?' So Instinct, wonderful product. It's growing like wildfire still, right? I'm getting texts from friends that are barely in tech being like, 'Hey, I'm inviting the product.'
He asked me to text.
Do you want to grab a beer?
But anyways, Instinct, very, very, very expensive to serve. And they're not charging right now. So they're losing a lot of money on a per-user basis.
They probably will be for some time. Muse is coming out and they can afford to lose money forever basically. And even for them, they're sort of like gating onboarding in some ways, like you sign up for the product and you're not necessarily ready to fully go live. And so even for Meta, Muse is going to be expensive to scale while they figure out these monetization strategies. Of course, Meta is in a really good position to figure out how to make money on this.
But very, very challenging for Instinct. I expect that Instinct's going to need to find like a real dance partner of sorts, right? You could imagine who...
Google.
Rufus.
Yeah.
Do you hear that? That's Rufus. Amazon...
Roffus. What is that?
Um, no, you could imagine, you know, a Google being like, 'Hey, we can't build products that people like, but we have a lot of products that people use.' Um, so maybe that's an option.
Um, you could see Amazon as well. Um, but for them to be competitive in this full onslaught by Meta, it's going to be a lot. And same, same with OpenAI.
Comm's advice for Noah. Should he be going toe-to-toe trying to ratio Alex Wang?
Oh, yeah. Cuz Alex is coming for it. He's like, well, yeah, it was crazy because Alex has started punching down. You know, there's punching up, there's punching. Punching up is when you're a startup and you're like taking shots at the hyperscaler. And punching down, we haven't really seen from a hyperscaler in a long time, right? But Alex is out on the timeline just...
But yeah, he's, you know, he's in a good spot. He's in the strong position. So, you know, let him know if you're going to try and come for me, you better not miss. I'm the king of this. And then of course, the other challenged competitor coming out of this is obviously Snap. Um, I think that...
I don't know how long the Specs project can really continue, especially because I don't think that... I don't think these VR glasses are going to be... I don't think they're going to be like a mass market hit where all of your friends are using them, right? I think it's going to be like a base hit, right? Another step in the right direction. Not as popular as the regular Meta smart glasses, right? But like...
A cool entertainment product. They're going to sell some units. Snap is going to sell... I would be surprised if they sold more than 10,000 units, right? And the market is just going to keep punishing Snap as long as they keep trying to force Specs to work when... unless they have some version that we can't see that is already way better than Meta. Like, I just don't see how they can really stay in the game after this, especially because Apple is going to try to come over the top with their own smart glasses as soon as possible.
I mean, they should do Visionaire. I... No, but I'm not expecting like Visionaire. I don't even think that's what it would be called. I think it'll just be called Apple Glasses.
I just think everyone is in consensus that Apple should basically wind down the Mac project and the iOS project and focus fully on VR. Like, probably that would be a good thing for them to do.
I don't think that... No, no, they're going to do glasses. They're going to look nice. I don't know. But the cool thing for Zuck is that he's beating them to the punch on like it functions as a hearing aid, right?
Uh, I mean, the AirPods already... Yeah, yeah, yeah. But in the glasses format, right? Like coming, being first is, that's the best chance he has.
And they've been doing some good marketing around that with a number of users that are hearing impaired. And it's just such a minor technology, but it's huge for someone who's actually affected by hearing loss. And then they tell a human story and it's, oh, what is AI doing? It's just slop. Well, it's helping people that are suffering from hearing loss. Here, like, that's magic. That's amazing. That's actually a medical... I mean, it's FDA. It's a medical innovation. So very, very good stuff.
Uh, 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.
Uh, what's going on with Australia and OpenAI? Tyler, you dug into this. So Jordy shared like the high-level news, which is that as part of like the Hugging Face rogue AI agents going all over looking for everything on the internet, that Australia had been hacked. But what actually happened here?
Yeah, I mean, I think a lot of the details are still not fully revealed. Um, the Australian prime minister was really up in arms like, oh, they hacked and they didn't tell us for weeks on end. Um, but it seems like the consensus now is that... well, like the hack was essentially that there were some unlisted files that were still publicly accessible.
Yeah, but they're just not listed anywhere and then the agent found those. And so they're saying like, oh, they hacked into us, but you know, there's no real personal identity revealed at all. Um, and like these are private in the sense that you can't just Google search for them.
Navigate them. There's no link if you just click a button. But if you're scraping, if you're doing some kind of more...
I have a funny story about that. In 2012, I had a buddy who was, you know, I don't know, hacker-ish software developer, not actually like hacking into things, but just capable with computers. And he was scraping the... I think it was maybe Boston Dynamics, but it might have just been like a humanoid robot company at the time. And there's a tool where you can go and get like a list of files and URLs on a server. You can just ask for it as a web crawler. You don't actually need to click through all the links and look at the HTML. You can just see what does this server actually have? Let me go browse it. And one of the images was like an artist rendering of the humanoid with a gun, like full murderbot mode that never made it to the publication. It was never like, yeah, we're going to do that, because it was like, we're still in friendly mode. We were like, wow, that's wild. So this stuff has happened years ago, and so not quite a hack, but...
Certainly hacking equivalent of elbows over the table.
Yeah, I think that's accurate. Um, well, Jordy, did you have another one that you want to run through, or should we go to the all-important new car launch from none other than Bentley?
Uh, let's go over to this notification from HP Smart.
What happened? Oh, yes. Says you're low on ink and then never run low on ink again.
It does sound like a threat.
Tyler Wolf says, 'Is this an ad or a warning?'
It sounds like a threat. Just never do it. 8,000 likes. FO. Very funny.
Uh, Flock Safety is weighing whether to sell itself. Um, very interesting. I mean, obviously insane public backlash for that company. Uh, we've talked to Garrett about it. Uh, very interesting technology, but very, very controversial and a lot of cities are kicking...
What kind of acquirer would solve the problem?
Um, I mean, a rebrand and then also like a Taser company or something, you know, because there are companies that compete with them directly. Um, like there are plenty of Axon, military and police product companies. Uh, somebody was saying, yeah, Zach Weinberg was saying honestly Flock should just rename itself to the license plate scan company since that's actually what they do and the problem might just go away. I don't know if it'll actually go away. I think people are pretty deeply embedded. They'll chase this around for a while, but...
Well, that's the thing. I just feel like Flock is having the most insane comms crisis of any company that I can ever remember.
Yeah. Right. No, serious.
And who wants to inherit that crisis and get that attention pointed to them?
Totally. Yeah.
Um, well, Bentley's first electric car is set to be half the price of the Ferrari Luce. The launch of the luxury car maker's Torcal SUV is aimed at younger and first-time customers. What do you think? Are they over their skis with an EV at a time when people are sort of moving away from it, or does it work for this particular luxury SUV buyer?
If there's any place where having the silence of an EV, a Bentley sort of fits. This was the thesis with the Rolls-Royce Spectre. You take the engine out. Yeah, there's some enthusiasts who might be upset, but most people aren't buying it because they want to redline it. So it can potentially work. Do you think we'll be seeing a lot of these, or will it be passed over for the Urus and the new McLaren SUV potentially?
Well, I don't think we're going to be seeing a lot because Bentley is just not selling a lot of cars right now, but yeah, I think the car looks great minus the grill.
Minus the grill. The grill is completely... It looks a little bit more squared off than what I'm used to with Bentley.
Well, it's specifically like whatever lighting pattern this is.
Oh, so it might look different in sunlight, you mean?
Well, no. I just think this looks horrible.
Okay. Well, it's smaller than the Bentayga, which was the first. Like you have this sort of very clean, sporty feeling car.
This is the company's signature diamond pattern. Okay. Put some respect on it, apparently. I'm just learning this.
No, I know. But just imagine how good this car would look.
How would you describe the design overall? Would you say it's muscular? Because that's how they're describing it. Is it a muscular design?
I think so. Okay. I would agree with that. I'm just saying I think the car looks amazing minus this light.
Do you think it should have some more vascularity?
Maybe in a wrap you could add some rippling veins to it to give it more of a muscular design.
No.
Instead of recreating the V8 engine noise, the EV has a propulsion sound handcrafted by musicians with a unique blend of sounds from percussion, crystal glass, birds, and the tapping of leather. So as you drive it, it sings you a song.
The Volkswagen-owned luxury brand said it had also completed a $350 million investment to design and build the Toral at its plant in Crewe, bringing the total investment in the UK automotive sector to more than $1 billion.
The EV launch comes at a challenging period for Bentley as it struggles with slowing demand in China, fallout from Middle East conflict, higher US tariffs. As with the broader VW Group, Bentley has slimmed its workforce as profits and vehicle deliveries have declined.
It's serious. Uh, we have to bring back the competitiveness. We have to be quick. We have to work harder. You can hear this is the sound it makes when you drive it. I suppose you hear the... Oh, it's the finger on the edge of the glass. Can you hear that?
Imagine trying to drive with that sound.
Let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches. And we are ready for our first guests of the day from a stealth AI security lab. Welcome to the show. How are you both doing? Oh, we are going back to the timeline. Back to the timeline. There's other news.
Oh, okay. We are good. Welcome to the show. How you doing?
Hello. Hey, thank you for having us. Nice to see you.
Thanks for hopping on. Uh, please, since this is your first time on the show and first time talking about this topic, uh, please introduce yourselves and the company.
Have a good one. Goodbye.
Up next, we have the founder of Typesafe AI. You saw Jeff launch, massively viral launch. And we have the founder and CEO here to break it down for us. The new model's faster, structured decision software can use it directly. A lot of excitement, a lot of questions to answer. So let's bring him in. How are you doing? Welcome to the show.
You're doing so great.
So sleep-deprived and so excited. Yeah.
Sleep-deprived because you're... no time to sleep. Everyone's onboarding and demand is through the roof. Is that what's going on? Or is it fundraising?
Everything. Everyone wants to talk. Everything, every customer wants it. Like as far as I know, all the companies I've ever heard of are using us and the systems are on fire on dimensions I've never even known were possible. And I'm like a backend guy.
Yeah. So, uh, I imagine... Wait, but back up a little bit. When you launched this, like where, what was your expectation? I'm assuming you felt like, you know, you had something pretty cool. Polish launch. It's going to be a hit, right?
I don't think that this launch could have been expected if I'm like totally frank. It's like I've planned for a lot of scenarios. It's above the 100th percentile. People are calling it like a ChatGPT for developers and that shouldn't even be possible because ChatGPT is a consumer product, right? Like, how is this happening? And I have my own hypothesis, but it's insane.
Yeah, I imagine you've at this point honed the elevator pitch, but just for anyone who is not familiar with Jeff and the product and the value, how are you breaking it down for people these days?
Oh, I have not honed the elevator pitch, but I will try. I will try. People just tolerate my overly intense teching. Sure. But I would say the simple version is make AI actually useful for automation. I hope I can curse. But I think AI under-promise, over-deliver, over-promise, under-deliver has been a major issue for years and it's been like automating almost nothing and I think people are like super frustrated with that.
Okay. Uh, I saw people, uh, like there are trade-offs when you go for speed and reliability and you want to over-deliver. What are you giving up? What can Jeff not do? When is it not the right tool for the job?
Oh, there's a couple of dimensions for that. One of them is it doesn't generate strings. So like it will not write code for you and it will not address the largest market for LLMs right now, which is coding agents. I think that's perfect because actually what I want to do is like not the superhuman math, but the really boring automation, you know, like why can't simple data entry be solved and accounting and customer service. Uh, another trade-off is it doesn't do what we call system two thinking, like mathematical reasoning. It's really focused on instinctive judgment.
Okay. So then instinctive judgment can be subjective. One company can have a policy, they want to automate their accounting workflows one way, they have a definition for what is GAAP profit. The other one has a different one. So how much, like, you know, one-shot or preloading into like a context window is happening versus like talking to you, doing some fine-tuning? Like what terminology actually does carry over from what we know and what is entirely new?
So I would say like the former is the most accurate thing, but the real thing that's happening is this logic actually would go in code. So LLMs don't play nicely with code due to like a lot of properties about LLMs. And the logic you're describing should happen in code instead of like in the system message. The classic example of this is LLM does something bad and the way to prevent it is say pretty please don't do that in the system message on top and you just have to pray it doesn't get context-rotted out of it and it keeps on listening.
Yeah. Yeah. A lot of these things you're sort of like begging for JSON, begging for it to be structured the right way and it's not guaranteed to be that way. It might be that way very reliably and people have worked around that, but it's not 100%. Right.
Exactly. And we are not 100% either. That would be like crazy to guarantee intelligence. But what we try to do is we try to encourage the best practice and we try to be calibrated. So this is another problem with AI, like it might be able to be like Einstein 95% of the time, but Mr. Bean the other 5% of the time. But if it doesn't tell you if it's in like Einstein or Mr. Bean mode, like how are you going to automate anything, right? But like calibration...
It's just a funny reference, but I love it. It's really funny.
Thank you. Yeah, Einstein shot at Salesforce. Mr. Bean AI. We need that as well. Um, so yeah, where... so you mentioned like the servers are on fire. Are you compute constrained in the same way? Are you looking for like large GPU clusters and systems, or do some of the knock-on effects of like the speed of the system that I've seen on display, can it also run on like more commodity hardware? Are there any different elements to how you should think about your compute scaling, or is it going to look like any other NeoLab?
Uh, extremely different from other NeoLabs. Like, you know, people know that about a year and a half ago you raised less than $40 million. You can't get NeoLab compute with less than $40 million. And also you can't build a profitable AI company with that. Yet we have because we are just innovating, you know, we're just doing different stuff than everyone else is. So, uh, we have extremely different compute. Um, actually the bottleneck changes every single day. I think it's just growing pains of scaling a system. Like you can't grow at the rate that we're growing without realizing like, hey, our system was not designed for this level of stuff.
Got it. So it's a constant battle there.
Uh, I saw some pushback that I didn't really fully understand, but I'd love for you to sort of break down what people are saying and how you'd respond to it. This idea that like this is innovation, but it's also a return to a prior era, a classifier model. Some people are like, well, we've always had this. And like personally, from a business perspective, I don't care because if we need to pull something old off the shelf and it's amazing and it's delivering value, that's great. But is that real? Is that a good criticism? Is that completely misinformed? How do you process that idea that this is actually building on the shoulders of giants from years ago or a lost art?
Or a lost art. Yeah.
Yes. So, uh, I can curse, right? Uh, please. We tend not to, but do whatever you need to do. Just rip it.
Heck yes. It's fine. It's fine. I'll try to be controlled. I'm sorry if it gets off the rails. I love this question so much. Heck yes. Heck yes. It's a classifier. Is it just a classifier? No. But classifiers are awesome. Classifiers were not constructed because like a mathematician would think it's like a fun thing to do. Um, a classifier is designed to make intelligence useful in systems, right? That is why everything was classifiers before Gen AI because it was made for usefulness, not for demos or playing around.
And actually, we don't take all of the ML 1.0 concepts, right? Some of them don't apply. What we do is we try to unlock the best that like AI of today has. And some of that doesn't apply, but like the other stuff definitely does. This calibration thing of like telling you when it doesn't know, that is table stakes for simple automation. How can you have it like run in the background and process something that would normally require an accountant if you don't have like extreme guarantees on how smart it is? So absolutely it's a classifier and this is the kind of thing that I think ML people really don't get with our launch.
You know, like this is a launch for developers.
Uh, yes. Yes. I totally get that. Is there a version of this? You know, the story, and I mean you would know whether or not it's apocryphal, of like Google had LLMs, they invented the transformer, they had a product that was similar to ChatGPT and they didn't launch it that quickly. Is there a world where a technology like this maybe exists inside of a hyperscaler and they just haven't actually gotten it out into the world, or do you think that that's less likely to be the case?
I would guess that it either I'm a clown or it's extremely unlikely. We worked really hard on this. Like when I started, I was one of the best in the world at pre-training, uh, post-training, sorry, and I thought it would take a week.
So, um, I would be... and actually when I first... Yes. When I first started, I thought this is such obviously the way to go.
Okay.
It's so obvious that for sure Anthropic is working on this already. OpenAI is cooked. I was at OpenAI when I first started like thinking about this direction and it just seems so obvious to like someone who's at the intersection of developer and researcher, but so non-obvious to pure researchers.
Mhm. Yeah. Uh, wild. What's this thing that you're wearing?
Oh, uh, pink or the shirt?
I guess both.
Both. It's our Tist shirt. Anti-language language model. Okay.
Um, technically it's not a language model, but it is a sick brand, so we wear it. And this is my classic Teu $20 pink for... There you go.
Very good. Well, thanks so much for coming on the show and breaking it down. Congrats on the project.
Yeah, I can only imagine how busy you are. Good luck with, I'm sure, a new round serving all that demand.
We'll talk to you soon. Have a good one.
Heck yeah. See you around. Bye.
Goodbye.
See you.
A lot of fun. Great energy. Uh, up next we have the co-founder and CEO of Silca Industries, Hooman Reza Nazad. Let's bring Hooman in to the TV for Ultra. How are you doing? Welcome to the show.
