Loading summary
Transcript824 lines
- [00:01]
A
You're watching TVPN.
- [00:02]
B
Today is Thursday, January 29, 2026. We are live from the TVPN Ultra Dome. The temple of technology, the fortunes 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, a whole lot more. They got super bowl ads too. In just a few days. We're counting down the days to the ramp super bowl ad. We're very excited for it.
- [00:24]
A
Also not the Super Bowl. Counting down the days.
- [00:27]
B
Yeah, I learned, I learned a fun fact. I think I put it in here. Apparently they, they had a Super bowl back in 2001.
- [00:36]
A
Okay. Have you heard of this? Tell me more.
- [00:38]
B
So I got followed by this account that's called 2001 live. 2001 live. It's 25 years ago live. It's a cool account. And in here they, they flash back to like what was happening, you know, 25 years ago. Every day they post what happened 25 years ago. It's kind of cool account. Followed me, I followed it a while ago but if you followed me and I saw a post in here that said 25 years ago they did a Super bowl in Tampa, Florida. The Baltimore Ravens beat the New York Giants 34, 7. This is the Ravens first Super bowl title. I guess people are not happy about that. I don't know. But anyway, very excited for the Super Bowl. In other news, let's run through the linear lineup.
- [01:19]
A
Pull it up.
- [01:20]
B
Meet the system for modern software development. 70% of enterprise workspaces on linear are using agents. And we have a fantastic show. We have Anton from Lovable coming in studio in the ultrasound. Then we got Christian Delian from Hill and Valley breaking down what's going to happen at the Hill and Valley conference in just a few weeks.
- [01:38]
A
And then we got Eric Sufinator.
- [01:41]
B
We need a nickname. No, we love Eric of course from Mobile Dev Memo, Heracles Capital. And he's coming out at one. We're going to talk about ads, ChatGPT ads. How's rollout going? Also obviously earnings makes a ton of a sense. And then Kevin from OpenAI is coming on again.
- [01:57]
A
Returning back science and science, AI and science.
- [02:00]
B
And if we're seeing a clawed bot type moment, feeling the AGI in science, what we should be looking for there in 2026. In other AI news, of course we will go into earnings and whatnot today. If you're tracking earnings, make sure to head over to public.com investing for those that take it seriously. Stocks, options, bonds, crypto, treasuries and more with great customer service.
- [02:22]
A
So in the news today, Apple said this morning that it has acquired Q AI, an Israeli startup working on AI technology for audio. Apple didn't disclose the terms. The company was backed by Matter Venture Partners, Kleiner Spark, XOR and gv. So nice little lineup and great outcome. The Financial Times reported it was worth nearly 2 billion. So pretty meaningful deal. Apple is not known for really paying up on a lot of different M and A that they're kind of folding in to their roadmap. We had Mark Gurman on yesterday talking about how Apple uses M and A effectively to accelerate their roadmap. So according to Reuters, Apple did not say how it will use qai's technology, but said the startup has worked on new applications of machine learning to help devices understand whispered speech and. And to enhance audio in challenging environments. QAI last year filed a patent to use facial skin micro movements to detect words mouthed or spoken.
- [03:29]
B
This is, this is. We've seen a couple startups that are, you know, doing like the whisper. Like what do they call it? It's like telepathy almost. It tracks your mouth movements so you can just. Then it like will track what you sound.
- [03:43]
C
Really, really crazy.
- [03:44]
A
So facial skin micro movements will be used to identify a person and assess their emotions. Heart rate, respiration rate indicators. Very sci fi. Qai's 100 employees, including the CEO and co founders, will join Apple. So it seems like founders, yeah, you can. Aviad Meisel founded three dimensional sensing firm Prime Sense and sold it to Apple in 2013.
- [04:08]
B
Absolute dogs, they're doing it.
- [04:10]
C
They call it double kill.
- [04:11]
A
The Prime Sense deal eventually helped Apple move away from fingerprint sensors on its iPhone and toward facial recognition technology.
- [04:18]
B
Interesting.
- [04:19]
A
Nell said joining Apple opens extraordinary possibilities for pushing boundaries and realizing the full potential of what we've created. So very, very wild to be going back to Apple for a second time.
- [04:32]
B
And pop quiz for Tyler, why does Apple acquire companies?
- [04:36]
D
I mean, why? Questions are usually pretty.
- [04:39]
B
Okay, good answer. But Mark Gurman told us, why does Apple acquire companies to accelerate this. Accelerate the roadmap. Yes. And it's funny because I asked him why are we gonna hear Tim Cook say that? And he was like, oh, because earnings. But we're basically hearing him say it today with the surprise acquisition. That's obviously the line from Apple and it makes sense. This is something that is uniquely acceleratable because of the Apple hardware ecosystem. They can Deploy this through AirPods. They can deploy this through the phones just like they did with Face id. If you have that technology and you're like, oh well, in order to log into your computer with your face, you're going to need a third party device that you plug into the usb. Like no one's going to do that. But if you're just like, yeah, it's going to get baked into the hardware. We have the patents, we have the technology.
- [05:27]
A
And haven't you been wanting them to make more moves in specifically audio and transcription for a while?
- [05:33]
B
I've been like an audio interface bull for ever. Like since I worked at a startup that competed with Siri in 2010 or something like that. It was actually a good outcome. I didn't get an equity because I was just an intern, but it was actually a great outcome. They sold to Dragon, naturally speaking and everyone made a bunch of money but Siri. Here we go. But Siri was like the hot kid on the block back then. And Siri came out of Stanford, so they were much closer to Cupertino. These guys came out of mit. Very similar natural speech recognition, machine learning pipelines for understanding speech, doing dictation. Their main platform was BlackBerry because BlackBerry was really, really big. They had a BlackBerry app store. They did a deal so that you would get pre installed so you'd get a big check from rim. They did a very interesting out of home campaign where they bought billboards right outside of Research in Motion, the maker of BlackBerry's headquarters, because they were advertising. Yes, Adquick didn't exist then, but they should have used Adquick because they were basically, they only had one customer effectively. I mean they had every customer that could go buy the app. But then all the other apps that were on either Google and Android quickly rolled out competitive features with Google Assistant. Siri got acquired and so it was very clear that like Apple was not going to be a really fertile ground for playground for them. And so they needed to get sort of like the attention of the BlackBerry execs. And so they put up a billboard right outside their headquarters, which I still like as a strategy. I think it's very fun. But yes, I've been very bullish on this. I was thinking back then that people get their wisdom teeth out. This is very weird in cyberpunk, but if you got your wisdom teeth out, you could potentially create, add a port for storing a microphone.
- [07:29]
A
Tyler, did you have any of these?
- [07:30]
D
I should have got that.
- [07:32]
B
Yeah, exactly. Because a lot of people get them out and then you could just put a port there and you could insert basically the tip of an AirPod. Very, very small device that you would charge and then you put it in and then when you're whispering, you can just hear it can only hear that. And it goes dictation straight to your phone. Obviously that's unnecessary because you can do bone induction. You can just also just have an AirPod in and just talk out loud and not care what people think. There's those crazy muffled things. You've seen that where you put on the headphones and then it puts like a cover over your mouth so you can talk in public. Oh, yeah, it's silent around you.
- [08:10]
A
Kanner says, I want my iPhone to unlock only if I silently mouth the words open sesame. Think. Could be in the future.
- [08:18]
B
I was wondering about.
- [08:21]
A
That would actually be nice if there was some other element because you still have this, like wearing sunglasses, trying to unlock your phone sometimes some pairs of sunglasses. I have it one shots it others. It's like, you know, pretty annoying.
- [08:34]
B
I was wondering about the quality of Whisper transcription these days. Like, if I just open up whisper on the ChatGPT app, put my app, put my phone in my pocket and just talk, can it just hear through my pants pocket and just dictate perfectly now because, like, AI is so good at transcription that it can be really muffled. Like, you can have music playing in the background and talk to Whisper and it just won't. It just won't mess up. So there is a world where you don't actually need like a separate pin. You just have something anywhere on your body and it knows, okay, this is what John sounds like. Let's kill all the background noise. Let's go kill all the other people talking. Let's isolate that with AI. It seems pretty, pretty good for that. Anyway, let me tell you about Lambda Lambda is the super intelligence cloud building AI supercomputers for training and inference that scale from one GPU to hundreds of thousands.
- [09:26]
A
One more thing on this QAI acquisition they're using, going back to their patent, using facial skin, micro movements to assess emotions, heart rate, respiration rate, and other indicators. You could imagine where a world in the future where you have Apple smart glasses that effectively have like health monitoring, tracking, respiration, heart rate, all these different things, and just basically integrating the features you're getting from the Apple watch today. Seems.
- [09:56]
B
But I wonder what it's really unlocking that the watch can't do. There's always the question when you, when you pitch a new device, it's like, well, your phone does have a camera on it, so you're not. This is like the smart glasses aren't the first.
- [10:08]
A
Yeah, but it's a new device. But it's a Lindy form factor, right? True, true. Like this something we've talked about in the past. It's like the new hardware that's taken off is an existing form factor. It's like headphones, eyewear, you have watches, right. Creating these pendant things so far hasn't hit right.
- [10:27]
B
It's fair. I just think on the heart rate issue specifically, there was an app before the Apple watch existed where you could put your finger over the camera of the iPhone and it would use the light, the flashlight which was right next to the camera at that point to light up your finger. So your finger would turn red and then it would take the sensor data from the camera and measure the pulses of the red and give you your heart rate just from touching your camera. It was pretty cool. It was completely outside of the Apple ecosystem. Just, just an app that you could download or pay for. And so the phone can take your heart rate, the watch can take your heart rate. If you give me glasses and you say those glasses can also take your heart rate, I'm like, I got enough heart rate measurement. I can also just go like this and estimate it. There's a bunch of ways to know that your heart rate's spiking, maybe tracking it's in different. But it feels like they need to go farther and the real opportunity is something more around. Audio interfaces, link it to Siri, have more ways to triangulate what the person's trying to say, what they're saying in a noisy environment, whether they're trying to be quiet and you still want to isolate what they're saying. Having a back and forth, reducing latency, all of those things are very critical to success in that category. Anyway, we are going to be in San Francisco on February 3rd next week for Cisco's AI Summit. It's bringing together leaders from Nvidia, OpenAI, AWS and more to discuss the future of the AI economy. The whole thing will be live streamed and we'll be there for a gigant stream. Did the fireworks get longer?
- [12:01]
A
I think so.
- [12:02]
B
I think they got longer today.
- [12:04]
A
There's a lot to celebrate. John, let's go into Microsoft earnings.
- [12:08]
B
Microsoft. So Microsoft shares have taken a dive as data center spending overshadows earnings surge. Let's give some numbers here. So Microsoft's Q4 revenue was $81.3 billion, which was higher than the consensus estimate of 80.23 billion. So they're making money. There's no lack of demand For Microsoft services Q3 2025 revenue was 77.1 billion, up almost 5%. And year over year growth was 17% for Microsoft as a whole. But the stock sold off by 12%. And Microsoft is now just a tiny little $3.15 trillion company.
- [12:51]
A
Not bad, but still only down what, 6% over the past five days.
- [12:56]
B
There was a rise, yeah, I mean, people get excited about the OpenAI investment, which did show up in earnings. So OpenAI owns roughly 27% of the new for profit entity.
- [13:08]
A
Sorry, Microsoft.
- [13:10]
B
Microsoft owns 27% of OpenAI's new for profit entity. And that value is actually reflected in Microsoft's earnings. And of course, more importantly, the GPT models are truly frontier. Like we've seen it again and again. There's like this horse race over this model's better at this thing, this bottle is better today. But it's just very clear that OpenAI is on the frontier and in the conversation for pretty much every application possible. And so you can just imagine, you take GPT 5.2 Pro, you vend that into knowledge work pipelines for Microsoft users. That sounds really useful. They aren't behind on coding. We've heard great stories about how Codex is a great model. Maybe it's a little slow, maybe they need to speed it up, but people are having a lot of luck with the Codex. And so you can imagine that Microsoft is capable of integrating Codex into all sorts of different pieces of the Microsoft empire to create agentic workflows. And then they also have a deal with Anthropic. They also made an investment. So they are multimodal, multi platform. But the problem is that Microsoft seems to be constrained on the data center side. So the CFO Amy Hood said on the Microsoft earnings call that limited availability of artificial intelligence hardware is affecting how quickly Microsoft's cloud cloud business can grow. And it's capping Azure's revenue potential. Not good. They need to take another trip to Abilene. They need to build more data centers. But it's tricky, it takes time and right now maybe it's just a data center capacity issue. What's next? Is it going to be an energy bottleneck? Is it going to be a chip bottleneck? These are stories that we're tracking. Ben Thompson is starting to sound the alarm bells around tsmc. TSMC capacity constraints. And what is TSMC doing on the capex front? Are they investing enough? Are they at risk of holding the bag if AI stalls out? Are they taking enough risks that if the AI boom continues that they can continue to deliver and ramp up or will Samsung and Intel need to step up and will companies like Microsoft need to put pressure on those companies? So that's another point of debate.
- [15:22]
A
Yeah. So David in the chat is referencing this, but basically, obviously it's amazing that Microsoft owns such a big slug of OpenAI. That's great. But the challenge is so much of their backlog is OpenAI and they're actually getting less credit for that.
- [15:36]
B
Right.
- [15:36]
A
The same way that Oracle had gotten punished for it. Now it's Microsoft's chance to actually get punished. And Microsoft, if you actually zoom out a little bit and just look at the last six months, down 17% in the last, in the last six months and down 4% over the last year. So it's funny like in a year where it feels like the last year Satya and Microsoft have just been on this insane run, they've fully round tripped.
- [16:05]
B
Totally.
- [16:06]
C
Totally.
- [16:06]
B
Yeah. Before we move on to Meta, let me tell you about Sentry. Sentry shows developers what's broken and helps them fix it Fast. That's why 150,000 organizations use Keep their apps working. So Meta also reported earnings yesterday. 59.9 billion in revenue in Q4 of 2025, beating expectations of 58.5 revenues, up 16% from Q3 of last year, which was 51.24. The company's growing its revenue 21% year over year. That's higher than Microsoft's 17%. Top line growth stock popped 10% after hours and the market cap is now 1.4 trillion. So mark Zuckerberg told analysts on the earnings call in 2025, we rebuilt the foundations of our AI program. That should be obvious. So many, so many acquisitions, so many hires, so much, so many experiments, so many different, so many different strategies and discussions and changing of the guard and restructurings, layoffs. In Reality Labs, it's the compute desk. There's been so many stories about Meta really rethinking their AI platform, re architecting the foundations of it. So he continued to say, over the coming months, we're going to start shipping our new models and products. Very excited for that. He says he's I expect our first models will be good, but more importantly, they will show the rapid trajectory we're on. And I think that we all have high expectations. I don't think anyone's expecting them to jump way, way out in front of everyone else. But if they can just be in the conversation with DeepMind, Anthropic OpenAI, I think that will quell a lot of the concerns.
- [17:55]
A
That, but also what are they doing with them? I think that matters more. Really, really important because you look at. This is why I'm excited to talk to Eric later today. But you look at Meta's business and the way in which Gen AI can accelerate everything from generating more content on the platform to having better ads to better targeting. All these things. Right. And not to mention, like, where they can just vend it in at the product level. Right?
- [18:22]
B
The product stuff is tricky.
- [18:24]
A
Yeah.
- [18:25]
B
I mean, like, I've bumped into Meta AI in Instagram many times and you do get reasonable, you know, natural language responses, but it clearly still has the knowledge cut off. It's not searching the web as effectively. It's not pulling together. It doesn't feel completely native to the platform. Like, if you go into Meta AI in Instagram and you ask it to go and hunt around in Instagram for a specific creator or reel based on some clues, it's. It doesn't feel like it has the hooks to really go in and understand. Okay, based on what you've watched and what I've showed you in the past, this is probably what you're thinking of. That is sort of a superpower where there's so many times when you're on the timeline and you're like, I saw this post, I didn't bookmark it, didn't like it, what was it? And you want the search products to be empowered magically. What are we laughing at? What are you thinking? You gonna flashbang me again?
- [19:28]
E
Yeah.
- [19:29]
A
Close. Continue.
- [19:32]
B
Anyway, I think that there's the basic case of they gotta get an LLM that's Frontier, that has the big model smell, fun to talk to, good vibes. Then they need video and audio models that are rock solid. And then they do need to ven those in. I think just having an API or just having a place where people can. Can generate photoreal videos, or even things like Sora where it has the aesthetics and pacing and cuts of an Instagram reel that doesn't feel like it's enough. It feels like to really empower the Instagram creator, it needs to be built into the platform. Like the captions are, or those, the face filters on Snapchat that started there. Augmented reality stuff, replacing backgrounds, just letting people still bring what's personal to them. Their family, their experiences, their car. But take a couple photos of their car and turn it into a really, really awesome drone shot of them driving their car. I've seen a lot of really sweet edits where people will fly a drone over a car. Then they'll have a first person GoPro on their chest while they're driving their car. And then they'll use AI to interpolate between the drone shot and their first person view because they can't actually like if you have a multimillion dollar Hollywood budget, you actually can fly the drone into the car, have someone sitting in the car, they grab the drone and then they hook it on a crane and the crane takes it out and does a different shot with it. But that's like a multimillion dollar experts, tons of equipment. There's all these crazy things where someone will be on a Steadicam, on a cherry picker, this, this, this crane comes down, then they start walking into. You can do these really elaborate shots but with AI you can, you can, you can interpolate and like make those transitions. And so even just like AI powered transitions would be a really, really cool thing to bring.
- [21:26]
A
Yeah. Not to mention they, they own Manus now, which is a fantastic product team. They've built some great agents. You can imagine them integrating like, basically like prompt to short form video. Right. Where you can just describe the video that you want to make. Insert real like basically like generate B roll for this, pull footage from around the Internet, whatever. And so there's so many things that they can do. And again, they've just been using Meta AI as a sandbox for the most part. But I'm just excited for them to start shipping across.
- [21:56]
B
The remixing thing is so underrated because there's a lot of people, the number of people that have true inspiration for new formats is pretty low. People always do the same thing.
- [22:06]
F
Yeah.
- [22:06]
A
Think about the remix functionality. If you see a funny video and you can just like basically do a character swap in there.
- [22:14]
B
Exactly.
- [22:14]
A
And it's like that's an interesting content that's going to be shared and a bunch of people engage with.
- [22:20]
B
Tyler, what's your take on Meta's new plans for 2026?
- [22:23]
D
Yeah, I mean I definitely think the image and video models are much more important to get right than the LLM. Mainly just because like it seems like very natural them to vend it into everything compared to OpenAI or Google who are like right now finding it out in image or in video.
- [22:38]
B
Yeah.
- [22:38]
D
So it seems like if they can kind of do really well, you can get OpenAI and Google out of that race basically.
- [22:44]
B
Yeah, yeah.
- [22:44]
D
And then the lm, it's like it's, it's unclear what the actual like use case will be in the short term. Eventually you want some Some cool, like, you know, Cloudbot style agent somehow vended in, but it's unclear how that's going to work out. So I think in the short term I'm very excited on the video model.
- [22:58]
B
Yeah, I mean, I could, I really like that model of the LLM going around and just seeping into the cracks of all the different product experiences, but in these really subtle ways. YouTube now has AI generated summaries on videos, and you can chat with a video. So if you're watching someone build a PC, you can ask Gemini on YouTube, hey, just print out the exact list of parts that the person used to build build the PC. And there might be a parts list at the end of the video. There might be a parts list that's randomly mentioned throughout. Sometimes a creator might actually link to a real parts list, but Gemini allows you to scrape the transcript and then just get that however you want and then transform it or add prices to all of it or see if it's available in Japan, because I'm in Japan. All these interesting things. I can imagine on Instagram being able to go to a post that has thousands of comments and just say, hey, I want the LLM to kind of summarize the sentiment. Like, like, what are the facts? Like, you know, people were debating whether or not this is AI. Is there a consensus? Or people were adding context. What was the key context that people were adding? There's a lot of posts that are basically like rage bait or clickbait, where it's unclear what's going on in the video. And so you go to the comments, people are like, I don't get it. Or like they put up that sign, like, context needed, please. Right. And so you can kind of do that.
- [24:19]
D
Yeah, but I feel like that stuff can just like, I'd rather just have that be in the, like, recommendation algorithm, like sorted.
- [24:24]
B
If it's like comments, if everyone in.
- [24:25]
D
The comments is saying, this is clickbait, just don't put it in my recommendation algorithm. Right. It's like also on X, there's Grok right on every post. You can ask Grok. I don't think I've ever used that.
- [24:35]
B
You've never used that? I use that.
- [24:37]
D
I find a little like, you know, where you add a little bit of AI here and there. I like, basically use any of those. Maybe that's just, you know, I'm a fan.
- [24:45]
B
I do see posts where there's a list of companies that someone's talking about. Oh, here are all the cool companies that I like. And I can Click and just get a little summary of each company and I don't need to pull it out, go to a different model.
- [24:57]
A
I don't know, I feel like once a day Grok will share context on something that is actually valuable.
- [25:03]
B
Yeah, totally.
- [25:03]
A
Like in general because somebody added it. Like, I remember Eric Suefort yesterday was responding to something somebody was basically saying, like, shareholders voted against the Tesla XAI investment.
- [25:17]
B
Yeah.
- [25:17]
A
And then obviously they went ahead with it and Eric responded and he was like, hey, was this real? And then Grok just shared kind of the history on it and it's like somebody else could have shared that and provided the context. But yeah, I can do it instantly.
- [25:28]
B
Helpful to have a little bit extra context there. Applovin profitable advertising made Easy with Axon AI. Get access to over 1 billion daily active users and grow your business today.
- [25:39]
A
Suspended Cap says I gotta really respect Zuckerberg willing to spend over 50% of revenue next year when they still haven't delivered a single compelling AI product. Hell yeah.
- [25:49]
B
Yeah. I mean, the CapEx is crazy. So.
- [25:51]
A
No, the whole thing though, it's $200.
- [25:53]
B
Billion in revenue in 2025. That's founder. No, so much revenue. It's a lot of ads. But now they're gonna plow 135 billion, according to the Wall Street Journal. New York Times said 115, but either way it's like more than half of the revenues.
- [26:10]
A
I think posts like this are funny and I think you can definitely agree that Meta has not shipped a super compelling AI product yet. Even though Meta Vibes has traction. Not necessarily in our world, but certainly has some traction. But you have to. This is the guy that owns the world's largest trough or one of the world's largest troughs. Right. And so he knows that it's working. He knows he can see the future. Right. He has all the data. He knows that people say they don't like AI content, but in reality they actually do. They engage with it, they watch it, they make it.
- [26:46]
B
The engagement must be growing exponentially even though it's very small. It started at a base of zero and then you got Harry Potter, Balenciaga, and Now you have 10 videos a week that are going so fast.
- [26:56]
A
So I look at this different than some of the Metaverse bets just because Zuck is one of the biggest beneficiaries of Gen AI. And so it's totally warranted to say, like, hey, we should invest an obscene amount of money in this. This is clearly the future.
- [27:11]
B
Yeah, yeah. I mean, there's so many different ways that no matter how you play out the AI future, Meta is a beneficiary of that. So it makes sense that they're investing so heavily. But it is like a remarkable shift in the financial structure of the business. If you go back to thinking of previous years, spending more than half of revenue on Capex feels like a lot for a software company, right? It's like high margin and whatnot. Anyway, Vibe Co where D2C brands, B2B startups and AI companies advertise on streaming TV, pick channels, target audiences, measure sales, just like on Meta.
- [27:46]
A
Let's talk about Tesla.
- [27:49]
B
Tesla reported revenue at 24.9 billion and this beat the consensus estimate just slightly. Consensus was 24.78 billion, but down 11.4% from the year over year quarter when revenue was 28.1 and 3% decline year over year. And Tesla's down 2% this morning. There was also a 61% drop in profit. And they're shuttering the high end models of production. So they're stopping producing Model S's and Model X's, which I think the Model S and Model X features will be sort of rolled into trim levels like we've already heard about the. So I was debating Jordi about this earlier. I thought that the Model Y and the Model X were the same length. The Model X is in fact almost a foot longer. So it is a bigger vehicle. They both have the option to have three rows. They both have very similar specs, but the Model X is bigger. But in China we've heard that they're working on the Model Y L which is a long wheelbase version that will be, I believe, longer than the Model X. So you have the ability to get a big Model Y. And then they're already doing different trim levels with long range plaid. So if you take the plaid badge and you put that on the Model Y L and then you add different specs and you say, oh, I want to spec my Model Y with gullwing doors. Maybe all of a sudden you're back in Model X territory and you're dropping 100k. But there's just no, there's no denying that the Model S and X sales have slowed and there's a whole bunch more competition at the high end EV market from Lucid Rivian. And so Elon is fully thinking about what's next. He broke out subscriptions for Autopilot Self driving for the first time. He's talking a huge amount about cyber cabs and robo taxis. He's making that cash Investment in xai. And of course he's really focused on Optimus humanoid robots. And it seems like he could be scaling up production there very, very quickly.
- [29:54]
A
And so, analysts, million subs of full self driving.
- [29:58]
B
Yeah, not bad.
- [29:59]
A
And it's what, 89?
- [30:01]
B
I actually don't know.
- [30:02]
A
I think you can look it up. I think I remember it's $89 a month.
- [30:06]
B
Yeah. But analysts, how much is it?
- [30:09]
A
I guess $49 a month for vehicles with enhanced autopilot.
- [30:14]
B
There are a few tiers, but I mean, if you build up the subscription, what is it? How much is it?
- [30:20]
A
I think it's 100.
- [30:21]
B
100Amonth? Yeah. So you got a billion dollars a year coming in from that. Not bad. And of course, very high margin because you don't need to manufacture it. Analysts thought Tesla was going to be cash flow negative for the quarter, but they actually were positive. They generated $1.4 billion in free cash flow and this was down just 30%. So there's plenty of cash to keep the aggressive investments going, especially as Elon shifts the business towards autonomy. He's not shy about making the change aggressively. He said we would expect over time to make far more cyber Cabs than all of our other vehicles combined. So he's, he's fully, fully going all in on.
- [30:57]
A
Yeah, it's a little, I think it's a little jarring for some people just because historically car companies have thrived by creating the perfect car in each category for every different consumer. And Elon is basically saying, actually, I know what you want and I'm gonna give you it. Yeah, it's, you know, you don't need that many options where you're gonna be able, like you said, to just leverage the different trim levels and spec out the car to satisfy.
- [31:22]
B
And probably, you know, I do think over time, like less and less trim levels and then, and then eventually it's just cyber cabs and no one's buying cars anymore. If it really goes that way, I think this rollout will be slow. I mean, the original rollout of the electric car was 20 years. Right. For Tesla to actually get to saturation, where you're seeing electric cars all over the roads, it'll probably be the same with people giving up their cars and going all in on. I only use Waymo and Tesla Cyber Cab, or maybe I just use one. But Elon certainly thinking in decades and not afraid to cut a entire business line that is still popular with a lot of people. I mean, I was just talking to somebody yesterday who was singing the praises of his Tesla Model X and how much he loves that and how he would never get a Y because the X is so much more premium. Everything about it's better. It cost a lot of money when he bought it. Still loves it. But you know, Elon's thinking to the future anyway. Gemini 3 Pro, Google's most intelligent model yet. State of the art reasoning, next level vibe coding and deep multimodal understanding. I love that sound so much.
- [32:26]
A
I know you do. Tesla also announced a $2 billion investment into XAI.
- [32:35]
B
Yeah, they just did 20. Is this part of the 20, I wonder. I don't know.
- [32:40]
A
But yeah. So there was a non binding Tesla shareholder vote on authorizing an investment in XAI that failed although 1.06 billion votes were in favor versus 916 million against. Absten absentations. Abstentions.
- [32:57]
B
Like, like, like.
- [32:58]
A
Yeah. Not one or the other counted as no under Tesla bylaws leading to rejection. So they had run this as a vote. It got rejected. But Elon said we're doing it anyways.
- [33:11]
B
Doing it live. We should pull up reactions.
- [33:14]
A
Yeah, pull up this here.
- [33:15]
B
I don't know where you want to start, but take him. Has a good one if you want to start there with Microsoft.
- [33:21]
A
I wanted to pull up this video of Optimus learning it has to make up for S X sales after they.
- [33:27]
B
Were canceled this one. This is such a crazy video. Taking off the VR headset and just smash again.
- [33:34]
C
I love that.
- [33:35]
B
This is really a test Optimus, isn't it? At the same time, like the motion's remarkable and the force with which the Optimus just smashes the water bottle open is crazy.
- [33:46]
A
It's amazing. This thing is going to be super powerful.
- [33:48]
B
The insurance business that will be built around having a humanoid in your home is going to be remarkable. Remarkable. You're also going to need to secure these things. So let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops glitching.
- [34:05]
A
This was one of the standout moments of the earnings call for me, please. Elon and Tesla are transitioning their Fremont facility to make Optimus and they plan to scale that facility up to be able to make a million of these things a year on a relatively near term time horizon. So very, very significant. He talked about how the robot would be able to basically learn on the job it's going to be able to do a number of valuable tasks. And yeah, I think, I mean the big question for me is like what is.
- [34:41]
B
That's A good question for me.
- [34:42]
A
Will they have an ad supporter?
- [34:43]
B
You have the Tesla walking around your house and it sees you pull out some sort of random credit card and it's like, are you not on ramp? What's going on here?
- [34:51]
A
Or it sees you having like, sir.
- [34:53]
B
Would you like a smoothie from athletic greens?
- [34:56]
A
Yeah, exactly. But, yeah, we'll see. So the question I have is, at what point will humanoids actually be like, ROI positive for consumers in the home or even just everyday businesses? And I would just judge that based on, like, okay, you're buying this thing up front. Maybe a business finances it, maybe somebody finances or they just pay cash. But is it valuable? Is it, is it valuable enough to actually replace a human? Because you're competing with. The optimist is going to be competing with jobs that are maybe like 40, 50, 60 a year, like somewhere in that range.
- [35:34]
B
Yeah.
- [35:35]
A
And so that's a pretty high bar to clear. So we'll see.
- [35:38]
B
Yeah. And the risk reward calculation of, okay, you have this thing, it has cameras and audio. It's on all the time. Is it safe? If it falls over, is it going to crush my dog or hurt someone or smash something even? Just like doing the dishes. We've seen some incredible demos of humanoid robots doing the dishes, but you have to imagine that every once in a while it's going to break a dish. If it's breaking dishes at 10 times the rate of humans, it needs to be superhuman at not breaking dishes. Because if it smashes a wine glass while it's doing the dishes, and can I clean that up, is it going to be able to do that? On day one, there will be a slow ramp of people feeling like, okay, it's good enough. I'm getting a lot of value. Obviously, there'll be a lot of novelty. There's already a lot of novelty. The Optimus is deployed in most of the Tesla showrooms.
- [36:29]
A
Some breaking news. Yes, from Reuters 21 minutes ago. Exclusive. Musk's XAI and merger talks with XAI ahead of planned IPO. SpaceX and X. SpaceX and XAI.
- [36:41]
F
Wow.
- [36:41]
A
Had a planned IPO. Okay, so this was something that many people were talking about predicting months ago at this point. But so, so no, no huge surprise here. This always felt like it made sense. I'll read through the article. I don't know how much context it actually has. So they're in discussions ahead of a blockbuster IPO planned for later this year. The combination would bring Musk's rockets, Starlink satellites and X social media platform. And The Grok AI chatbot under one roof. Imagine owning X, the Internet's dive bar and space, in one ticker.
- [37:19]
B
I mean, the companies launched a couple years apart, I think like 2007 and 2005 or something like flashing back to 2007, 2010, being like, yeah, Twitter and that space company that hasn't successfully launched anything. They're going to be part of the same company one day. That is a wild future. Yeah.
- [37:38]
A
I am curious to see if they end up at some point just spinning X, the social platform out again. Like, does it really need to live?
- [37:45]
B
I think Musk holding companies just gets bigger and bigger and bigger and it all goes together into one because there's a lot of efficiencies there. Sean Puri from My First Million had a post about this, how he thinks Elon should IPO his personal holding company. So you get stakes in everything that he has stakes in. And he was proposing it like a Berkshire type structure, but instead of Berkshire Hathaway, it would be like Musk Holdings. Right, or something. Or Musk's initials. I think a lot of people would be buyers of that. But it'll be interesting to see the S1s getting more complicated by the day as the investment banks prepare for the June IPO for SpaceX. Now is June, that's the target. But it's late in June. It's June 29th or 19th or something.
- [38:34]
A
Yeah, yeah. So, yeah, we'll see. I mean, expected this to happen. I think it builds a. You know, again, some people will be frustrated with the narrative, the data centers in space narrative. But no, it's real.
- [38:49]
B
It's real.
- [38:49]
A
And we saw this sort of like, sort of organized narrative shift around SpaceX being the data centers in space play.
- [38:57]
B
Yeah, that was the bridge. Like without that, it didn't make any sense. And then, and then once, if you can get behind, okay, data is in. Centers in space is maybe possible, maybe.
- [39:06]
A
Maybe you want some on land too.
- [39:09]
B
Then it starts to make sense that the merger fits a little bit more. The other interesting data centers in space thing, Xai SpaceX thing. I just, I don't know. There is a fair, like, odd history of telecom companies owning media companies. So Comcast, NBC Universal.
- [39:36]
A
Right.
- [39:37]
B
Comcast was the owner of that history.
- [39:41]
A
Rhyming.
- [39:42]
B
And so there is this world where you have the pipes and then you need something to push through the pipes. And so you own a media company that pushes content through that pipe. And it sort of rhymes. Right? It's like you have starlink Internet connections and then you need something to go over those Internet connections. So X, the social network pipes over those. If you squint, it's not the craziest thing. The other interesting thing that I want to do, or we should talk to some of the folks who have done some of this data center and space analysis is how powerful are the Starlink satellites right now? They have some onboard compute. They don't have Nvidia GPUs, they don't have whole racks in space, but they probably have some compute just to run the network. And I'm wondering in terms of energy draw, are they one order of magnitude off, two orders of magnitude off, how underpowered are the current Starlink satellites? Because that could be an indicator of. Okay, well they need to scale them up by this much. They need to add this capability. They need to do X, Y or z really quickly. 11 labs build intelligent real time conversational agents. Reimagine human technology interaction with 11 Labs.
- [40:58]
A
Speaking of 11 Labs, 11 Labs is the new sponsor of Audi F1.
- [41:02]
B
Yes.
- [41:03]
A
John, you sent me a picture the other day of Audi out, I think testing and you were like it's insane.
- [41:10]
B
That they don't have a 11 labs a lot logo on there. Exactly.
- [41:14]
A
The sponsors were looking pretty sparse, but they're just adding them kind of in real time because it's fun to roll them out. Super, super exciting and exciting. I love the Audi, the new Audi livery. It looks very unique. MKBHD responded to the Tesla is discontinuing the Model S and Model X and said oh, roadster is so cooked. Elon just responded a little bit ago said new Roadster will be incredible.
- [41:40]
B
So maybe.
- [41:41]
A
And MKBHD is in the process of getting community noted. Elon said on the earnings call they are aiming to unveil in April and production will commence in 12 to 18 months. The shareholder deck notes the Roadster is in design development. Lars VP of vehicle engineering said the Roadster was definitely in development. So it will be pretty funny to be in like, you know, something that hopefully will look like a hype look and feel like a hypercar, but then you're just self driving in traffic. It's like effectively like depending on like pricing. It's not going to be like price. I think I'm hoping it looks like a hypercar but prices like a Turbo S or something like that.
- [42:20]
B
Yeah, yeah. I think the secret to success is probably 200 to $300,000. But it looks like a cyber car or cybertruck huracan, like low to the ground, super insane performance, self Driving. But it has, like, entirely new aesthetics that turn heads. Like, the cybertruck is kind of.
- [42:44]
A
Yeah. What was he saying on Joe Rogan? He was talking. He was like. Wasn't he alluding to the roadster potentially being able to take flight of some sort?
- [42:52]
B
That's exactly it, yes. And we were debating with Doug Gmuro from Cars and Bids and his own YouTube channel. What does the definition of fly mean? Like, we've seen the Xiaomi car jump. It can jump over a pothole. It can jump a couple inches off. There's that Mercedes Maybach GLS that can bounce. So what does fly really mean? Fly a foot, fly 10ft. I don't think anyone's expecting that in the next year it will just actually take off and fly you from LA to San Francisco or something. That would be.
- [43:27]
A
Never bet against.
- [43:29]
B
I mean, I would. I would want that so, so badly. But there are a lot of other cool things that we could do. I think the one that I wasn't. The one that I predicted, like, basically, it. You'll be able to pull up to a parking spot, and it will be able to use fans or something like that to lift off the ground and slide into a parking spot or some. It'll have some gimmicky.
- [43:51]
A
It'll shoot the battery out so hard that it sends the car.
- [43:54]
B
Shoots the battery down, which pushes the car up into the air, and then there's an ejector seat. You will be flying. The car will not be flying. Who knows?
- [44:05]
A
Insane.
- [44:05]
B
Hopefully it's available on Shopify. Shopify is the commerce platform that grows your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces. And now with AI rounding, rounding really quickly, Tyler.
- [44:17]
D
I mean, this is about the question of, like, can you run GPUs on Starlink?
- [44:21]
B
Yes.
- [44:22]
D
So like a current satellite. If you're looking at the V2 mini satellite, the peak power draw is like, around, like, 3,000 watts. Okay, so if you're running an H100.
- [44:31]
A
Right.
- [44:32]
D
So this is, I mean, maybe a couple of years old. It's only like 700. So you can actually put it. You could put a few H1 hundreds on a Starlink right now.
- [44:39]
B
Wait, wait, wait. Sorry, sorry. 300 watts currently. And an H100 is 3,000. 3,000. Oh, okay. Yeah, you could put a couple. That's like the peak power in theory, but you wouldn't be able to do anything else. But, like.
- [44:48]
D
Yeah, I mean, like, three H1 hundreds is, like, not useful really at all for, like, any for serving any big model.
- [44:55]
A
But.
- [44:55]
B
Yeah, but we're not four orders magnitude off from doing something. I mean, you can do inference on just a laptop, right? And that doesn't have the power draw anywhere near an H100 rack. Granted, it's gonna be slow, it's not gonna be functional, but like, we are in the ballpark.
- [45:10]
A
Somebody just gonna put. Somebody should put a Mac mini on a weather balloon.
- [45:15]
B
Just send it. You really could. Have you ever seen those trends of like, we sent X to space. It's like a proven viral format on YouTube. I think someone sent pizza to space. People send like a piano to space and it plays a song. Basically, YouTubers figured out that you could get a weather balloon that you could just buy for like, I don't know, not that much money and just release it and it would just fly up, up, up, up, up, and then you would reach the sky and you would take pictures and time lapse photos or video automatically.
- [45:47]
A
You can get a weather balloon, a 22 foot scientific weather balloon on eBay for 220 bucks.
- [45:53]
B
What should we send to space? It's kind of played out, but yeah. Garlic bread to space. Bobby. Thank you. Yeah, the garlic bread. The garlic bread went to space. I don't know why. I don't know why. But yeah, the sending multi bot. Multibot to space on a. Yeah. I also, I don't know if you can get Starlink working.
- [46:13]
A
I think, I mean, ideally, Tyler, what if we got a handful of weather balloons and we sent you up there with.
- [46:20]
D
Then I could. I could parachute that Red Bull video.
- [46:23]
C
Yeah.
- [46:23]
B
Yeah.
- [46:24]
A
Just one shot it.
- [46:25]
B
Just do it.
- [46:26]
A
Just one shot it. Last thing on Elon. Looking at prediction market on Kalshi. When will Elon become a trillionaire? He's got a 65% chance before 2027. Just in time for the midterms. It's not like that will be used. It's not like that will be used against him at all.
- [46:46]
B
There's still a chance that it takes the pressure off the billionaires because they're like, oh, there's a new thing we got to focus on. Like, no one spoke to millionaires right now.
- [46:53]
A
All the billionaires start ganging up on Elon.
- [46:55]
B
Let's actually just do a trillionaire tax. This whole billionaire. Why are we focused on billionaires? Trillionaires have so much more.
- [47:00]
A
Waste of time.
- [47:01]
B
Waste of time. Let's just narrow it down to trillionaires. Just the trillionaire tax. Anyway. Console consul builds AI agents that automate 70% of it. HR and finance support giving employees instant resolution for access requests and password resets. Where else do we want to go? Do we want to go? More reactions to earnings. There's a whole bunch of other stuff in, in here. Microsoft is on sale, according to Jay Catesby because Joe Weisenthal said incredible day. According to Barclays Alexander Altman, Microsoft has lost $441 billion of market cap today, marking it the second largest drop ever since Nvidia lost nearly 600 billion after deep Seek. Wow. Yeah. It was the worst day for Microsoft since 2020 since like the COVID sell off. Absolutely remarkable. But I mean I don't fully get it. It seems like Microsoft is very well positioned. I mean maybe they're facing the same like implementation question as Meta, but it seems like they have all the key ingredients. I mean they don't have, they don't have. They certainly don't have like the crazy team that Zuck just went and poached. But they got OpenAI, they have the partnership and so they have access to the models where in a way that Meta does not. And so Meta does need to build the team to build the competitive models.
- [48:17]
A
So worst sell offs in Microsoft history. The crash of 87 then in 2000.
- [48:24]
B
April down 24th, 30% in one day in 87.
- [48:28]
A
We don't know how to brutal. We don't know how to create sell offs like that anymore. Hopefully we don't. Famous last words.
- [48:36]
B
Everything candles down 40%. Teach you a lesson?
- [48:39]
A
Yep.
- [48:40]
B
Yeah. This is an interesting history here.
- [48:43]
A
Well, Microsoft is unfazed.
- [48:46]
B
When you look at it this way, it is truly the 15th worst sell off in Microsoft history. I think they're going to get through it, but Stan disagrees. Stan says it's over. Thanks for playing to 0 likes. How did this even get. I'm going to like this.
- [49:05]
A
We're going to like it.
- [49:06]
B
Good. Stan, I don't agree but very funny post. Andrew Curran is sharing a post from the information. Nvidia, Microsoft and Amazon are all ponying up.
- [49:19]
A
Let's give it up for talks.
- [49:21]
B
60 billion. As much as 60 billion in OpenAI.
- [49:23]
A
We don't know what kind of talks they are. Are they early medium, medium, late stage.
- [49:28]
B
Talks, advanced talks, intense preliminary talks? There's lots of different talks. But this is on top of the 30 billion that SoftBank is in talks for, which means that the 100 billion target for the next round is already almost met. And of course this is going to go back to is this circular, blah, blah blah blah. I don't know. It feels like the circularity narrative, we're just at a scale where if you want to raise $100 billion, like who are you going to? Who, who would be okay? Like it's governments and the hyperscalers. Like there just aren't that many funds that are saying like yeah, I'd love to lead $100 billion round. I don't have the money. Like it's just, it's just impossible.
- [50:16]
A
They're tapped out.
- [50:16]
B
Like yeah, all the, all the traditional financial players are very much tapped out.
- [50:21]
A
And so let's go to, let's go to company market cap.com while we pull that up.
- [50:26]
B
Let me tell you about figma. Figma make isn't your average vibe coding tool. It lives in figma so outputs look good, feel real and stay connected to how teams build, create code back prototypes.
- [50:35]
A
Okay so at the 830 billion that this round is being discussed at, JP Morgan is slightly worth slightly more at 832 billion but then below that OpenAI will be valued more than Samsung $0.1, Visa, Exxon Mobil, ASML, Johnson and Johnson, MasterCard, Costco, a lot of them. Yeah, you start being able to add up some of the biggest companies in the world and still not be worth.
- [51:10]
B
More than zero Hedge had some extra context. Q3 2025 Meta crashes on surging CAPEX forecast Q4 2025 Metasaurus on surging Capex forecast Let's go research and invest pushed back though said no it soars on growth acceleration guide Capex held it back or could have opened or it could have opened at around 880 and it's because of the profit. A lot of people are chiming in so we'll talk to Buco.
- [51:35]
A
Buco has continuously had great calls on Meta. He I remember very distinctly when he would he shared multiple times over a few day period like Meta will start touching 600 and you need to develop the mental fortitude to buy even though everything, even though everything in your mind will be telling you not to do it. Just hit the button.
- [51:59]
B
Suspended Cap says I gotta really respect Zuck willing to spend over 50% of revenue next year when they still on Capex when they still haven't delivered a single compelling AI product. Hell yeah. But Pythiacap chimes in and says the ads are the compelling AI product and this has been the under discussed narrative with our Capex. They do break it out and for a long time 50% of CapEx more than 50% of CapEx was going towards core AI just workloads for Instagram Reels recommendations, ad recommendations, running the core product different. I mean, there is a specific chip for YouTube, there's capex that goes into just running YouTube, just running Instagram. And so it's not all AI, but it'll be interesting to see how that shifts. How much of this is okay? This is generative AI workloads going forward. Is it more than 50% of this new boost?
- [52:52]
A
Brother Joe Wiesenthal shares One of the best headlines that I've ever seen, which is Blackstone nears deal to become New World's largest shareholder.
- [53:02]
B
Just in general, no.
- [53:04]
A
New World is a company. Blackstone is in advance talks to become the single largest shareholder of New World Development Company according to people familiar with the matter. A move that would see one of Hong Kong's richest family families relinquish control of a major asset. Under the proposed deal, the US company would be able to restructure the embattled developer and New World could continue to try to offload assets to shore up liquidity. The family of Hong Kong tycoon Henry Chung currently holds about a 45% stake in New World. We need to bring the word tycoon to the west coast. Sam Altman is an AI tycoon.
- [53:41]
B
Tycoon is good. We should do a list of the top tycoons.
- [53:45]
A
Jensen is a chip tycoon.
- [53:48]
B
Yeah, tycoon's a good one.
- [53:50]
A
Satya is an enterprise tycoon.
- [53:52]
B
We had tycoons during the railroad barons. It got it turned into a bit of a pejorative. But I think we can turn it down or turn it around. I think we can turn it around.
- [54:00]
A
It's such a fun word. Why should it be negative?
- [54:02]
B
Yeah, it sounds like Yahoo. Or like, you know, it feels like you're wearing a cowboy hat on a horse. If you're a tycoon.
- [54:09]
A
That's right.
- [54:10]
B
You're running all over the economy. You're a tycoon.
- [54:13]
A
Cowboy hat on a horse.
- [54:15]
B
Yeah, you love it. Well, let me tell you about MongoDB. Choose database built for flexibility and scale. With best in class embedding models and re rankers. MongoDB has what you need to build. What's next?
- [54:26]
A
We have to talk about genie. Genie.
- [54:29]
B
The genie is out of the lamp.
- [54:30]
A
Logan says. Introducing Project Genie, a frontier world model product powered by Genie 3 and available to G1 Ultra users in the US starting today.
- [54:39]
B
Are you a G1 Ultra user?
- [54:41]
A
This morning we were playing around. We were playing around with this this morning. It is absolutely wild. You can basically prompt an entire world. It instantly Turns into effectively a simple video game. And you can create some really funny scenarios. We will show you.
- [54:59]
B
They added the jump button. They added the jump button. You get to pick if it's third person or I guess if you don't check that it's first person. But sometimes you. Even if you do check, even if you don't check that, you can still wind up in a third person game if it's obviously a third person request. Look at this dog.
- [55:15]
A
They need to add the crouch button.
- [55:17]
B
Crouch button next. And then the flashbang button probably.
- [55:20]
C
Whoa.
- [55:21]
B
Yeah, you're jumping. This is so. It's so fast. I mean, the previous GENIE launch was still called Genie 3, right?
- [55:29]
D
Well, no, I mean, yes. This is not like a new product. This is just making it public.
- [55:33]
B
Yes.
- [55:34]
D
So I think it was In August when Genie 3 was originally released, but it was basically just the paper. There were some demos, but no one got so.
- [55:41]
B
So cool. Yeah, I mean, it feels like more directable than VO3 in some ways. And it's certainly more stable as you move around.
- [55:51]
A
Well, it's just way cooler too, because it's a world you can move around in. It's not like VO3 where you're just creating a video.
- [55:59]
D
Yeah, the memory is really good too.
- [56:01]
B
Oh, you can upload an image and turn back. You can upload an image. I mean, get ready to play dinosaurs.
- [56:06]
A
Do you have the clip of.
- [56:09]
B
So we got access. We generated some. Now it's in such high demand that you might not be able to generate these worlds for yourself immediately. There might be some rate limits going on. I'm sure the GPUs are on fire. It is going to be Genie 3 day on the timeline for sure. People are going to get crazy creative with this.
- [56:28]
A
Do you have John's first prompt?
- [56:30]
B
You can't access the videos. No, because, oh, no, we didn't download them.
- [56:35]
D
I think the site is being overloaded so much.
- [56:37]
A
Shane. Hey, Tyler, take some responsibility. Take some ownership. You're 21 years old now. Take some ownership. You could have downloaded the video.
- [56:47]
D
As the show goes on, I'll try to make a new one.
- [56:49]
B
Yeah. I mean, yeah, yeah. This is a good point by Noah in the chat. Wait, can we just take a GTA 6 leak and generate the game so we don't have to keep waiting?
- [56:57]
A
So pretty much like first.
- [57:00]
B
I mean, it's incredible that even a product at Google scale, they knew that there was going to be demand, that they are like seeing rate limits. That's wild and incredibly bullish for AI and Shows that there's so much more demand and we need more chips and energy and Data centers and GPUs, obviously. But on the flip side, we are going to move the goalposts. This is AGI, but it's not sufficient AGI because my definition is not just the jump button. I want mechanics. We generated a video of Maybach driving on the Nurburgring. It was remarkably high fidelity. It was a little sluggish, but that might just be the dynamics.
- [57:38]
A
I think it was the driver.
- [57:38]
B
I think that's just. Oh, you think it was the driver?
- [57:40]
A
I thought it was a lot of body roll, this thing on the road. There was body roll.
- [57:44]
B
Yes.
- [57:44]
A
I was like, just put it in a straight line, Tyler.
- [57:46]
B
Yes. But I want. I'm waiting. I'm moving the goalpost because I want Genie 3 or Genie 4 to be able to generate game mechanics. If I say I'm racing on the Nurburgring, I want a track timer. I want to be able to stop, change my tires. I want to be able to get a refuel overtakes. I want overlays, I want boost pedals, I want. I want drs. I want shifting. I want the whole Forza simulation. The whole. I want. I want. What's the one that people actually use for the simulators. I forget what it's called. Assetto Corsa racing. I want to be able to generate Assetto Corsa in two seconds. So mechanics are clearly next. This is obviously.
- [58:26]
A
Yeah, we can pull this up.
- [58:27]
B
So Tyler said six months. We'll see over.
- [58:30]
A
So pull up this video from you.
- [58:32]
D
In six months, you'll have full games.
- [58:34]
B
This is the new AGI benchmark.
- [58:37]
A
Move it, move it, move it. There we go. Pull up this video from Ethan Mollick. He had early access to Genie 3 world modeling. Huge leap forward in modeling physics. But some issues remain. Here is a bit of an otter airline pilot with a duck on its head walking through a Rothko inspired airport. And an otter in a wingsuit flying through a city of gothic towers. So this is why this video is why we have $2 trillion of CapEx for AI. We just simply before AI, we would need a team of advanced motion graphics artists to work for.
- [59:23]
B
You could probably vibe code something like this in Unreal Engine in a Day if you were strong, John.
- [59:30]
A
Don't try to pop the bubble.
- [59:32]
B
You could. I like the wingsuit, though. This is fun. In a gothic tower. It looks sort of like a Harry Potter world.
- [59:39]
E
Wow.
- [59:39]
B
It's really a full city of gothic towers.
- [59:42]
A
Yeah. And so I like the thing that's wild is that it takes 20 seconds.
- [59:50]
B
Yeah.
- [59:50]
A
To go from idea to this world.
- [59:53]
E
Yeah.
- [59:53]
B
Super cool. So super, super cool. And you're just gonna see this, like, unlimited. Like, the people are gonna go and just generate Mario, and you'll just be like, well, I could just download Mario. Like, it's not like I'm okay with that. It's like when you go to, you know, midjourney, you say, make me a picture of a dog or. Or Nano Banana. And it's just like, that's just a picture of a dog. You just find a picture of a dog. What's cool is when it's like your dog in your house doing the specific mechanic that you wanted, and it's a reenactment from a film that your friends that you're in love with, and you're a mechanic, and you're creating this unique thing, and you're mixing together these interesting ideas. We're going to see a ton.
- [60:32]
A
Okay. So Tyler took full ownership and turned it around.
- [60:36]
B
Okay.
- [60:36]
D
We made a new video.
- [60:37]
B
While we pull that up, let me tell you about Plaid. Plaid powers the apps you use to spend, save, borrow, and invest securely connecting bank accounts to move money, fight fraud, and improve lending. Now with AI.
- [60:48]
A
All right, look at this.
- [60:48]
B
Oh, yeah, this is.
- [60:49]
G
Wait, this is a new one.
- [60:50]
B
This is a new one.
- [60:51]
D
I just made this.
- [60:52]
B
The original was not two tone.
- [60:54]
D
Yeah, this is a new. I just made this.
- [60:55]
B
Okay.
- [60:56]
A
Look at you barely keeping it on the road.
- [60:58]
B
It's really hard to keep still issue. You gotta. So if you zoom in on this, the license plate does look a little AI generated. There are some artifacts, but this, at this distance, this looks. This looks.
- [61:09]
A
Can he hit the apex?
- [61:10]
B
Turn. This looks photo real.
- [61:12]
A
Can he hit the apex?
- [61:14]
B
Can he hit the apex?
- [61:15]
A
Oh, the body roll.
- [61:18]
B
The body roll. You're actually going pretty fast. You're getting some speed on this thing.
- [61:22]
D
I think I fall off. I get off the.
- [61:23]
B
Oh, no. The grass.
- [61:25]
G
Oh, no.
- [61:26]
B
Wait, what happened to the.
- [61:28]
A
Oh, no.
- [61:28]
B
You're destroying the side rail. How'd you get on the other side of the side rail?
- [61:32]
H
It says.
- [61:33]
A
Looks inebriated.
- [61:34]
D
Yeah.
- [61:34]
A
Crazy timing. Tyler turns 21 and he can't even keep a car.
- [61:38]
F
Wow.
- [61:39]
B
That was very.
- [61:40]
A
I don't see any happy dads over there.
- [61:43]
D
Yeah, but it does. Like, when it was working so earlier, I think there were just too many people using it.
- [61:48]
E
Yeah.
- [61:48]
D
This took like maybe 15 seconds.
- [61:50]
B
That's remarkable.
- [61:51]
D
First it, like, makes an image, and then you can edit it.
- [61:53]
B
Yeah.
- [61:53]
D
And then from there you can make the actual world.
- [61:55]
B
Yeah, yeah.
- [61:56]
D
And it's like a minute long. Yeah, it's like, yeah. Incredible.
- [62:00]
B
Josh Woodward has another clip he shared nanobananapro. He's the VP at Google. Google Labs, Gemini AI Studio, nanobanana Plus Project Genie Low Poly cowboy dreams. They've been fulfilled. Let's take a look at what Josh Woodward over at Google built with Genie 3. So he takes the image he's generated with Nanobanana Pro, drops it in, summarizes the environment, summarizes. And it's interesting seeing like what does prompt engineering look like in this? In this, like how. What type of prompt do you want to put in? Because I think we just put in like two words. We put in like Nurburgring track Germany. What happens if you add a little bit more there? How far can you push this current state of the art? The horse can jump. AGI achieved.
- [62:49]
A
AGI achieved. Do you think this is bullish for platforms like Roblox and Fortnite that have the existing network and they can integrate world models so that the players can. That are already a part of these ecosystems and these economies can generate new worlds quickly, generate new games, new characters, etc. Or is it as these world models get better, do they become competitive, a bigger threat? Because anybody can just. I'm sure there's infrastructure providers that can say, like, yeah, we're going to handle everything from account creation to in game currency to things like that.
- [63:22]
B
Yeah, I mean, definitely competitive in the long term, in the medium term, super good for prototyping and communication. And this is a key flow to, okay, you have an idea and you don't just want to generate a basic image of the game that you're trying to build. You generate a demo, a prototype and then you go from here into, okay, let's wire it up in Unreal Engine or Roblox or Minecraft or whatever we want to do. And then you have the full gameplay.
- [63:52]
A
I think Roblox and Fortnite will prove that they have real network effects and it's going to still make sense to create new games within these existing ecosystems.
- [64:02]
B
Yeah, yeah, but it is a very different architecture. If you. But yeah, I don't know. I don't know how it, how it interact. Like if you're in Roblox and then you want to go into a fully generated experience, the persistence doesn't exist. You're not in necessarily the same world. Just building like multiplayer capabilities in here seems very difficult. That seems like a real research challenge.
- [64:23]
D
Yeah, I think broadly, like the World model like World Labs seems much more easy to integrate, at least in the short term, to something like Roblox, because, I mean, this is literally just like generating frames on the fly. You're not actually making a, you know, 3D, like rendering.
- [64:36]
B
But it seems like it would be useful for training Data as well. Yeah, lots of, lots of opportunities there. Well, speaking of training data, let me tell you about Label Box, RL environments, Voice robotics, evals and expert human data. Label Box is the data factory behind the world's leading AI labs.
- [64:53]
A
Well said, John.
- [64:55]
B
There are a lot of those. Google DeepMind created this short film. AI is going to disrupt Hollywood sooner than most might expect. It's their short film, Dear Upstairs Neighbor. It's previewing at Sundance Festival. It's a story about noisy neighbors, but behind the scenes, it's about solving a huge challenge in generative AI control. Developed by Pixar alumni, an Academy Award winner, researchers and engineers. Here's how it came together, says DeepMind. And there's a community note on here. Let's see what this said. The quoted video explains that the short film was driven through human made art with AI assisting and speeding up the process. OP is engaging. Poor Chubby, always getting community noted. We've got a number of these. I think that's like, at a certain.
- [65:41]
A
Point, maybe that's his bit assume that.
- [65:43]
B
Things are happening, but, you know, the point still holds that, you know, this is disruptive. Who knows if it's completely disrupting there certainly, I mean, there's demand for movies that are shot on film still. So how quickly will all this roll out? But if you have a vision these days for an animated movie, you should just go try and make it. At least you have to imagine that even if you want to use a more traditional process and go through the traditional Hollywood pipeline, showing up to a pitch meeting at an agency with a pretty much polished AI version of your film is going to resonate in a way that a script might just get sent back in the mailroom. So I don't know, I could imagine this in like the prototyping, the storyboarding phase.
- [66:33]
A
Really, really taking off, staying in Google land, please. Addie Osmani over at Google Cloud shared some big changes to Gemini in Chrome Agentic browsing with Auto browse, Nano Banana and more. Let's pull up this video. You can see it here.
- [66:52]
B
It's a huge day for tech news today. We had so much earnings. I mean, all week, all week there's been a lot of stuff, but the other stuff was like planned the cloudbot stuff earlier was not. Is there audio on here? Can we play this audio as well? Let's see.
- [67:10]
I
Chrome.
- [67:11]
B
Okay, let's see. This is Gemini in Chrome Agentic browsing with auto browse, nanobanana and more.
- [67:21]
A
Does it work?
- [67:22]
B
Let's see.
- [67:23]
A
It's a new state of the art model for in context image transformation. And traditionally, if you wanted to see how furniture would look in a room, you'd have to download a photo, upload it to an editor and hope for the best. Here with nanobanana, you just point Gemini to an image already already on your screen and in this prompt to show.
- [67:38]
B
Me AI in every feature. You actually like downloading the images and round tripping them. That's your preferred workflow. And every consumer is just like you. Right, Tyler? Right.
- [67:50]
A
Stunned and decided, all right.
- [67:51]
B
I mean, hit it with the flashbang.
- [67:53]
D
Oh, no.
- [67:54]
B
Growing Flashbang. Surprised there's now big changes to Gemini in Chrome. We got flashing lights over here now.
- [68:04]
D
Okay. But this is still like making the product from like a very like, AI first view. It's not just like taking the normal thing and then just adding one extra AI button.
- [68:13]
B
I mean, I'm not a product manager at Meta, but you have to imagine that there's little features like this that can seep their way into Facebook and Instagram where you see a photo of like, you see a photo of a restaurant and you're like, I think that'd be great for a team off site. Let me just not download the photo and put me, you and Jordi at the dinner table at the restaurant. You know, envisioning what we're like, I just hit share and then I say, hey, put, you know, put the group chat that we're in on Instagram, DMs in this photo. That'll be funny. It'll be more inspiring for them to be like, yeah, we should go to dinner there.
- [68:52]
G
Right.
- [68:52]
A
I guess the question, yeah, the feature in the video actually got shown. I wonder if that should actually live. So in that example, you're looking at an apartment online and you're in your browser being like, make it look like Kelly Wearstler styled it.
- [69:08]
B
Yeah.
- [69:08]
A
And it adds a bunch of furniture. Should that live at the browser level or should it just live in Zillow or Redfin or Compass, Right?
- [69:16]
B
Yeah.
- [69:16]
A
It's kind of a question on where this OS layer.
- [69:19]
B
I know what you're going to say. You're like, the OS layer, your OS build now, right?
- [69:23]
A
No, I mean, in this case, I feel like it should be like, actually in Redfin and you should be able to cycle through a bunch of different interior design.
- [69:32]
B
We have seen that every single layer of the stack is adding AI features from top to bottom.
- [69:38]
A
Interesting context here from compound248. Google with an absolutely murderous response to OpenAI's app just browser. Highly strategic for Google launching a browser, which was a mistake by sama and the OpenAI team. OpenAI was never going to gain big browser share, but launching gave Google the needed cover to launch its Chrome AI embedding without violating antitrust. Because now it's a competitive response. Would they have had. It's hard for me to believe that Google wouldn't be allowed to add AI features.
- [70:08]
B
I don't necessarily buy that, but it's an interesting take and it's one position Google's using its browser incumbency as a distribution and learning angle for AI is going to be exceedingly powerful if you allow it. The browser is going to learn from everything you do online, creating differentiation for its position as a personal AI.
- [70:27]
A
Wow, big day today. We're not getting any fines from the chat because we're both suited up.
- [70:31]
B
Yes, no fines.
- [70:32]
A
Dude, it's been a minute.
- [70:34]
B
I like that.
- [70:36]
A
Thanks for holding us accountable.
- [70:37]
B
Turbo Puffer Serverless vector in full text search built from first principles and object storage. Fast 10x cheaper and extremely scalable. There's so much new TikTok uninstalls surge. 150% after apps US takeover. That is a crazy surge.
- [70:55]
A
Okay, yeah. So there's a new king of the app store. It's called upscrolled. Yes, it is a TikTok Instagram clone, clone competitor, just getting off the ground. But I guess, yeah, people. I think people had the perception that TikTok was immediately censoring a bunch of different topics, political topics, and then just decide upscroll. Use that as an opportunity to say, hey, we're the alternative.
- [71:25]
B
Yeah, it was interesting because I know people were tracking the TikTok changing ownership. Is TikTok going to get banned? There were a lot of TikTok creators that took that as an opportunity to diversify to reels and shorts and it just became obvious that you should be a multi platform creator around that time. But I wasn't aware that the TikTok community was going to be tracking like when the servers actually shifted over to Oracle. It feels a little wonky and not something that would go super viral, but.
- [71:54]
A
Yeah, but people noticed because videos stopped being served.
- [71:56]
B
Yeah, yeah, exactly. So there was an outage and that.
- [72:00]
A
Certainly drove deletions and Sean highlighted something.
- [72:05]
B
Yeah, but again, to be clear, it's. It's like there is a base level of TikTok uninstalls and then that serves 150%. That doesn't mean that they're declining by 50% or anything like that. Like there's still, there's still new installs.
- [72:18]
A
Calling up scroll. Yes, the, the rumble of TikTok.
- [72:22]
B
Really? The rumble of.
- [72:24]
A
Sean Frank says TikTok views are down. People are blaming the new owners. I think this is just proof that TikTok was botting views the whole time. Your 100,000 view video is probably reaching 25,000 real people. So no surprise here for me, I always felt like it was always obvious that there were very real people on TikTok. I know people that went over to TikTok, they'd start making videos about startups, they'd start getting real inbound from people on LinkedIn. So they very clearly was always a bunch of real people there. But TikTok had every incentive to just bought all the views because what happens if somebody's getting way more views on TikTok versus Instagram? They're going to lean in, they're going to say, like, I have more followers on TikTok. Yeah, I should be creating content there. And that created a flywheel.
- [73:11]
B
Yeah.
- [73:11]
A
And so you can imagine as things shifted over, who knows, right? Like the, the new, basically the new, the new product.
- [73:19]
B
Yeah.
- [73:20]
A
I mean they had some system that was.
- [73:22]
B
I like different formats. I liked vine back in the day. At one point I did set up a TikTok and I uploaded like two or three videos. I was just trying to see like what it felt like to use that platform. And I noticed even though I came to the platform with zero followers, the three or two or three videos that I uploaded immediately got 500 views each. And I thought that the model was you get more of an opportunity to like sort of audition your content in the algorithm and then if it works, it can blow up very quickly. And I think that that's somewhat true. Like when I started my YouTube channel, the first videos that I put up got 100 views. And for like a year, if I broke a thousand views a video, I was like, this is amazing, like crazy.
- [74:06]
A
You're really grinding it. Really grinding.
- [74:08]
B
Yeah. But on TikTok, you post and you immediately get 500. And I've talked to some folks years ago who would set up a new TikTok account for a brand new. And they would launch one video that was so polished and so designed to be go viral. Like they'd blow up a car and they'd spend all this money shooting this and it would actually just. They knew that it would go viral and it would just immediately go out even though it's a fresh account and there were no followers because they just knew it was good content, it would get shared. TikTok would audition it to like 100 people.
- [74:36]
A
Remember when TikTok launched, it was a period where it was so difficult to grow on Instagram.
- [74:43]
B
Yeah.
- [74:43]
A
Like that there was a huge challenge. If you're a new creator, you go on Instagram and be really frustrated because your stuff just wasn't getting shared with people that didn't already follow you.
- [74:51]
B
Totally.
- [74:52]
A
Well, we have some comments under here. Last thing on the TikTok front, Turner Novak commented, maybe they were views from Chinese users which are gone now. It also could be international users. Remember there's like, if you have a new US app under US ownership, is it getting shared with international users that are using other versions of TikTok? Unclear.
- [75:13]
B
Well, let's bring in our first guest of the show, Anton from Lovable. While he comes on, I'm going to tell you about Gusto, the unified platform for payroll, benefits and HR built to evolve with modern small and medium sized businesses. Welcome back to the show. Welcome to the TVP and Ultradome. Thanks for taking the time to come and hang out with us.
- [75:31]
A
West coast tour.
- [75:32]
F
It's cool to be here in person.
- [75:33]
B
Thank you.
- [75:34]
F
It's been more than six months, but it feels like much more shit. Ton has changed my side. The business is doing better than ever, I can imagine.
- [75:44]
B
Every time I. Every time I see a post about you from you, there's a new biggest number. How is growth going? Is it still accelerating? How's the business doing?
- [75:53]
F
We are doing great on the growth side. We hit 300 million AR two months. After 200.
- [76:05]
A
The gong is even louder in person.
- [76:08]
F
But more interestingly to me is the builders who are doing real things. I was on the airport from San Francisco here and adam@owner.com who helps tens of thousands of restaurants, he was super excited and raving about he had built out a new product line over the weekend in eight hours where he was solving a new problem where all the customers were saying, we keep getting these phone calls, it's not always someone picking up. And he built an AI who answers the phone for the restaurants. And it was all working. You could configure it just like you were training a human. And he was now building it, releasing it next with his team. And we're seeing all of these new opportunities where people who use Lovable as their technical co founder and build businesses and maybe you saw I was super shocked when someone's shipping AI generated dog pictures in frames to pet owners and they're making $300,000 per month on building that out without any technicals.
- [77:08]
B
How are they actually doing that? Obviously I think maybe you can describe the technical co founder process of building a new application that someone takes payment but then I want to know about the top of funnel for that.
- [77:22]
A
Yeah, part of, I mean I always saw Lovable as growth, as millions of people out in the world that historically were just perpetually searching for a technical co founder.
- [77:33]
B
Yeah.
- [77:33]
A
Problem is a lot of people that are technical have a bunch of great ideas. They don't necessarily need an idea guy. And so there's all this like sort of pent up creative entrepreneurial energy that just never really goes anywhere because somebody doesn't take the time to actually learn to code. They never find that right partner.
- [77:49]
F
Yeah, that's what we're seeing. And our mission is about human creativity which is very fun to empower. But it's not just new businesses. There's a lot of companies and I keep hearing from how they build out AI applications and agents using Lovable. And once you get into it, especially if you look back compared to six months ago, this feels like to me when I use it, there's no limit almost of what you can do. And the thing where we're lovable, what I'm hearing is that it's a bit unique, is that there are a lot of tools that are coming from developers and then more people keep using them. We started with the non technical people and that makes it just super simple. But it still has the same capabilities that more sophisticated users are looking for in the same way.
- [78:36]
A
How are you thinking about working with companies like Owner that I'm not super familiar with onr we got to have Adam the founder on but my understanding it's somewhat of like a system of record for a restaurant owner. Is that right? And so in theory he could also offer Lovable as like an integration so that individual businesses that are using Owner can expand kind of and just build products that they themselves want, not necessarily have to rely on Owner. Is that right? Or are you seeing opportunities like that with other companies?
- [79:05]
F
I think we have these partnership programs and that's a good idea. I'm going to bring that back to talk to to Adam specifically.
- [79:13]
A
But what we're seeing is theoretically somebody is On Owner, they have a feature that there's a feature that they want that's not being offered and they can just build it. They don't even have to request a feature.
- [79:24]
F
I mean, that's what the future looks like. And what we've seen is that Lovable is first used as a communication medium. And there are a lot of people that have stopped using Google Slides and PowerPoint. And now they just prompted away because it looks so nice. And I think it's a better medium than the animated videos you were talking about previously. But then now a lot of companies and us is one of those. We're switching out SaaS providers to using tools that we build ourselves. And we tell Lovable, can you connect these two different SaaS tools, internal tools that we built, and add AI in between of those. And that I think is more of the future of how companies are going to be run.
- [80:09]
A
Okay, let's talk about. Let's talk about kind of Vibe coding debt and tech debt, because we've gone through this where we'll build a product. We built like a guest directory, very simple product that would basically every time an episode would get published, we had an app that would effectively scrape the episode, figure out who was a guest, link everything out and publish it. And it was built in a day by one of our interns last summer, and then he had to spend the entire summer just maintaining it and shipping incremental updates. So are you seeing any changes on that front? Obviously, if it's easier to make products, maybe it's easier to maintain them. But I think the concern is if you've created 200 different SaaS tools and internal tools as part of your organization, then you end up spending really valuable resources at the company, you know, human labor, time, just kind of maintaining these systems and making sure they keep running.
- [81:11]
F
Can I talk a bit about the recent updates we did yesterday? Lovable. So there's a much more advanced planning mode that helps asking the you or your intern questions about what it is that you're trying to achieve. And that makes the AI be able to find a better solution that doesn't need that type of maintenance. And that's one of the many improvements that we shipped yesterday. Another big one is so basically, if.
- [81:37]
A
You Vibe code poorly or if you prompt poorly, you're adding a bunch of Vibe coding debt. And if you spend a little bit more time prompting.
- [81:46]
F
Yeah, I think people should just move fast and be. Not be like, oh, I have to think through this. So it's maintainable. It should move really Fast to get something that works and that's like useful. But then you should sit down and you should reconsider talking to lovable and asking should I restart from the beginning I'm having these problems with the maintenance. Is this, should I just be keeping shipping incremental updates or is there something else that could actually solve this on a deeper level and keep having that conversation and be creative in terms of what could be the right solution for this? Because the systems are, they're extremely intelligent now and they're just getting more and more intelligent. So that should be your status quo. Explain your situation so that which the AI doesn't know and then it will help guide you through what's the best path.
- [82:31]
B
Yeah, it feels like the solution to vibe coding is more vibe coding or more code. Right.
- [82:37]
A
But the question says the vibe coding.
- [82:39]
B
Dealer is like actually you need more. But I believe that, I believe that. But I think the interesting question is the shift to mobile and I want to know how you're thinking about that transition. Because if you go and vibe code an application and you sit down, you really think about it, you work with the tool to plan and test and you're giving feedback and you spend a full day or a couple hours and you got it in a good spot. But then there's that one little oh, actually this button isn't quite working. Can you just go check out. It doesn't work in this scenario that winds up just being normally like, you text a developer, you slack them. How are you thinking about long term integration with end users who might not be jumping into the tool themselves and might not even be have the time to actually sit down and vibe code.
- [83:26]
F
I think the future of how teams operate is of course changing and I think there should be someone who is ultimately responsible for the quality of the application that might not be you. Even though you can throw in ideas of changes and then someone looks at it before it goes out to your customers.
- [83:43]
E
Got it.
- [83:44]
F
So that's high level how I think about it. There's a lot of new capabilities that we are looking at to make sure that quality is always maintained. And we, over the last nine months we've been super focused on security, which is this. Like if you're putting something in production, that should be your first priority. And there's a lot built in that makes applications secure from the ground up in how we approach.
- [84:08]
A
Yeah, the challenge right now is if you have a hit product, you could go to 100 million users in 48 hours. We saw this. What was the company that was like Palantir for dating.
- [84:20]
B
They really go that big?
- [84:21]
A
Yeah, no, no, no, they went the T dating app, remember? Oh yeah, they went.
- [84:26]
B
They don't Palantir for dating. But yeah, yeah, the tea dating app really did scale. But yeah, they have a security issue.
- [84:31]
A
Went to the lovable.
- [84:32]
B
Yes, yes. To be fair.
- [84:35]
A
But yeah, the founder was just kind of coding.
- [84:38]
B
I think he had some team members too.
- [84:39]
A
Yeah, sure. But clearly they were moving quickly and some mistakes were made.
- [84:45]
B
Yeah. Has anything changed in your mindset about the level that you want to operate at or integrate with versus being a web app with cloud hosting, operating into the browser level, into the OS level, launching a dedicated app on a phone or on a Mac or command line tools. How are you thinking about the different unlocks that you get at different levels?
- [85:15]
F
Yeah, great question. So first of all, I use Lovable on the phone all the time and it's really a good mobile experience. That's why we are seeing usage there. I know that a lot of entrepreneurs, they're building out complete work OS systems. I talked to Gwyneth Paltrow yesterday and they're building a lovable brain that accumulates all the information that they need in running their company. And what that gets some more capabilities from is that it connects to your local file system and it connects the to. I mean in the future, I don't know exactly what the mobile device is going to look like, but it's going to. The AI has to have context on everything you do on your phone for it to be fully unlocked, all the capabilities unlocked. And that's something we're looking at. Exactly how does that future look like as we are building towards the best interface for humans to use technology and use AI? Because that's the vision we're on.
- [86:17]
B
Yeah. How do you think about the trade offs between consumer and enterprise? Obviously you can use it to make something that replaces a slide deck, just communication, external marketing materials, landing pages. You can also use it for building an internal tool and just spin something up internally to a company that's never going outside. Both are very interesting markets. Do you want to serve both? Do you think that there's a value to focusing on one over the other?
- [86:43]
F
Yeah, great question. So lovable started out from the early AI adopters and that's where we're seeing now growth in all directions because people love how there's maybe some Swedish design sensibility. Things just works and we put a lot of hard engineering in making it simple. Right. And one of the directions is spreading is in the Enterprise people come to the work and they show like look how much we can get done. Use glovable at work. So now we have all of these required features for enterprises, SCIM and so on. And specifically I'll tell you an example of a really large company. They have 80,000 plus real estate agents globally, dozens of countries, hundreds of websites. They had the offer to replatform all these websites because there's always this pain for marketers changing website and the VP of marketing there, that would take a year. The VP of marketing found lovable and he was like wow, I can just build one of those websites in a few minutes. And then they sat down and they looked at how can we.
- [87:48]
A
These are websites for individual homes, for.
- [87:51]
F
Different brands or real estate agents across different geos. And then over three weeks they re platformed and they saved $2 million in a on a solution which was part paying for AI chatbots on the website. And their chief innovation officer just says this is the biggest change that the company has experienced in how fast it can move. How they could build out an AI chatbot just by prompting Lovable. And that's a large company and that's them building product lines that are facing their end customers. What they're seeing even more volume of is of course internal application applications and applicant tracking systems built from the ground up by the people who are using those applicant tracking systems. With AI, with thousands of people being.
- [88:38]
A
Served, how are you thinking about headcount planning specifically on the engineering side?
- [88:45]
F
Yes, it's a good question. I was at the Anthropic office and I think they're hiring a lot of people just like we are doing. Because if what you're actually seeing is that if you use AI correctly as a business, which is one of my personal top goals or the entire company's top goal out of three goals is us learning how to use AI as productively as possible. And that means that each human has more output, right? If you do that really well. And as we are learning that and figuring that out, we're putting that into the product and giving our users all the best practices in how to connect all the internal tools that your company is running on and share context across everything you're doing. When you're doing research on a podcast, you should all be able to just ask an AI. What question should I ask Anton? And all of that infrastructure, all that tooling, you should be able to build on lovable or just get someone else tool and copy it into your workspace.
- [89:43]
A
How much do you guys pay attention to job descriptions at various companies in terms of understanding market pull. I could imagine people, companies that wouldn't necessarily hire software engineers saying like we want engineers that are can, you know, use lovable and other relevant tools. Because I think one of the, one of the interesting things that we're trying to understand from just kind of understanding like how job, how we'll see job growth in some areas, maybe job reduction in other areas is I think people are underestimating how many companies there will just be a net new role where somebody or multiple people are creating and maintaining software that the company is creating. And we're an example of that. Right? Historically a media company like if you, if you had a fledgling media company. We brought Tyler on last, I think like Q1 of last year. We'd only been in market for a handful of months and normally if somebody was building a media company and they were telling me they were hiring or a podcast right and they were telling me they were hiring a software engineer in their second quarter in business I'd be like dude, you should focus on a million other things, just make good content. But in our case we realized hey, we can build quite a lot of internal tooling for what we're doing to make the show better. And so I think we're going to see a lot of job growth from companies that again wouldn't normally hire anybody internally doing software engineering.
- [91:06]
F
It's a good question. I don't look at the external other people's job offers. My background is in physics. So I just break things down into like what is it that what are the governing principles for making a team work really well and people work humans working well together and getting output as a company. And what I'm seeing is that you want the team of people who they don't have to be software engineers but they're very curious and they get the it, they get new things, understand them very quickly and that's one of the things that we're fundamentally hiring for. I mean I can take another example because I love, I'm so inspired every time I meet someone who has like we've changed their lives with lovable. So it's a guy who Alan, he built a healthcare AI staffing company in five months there is a million dollars in ARR and he couldn't have been able to build something like this.
- [92:01]
A
And then yeah there's so many niche labor marketplaces. These were, they used to be like more popular like venture backed businesses. And then a lot of VCs figured out like hey, you can Build a great business here, but you end up kind of capping out. There's some software component to do with like matching and kind of booking payments, stuff like that. But ultimately a lot of these businesses should just remain like somewhat niche. Right. You could have a $10 million a year business doing staffing and have it like the people running it. Right. You have a lean team. But historically it was a challenge because there was a number of companies that would build out of the box marketplace software, but usually they were. You had to do so much work to actually adapt the software that it was kind of this trade off of like, do I build it in house, do I use.
- [92:48]
B
Some people used to abuse WordPress like crazy. They built a whole social network on WordPress or try and modify it with e commerce stuff. And it would take a year and millions of dollars to build something that now it takes a weekend.
- [93:04]
A
How do you lead the team around dealing with competition? And you guys are now competing with pretty much every company in the whole world, it seems like from hyperscalers that are building vibe coding tools to other startups. Like what's your philosophy around dealing with competition? Just kind of like emotionally, like in the organization.
- [93:28]
B
Yeah.
- [93:30]
F
First of all, what we're seeing is that people who come back to lovable and compare it with other tools, they say like, wow, this is such a wonderful simpler experience that works better for them that we're actually working. It's more capable in many, many dimensions. So that just gives us a lot of belief in what we're doing. And the way we think about it is to not look at what others are doing is just to look at what do our customers want us to do. And being very grounded in that and that we have.
- [94:01]
A
You banned the competition Slack Channel because it would literally be every day it's like, hey, we have a new venture back. But it's a sign that we fundamentally.
- [94:13]
F
Design, say we're building the best product based on first principles, what we think is the best way to do these things. And it often means not adding new features, adding new capabilities, but thinking very long term. How is this whole experience going to be for us humans as we interact with superintelligence basically in the future? And that's what we're driving us. Building that product is the motivator for everyone.
- [94:37]
A
What about international expansion? You're obviously in Europe here already. Are you opening offices in other continents?
- [94:45]
F
Yeah, we haven't announced more offices than Stockholm. And then a great go to marketing.
- [94:50]
A
Have any to announce?
- [94:51]
F
What's that? I can come Back we have more announcements, but we have people hired. A few people are hired remotely, including here in the US in various cities. So that's what it's like. But we getting a lot of value of being in person. People who were very long term thinking, very bought into the mission that we're on, based in Stockholm.
- [95:13]
B
How do you think about retention? There's so many amazing tools that come out. I remember those. It was Lensa magic avatars. Everyone had to take their photo and do the studio Ghibli precursor like a year earlier, the style transfer on their image. And then people sort of like they tried it and then they churned out. But at the same time, like website builders have incredibly low churn because once you build your website, even if you're not using the tool every day, you're using it as hosting and you might be dumping be jumping back in to make changes. How are you thinking about the different retention dynamics, the different cohorts? How are things looking?
- [95:51]
F
We're seeing very, very healthy retention, net dollar retention. It's been more than 100%. But we are also of course seeing some AI tourists that are just, yeah.
- [96:03]
B
Yeah, they must exist and they just stop by and then they move on to something else. Right?
- [96:07]
F
Yeah. And what we're doing is that we're splitting out the different segments and seeing like what, what are the jobs to be done for the segments that we could do better. And fundamentally what we, what we're noticing is that for most things we're adding, which is about making the whole experience to do more things more seamless, like more capabilities more seamless without making it a more complex experience that's helping the transition across the board.
- [96:35]
B
And less of a point solution, less of, yeah, I use that for slides this week, but then there's a new one over there. I got to try the new thing more of like actually run your entire business, build a new business on this and then stay with that for a long time. That makes a ton of sense. What about recommendations for other software products, databases? I've seen a bunch of charts where the default recommendation for Supabase, for example, is just climbing like crazy because it's recommended by you and by a back end and by a number of the models. Does that business evolve into an affiliate model? Do you have demand from your customers to educate them about the different trade offs when they're building, when they're picking something? Or do you want all of that to be abstracted away so that the end user doesn't even know what a database Is, for example, and database is just one example. There's a million things that you probably plug in with.
- [97:32]
H
Right?
- [97:32]
F
This is a great question. We have some really good product integrations you mentioned. Superbase is one of them. We are covering more and more area of payments, email and communications. We want more of those amazing partnerships. I think there shouldn't be any limits of those integrations. You should be able to ask Lovable to get cleaning for your podcast studio or whatever. And that shouldn't be done by us. It should be a partnership where we're currently just making more of those partnerships successful. And there are multiple companies that have seen absolute majority of their growth coming from. From us actually. But then at some point we need to let our users get the best product integration if there are many ones.
- [98:22]
A
Feel like other companies should be a little bit worried if the majority of their growth is coming from you guys, given that the whole purpose of the company is to make it easy to build software and eventually you guys will just naturally do some of those things.
- [98:36]
F
Yeah, I mean the integrations, I'm happy to not do them. We want to focus on the interface, the app creation layer, and then underneath there needs to be a lot of things that connect to very, very technically challenging systems to build infrastructure and so on that we don't need to put like our efforts into building. So it's. We want to do a lot more partnership this year.
- [98:58]
B
Yeah, makes sense.
- [98:59]
A
Talk about the dynamics between Lovable and all the different labs. I'm sure that you have very open dialogue with them around. Just like integrating with new tools, getting previews of models, all that kind of thing. And I'm sure part of why you're on the west coast is just to spend time with the different teams.
- [99:17]
F
Yeah, those relationships are really good. We are fundamentally listening to the users more than them. We're building for the 99% while they're.
- [99:29]
A
Building developers, please listen to your users less.
- [99:32]
C
You really want to?
- [99:33]
A
No, I'm kidding.
- [99:34]
C
Yeah.
- [99:34]
F
No.
- [99:35]
C
And.
- [99:39]
F
They have to spend so much of their focus on the strategy on building these models and making pumps, squeezing out more and more juice from those models. And we're in this place, which is very favorable for the end users, where we're not tied to any specific model. We use a combination at level of different models and that gives a better user experience in the end with the trade off level of speed and the highest quality for each specific task.
- [100:06]
A
How did you and the team react to claudebot? Was there any kind of like new kind of user paradigms that you were inspired by anything to take away and kind of integrate. I mean, we had Peter on the show Tuesday a couple days ago, and he was very open. Like, he's just like, in it. He's clearly not in it to make money. He's in it for the love of the game. And just kind of. We asked him, do you think that people will clone what you're doing or copy it or build on top of it? And he was like, yeah, I do. It's cool. So kind of an unexpected answer, but I'm curious if there's anything that you thought was particularly interesting that you can bring back to Lovable.
- [100:46]
F
Yes. I love Peter. Happy you had him on the show. I think me and many people in the team have been building our own versions of this personal assistant that you just talk to, record audio to and.
- [101:00]
A
Talk on your phone.
- [101:02]
F
Yeah. In my case, on my computer. And then what we're of course thinking about, which is like the other top goal for this year is to provide the best workspace or the best way to collaborate as a team. And that, of course, course, a big part of that is being able to talk to an agent that has access to all of your different tools that once you're building with Lovable and can collect, can pull out the data when you need it. And I think I'm very happy about this open source project in particular, getting a lot of traction. It's between my roots as well.
- [101:39]
B
Yeah. It feels like it's just waking everyone up to what's possible. It was real feeling, the AGI moment generally. And maybe not everyone reaches for open source and all that's entailed with that, but it's sort of just like a new call to action for everyone to try new tools. So overall, very bullish.
- [101:55]
F
Twitter likes this completely new things that you can try out. In the end, what users want is something that's simple and that's very, very trusted. High security, high reliability. And when you're building a strategy like us, that's what we're focusing on. What is it that most people want and what's the sustaining product that you want to build?
- [102:15]
A
What social platform is most important to Lovable's growth at the moment?
- [102:22]
F
I think I'm not sure if my team likes me sharing data here, But Twitter and LinkedIn is what the ones I know are the big ones. But we're also seeing Instagram.
- [102:33]
A
Yeah. Because I feel like products, when they actually go mainstream, like going mainstream looks like going over to Instagram feels like.
- [102:42]
B
Beyond a niche, which I think of LinkedIn and X as. But yeah, I mean, it's a big community, so it makes a ton of sense, especially with some of the larger enterprises that you're working with the business side. That makes a ton of sense. What does the top of funnel look like and growth function look like? Is there a flywheel where someone builds and then shares? And then I imagine that there's links built with lovable in the footer. But is that a real engine of growth or is our advertising what's driving growth overall at the high level?
- [103:13]
F
Yeah. So we're seeing about 300 million visits to lovable applications per month. So that is a real effect. People are like, oh, this is this great website. And most of the time that leading to conversion is they just talking to the person like, oh, yeah, you built it with lovable.
- [103:31]
B
Sure.
- [103:31]
F
And that is a part of the biggest channel, which is just word of mouth. And that continues to be like, absolutely.
- [103:37]
B
They see something cool, even if it's just a small demo, even if they're not paying for that business. They see it and they're like, I want to do that, but for me or mine.
- [103:46]
F
I mean, they hear about us from the friends or they see about us on the Internet. And yeah, I'm looking at investing more into growth, of course. And then I think Instagram and getting growth marketing, really, really strong. It could be the next big channel.
- [104:04]
B
Super Bowl's coming up. It's not too late to throw something together. Maybe next year. Anything.
- [104:09]
A
Anything.
- [104:10]
B
Maybe next year.
- [104:12]
F
Oh, we'll see.
- [104:12]
B
It's not enough breaking. Thank you so much for coming on the show. This was fantastic. Have a great rest of your trip. We really appreciate you coming by. And we will tell you about FIN AI, the number one AI agent for customer service. If you want AI to handle your customer support, go to Fin AI. And up next we have.
- [104:33]
A
I'm expecting a Super bowl ad.
- [104:35]
B
I want a Super bowl ad. They could do some really fun stuff. I mean, even some of the customer testimonials would be great. Even if they go with something more, they should build the entire ad in lovable dog food.
- [104:47]
A
The product.
- [104:47]
B
We have Christian Garrett and Delian Asperuhav in the Restream waiting room. We will bring them in to the CBPN Ultradome. Dellian, Christian, how are you doing? How are you feeling? What's going on?
- [104:58]
I
Go, baby.
- [104:59]
A
Live from New York.
- [105:00]
B
You guys look good to see the boys.
- [105:01]
A
Great to see you guys.
- [105:03]
B
Take us through the news. Break it down for us. This is year four, correct? Year Five year. Five.
- [105:10]
A
Whoa.
- [105:10]
I
Third year of doing the Daytime Forum. So we started it back in the day just as like a small dinner series. And then three years ago kind of experimented with Daytime Forum. That went great. And then this could be our third time doing it.
- [105:22]
B
Okay, what's special about this time? Who's on the lineup?
- [105:25]
H
You want to start? I'll start. I mean, obviously the 250th is great timing for us to have another event here. We're very excited about that. I think the key focus this year, outside of continuing to build on what the forum has always been about, which is building bridges between technology and government, the technology sector and government is going to be increased focus not just on technological leadership, but industrial leadership and then as well as looking at having leadership among our allies. So I think this year you'll see a bigger expansion of the industries where technology is having an impact and playing a critical role. And you'll see leaders from companies that are leaders in those industries. And you'll also see a lot of folks coming from abroad and participating as well. And so I think you're just seeing that the relationship of the bridge between technology and government is actually a key framework for building deep relationships between the US and our allies, as well as building deep relationships and strengthening other industries.
- [106:15]
I
And we didn't say but. March 24, Washington D.C. for now, we've only announced some of the, like, the private, you know, basically speakers. We haven't announced any of the gov speakers yet, but it's folks like Trey Stevens from Andrew Founders Fund, Vinod Khosla from Khosla Ventures, Brad Lightcap from OpenAI, Shyam Sankar from Palantir, a bunch more that we'll start to, you know, basically release. But yeah, what I like about this year is we're also going to start to expand themes a little bit. There's going to be a strong, basically biotechnology sort of focus. And some folks both within the administration, within private industry, talking about basically how we need to make sure that we don't lose that entire like, supply chain and industry to China the same way that we have with like consumer electronics or Humanoids. You're starting to see some like, early signs towards that. And there'll even be some joint policy put out that HVF will be, you know, sort of endorsing. And so what's been cool is as the like forum has been around for several years, we're now not just like acting as a gathering place, but it's even a place for people to start to think about you know, not just meeting, but even like publishing policy together that align sort of like private public incentives in a way that is best for the nation. And then throughout all of this, you know, like I said, we started this five years ago, obviously under the Biden administration, continuing to maintain, I think, you know, sort of true bipartisanship. You know, Vinod Khosla obviously has a certain set of, you know, sort of thoughts about, you know, sort of policies, regulations, etc. I may not necessarily agree with all them, but we try to, throughout both the, you know, sort of forum, the events that we host, everything is pretty consistently, you know, sort of, you know, even and across the aisle. And we try and stick with sort of centrists on sort of both sides that can actually reach across the aisle and produce productive policy together. You won't see a ton of the sort of extremists from either party there.
- [107:43]
B
That's great. I noticed young Liu from Foxconn is attending interesting company. Obviously, everyone's very familiar. Do you think there's going to be more energy about let's bring more Foxconn manufacturing to America or just startups trying to work with Foxconn? There was this leak that that OpenAI might be working with Foxconn to make the new sweet pea earbuds. And it feels like more and more companies, as they think about hardware, they're going to be wanting to establish relationship with Foxconn. That's sort of the beauty of this, at least from the business side. What's interesting about Foxconn, why is he important to the conference?
- [108:23]
H
Well, I think first off, you know, you have Foxconn, Qualcomm, you'll have some other names that we announced across these other kind of categories of companies that have maybe been in industries that weren't viewed as technology industries, but now increasingly are. And I think a lot of leaders that are coming there are really just representing a trend that you've seen, which is once again, technology industry is a great way to build deepening relationships and partnerships with our allies. And to your point, it works both ways, right? Yes. You'll see, just like TSMC has made a huge investment in the state of Arizona, and you're seeing that from other companies as well here domestically. You're also seeing companies abroad continue to invest in other countries and scale there through partnerships like these. An example in another country, you know, we've seen Anduril, for example, invest heavily. Andrew is a portfolio company for founders fund and 137 for us, respectively. Andrews invests heavily in countries like the U.K. australia and Japan just recently opening their office there. Japan is another country as well as many others like South Korea and Europe, that are increasingly opening their doors to become manufacturing hubs for the reinvestion here in the US as well as supporting their own re industrialization efforts. And I think you're going to have to see many companies like Foxconn play a critical role as the west tries to shift off of reliance on Chinese capacity across critical industries.
- [109:40]
I
We, as much as possible, obviously, try to make it so that it's bipartisan from the elected officials and Congress. But we do try to, as much as possible, whichever administration basically is in office, try to reflect some of their priorities. And what you've obviously seen under this administration is, you know, them leaning into, you know, making sure that our allies are, you know, sort of contributing their fair share, but also re establishing their own re industrialization efforts. And so I think a significant theme of this year's forum is going to be having a lot more international presence and there to both, you know, sort of learn from how the US Is re industrializing, but also partner with some of these companies to bring those technologies locally, partner with some of their local sort of primes and re industrialization efforts to accelerate those in some ways, these trends that you're seeing in the U.S. the United States is much better off if France, if Germany, if the United Kingdom, if Australia, if Japan are mimicking basically those same technological patterns. And so that's definitely going to be a major topic of this year's forum.
- [110:32]
B
Yeah, I know you haven't announced the government attendees, but is there a desire to have international leaders attend as well? Is this going to be the next Davos in a couple of years or. Or do you want to keep it focused on domestic policymakers, U.S. policymakers, senators, representatives?
- [110:49]
H
I mean, obviously the event being in D.C. and the core focus is on the technology sector, the government sector in the U.S. but the event's always been international. You know, from the beginnings. We've had foreign ambassadors from other countries attend. We had the VP of Taiwan speak multiple times. I think you will see a lot more international leadership attend as well. I think it's reflective to Delian's point that this really is about the west being successful. And, you know, the concept of it's not us alone.
- [111:15]
G
Right.
- [111:15]
H
And so I think, yes, you will see that represented most of the leadership, of course, is still going to be heavily domestic. Right. And so leaders from Congress, leaders from the administration, et cetera, but you will definitely see a lot more Leadership from from countries abroad and our allies. And that's going to be very exciting. And that's going to be from the Middle East, Japan, South Africa, Korea, Australia, Taiwan and Europe.
- [111:36]
I
Yeah, you'll see like in the long term, I think the way that like this will always break down is it'll probably be like 85%, 90% domestic focus, but there'll always be like a twinge of international. And so I think that's what it's been probably historically on the attendee side of things. And this year for the first time there will be a few more sort of like internationally focused panels. But it's not going to quite go to like Davos level, which is like kind of global elite in some ways. This is always going to be primarily sort of United States States focus with a twinge of international.
- [112:00]
A
What are all the ways that people can participate, watch from home, be a part of it, whether as a sponsor, all that stuff.
- [112:10]
C
Yeah.
- [112:11]
I
So the event is invite only to participate during the day in Washington D.C. but it is going to be about a thousand people. So it's a relatively broad audience that will be bringing in there. We'll also be live streaming all of the panels on our YouTube and X accounts. And then there will be a couple of media groups that are live streaming from there. Some traditional ones like Bloomberg and some super exciting novel ones like TVPN on site there. So there'll be a variety of different live streams to tune into that I think will give you a feel of the event, both the official panel as well as obviously the side conversations that you guys will be doing.
- [112:43]
B
Yeah, that's great. What is the business structure of Hillen Valley now? Is it Its own nonprofit?
- [112:50]
E
It?
- [112:51]
I
We are officially a 501C3 so this is the first year ever we are welcome to the board of the Hillen Valley Forum Foundation. 501C3.
- [113:02]
A
Are you guys just being memetic with OpenAI?
- [113:04]
D
Yeah.
- [113:05]
B
Are you considering a for profit conversion? Because I feel like the ads really, we could do a lot with each other here.
- [113:11]
H
There's a lot more to be announced like Delian said, so just stay tuned.
- [113:15]
C
Yeah.
- [113:15]
I
But we do have full time staff.
- [113:16]
C
Which is really exciting. So that's cool.
- [113:18]
I
Yeah, we, you know, Christian and I were starting to get a little too stretched thin between you know, both of us having too many jobs. So it is nice finally.
- [113:24]
H
And we lost the key leader as well. So we really stretched that.
- [113:28]
I
I believe now the youngest undersecretary in United States history.
- [113:31]
H
So handsome and most handsome.
- [113:33]
I
Goddamn.
- [113:33]
A
Some of those photos on the official most looks. Max.
- [113:38]
I
I wish I looked like that.
- [113:39]
B
Yeah, yeah, it's true. Maybe I'd be an undersecretary then.
- [113:42]
I
His undersecretary.
- [113:43]
B
So yeah, talk to talk more broadly about the 2026 calendar. 2027 calendar. People know the annual event in D.C. there was that one off event with all in. There was a special Hill and Valley sort of collab. Is there a plan to do a quarterly series, have other dinners, go off the Hill and have something in the Valley? What are you thinking?
- [114:04]
H
I think it's something that we're open to opportunistically. It's not something we're committed to or it's part of the plan. We very much are focused on the main event every year, but. Absolutely. You know, we actually just did something, something recently with fii, who is also. They do an incredible job. Their conference is right after the Hill Valley. So I recommend most people, which a lot of us are, will end up flying straight to that right after. And you know, Richard has done an incredible job at really gathering a lot of world leaders and running an incredible event in Saudi and they've done events globally. We did a collab there and brought a lot of Silicon Valley leadership. You know, Ruth Porat from Google attended Sriram. Right. From the White House and the administration attended and spoke and a bunch of other folks. And so I think, you know, we will look at these opportunities where we can try to help bring more of really Silicon Valley to these other events internationally, particularly, you know, when it aligns with countries that are our allies and they're focused on building deep bridges between technology and government. And so, you know, I think you'll see some of those maybe internationally or potentially obviously domestically. The, the one off event with the all in guys with, with J. Couch Moth and, and Freeberg and Sacks obviously in his role organizing that along with Jacob. That was incredible. We would love to do more things like this. Just opportunistic. So if it happens multiple times in a year, great. If it doesn't happen, also great.
- [115:13]
I
Yeah, I think one of the ambitions that I have for the event is starting to become, you know, sort of more deep in particular sectors like you know, this year very intentionally trying to go a bit sort of deeper into biotech. I could imagine eventually where you know, you have the marquee event, you know, on a particular day of the week, but there may be some like subsector tracks, you know, over time where we do like Hillen Valley space, Hillen Valley Biotech, Hill Valley Industrialization, Hill Valley rare earths. And those are maybe like sub tracks. And again, with the goal of this being like, I don't want to just gather people for gathering people's sake. Maybe that's what they do in Davos. But like, you know, the goal is like, I would like to actually make sure that, like, policy.
- [115:45]
A
I thought you were. I thought your whole brand was. I'm the networking guy. Here's a 20 point thread on how to. How to network.
- [115:52]
B
Yeah, yeah. This is a mastermind, right? We should reband.
- [115:56]
H
Neither of us were at Davos, so.
- [115:57]
I
That'S my favorite types of venture investing to do, is just follow the heat. That is how you generate returns. You just identify the heat and invest into the hottest deals. And other people are marking it up. You just got to mark it up even faster at higher prices and chase momentum.
- [116:12]
B
Well, speaking of exciting deals, Bridget Mendler was on the show yesterday. The chat. We just want to relay this. The chat is pulling for a Bridget Mendler attendance. Is speaking an invite. Something like that. We'd love to see her there.
- [116:23]
H
Bridget's Bridges event.
- [116:24]
B
Bridget's been there. Oh, fantastic. Okay. Chat's a lot of fun.
- [116:27]
H
I mean, Kevin Whale also, I have to shout him out. He's been a veteran. I know he's coming on later.
- [116:31]
D
He's been.
- [116:31]
H
He's attended every single forum.
- [116:33]
D
Wow.
- [116:34]
H
And has been an incredible partner for us. And obviously now, you know, alongside OpenAI having great partners. And so, you know, the stuff that he's working on, on science, by the way, is awesome. I'm very glad he gets to talk a bit more about that later. So, yeah, it's, it's. Bridget's actually a veteran, so we need to find a poll for someone that has not attended before.
- [116:49]
B
Okay.
- [116:49]
I
I know what I like is like there's some pretty. We have like a. Just a really solid veteran squad. It's like Alex Karp, been there every single year. Vinod Khosla, I've been there every single year. Trey Stevens, I think went to the second one coming back this year to speak. And so I think what's nice.
- [117:02]
H
That's a big deal for Trey.
- [117:03]
E
I know.
- [117:03]
A
Yeah.
- [117:04]
H
Trey's hard to be social and travel is not easy.
- [117:07]
C
Exactly.
- [117:08]
I
I just peer pressured him a lot this year, thankfully. So, you know, got him up there.
- [117:11]
H
Yeah.
- [117:11]
B
You mentioned a bunch of different categories. Is energy top of mind for you. How are you thinking about the. All the work that's going on? We see a lot of energy or a lot of excitement around nuclear. Been tracking A little bit less in solar. But who are you interested in? What conversations are you excited about in energy?
- [117:26]
I
Yeah, there'll be like two major areas that you'll see energy come up in the forum. The first is there is going to be like a nuclear sort of focus panel. Haven't announced those, you know, sort of, you know, panelists yet. But, you know, let's just say there may be one that's a general.
- [117:38]
B
Yep.
- [117:40]
I
As one of the panelists. And then other one, I think where you'll see an angle on it is a little bit in relation to basically like the build out of the physical infrastructure behind AI and how much of, you know, energy is a limitation there. So those be kind of the two, you know, topic areas that we focus on there. And again with both those trying to make sure that by the time we get to the forum, there's particular sort of policy and things that are coming out of those panels that aren't just discussions, but ideally legislation that's going to be passed. So we always try and make sure the forum is in a week where both Senate and the House are in session. And so in some ways, Congress basically chooses the dates of the forum more so than we do. But as much as possible, I'll try and make it so that sort of policy immediately sort of comes out of it. And sometimes we also react live.
- [118:15]
E
Right.
- [118:15]
I
In one of the forum's years, it was the main event was the day before the TikTok CEOs basically, you know, what's called briefing in front of Congress. And so obviously there's a lot of discussion that year and you know, policy papers and recommendations from that, you know, sort of group around, you know, what needs to be done with TikTok.
- [118:31]
A
Yeah, don't dress code. I mean, you dress code should. Should people wear different colored shoes? Deliance suit. What do you guys recommend? There you go.
- [118:42]
H
It is D.C. so you know, probably not full Silicon Valley casual, maybe more business casual or suit and tie. Definitely.
- [118:50]
G
Yeah.
- [118:50]
B
No, it's the best dressed. It's the best dressed event in tech. On the tech circuit. It's the best dressed one. It's fantastic.
- [118:56]
A
Yeah, we went last year.
- [118:57]
H
We're like the best people in trust, but it is the best dress event in tech.
- [119:00]
B
It is fantastic.
- [119:03]
I
I have been told that my head of the head of comms at Hill Valley is unhappy with the fact that I keep wearing the same damn gray suit every single year. He's making me shop, shopping, you know, later today to get.
- [119:15]
A
It's also the set you also Wore the shirt last time you came on the show. I don't know if you know, I.
- [119:20]
I
Actually wear this specific quarter zip every single time I'm on tvpn.
- [119:24]
A
I know because it's the green. It's the green.
- [119:27]
I
I make sure that it's washed every single week so that I make sure that I can wear tv.
- [119:30]
B
Fantastic. Well, we appreciate it. Thank you so much.
- [119:33]
G
We can't wait.
- [119:34]
A
Can't wait to be there.
- [119:35]
B
Congrats on the list. Everyone will be staying tuned. Cheers to the new guest.
- [119:39]
I
See you in Washington, boys.
- [119:40]
A
See you there.
- [119:41]
B
See you later. Goodbye. Let me tell you about Okta. Okta helps you assign every AI agent a trusted identity so you get the power of AI without the risk. Secure every agent. Secure any agent with Okta. And up next we have Eric Sufert from Heracles Media. He's been on the show before from Mobile Dev Memo, if you're not familiar. He got us absolutely fired up about advertising. We're huge fans of him. We're huge fans of advertising and he's in the Restream waiting room. So we will bring him in to the TVP ultradome. Eric Seufrit, how are you doing? Good to see you.
- [120:13]
G
How's it going, guys? Thanks for having me.
- [120:15]
B
Thanks for hopping back on. We've been following the ChatGPT ads rollout, the story. You've had a ton of great analysis. We've talked about it a little bit on the show, but we wanted to hear it from you and dive way deeper so maybe you could set the table for us on what the launch is looking like right now, what the product is looking like, what your expectations were just maybe begin at the beginning of this saga for the ads in ChatGPT.
- [120:43]
G
Well, the saga begins where ChatGPT begins. We all knew we were going to end up here. We all knew ads were coming. I mean, I wrote a piece May 2025. I said, obviously ChatGPT will monetize with ads. Why did I feel so confident in that? Because they had just hired Fiji Semo. She is kind of known for having led the mobile newsfeed ads product to Facebook. That's obviously been one of the most successful ads businesses in the history of mankind. And so it seemed natural that, that they would, they would bring, given the domain expertise they had acquired, that they would bring ads to bear for ChatGPT. And they have now the launch looks like the launch of Netflix ads.
- [121:20]
E
Right?
- [121:20]
G
And if it remains that way, it's probably doesn't bode well for the. For the business. But I think they're going to evolve it over time. What they've said now or what's come. The information has had a lot of good reporting on this. They're charging on a CPM basis, which is not standard for direct response ads. It's more of a brand advertising formulation. So it's $60 CPM, which is exactly what Netflix came out of the gate charging. They're asking for commitments of less than a million dollars. So this is clearly like a testing phase and they're going to offer very little in the way of measurement or targeting. So this feels like a very sort of primitive mvp. But the question is how quickly can they evolve it into something that looks more like the Facebook ads platform?
- [122:07]
B
What do you think got us to this place where you get Fiji? Simo, she's done this before. She has massive connections and skill set, but it still takes eight months to roll something out that feels not as mature as something else. Is it a talent issue? Are there other regulatory concerns or do they need to worry about certain edge cases that people might not be aware of? Or is it specifically strategic to actually. Is there some secret reason why they're taking this approach?
- [122:38]
G
So here's my hypothesis, right? And so it's not just about Fiji Simo, she's the CEO of Apps, but they acquired Statsig, which is staffed by a lot of ex meta people. That was an experimentation platform. That's the exact same exact kind of talent that you'd want to ingest into a company if you're going to roll out an ass platform which needs to be aggressively experimented on. Here's, here's what I think happened. I think, you know, chatgpt so OpenAI kind of evolved over time. It was a research lab then, you know, obviously they've had the drama around trying to pivot into a for profit corporation. My guess is the investors wanted them to bring in somebody that was a little bit more sensitive to, you know, the commercial sensitivities. Right. And so, you know, Fiji CMO would be a very good fit for that profile. They brought her in and my sense was there probably was a little bit of animosity internally towards the idea of this being an ads driven business. Right. Primarily an ads driven business, which I.
- [123:33]
A
Think it ultimately will be. Only like two years ago, Sam said I look at ads as like a failure state for.
- [123:41]
B
Yeah. And then he started pivoting the rhetoric and started saying, well, I like some of the Instagram ads when they're targeted. Well, they're actually delightful and in the right context, they can be good. And he was coming around to the 3 of our worldview, which is that advertising is great, but he did have to message that externally and then obviously had a whole bunch of work to do, messaging it internally. And maybe that's why there was a little bit less energy internally to really push towards that. That makes a lot of sense.
- [124:07]
G
Well, he may have been the one that was messaged too. He may not have been the one controlling the message there. He may have been the one that was with messing message too. We need ads because ads are inevitable. It's not that ads are great. Ads are inevitable. If you want to reach what I call humanity, scale ads are inevitable. If you want to monetize that at the scale it can be at its total potential, you need ads. That's how you do it. And so he may have been the one that was convinced and bringing someone external in who had been there, done that may have been part of an investor strategy. But putting aside my hypothesis, look, I think this has the potential to be an incredibly successful initiative. The question is how quickly can they pivot into something that looks like the meta model, Right? So if you look at it right now, it's, you know, and I think look to everything. Everything you just said mirrors what happened at Netflix. For years, Netflix said we would never do ads. Ads. Ads are offensive. They're an affront. They're front to, to, to. To my sensibilities as a consumer. That's, that's what the leadership said. And then one day they weren't right. The exact same thing happened with OpenAI. The question all in six months.
- [125:03]
A
All in. Yeah. All in was. And we, we begged them, we begged them, we begged them. And now, and I heard they made the decision independently, of course, but they're running ads now. They always.
- [125:16]
G
Yeah, I've been banging the drum since May 2025. Yeah. But I think the question is, is it six months, is it 12 months, is it 18 months to get something that looks like meta ads? Right. And that means targeting, right. Based on behavioral profiles. That means measurement, that means allowing. That means explaining to advertisers the sort of impact of what that you've delivered on their behalf. And that means like a whole lot of tech that needs to be built out. Right. That means, you know, creative optimization, all these things just conversion optimized advertising is very different from what they're offering. And is it six months, is it 12 months, is it 18 months to get there? That's a big question. Right, but that'll determine the success. And if you say, look, this mirrors what Netflix did, well, that's not really a playbook that you want to draw inspiration from, right? I mean, they had to pivot. They had to do a hard pivot. I mean, the ads tier has been live for three years. They did 1.5 billion in 2025. Right. So, you know, it's not a significant chunk of the revenue yet. I think it could be a much larger chunk of the revenue. But Netflix has other concerns, right? They've got quality concerns, they've got, you know, consumer sentiment concerns that where they can't just like fully embrace ads. I don't think ChatGPT is limited in that way.
- [126:19]
B
Yeah, yeah. Take me back to that initial launch of Netflix. You said similar CPM$60 per thousand views. But what, what was the response like? Like, were there any advertisers that were like, yes, this is great. I got my money's worth it. Vague and unattributable problem.
- [126:37]
A
My perception is like, great performance marketers are always excited to try new channels, especially a channel that's driving high intent traffic. Right? So if you tell a performance marketer, hey, I've got a billion users and I'm going to let you advertise to them, they're going to be excited no matter what the general setup is because they just want to get in there with a test budget and like, you'll talk to marketers and at a certain scale, they're like, yeah, let's try 100k here, let's try 100k there. It's just like, like you're just kind of throwing around money and then you're going to double down on what really works.
- [127:09]
E
Right.
- [127:09]
G
For the initial rollout. This is how you launch an ads platform. Don't get me wrong, this is how you launch an ads platform. That's why I'm talking about six months versus 12 months. This is how you do it. This is how you bootstrap the data. This is how you get people onboarded. This is how you get feedback. You have to launch it like this. There's no other way. You can't launch a conversion optimized ads platform because by definition, you don't have the conversions data yet. That's another reason why I think they launched instant checkout. I think instant checkout is a stalking horse. I think that's a method of gathering conversion data that they can use to then target against with ads. I think that was the whole purpose of that. I don't think that was seen As a long term revenue opportunity, I think that was seen as a way to bootstrap the data. But this is how you launch in an ads platform. It's got to evolve into a conversion optimized ads platform. But you can't be that at the start. You can't intuit the settings, you can't intuit the tuning that you need to implement to make this work. But the question is, how long does it take them to get there?
- [128:00]
A
Right.
- [128:00]
G
And so that, you know, that we'll just see. But my sense is they've got the DNA, they, you know, the information reporter, they have 700x meta employees. Like essentially everyone at Meta is working on ads. Whether they're working, you know, as an account rep or they're working as an ML engineer, they're all sort of rowing in that direction.
- [128:16]
A
Yeah. What was, what's been your reaction to instant checkout? There's been a lot of debate on the fee structure. Some people saying no retailer can afford this. We were pushing back on that in a number of different cases. There's plenty of brands where you say, like, would you spend like 4% of AOV to acquire a customer? I can also see the other side of it. A concern that I have is like, you know, you're running an ad on a Meta platform to find a customer. The customer then goes to ChatGPT, if that's where they're searching for products. And then the brand is effectively paying twice to like actually find the customer, drive them, get them interested, and then at the bottom of the funnel they're paying like another 4% fee. That feels like not great, but what's been your reaction?
- [129:02]
G
Yeah, so I was in the former camp. I said that that's not a workable fee. Let me come back to that though, because you asked a question about what happened when Netflix offered ads. They had too many advertisers sign up. They had too many. They were supply constrained, not demand constrained. They had to return money. I think that's exactly why OpenAI is saying, no, look, we're setting an upper limit on this. A million is the maximum.
- [129:19]
C
Right.
- [129:19]
G
Because they won't have enough people to advertise to. This is just a way to test it out and to start scaling. The problem Netflix made was they never act. So the problem that Netflix had from day one was they had the partnership with Microsoft, with Xander. They were using external tech to run the ads. Right. Xander's an ad platform and so that's the problem. They couldn't implement custom placements, they couldn't do a lot of measurement stuff. They couldn't do a lot of targeting stuff because they weren't operating the actual pipes, the actual plumbing that was delivering the ads. There's a lot of like data, data issues, just access to data issues. That, that's not what OpenAI is doing. Your opening is starting out by building their own tech and Netflix, by the way, pivoted to that, but they pivoted to that very late. That's why I think the ads initiative at Netflix was not successful and hasn't really delivered that much revenue. But going back to the 4% fee, first of all, here's what I think is going to happen. I think they're going to drop that to zero or very low. It'll be tiny because I think the whole point of this is to just drive conversions. And the whole thing is like when you see performance in ChatGPT as a retailer who's just basically paying for the conversions essentially with the fee, you're going to want to actually advertise against that. You're going to want to control it because you have no control if it's just surfaced based on what ChatGPT believes is relevant. And at some point you're going to have too many retailers to place into a single placement. So you're going to have to mediate that by bids. Ultimately, I think this gives way in the main, this gives way to advertising. But whether the 4% is sustainable or not actually comes down to how you view what those purchases are. Because it's not user acquisition, you're not acquiring a user. If you were, then you'd compare that to ltv. But since it's not User acquisition, you're not acquiring that user as a sort of like with a long term relationship in mind. You're essentially trading a percentage of the checkout for the checkout, but that user doesn't interface with you, right? And in a lot of.
- [131:05]
A
So there's going to be. No, there's not going to be pass through to the underlying vendor. You're saying like, well, no, I mean.
- [131:12]
G
Look, you're the merchant of record, right? And you get their information because you have to fulfill the order. But chat GPT or OpenAI specifically says in the terms that you can't use your email address for like remarketing, right? And you know, Amazon says this too.
- [131:24]
H
So.
- [131:24]
G
So the thing is you're not capturing the customer, you're just getting. You're giving 4% for that one off transaction, but you have no control over whether you reach that customer. Again, whether that customer is brought into your orbit, there's very, that's a very different proposition from advertising to someone and acquiring a customer. You acquire that customer, you get a whole stream of opportunities to remarket to them. If you're just trading 4% for the transaction, that's all you get. So it's not actually customer acquisition. If you think about 4% per transaction. Now consider that if you thought about, people talk about when that number came out, they talked about, well, some people spend 50%, 60% on Facebook ads. Why is that different? Because that's over the life cycle of that customer.
- [132:01]
A
Not for the ones that or even one time they pay for the customer one time and then they come back 15 more times or they end up subscribing and they've actually owned the customer relationship.
- [132:11]
G
Exactly, exactly. Think about it also this way. That 4%, that might be all the margin you're getting as an E commerce operator. The thing is, that's not preventing you from advertising. That's not saying, okay, you get to save your entire advertising, but you're going to still have to advertise everywhere. This is going to be a tiny sliver of your orders that you're not in control of. You can't control it. This could be lumpy, this could be up one month, down the next. You can't control that. You're going to have to continue to advertise. Now you have no idea to your point, whether this is cannibalizing purchases that would have happened from the people that you advertise on other platforms. This ends up just being a cost, a drag probably on your advertising expenditure elsewhere.
- [132:51]
B
Can you explain attribution and the evolution of Netflix's ad product? Because if I'm sitting on my couch watching a movie and I see an ad and I go to my phone, it feels really disconnected. Even in a pre ATT world, it feels extremely hard. How do they do attribution? How good can this get? And then with ChatGPT, what's the upper bound on attribution for OpenAI? Given that we are in a post.
- [133:19]
G
ATT world, upper bound is basically what you see with Facebook because it's going to be direct response, it's going to be click based. Right. With Netflix it's a little different. So with ctv, the way that companies do measurement and attribution in CTV is traditionally they set up what's called a clean room. I set up this environment. It's a centralized environment. I push all the ad interaction data that I have. So basically when I Know, I showed an ad to someone, I push all the data into this clean room. A lot of times it's IP address and device information. And then the advertiser pushes that information into the clean room and I match it up. That's attribution for CTV traditionally. Now there's other ways to do it. There's other ways to do measurement that, that's just totally probabilistic. This is like media mix modeling and stuff like that. But with attribution, when you want to use that term, traditionally it's done with a clean room like that. So I'm just linking the data sets together based on some key. A lot of times it's the IP address. But if you think about clicks, clicks out, that's like I'm viewing, I'm on second screen. I can look at time, there's time elements. You look at temporal elements to do attribution. But if I'm clicking out, I've got the click. I've got the click. So that carries a lot of information. I can still look at temporal aspects of that, but I can do a lot with the IP address too. And so that's, you know, if you look at a capi, that's what capi is. A capi is just a way to get this data on the back end. So it bypasses the apps, you know, it bypasses the mobile device, it bypasses the browser. So there's no way for, you know, Apple to interfere with that with ITP or ATP or att. And so you're able to get, you know, it's still probabilistic in that way. If you consider, you know, the decay of the validity of the IP address over time to induce probabilistic measurement. But that's. Traditionally, I've done it in a pixel as well. But I mean, that's what OpenAI has at its disposal. It's got the cap and the pixel because it's going to be clickbait. So there's a high degree of fidelity that will come with the measurement there, the attribution. That's, that's way higher than what you can do with ctv.
- [135:08]
A
What do you think the prospects are for the number of companies that are trying to build LLM ad networks? I got pitched at least a few companies that were, you know, maybe a year or so ago saying, like, hey, LLMs are going to be free. They're going to insert ads. We're going to build this kind of network to serve ads across a bunch of different apps. I was super bearish on those specifically just because I thought that OpenAI would want to own the full stack. Obviously Google would own the full stack and I think we're heading towards an oligopoly of sorts in consumer AI. And where's the scale really going to come from but. But any kind of takes on that front?
- [135:47]
G
Well, there'll be a long tail of agents that are going to want to monetize with advertising. So I think there might be space for an ad network to exist that services them. I don't know why that wouldn't be an existing ad network. I don't know why that wouldn't be. Or just sort of dsp. I don't know why that wouldn't be the trade desk say. But yeah, I mean I think there's probably an opportunity to monetize some of those long tails of agents with ads. But if you think about a ChatGPT, of course they're going to build their own. Perplexity is apparently restarting its ads initiative. I think these companies, it would be silly to not build their own. You want to have that ad stack so you can make everything bespoke and maximally perform it. If we look at Meta earnings, that's exactly what delivered that big beat. I think what people forget about the prospects of this AI enablement with advertising in digital advertising is that these effects compound over time, especially when you're talking about direct response based ad platforms. Right. So like with Meta, I've been banging this drum since Q1 2025. These effects compound like people point to like oh 3.5% or 5% or whatever increase in click through rates or increase in conversion rates from from Andromeda or from Gem or from Lattice. Who cares? These are tiny. No, but first of all that's every quarter they're noting these performance improvements but also they compound over time. If I'm targeting 90%, a 90 day recoup on my ad investment and you know, 120, 110% ROAS what am I doing after 120 days? I'm reinvesting the 110% in advertising on your channel. And so if you're giving me 3.5%, 5% bumps every quarter, that's just going to end up getting reinvested. That's why we're seeing the growth re accelerate. Growth is re accelerating going into Q1 2026. That's amazing. That's why Facebook was up 10% last night. Because growth is re accelerating. They're going to see growth rates they haven't seen since the post ATT doldrums. That's incredible. That's incredible growth. And anybody who doubted the Capex going historically has to accept that reality now. And they did. They did yesterday.
- [137:42]
A
Yeah, yeah, yeah. What people like. The criticism is like, Zuck is being so ballsy with capex specifically on Gen AI. On Gen AI specifically. And the criticism is like, hasn't launched a hit gen AI product yet. Our take earlier on the show was like, look, this guy is serving more Gen AI content than almost anyone on earth. Even if he doesn't have like a breakout net new product, his product has massive tailwinds because of all this. Like, he's in a perfect position to decide, yeah, I'm going to, I'm going to be, you know, one of the biggest spenders in this category and plan, you know, however many years out in advance.
- [138:22]
G
They're making more money on AI than anybody except for Google. Look, the idea that they're not, they're not utilizing generative AI. Have you seen the ads on Facebook? Those are all generated. There was a revolt, there was an advertiser revolt. There was a scandal recently because Meta was being too aggressive with the ads generation. The idea that they're not implementing generative AI is absurd. I mean, first of all, generative AI is not just what you see in the output. I mean, they talked about pairing LLMs with AD ranking. That's really fascinating. Cutting edge research in advertising, using LLM to sort of give you feedback on the ad. Does this resonate? Is this something that someone would click on, like pretend you're this demographic? There's already research that shows that that has a really beneficial impact on advertising. It's not just like what you see in your feed, but also the stuff that you see in your feed is generated. The idea that they're not utilizing generative AI in advertising, that's absurd. You can see it just go on Facebook right now.
- [139:19]
B
I want to move on to other companies, but I want to have one more question about opening I Do you think that there's any, any sort of risk to push back of the creepy ads, like being too well targeted? Facebook went through that with the T shirts that said your name on them basically, and your whole career path.
- [139:36]
A
My assumption is that eventually you're only getting ads that were specifically generated for you. This idea of historically you'd have one ad that a brand would make and they'd run it at the Super Bowl. They're like, we're gonna send this to all of America and then eventually with, you know, on platforms like Meta, I would assume that every ad is like a one off generated and it knows, it knows exactly how to position a product, what color, what environment to put it in. And it'll be insane that we used to have ads and we'd make one piece of creative and then run it to like 10 million people because it was like good. And it's like, no, this will all just niche, niche, niche, niche, niche down more and more.
- [140:15]
B
What do you think?
- [140:15]
G
Look, I mean, it's like the idea that everyone would see the same feed when they opened up Facebook, Facebook is absurd now, right? I mean, like, look, if, if people truly hated creepy ads, Facebook user numbers wouldn't monotonically increase every quarter, right? And you know, Europeans would be lining up to buy the subscription in Europe and when in fact only 1% of users in Europe has taken that offer, right?
- [140:42]
A
Is that an offer to eliminate? Is that like an ad free, like meta offer? Is that what you're talking about? Because I know they're talking about rolling out.
- [140:49]
G
So they've got the less personalized ads option in Europe as a result of the DMA compliance. So they have three options. They have full on subscription, no ads. If you get the less personalized ads and they get the normal experience, 1% of people opted into the subscription. Like people hate ads. They just hate them less than every other monetization model, right? People love ads. And if you look at demonstrated behavior, people love ads, people love access to free tools, people love access to, to sort of like endless, endless capacity, right? You start capacity constraining things or you start putting up paywalls, obviously by definition those are going to get used less or most likely by definition. The thing is, the idea that ChatGPT is going to run into this problem that's unique to chatbots. I just don't buy that. And first of all, maybe they will because maybe they'll design this the wrong way. But I give them the benefit of the doubt. I think they can do it in the right way. And one way to do it, it, it's to just not tether the ad at all to the chatbot context. You could just say, look, this is a display ad for what we know you're in market for, based on capi, based on the Pixel. And we're going to make sure it has nothing to do with what you're talking to the chatbot about. So you never have to even question whether the chatbot has your best interest at heart. That's a very easy way to Sidestep that problem. But I think they've got very skilled people there. They know exactly what that trap is and they're going to avoid it.
- [142:04]
B
Yeah.
- [142:05]
A
Do you think we're going to get ads in Siri in the next 24 months?
- [142:11]
C
No.
- [142:11]
G
Because of. Just because of Apple's. Because of the optics of Apple doing that. I think Apple would probably love to do that. I think there's probably people inside Apple that are pitching that, but they won't because it's not in alignment with their optics around advertising. But nonetheless, Apple is increasing.
- [142:25]
A
We were talking, like, I think people will be normalized to ads in Apple products. They obviously are. In the App Store today, Gurman was talking about ads in Apple Maps coming in the near future. And then I think from that point, if people are starting to use Siri as a search engine, there will be ads. Eventually it will just be too. To be able to serve ads like basically natively in the UI to iPhone consumers would just be too tempting.
- [142:54]
B
They gotta do ads in the alarm clock app. I wanna wake up to a different ad every day. Get out of bed and start buying. Get online. It's time to check out.
- [143:06]
F
Yeah.
- [143:07]
G
Morning. Tonight you should be consuming. I don't know, maybe. I think Maps podcast for sure. But the thing is, like, I wrote this piece when Apple introduced att. I said, apple robbed the mob's bank. I said, look, this is just an opportunity for them to shift some of the budget that's going to meta into their own ads ecosystem. They did that because they expanded their Apple search ads. They just doubled the number of impressions per search. They've got two placements now in Apple search ads. And they're also blurring the lines between ads and organic results there.
- [143:35]
A
Right.
- [143:35]
G
So the idea that Apple hates ads, I think is mistaken. I think what Apple needs to do is they need to sort of walk this tightrope of appearing to hate ads while also benefiting from ads because they are such a, you know, an accretive sort of margin, expansive way to make money. Right. So Maps for sure. I think podcast is another, like, prime piece of real estate for ads to be inserted into. I don't know about Siri. Siri. That feels like maybe a little bit too on the nose, but who knows?
- [144:03]
B
What about Gemini? Did you. What was your reaction to Demis talking about ads in Gemini?
- [144:08]
G
That makes total sense because they're already monetizing Gemini with ads. They're doing it with AI overviews and AI mode. They're monetizing ads and AI overviews at parity with search. Why would they rush to put ads in Gemini the chatbot? Keep in mind Gemini is two things. Gemini is the family of models and Gemini is the chat bot. Why monetize with ads in Gemini the chatbot? You never need to. You could just drive adoption of that. They're already monetizing Gemini, the family of models through overviews in AI mode. Overviews reaches 2 billion people a month. That's the biggest single LLM output ad surface that exists. They're already monetizing Gemini through there.
- [144:47]
B
Yeah. And is that product just what's the shape of that product? Is that just the exact same as just branded keyword search on Google? Like what? Are there anything different or any learnings from the ads in AI search overviews or AI powered search?
- [145:04]
G
So AI mode is essentially the chatbot AI overviews is the results that basically takes up now the whole above the fold on Google search. But it's better, it's better than search. You know why? Because the whole point with search is I want this to be one shot. I want to type in my keyword and I want to get the best possible result in that first list of results. And if I don't, then I think your product is bad ad. Right. And so Google is incentivized to make sure that there was one click and that's it. You get one chance to show ads with overviews. The whole point is to then prompt further queries. Right. So you get a bunch of opportunities to show ads. Right. So if that's monetizing at parity is probably also driving impressions up. I think that's a much better ad surface area than traditional search.
- [145:44]
B
Yeah.
- [145:45]
A
How should startups, business owners be thinking about TikTok under new ownership?
- [145:53]
G
TikTok is funny because they're going all in on commerce again. It seems like they had a really big kind of Black Friday whatever event revenue pop, revenue spike. But they seemingly walking back the TikTok commerce opportunity now they seem to be leaning into it. I'm not quite sure what to expect there. TikTok, I think they were in a little bit of the stasis for a while just because of the uncertainty around what was going to happen with the spin out. My sense is now that they have the certainty they'll probably lean into where they feel like they're best positioned and that might be commerce now that meta has completely abandoned it. But I just don't think that that's the right approach. I don't think western audiences really appreciate that as much as Just the ads driven model and so the pure play ads driven model, we'll see if that ends up working for them. But yeah, I mean, I think TikTok, if Meta can do what they did to snap with TikTok, they're going to be in a pretty difficult position and they may have to just adapt in ways that Meta won't. And that would be leaning into commerce more heavily.
- [146:56]
A
Yeah. I always assumed maybe naively that they would lean out from commerce because it seemed like it was just a money pit. They were just lighting money on fire and the new owners would say, okay, we just bought this, we have to pass back 50% of our revenue to ByteDance. The whole economic model of the business seems super upside down. Let's just serve ads and try to run the business lean and maintain the user base. But leaning into commerce just feels like, I don't know, there's so many brands that I've heard about over the last 12 months that are scaling into the hundreds of millions of revenue and it's almost all on TikTok shop and it just feels like that was basically subsidized by ByteDance for a long time. And will the new owners want to subsidize these brands that are driving real volume? But ultimately, I don't know if they would be able to stand on their own two feet.
- [147:49]
G
Yeah, I mean they did walk it back. I mean they did two rounds of layoffs in the commerce division. Yeah. I mean the thing about one thing you have to remember about TikTok and the ByteDance and the whole like the parameters of this deal is that they're still licensing the algorithm. Right. So I think that impacted the kind of the valuation and that may impact some of these other economics. But yeah, I mean, I just feel like they maybe sense that that's an area where Meta retreated and so that's where they could potentially flourish. But it was odd because they did walk that back and now they seem to be embracing it again.
- [148:21]
B
Yeah. What's your current thinking about the Netflix Warner Brothers deal? It feels like we're in a lull. It's moving forward, but would be interested to see.
- [148:32]
A
Yeah, I'm also interested just how do you think about Paramount's prospects as an ad platform without all of that ip?
- [148:40]
G
Yeah, yeah, well, so it's interesting. So the thing is the IP is so important. Right. So one of the aspects that I really interesting about the Netflix deal was that they essentially valued WB Games at 0. Right. So WB games makes, whenever they're a big hit games and makes their Game of Thrones game on mobile and they've got Mortal Kombat games and a lot of different console games. Batman, Arkham, and at one point it was for sale in the multi billion price range. Right. And so Netflix essentially valued at zero. Why? Because ultimately there'd have to be like an IP licensing agreement that allowed them to continue to monetize those games. Right. So, so I think the same is probably true on the IP side. Just generally with wb. My sense is what this indicates to me is said explicitly in the earnings call. It's like we're going to pay for content one way or the other and we need this high end ip. If you look at everyone saw the Joe Rogan clip with Matt Damon and Ben Affleck talking about Netflix makes us restate the plot of the movie every couple minutes in dialogue. Because if people don't hear explicitly, they just lose interest because they're on their phones. Like you wonder, okay, well is that a result of people's attention spans declining or is that just because the stuff you're pumping out isn't that interesting and people are not really paying attention because.
- [150:01]
B
They'Re kind of bored.
- [150:02]
G
Right. And so I think maybe it's the latter. And so if you bring in a bunch of top notch ip, maybe that keeps people more engaged and then hey, maybe that boosts the ad CPMs or maybe ultimately boost subscription numbers. But, but my sense is like they're gonna pay for content one way or the other and maybe just buying a lot of it in bulk at once is the best way to do that.
- [150:21]
B
What else are you tracking this week across earnings? Microsoft and Tesla? I don't know if you follow those stories, but what else is interesting you this week?
- [150:31]
G
Well, we got Apple right now. Yeah, I mean I think like you had Gurman on yesterday and he's just an amazing wealth of leaks. It's pretty impressive that he's able to get those. Apple's pretty buttoned down. But. But yeah, I mean I think just the AI story there is is kind of embarrassing. And you know, like I wrote this piece a while back, Apple is besieged on all fronts and I was talking about it like, look, if you look at AI, if you look at some of the threats to ATT and Europe, you know, there's a lot of issues that they face. I think, you know, one of the big questions for me when they announced the deal with Google to use Gemini for Apple intelligence, which among other use cases include Siri, was how they would just manage the privacy aspects of that because they were so adamant, they talked about the private cloud compute. They were so adamant that when we do this, we're doing it the Apple way, we're doing it the privacy centric way, then they do the privacy washing thing, which is just then to partner with Google and allow them to do all the dirty work of sourcing the data. Right. Sourcing the data to train the model.
- [151:40]
F
Right.
- [151:40]
G
Or of just monetizing with ads like, and they just get a cut on the back end. And in this case Apple's paying, but at some point, but at some point Google be paying for penetration into iOS and the exposure. And I think, you know, that's just a question of like, is that really the reason, is that part of the reason why they never invested here? Because they couldn't really do it in the Apple privacy centric way. And if you want someone, you know, if, if you understand that the dirty work of sourcing this data, data in ways that maybe you're not completely above board or just monetizing with ads. I mean, Apple makes so much money from advertising, they just don't do it directly.
- [152:11]
E
Right.
- [152:12]
G
They do it through the partnership with Google. Now they're gonna be making money through AI interactions, but they're not actually the ones training the models and sourcing the data in ways that could be questioned. Right. So I think that maybe plays a part in it and I think maybe when you get to a point where all of this can be done on device, Right. When the device is actually powerful enough to service most of these use cases, maybe a smaller models, you don't have like foundational models. We have smaller models that can all run on device and maybe Apple be more interested in that. But I mean like you have to look at the AI thing as either as a massive whiff, which seems to be what Gurman's take was, or it was just a sense of like, look, this is actually going to be really difficult for us to reconcile with our brand imagery and the sort of messaging that we've, you know, we've relied on for years and years, like, you know, privacy, that's Apple. Maybe we'll just outsource this in the same way that we outsource search and advertising to Google until we're able to all process everything on device and ensure that we can maintain that messaging.
- [153:06]
B
Yeah, my take is I agree with the WIF narrative, but also it does feel like there's the seeds of a rebuild. Between this new acquisition that they're doing, they're starting to open the pocketbook a little bit. The hardware really is best in class. We saw that with the Mac mini boom. To some degree people are excited about the silicon, Apple silicon and then just the partnership with Gemini. Like they're going to have access to something frontier so they can start to vend it into different places. But it's just on Apple's timelines, which have been slow lately. So.
- [153:39]
G
Yeah, yeah, but the iPhone air was, the iPhone air was a flop, right? I mean we'll see what happens with the foldable. But like, I mean I think they are facing, you know, obviously the replacement cycle is really long now. I think one thing that's like kind of under discussion us is like I wrote this piece a while back, like years ago where I said, you know, what are the implications of a billion iPhones? And this is well before there were a billion iPhones in circulation. And my point was the implications are that a lot of those are in the developing world, right? They're secondhand third hand iPhone devices. And that's kind of like there's this strange dilemma that Apple faces. Like the more powerful the devices get, obviously the longer you can go without replacing them, right? Because the demands of Spotify aren't changing, the demands of, of Facebook and Instagram aren't changing from like a hardware perspective, right? And so the question is like how, how, how central can you make AI to the core use case, which is a lot more compute intensive? If you do everything on device, how central can you make that to instigate a need to replace the device more frequently, right? Not just like because the power of the device at some point doesn't even matter. Like okay, you tell me like the compute capacity of the next iPhone is 10x what I have now, okay, what are the demands of Spotify 10x? Because that's the only app I use for hours and hours every day, right. And so that's kind of the question. And so it's like if all of that functionality can sort of like percolate down to like the iPhone 7, right? And no one really sees a difference in sort of like the, in sort of like lifestyle between like the brand new iPhone and the iPhone 7 with they're only using Instagram and Spotify, how do you actually convince people to upgrade on every cycle, right? Because keep in mind as that sort of install base, that iOS install base grows, most of that's going to be through secondhand.
- [155:24]
A
Yeah, yeah, Gurman, it was interesting yesterday. He shared that a lot of Apple's advantage on the hardware side in AI is Because of the self driving push of the car that ultimately failed allowed them to do the R and D that that is now being rolled out in devices. Apple had a huge quarter beat estimates on the revenue side by 4 billion with China specifically and up after hours. The other news that's kind of interesting. I'd be curious before you leave, Eric. ChatGPT is retiring 4.0 as of today.
- [156:04]
B
Yeah, that's interesting.
- [156:06]
A
Putting it to bed. I'm sure that'll make a lot of.
- [156:08]
B
People mad but simpler process for them.
- [156:12]
G
Well, yeah, but they also launched go, which is the sort of like lower price point subscription tier. So my sense is like, you know, if you're trying to push people to adopt that, which I think they are, you maybe have to get rid of some like the older models that the free tier relied on.
- [156:27]
B
Well, thank you so much for taking the time to join. This is a blast.
- [156:31]
A
Yeah, really enjoyed it. Eric. Come back on soon.
- [156:33]
G
Yeah, we'll talk to you soon.
- [156:34]
B
Have a good one. Goodbye. Cognition, they're the makers of Devon, the AI software engineer. Crush your backlog with your personal AI engineering team. And I'm also going to tell you about Vanta Automate Compliance and Security. Advanta is the leading AI trust management platform. And without further ado, we have Kevin Wheel from OpenAI, the chief production officer in the Restream waiting room now.
- [156:58]
C
Hey, what's up guys?
- [156:59]
B
How you doing? Great to see you.
- [157:01]
C
I'm good. It's good to see you again. Thanks for having me on.
- [157:03]
A
It's been a minute.
- [157:04]
B
Yeah. What's new with you? Since we've had you on the show, you've changed duties, you've changed focus, like reintroduce yourself, I guess.
- [157:11]
D
Yeah.
- [157:11]
C
By the way, I love that we are talking about science today. I think we need to bring more of this kind of coverage to science and get people excited about science because there are a few more important factors in all of our lives and AI is in the process of fundamentally changing science.
- [157:29]
B
Yeah. What is your focus? Because science and health can mean a bunch of things. It can mean D to C advice. And one of the ways to cure cancer is just to get a lot of people to stop smoking cigarettes. That's something they could learn by chatgpting should I be smoking? And it makes a very convincing case for why you should not. And maybe it helps you not smoke at the same time. There's researchers that are gonna spend decades developing cancer or drugs getting them through FDA approvals. There's a whole bunch of different ways that Even Microsoft Word and PowerPoint and Excel helped the previous generation of biotech companies. You can imagine LLMs vending in there, and then you can imagine the fully automated AI scientist. What's been your focus? Where do you think we're going to see the most progress in 2026?
- [158:15]
C
Yeah, the mission of our team is to accelerate science. And the reason that we are starting to do this, I mean, it's obviously been something that OpenAI cares about a lot, but when you go back a couple years, we were all amazed that GPT 3.54, whatever could do well on the SAT. And then you start seeing it solve graduate level problems and then it does okay on math competitions. And then you fast forward a bit and our model can win a gold medal at the IMO at the top math competition in the world. And now, now we're seeing it solve open problems in mathematics. So things that humans have not solved, suddenly AI models are contributing solutions to. And it's not everything, it's not every open problem. Science is far from done. Right. But the fact that AI models can now contribute at the frontier of science and actually push beyond where humans have gone before is incredibly exciting. Because if we can, if you can replicate that across math, physics, biology, chemistry, material science, and you can do the next, say 30 years of science in five years. Instead, if we can be doing the science of the 2050s in 2030, the world is a better place and the models are there now. And so this is a huge focus for us.
- [159:36]
B
So, yeah, talk a little bit more about your work specifically, because not to discredit you, but I feel like part of the magic of these large AI models is that you get a lot of this stuff for free. Oftentimes you don't need a specific SAT model, you just build an amazing model and it learns how to code and also learns how to write poetry and tell jokes and it can do some science too. So how are you working on fine tuning? Is there a reinforcement learning project? Is there specific compute allocated towards this? You have a team. Are you listening to certain customers?
- [160:08]
A
Yeah, I imagine my guess was just working really closely with customers, understanding, making sure they're not just signing up for ChatGPT off the shelf and using it, but actually building a bunch of specific functionality for them and innovating on the product side.
- [160:21]
B
Yeah. How do you think about that?
- [160:23]
C
Yeah, well, it's not about building offshoot models. This is still about improving our core models. But if you think about science, science has a huge surface area.
- [160:35]
E
Right?
- [160:36]
F
Right.
- [160:37]
C
So there's a lot that we can teach models at the frontier that, you know, there's, there's work that we can do to make, to make our models even better at performing frontier science. And that crosses from pre training into post training and reinforcement learning. And so we get to solve, you know, frontier problems in science to make our models better. There's also a whole lot of work that we can do on the tooling front. You know, scientists use a lot of different tools and you want to bring those to bear so that the model is using the same kind of advanced scientific tooling that scientists are, because then it can be an even greater force multiplier for the work they do. So there's a whole bunch that we're doing on the pure model front. That's not the only thing though, because I think if there's anything that we've learned from the way that AI has completely revolutionized coding over the last year is it's two things. It's great models and it's also integrating AI into the environment where engineers operate. Right? You don't, you don't spend your time going to ChatGPT and copying and pasting back and forth. You use codecs in your ide.
- [161:49]
B
Okay.
- [161:50]
C
And bringing the model into the environment where you're working is a huge part. That's why we launched Prism yesterday. The idea is, or I guess on Tuesday, but the idea is to bring great models that can help with AI into the workflows that scientists are using every day. In this case, Prism is about scientific writing and collaboration. And if you can help people communicate their ideas faster, if they can use AI to express the research that they've done, collaborate with other scientists, that's its own form of acceleration. So we both want to make models smarter and we want to bring AI into the environments that scientists are operating in day to day. Both of those are parts of accelerating science.
- [162:30]
B
So talk about some of the integrations. ChatGPT at one point just got access to a Python REPL and could sort of write code and execute it in the query for just a GPT5Pro query. But ChatGPT also has an integration to Gmail and there's some sort of business development relationship. What is your as you're bringing more tools to bear in the scientific workflow, is there some sort of balancing act? Do you just need to integrate with everything? How do you think about actually wiring up and unhobbling the models?
- [163:06]
C
Yeah, I think it's about both a handful, a set of tools that we think are going to be broadly useful that we can integrate ourselves. As an example, if you're in the process of trying to solve a hard scientific problem, you need to solve a differential equation.
- [163:23]
B
Yeah.
- [163:24]
C
Models are actually smart enough these days that they can solve a differential equation just by reasoning through it. But you also can integrate a computer algebra system, a system that will deterministically and very quickly solve differential equations. Why not let the model use that and go back to doing what it does best, which is reasoning broadly about how to solve hard problems. So you have that. You have, like, protein databases in biology, you have so many tools like this. And I think it'll be important because I might be a scientist studying the evolutionary biology of snails, and I might have my own set of tools that I use or maybe small models that I've trained. They're good at doing very specific things. So it'll be both about us teaching the models to use certain tools that'll be broadly helpful and enabling scientists to bring their own tools to bear so that the models can very deftly adopt them into the process.
- [164:19]
A
What does a ChatGPT moment look like for PRISM within the scientific community?
- [164:27]
C
Ooh, that's a good question. I think it's going to be a.
- [164:36]
B
More.
- [164:38]
C
Process of incremental compounding, where you've got a product that helps people work and operate faster. You've got models that are going to be increasingly useful, and we're going to see kind of two exponentials, if you will. One is just the exponential that the models are on and their ability to help any scientist who's adopted them do what they do faster. Right. To aid in their thinking. I was just talking to a scientist earlier today who called GPT 5.2ametal detector for hypotheses.
- [165:11]
E
Right.
- [165:11]
C
So he's thinking about. He's got so many different ideas in his head. You can only run so many experiments, you only have so much time. And he uses GPT5 as a thought partner in helping him hone in on the most valuable ways to test his ideas. So you've got that exponential and then there's a separate one that comes from scientists beginning to adopt these tools, because a lot of scientists still haven't adopted them. And the more that they do, the more that they individually move faster and the more that the entire field of science accelerates. So there's a lot to be excited about here.
- [165:44]
B
A couple weeks ago, we talked to Andrew from Cerebras, and I was asking him about where wafer scale computing is most exciting to him, and he actually cited Science as a particularly valuable place. And I was wondering if you had thoughts on the value of speed or the value of different chip architectures in the scientific workload. I was kind of coming to it like, well, you know, developing drugs takes a long time. It's probably fine if the model goes off and cooks for an hour. But he was sort of pushing back on that. And I'm wondering how you think about the different parameters. Obviously, we're on an exponential with intelligence and capabilities, but. But there's also latency and usability and flavor, and there's so many other knobs that are being turned as the models progress and as the projects progress.
- [166:37]
C
Yeah. I think one of the interesting things about scientific problems in particular is that when you're solving the hardest frontier science problems, you need to do a lot of thinking. Right. If these problems were easy, really smart humans would have solved them long ago. And so the kinds of problems that are left often involve the model thinking not for five minutes or 20 minutes, which would be a long time if you're inside ChatGPT, but maybe an hour, two hours, 12 hours, two days. And that is where we're going. And if you have really fast inference that can take that two day rollout and turn it into a six hour rollout or an hour thing only takes 10 minutes, then again, that just. It's more opportunity for you as a scientist to maybe Instead of testing two hypotheses, you're testing 20 in the same period of time. Again, it's acceleration.
- [167:31]
B
Yeah. To go Back to the ChatGPT moment, how are you thinking about actually seeding your work into the scientific community? Because there's one angle where it's just chatgpt.com, it sort of goes viral. Everyone's playing with the. That's part of what the ChatGPT moment was, was just anyone with a web browser could use it. At the same time, there's only so many real scientists. A lot of them are in labs or in academic institutions. And I could imagine you doing partnerships or deals or anything to actually get something deployed into the most elite scientific environments. Have you thought about the different trade offs there?
- [168:12]
C
One of the cool things is we're seeing so much organic adoption.
- [168:16]
E
Yeah.
- [168:16]
C
You go on Twitter these days, as you do, and you're like, every day I feel like I'm seeing new examples where someone will say, you know what, I just solved this problem with GPT5. Or I gave this problem to my grad student and they were busy. It took them too long. I wanted to make worker progress. I just gave it to GPT5 and now I have a solution. And so there's this incredible organic adoption because of course, course, when other people see that, other scientists see that, they go, oh, wait a minute, maybe the last time they tried the models was a year ago when they weren't at a place that they were going to really meaningfully contribute to scientific research. They could help with other things, but they weren't going to help with your hardest problems. Now they can. And so there's this groundswell of scientists that are adopting GPT 5.2 especially, I think has been kind of an inflection point. So it's exciting to see and there's a lot happening even without us with people just discovering this and talking about it.
- [169:17]
B
Yeah, a lot of people experience that with like Andrej Karpathy's journey with Vibe coding. And a lot of people were like, okay, yeah, like if it's good enough for him, I got to jump back in.
- [169:26]
C
It's funny joining that man is an incredible communicator.
- [169:29]
B
Yeah, yeah, really?
- [169:31]
A
What's the update on how? So like forward thinking labs are thinking about integrating with a product like Prism. Ideally there's a future where a scientist could be at home, have an idea, maybe they're working in pharma or something in biology and they can just start running. You can imagine somebody starting to run an experiment just based on a prompt, and then somewhere in the physical world there's actual the biological processes actually happening. How much progress is there on that front?
- [170:03]
C
Yeah, I'm super excited about the world of robotic labs. I think it is 100% likely to be the future that we're moving towards because you can do so much more in parallel again to the idea of accelerating science and moving faster the world where you can have a hypothesis, maybe that you've honed with back and forth with ChatGPT. In this case, it may also be running simulations. If you're doing something like fusion, where you want to do heavy simulation before you run an experiment because your experiments are expensive, then you have the model thinking, running a fusion simulation, looking at the results of that, refining its thinking, running another fusion simulation, and you do as much with the compute that you have in advance so that when you do something in the real world, it's that much more likely to be successful. You can look at the same thing with respect to biology. There's no reason at this point that you need to have grad students, you know, pipetting one thing into another thing.
- [171:08]
A
The idea That I think like a scientist in a lab coat could easily fade away where they're just like in a normal office, you know, or at.
- [171:15]
C
The very least doing the kinds of things that are incredibly hard for models, for robots to replicate. Yeah, but there's a lot of science that can be totally automated. And the idea of robotic labs that are, you know, 24, 7 online, that you can scale in parallel as far as, you know, you can make it efficient. And you have models thinking, you know, reasoning for two days to find the most efficient experiments to run, maybe running simulations to test that. And then once they get to a good point, passing that to a robotic lab which can experiment in parallel at high volume, the results pass back into a model which reasons about the results and then goes out and runs a different set of experiments. You know, you're doing reinforcement learning with a loop through the real world. And that is absolutely.
- [172:07]
A
Something that OpenAI would ever do. Or is that best suited for an external partner? How do you think about that point of the stack?
- [172:15]
C
I think it'll be both. We want to partner broadly with, with scientists that have their own labs that are already doing anything. Like I said, science is an extremely broad, has huge amounts of surface area. There's no way that we can do even a tiny fraction of all of science. And I think there will be a lot of opportunity for us to learn from things like robotic labs to make sure that we're working really well with those scientists. So I think it will be both.
- [172:42]
B
Can you clarify the business model? There was some confusion about how this might be monetized.
- [172:46]
A
Poor Sarah Fryer, she keeps getting, I'm.
- [172:49]
B
Pulling for ads in science. I want a pill that I take, it helps me lose weight, but then I wind up loving Ford F150s. I'm willing to take that trade ad supported medicines. That's what we need. But no, what are you actually thinking? Is anything changing?
- [173:04]
C
This is why I love you guys.
- [173:07]
B
We're your strongest defenders, whatever you do.
- [173:10]
C
Yeah, Sarah said something when Sarah was talking about, about. Oh, maybe, maybe there are ways to like monetize ip.
- [173:18]
E
Yeah.
- [173:18]
C
She was speaking specifically to the idea of us doing partnerships with large companies. You know, pharma companies, people like that.
- [173:27]
B
Yeah.
- [173:27]
C
Where there would be a specific partnership that was developed with the idea of sharing royalties in the future. It was not meant at all about normal.
- [173:37]
B
And this already happens with the OpenAI investment fund. There's startups that take that, take capital from OpenAI. They build something on top of ChatGPT or GPT APIs and like, I think most people inside sort of understood that, but thanks for clarifying. So in general, most people will just be on sort of some sort of like API consumption based pricing or subscription fee, mostly the normal.
- [173:58]
C
The nice thing about Prism is you log in with your ChatGPT account, so you just bring your existing ChatGPT account along with you. Our. This is not, you know, there are not a billion scientists in the world. This is not an effort to like build a brand new business model. This is about accelerating science because I think it's one of the most impactful and mission oriented things we could possibly do.
- [174:18]
B
Yeah, that makes a ton of sense. Give us the update on Detachment 201. How's it going? Are you sold?
- [174:23]
C
Oh, it's been a blast.
- [174:24]
B
Working out a lot.
- [174:25]
C
It's been an absolute blast. I've been in D.C. to various spaces, things like that.
- [174:31]
F
It.
- [174:32]
C
I think it is massively important that we bring Silicon Valley and D.C. closer together because we have an incredible tech community, we have a country and a set of values that are worth defending. And it's a dangerous world out there. The more that we can do to strengthen who we are and what we do, the better.
- [174:53]
B
That's great. Well, thank you so much for taking the time. Come talk to us. Very fascinating stuff.
- [174:58]
C
Yeah, thanks so much for having me on, guys. Good to see you again.
- [175:01]
B
Good to see you too. We'll talk to you soon, Kevin. Goodbye. 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. Up next, the New York Stock Exchange. Want to change the world. Raise capital at the New York Stock Exchange. And yes, we do have a surprise guest who wasn't on the lineup, but he's joining.
- [175:33]
A
We have Mahool. From Matic.
- [175:35]
B
From Matic.
- [175:36]
A
What's going on?
- [175:38]
B
Hey guys, thanks for the robotics mastermind.
- [175:40]
A
You're gonna, you're gonna love this. I. The other, I think it was two nights ago, my youngest one and a half walked in the other room and I was pretty quick to chase after her, you know, because she's very, very active. She can destroy a lot of things. And I go in the other room and she has the like, you know, soap dispenser of the Matic and she has it fully vertical. Trying to drink out of it. I was so appreciative of your design because nothing came out.
- [176:15]
B
Wow.
- [176:16]
A
So I was so worried, worried for a second because it like looked like she Was just trying to chug.
- [176:22]
B
No, that could end up very messy, potentially harmful, who knows.
- [176:25]
A
Anyways, so, so excited to have you back on. We've been loving our Maddox.
- [176:29]
E
Thanks for having me.
- [176:29]
A
And I know you have some big updates.
- [176:32]
B
Yeah. What's new in your world?
- [176:34]
E
Well, we just announced today that we've raised $60 million in the next round. So. So led by.
- [176:42]
A
Yeah, Boom. Led by who?
- [176:48]
E
Led by hi. Sutter Hill Ventures and Vic Miller. He's joining overboard. Thank you. And we've had some amazing investor in Gavin Baker at Antonio, Gracias at Valor, a bunch of other joining on their own. So it's great. And then for us this is sort of the output of the product we built and shipped. Yeah, I don't think this would have happened without us shipping. So that's been the good part of it, that it's not about the demo. It's based on real product, real usage.
- [177:15]
A
Real customer service, real robots, thousands of them. Real robots out in the real world doing real work.
- [177:23]
E
Yeah, we've. Sorry, jump in. We've shipped more than 6,000 now and have cleaned about 110 million square feet.
- [177:31]
A
There you go.
- [177:32]
E
Which results in 80,000 miles travel. So part of it.
- [177:35]
B
That's amazing.
- [177:35]
A
That's wild.
- [177:36]
B
Talk to us about pricing and willingness to pay. Obviously with modern AI robotics, everything that you're doing, you're creating more value for the customer. And I mean the sales show people are willing to pay more than the previous generation, which I feel was like more of a three digit price point. And we were talking about like will people really pay $50,000 for a humanoid? 100,000. That feels like a big jump. But what are you seeing in terms of just willingness to pay for robotic process amongst consumers?
- [178:11]
E
So great question. We actually launched first time in April of 2023. No one cared. But we started as a robotics as a robotics as a service and we tried to launch it as a subscription and that was like an organ rejection that didn't work at all. So people are definitely not interested in paying yet another subscription. That's one thing we learned. Then we price it at about $1,500 plus plus subscription as an attachment in November 2023 and learned again that that didn't work as well. That that was the price point at which a lot of people weren't yet convinced that at least the robot vacuum or this could do a job. So we really took that feedback to heart and then eventually priced it much, much lower introductory price of what, 995 and now it's because of tariff, we've had to raise it a little bit. But really what we learned, and this was our thesis to day one, that if you take a step back and think about it, there is literally zero ubiquitous consumer electronics device priced higher than $2,000. Beyond $2,000, it's prosumer or you pay for cars, right? There is literally nothing. Even my.
- [179:18]
A
So yeah, that's why, yeah, that's why I think the humanoid robot rollout is going to be just really brutal because you're trying to get adoption of a new product that is still and very much like gonna be in an R D phase. And you're asking people, you're asking people to pay the price. Car money.
- [179:38]
B
But counterpoint piggyback rides. You could take it to work, replace your car, you hop on the back.
- [179:45]
A
It'S a car that can clean your house, it's the car that can do the dishes.
- [179:49]
B
And I'm hopping on the back, give me a piggyback ride to work. And then while I'm working, you're gonna go around and clean, do all sorts of, I don't know.
- [179:56]
E
Well, AGI is coming. Maybe it will just do the work for you. Maybe we'll go there, there will be.
- [180:01]
B
Human eyes on tvp, no more commutes, but you will need a clean house. So you will buy a MATIC robot in the future.
- [180:08]
E
That's exactly right.
- [180:09]
B
Well, I mean talk about the AGI process. The process and the models. We're seeing like incredible advancements all over image, video models, reasoning models, agents. Are you seeing just, just natural improvements to the performance of your fleet? Is that just something that rolls out for free? Are you doing your own training runs? What have you learned or integrated from the progress in just fundamental AI research over the past year?
- [180:38]
E
That's a great question. We started with this idea that AI is coming and AI would make progress and that has always been part of it. And we are taking advantage of VLMs and SAM3 and all the open source model that are available to us. But at the same time, one of the things that we had realized on day one, that at least when it comes to home, getting data, given the privacy that consumers want, especially at this price point, was going to be a challenge. So we knew, so we knew that just like Tesla sold a car, collected data, and now is moving on to autonomy and robotics in the same exact way, the robotic vacuum was our badge. That's the way we wanted to get inside, home, earn customers trust, build a brand and also a data veg and we knew that if we build it the right way by doing everything on the device and earn customers trust, they will share data. And that's actually one of the most interesting thing that the long tail with which we actually need all that data is coming in and customers are willingly sending it to us and they are curating it for us as well because we know this data is actually error oriented. So that's incredibly helpful. And if you take a step back and think about it, even for humanoids, if you had humanoid in your home, you don't want it to step onto that dog poop or cat vomit or get tangled up in wires or rug tassels or maybe squeeze your or break your kids toys or dolls, right. So every single thing that we do actually directly leads into it. So our thought was always that, that, hey, instead of sort of building a humanoid in day one, let's grow a robot. Let's grow Rosie the robot and we'll solve perception first just using floor cleaning space. And then we can worry about manipulation and kind of think about what would kid be able to do between 5 to 10 year and can we enable that sort of functionality and make product useful and then ultimately get to.
- [182:30]
A
It's also kind of insane to think about like the actual reality of a humanoid using a regular vacuum, you know, just like dragging around a vacuum or using a Dyson. It's like, is that really the most efficient way to like clean a floor is to have this huge robot, like human sized robots?
- [182:47]
E
Actually it was in that movie Bicentennial Man. I remember that with Robin Williams and he was using the regular newspaper and he was doing it and we're like, that doesn't make sense. Like, you know, do you like. We have this humanoid really amazing intelligent robot, but it's our appliances never change. So there was always this idea that it has to evolve in both ways.
- [183:06]
B
Yeah. Any traction or plans on the enterprise side? I'm thinking like hotel rooms, they need to be incredibly efficient. Every minute that someone is in there is a minute they couldn't be somewhere else. Do you have plans there or is that just a total distraction at the moment?
- [183:26]
E
Hardware that we've built is really just for homes. I did put one in hair salon and it works fine. We did put some in hotels and areas. So perception and algorithms wise, it is working across the board. We do have people using it in their daycare centers, churches, schools. Animal hospital is another one that we've seen use cases. So there is an opportunity there. But I think that will require a little bit of a hardware tweak. And then we just have to question is it really about savings time, which is really what we want to do, or is it about just saving costs? And if it's saving costs then it may not be as interesting of a problem. But if it turns out that it's a labor shortage oriented one, then we really need to go and look into it.
- [184:05]
B
What about lawnmowers?
- [184:08]
E
Lawnmowers is something we want to stay inside home for a while.
- [184:13]
B
What about advertising the reason lawnmower.
- [184:17]
A
Think, think about it. It's like you guys would make it a fantastic lawnmower. But I've never once thought like in between when the like the landscapers come to my house, I've never once thought like I wish my grass was just like a little bit shorter today.
- [184:30]
B
Perfect.
- [184:30]
A
Every like because they're just coming every week. Whereas in the home. In the home it's like your house can always be cleaner.
- [184:35]
B
Yeah, every day. Every day. What about.
- [184:37]
E
That's exactly right.
- [184:38]
B
What about an ad supported tier? It says, hey, I'm on linoleum. Why don't you upgrade to hardwood? We'd love to connect you with this.
- [184:45]
A
Have you considered NordVPN? I see, I see you're spilling. You're spilling a lot of vegetables on the floor. Why not athletic greens?
- [184:52]
B
We've been talking to Eric sue for too long. Today. He got us thinking on ads and everything.
- [184:56]
E
What?
- [184:56]
A
What, what?
- [184:57]
E
Some great idea.
- [184:58]
A
What's the plan for this year? Like just scale sell a lot more robots.
- [185:02]
B
Do you need to be in Best Buy and retail distribution at some point? What do you think?
- [185:07]
E
Not at the moment. Luckily for us at the moment we are still supply constrained. So we would keep selling and doing build to order model that allows us to keep working capital low. And the goal is really to scale. We scale about 20x. Last year they're selling from 300 to 6,000 robots now. Thank you. And try to go another 10 to 20x this year. That's the goal at the moment. And there is a second product in the work. I won't be able to go into much details yet, but there will be something.
- [185:36]
B
I tried to guess it and all of our guesses were way off. A robot that you can take a pig. He's like, no, you guys are not even close. Anyway, thank you so much. Congratulations.
- [185:50]
A
Great to get the update.
- [185:51]
B
Always great to have you on the show.
- [185:52]
A
Yeah, I love, I love.
- [185:53]
B
And we appreciate the product.
- [185:55]
A
You know we talk about robots a lot on this show.
- [185:57]
B
This is a good one.
- [185:57]
A
Love A robot company with a fleet. Yeah, with a real fleet out there.
- [186:02]
E
And I think Dalian's joining today telling that we're still growing without advertisement, although we will need some at some point.
- [186:07]
B
We will, yes. Well, have a great rest of your day. We'll talk to you soon.
- [186:11]
A
Great to get. Great to catch up. Cheers.
- [186:13]
B
Cheers. Goodbye. Let me tell you about phantom cash. Fund your wallet without exchanges or middlemen and spend with the phantom card. And I'm also going to tell you about Restream 1 livestream 30 plus destinations. If you want to multi stream go to restream.com the other news today is of course that rare metals are absolutely spiking. Gold's up 4%. Silver's up 7%. Copper's up 10%. Platinum's up 7%. Palladium up 5%. Oil is up 4%. Geiger Capital says don't worry. Powell said it means nothing.
- [186:46]
A
He looks worried.
- [186:47]
B
A photo.
- [186:48]
A
We're all a little bit very worried.
- [186:50]
B
We will be monitoring that. It is an odd, it is an odd, you know, dislocation in the market. Boring business is posting a meme saying realizing that our collective retirements now depend on whether digital gold coin created by an anonymous founder goes up or down. Is that bitcoin is referring to? Wait, what?
- [187:10]
A
No. So retirement accounts will be able to directly buy.
- [187:14]
B
This is from kaushi from the SEC chair. Now is the right time to open 401k retirement market in crypto.
- [187:21]
A
And so more into that one. Update some more updates on the Apple numbers. Apple earnings mark Gurman says massive Apple revenue beat 144 billion in overall revenue. 85 billion in iPhone revenue, 25.5 billion in China revenue and 30 billion in services revenue. Tim Cook gets to keep his job. Obviously a joke but. And we can close out.
- [187:49]
B
We gotta do some of these timeline posts.
- [187:50]
A
There's a lot of good ones. Well, let's pull up this video.
- [187:53]
B
Okay, let's pull up.
- [187:54]
A
AI is going to take your job. What job?
- [187:58]
B
This is great. We will pull this up while I tell you about graphite code review for the age of AI. Graphite helps teams on GitHub ship higher quality software faster.
- [188:08]
A
I pull this up and let's go to AI check in with the job market.
- [188:15]
B
Driving is the entire job. I've seen this meme template. I have not seen this exact video but the jump in that car you have to be going so fast.
- [188:27]
A
This makes me want to buy a beater so bad drift it just wait for a rainy day.
- [188:31]
B
Drifting is is next up on the bucket list. We got to learn how to drift like that. Maybe stay off the bottle.
- [188:38]
A
Maybe stay off the bottle, but maybe close down a road somewhere. Yeah, Delicious tacos with a Fantastic post says AI CEOs are like if Henry Ford said these cars are going to run over all of your children, they're getting I appreciate it. I appreciated this one. Just going back to the where's our AI Steve Jobs? We can't have every lab founder saying that everyone's going to lose their job and AI is going to kill everyone on a long enough time horizon.
- [189:12]
D
I do think Dario has done a reasonably good job if you're pretty tapped in. He has this great essay.
- [189:18]
B
Wait, I'm all watched over by Machines of Loving Grace.
- [189:21]
D
Yes. I feel like he's not that public yet. He's starting to do a little bit of a press tour right now, but I think if he was more in the public eye, I think he could be that person. Definitely.
- [189:33]
B
Yeah. I mean Demis has also done a fantastic job and is one person that's not really highlighted with those questions. But I mean this delicious tacos post like Dar is definitely in this crowd because he is talking about job displacement.
- [189:46]
D
I guess his more recent essay was about the catastrophic risks. Yeah, exactly.
- [189:51]
B
Which are important and it's good that he is talking about it, but it is a very different It's a different world than other innovations.
- [189:59]
A
Jeremy Fawn says the best way to evaluate a VC is by picking the one with the least bad podcast you'll be forced to go on. Of course, if you raise from a vc, you're probably going on their podcast. Just face it, you may as well pick the one with the best podcast.
- [190:15]
B
Buckle up. There are some good the Jocko Willink Venture Capital for that's what you got to do. You got to do Jocko's pod. That would be amazing. GPT5 apparently serving it is profitable. According to Epoch AI, gross margins were around 45%, making inference look profitable. But after accounting for the cost of operations, OpenAI likely incurred a loss. Of course they have a huge staff, there's a lot of capex, there's all sorts of different stuff, but just on the actual inference. This confirms what a lot of people were saying yesterday about just comparing the open source models to the private models. Inference gross margins seem healthy, which is of course good news.
- [190:57]
A
We went on Colin and Samir's show we did to talk about TBPN and media and John Palmer didn't just listen, he studied and he had a great take here he said. On the Colin and Samir show, John Coogan laid out a barbell thesis. Immediately value either accrues to the platforms or the individual personalities. Platforms being YouTube, Spotify X et cetera, or the individual people creating the content. He says. I'm guessing that software is headed to the same place. Value goes to the platforms with Monopoly distribution, Google, Stripe, can think of system of records, et cetera, or the individual Vibe coders. Shipping one man businesses at a thousand x. Lower cost, not much value in between. What happened to the media business is now having new software. When production costs drop a thousand X, remember to make a show like this historic and distribute it. Historically we would have needed to be on a major television network. Now we can do it with a, you know, tiny fraction of the cost.
- [191:57]
B
And restream.
- [191:59]
A
And restream of course. But anyways, interesting, interesting take and I think we'll certainly see some of this play out.
- [192:07]
B
Gary Vaynerchuk rebranded his agency for enterprise clients. Chuck Media.
- [192:13]
A
Chuck Media.
- [192:14]
B
He's just reinvented himself so many times. He was Gary Vaynerchuk, I think that's his full name. Then he was Gary Vayner or VaynerMedia and then Gary Vee and now just Chuck, which is very interesting. I've talked to some people who have worked with them. Oh wait, this is a different agency. Oh, interesting. Vaynerchuk small business agency, the Sasha Group is being repositioned for enterprise clients as Chuck Media. I think that's separate.
- [192:38]
A
Oh, so VaynerMedia, which is its own thing.
- [192:41]
B
Fascinating. In other news, Treehacks at Stanford is looking for new judges. If you have experience building a notable project, building a major company, you're a journalist, a creator, et cetera, please reach out to. I'm going to not pronounce that Thiz coding dev. He's a 19 year old CS Stanford. And I was thinking about this, like hackathons were huge in, I don't know, 2010, 2015 and then they sort of fell off or just became sort of like background noise, like they would happen but nothing was that exciting. And this feels like the best time to go to a hackathon because people are going to be vibe coding like truly fully functional products. And there's something that the pressure of meeting a bunch of people, spending two days, you know, staying up all night grinding, it's really, really fun.
- [193:34]
D
Yeah. I don't know if hackathons have necessarily declined. Like when I was at school, hackathons were like massive. Yeah, Kept getting bigger.
- [193:41]
F
Yeah.
- [193:41]
B
I guess just the Novelty wore off and they just became background noise. Yeah, that's right. But it's got to be the best time ever. I remember the first hack hackathon I went to. It was like oh cool. Like Twilio exists so you can like build something that texts you and now it's like, okay, like you can use agents to just code you everything and you're. And the pool of people. The first hackathon I went to, I was. I didn't really know programming at all. I learned a bunch was able to build some things but it took a long time and by the end of the hackathon you sort of just had a landing page that barely worked. Now you can definitely leave. Leave with a fully fledged product if you lock in for two days at.
- [194:18]
A
A long hackathon unicorn valuation, potentially.
- [194:21]
B
For what?
- [194:21]
A
Right?
- [194:21]
B
Oh yeah, for the company for sure. And yeah, this seems like a lot of fun. So head over to Treehacks.
- [194:28]
A
I have to get it. We both have to get on with Sydney.
- [194:32]
B
Oh, we do?
- [194:32]
F
Yeah.
- [194:32]
B
Okay, well, we will have to talk about sticker packs tomorrow.
- [194:36]
A
We can talk about this for a second. Plant the bottle from Dom, founder of Fast, the payments company. Launched a new company called Stickier and I was super confused because he's selling stickers with the TBPN logo.
- [194:54]
B
This is very weird. It's supposed to be a way to.
- [194:58]
A
Short overhype startups, but you're just buying stickers with other people.
- [195:02]
B
Very questionable where this will go, but we will have to check it it out more in the future. Anyway, leave us five stars on Apple podcasts and Spotify. Thank you. Goodbye. Nice work, brothers. I'll see you on the next one.