Rage Baiting is for Losers, Everett Randle’s 5x Controversy | Jordan Nanos, Brian Halligan, Dr. Fei-Fei Li, Scott Sanders, Jeremy Allaire, Vladimir Novakovski, Andrew D'Souza, Parag Agrawal

28:25And if you look at the recent return data, I think it suggests that. So I think they'll be able to make an immense amount of money on an absolute basis. But I think a lot of these LPs are in the business to make, or in venture to make high money on money returns. Like they had PE for the low return stuff and they probably get better liquidity from PE. They're here for the high money on money returns. And this is one of the reasons why I'm extremely excited about benchmark's competitive position in today's market, because we can go to LPs. We can say, hey, we're shooting for higher than 5X net.

28:53We have the historical track record to back it up and we have the fund sizes to back it up as well. I mean, you had Miles from Carnegie Mellon come on here and do the awesome math and the very clear math of, hey, do you know how hard it is to return Forex net on$8 billion,$10 billion? It is immensely hard and it defies the laws of physics. So I think there's a difference between are they going to make a ton of money and are they going to produce the returns that LPs really want this asset class to produce? Two very, very different things. But for now, the rubber won't meet the road because, as you mentioned, there's just so much global demand from LPs for exposure to private technology.

29:28And they are happy to take lower returns. And so I don't think there's any end in sight. But I think on a relative basis between all of these different constituents and all these different GPs, there's a huge, huge delta and a huge differentiation between who can actually produce venture-like returns. Boo! He's not HGI-pilled. Boo! 10x. You're going to 10x the fund. Just 10x it again. Just 10x it again. 10x it again. the fund, raise a$100 billion fund and 10X it. Just 10X it and we're going to 10X it again. Ev was on a little bit of damage control this morning. Wait, wait. So I want to hear what exactly did he say the first line?

30:04He says, they cannot go to LPs and say with a straight face that they can do 5X net. Can we play the actual clip? Because I feel like that's not quite right. What do you say? I don't think, Ravi or Hamant or even Ben and Mark at this point, I don't think that they can go to LPs, one of those legs of the stool, and say, hey, this basket of funds that we're making you invest Pari Pissu across, we're going to get you 5x net on that. I don't think they can say that. Because maybe they can go with a straight face and say, we're going to get you 6x net. Boom. We're going to get you 10x net. Maybe they can say that with a straight face.

30:34Taylor knows. Nancy Pelosi is not afraid to say, I'm going to 10x it. I'm going to 10x it again. I'll 10x it again. Yes. Ev is fighting for his life a little bit. But tagging big shots, quick edits, clickbait call-outs, giving me pick-me vibes. People are not happy about this clip hitting the timeline. People are taking shots back and forth. What actually happened here? Part of the reality is I don't think no fund manager can really go to LPs and you can share that you believe there's a chance we'll get a 5X net. But isn't it like 99 % of funds just don't come anywhere close to that? Yeah, that's for sure true.

31:21Also, I mean, I think the broader point that he's making is something along the lines of like, I saw some other post about like there will be fortunes made just by getting retail investors and the broader capital that's out there in the world into the private MAG7. So the Open AIs, the Anthropics, the SpaceXes, the Andurals. There's a whole host of companies. That was also an Ev Randall tweet. Oh, that was him? Was that him recently? Yeah, that was, I think, yesterday. Okay, yeah. So there is a world where you set up a fund that has lower return expectations, but also lower risk, and it's a great deal, and LPs love it.

32:05And so I don't know that it's that hot of a take, but it certainly put the timeline in turmoil. Christian Garrett here is sharing the gif of digging himself deeper into the hole, I imagine. So what do you have to say? Part of it is that Benchmark is feeling pretty good right now. If you look at their 2020 fund, they have Mercore, they have Fireworks. What's the other one? And they have a number that have a bunch of multi-ten baggers at this point in that fund. So they're feeling pretty good about going to LPs and saying, look, we still got it. Yeah. Still cooking. Yeah. And yeah, I mean, Everin Benchmark, newly at Benchmark.

32:49Benchmark has not become the platform fund, mega scale fund, has not 10x the LP base, certainly. Ev says, to be clear, I slash we love working with our friends at all of these funds. And this part was not meant as a slight slash commentary on their quality as investors, just POV on fund strategies and the unique value prop of Benchmark. I think it's a fair tweet, but it's amazing because everyone is coming out and just trashed. Spencer Peterson in the comments there was like, I took that personally. I run a big fund. I'm at Co2. We got$60 billion. What are you saying about me, buddy? You don't think I can put up a 5X?

33:33Watch me. Watch me. I take that personally. I'm going to put up a 5X. I'll call you when I have$500 billion under management. I think part of what made it feel super personal is that Harry tagged Robbie and Ben. Totally, totally, totally. And Ben. Yeah, we actively try to avoid tagging people if someone is talking trash about someone else or sub-tweeting them. We don't really try and hand-hold into drama. But it is an art, and it is delicate. Sometimes it's like, oh, wow, they're really talking trash about that person. Sometimes it's fun. Sometimes we've done this where, I mean, we've done this with Ev, right?

34:12We've had Delian on the show. He's called out Ev, and we've tagged Ev in a clip. But it's all been fun. We've all been chatting about it behind the scenes and then obviously had them all on the show to duke it out and stuff, and that was a lot of fun. This felt personal. I don't know. It doesn't feel that big of a deal. But let's go to Scott Kapoor, obviously speaking, for his former colleagues at Andreessen Horowitz. Now he is the director of the Office of Personnel Management. Scott Kapoor says, Since my former colleagues at A16Z are RIAs and thus cannot legally comment on what they can slash cannot say with a straight face to LPs, this sounds a lot like what crappy board members say.

34:55They think they understand your business in depth and make inane comments about what you should slash should not do when as an outside observer, they have zero clue what actually happens day to day in the business. Buyer beware. What? What? I'm so confused by this. Okay, so he's saying that Ev thinks he understands Andreessen's business, but in fact, he does not understand Andreessen's business. But I want to know, like, what is the misunderstanding? Because the steel man, the bull case on the Andreessen strategy is that Ev is making the claim that they will make more money, dollars, because they're investing out of a bigger fund size.

35:41They don't need to go and say we're going to 5X it, so it's a different pool of LPs. Also, part of the major appeal of investing in A16Z is that you pretty much know that you're going to get in every important company through just one single check. Okay, so yeah, maybe that's something that I didn't articulate fully. I don't know. That's my point of view. I'm not an LP, but I happily would be. And it's because you know you're going to get some exposure to pretty much every important company. Not necessarily always super early, but at some point in the company's life cycle, it's very likely that they will take a meaningful check from A16Z.

36:24Yeah, yeah, yeah. I wonder how the RIA dynamic is playing out at Benchmark. It seems like such an advantage in the vibe wars to be able to speak freely. If Andreessen can't defend themselves because of RIA rules and Ev can just go on podcasts and talk trash, you kind of win by default. You got to get out there, guys. You got to convert. You got to deconvert from the RIA. I don't know. You got to get on the timeline, fight it out. What does Ev say? He says, I think smaller constrained funds can produce higher returns in venture. Quote, this must be a shitty board member. Don't work with him. Incredible non sequitur.

37:10Thanks, Scott. Chad Beyer says, while I actually agree with Scott in general, my lived experience is most board members are useless. I've been on a board with Ev and he was consistently the most prepared slash founder raved about his contributions. And Alex Klein is saying, love you both, because they're beefing. Interesting. Yeah, smaller funds can produce, smaller constrained funds can produce higher returns in venture. That seems to be like a reasonable take, but I don't know what they're actually saying to LPs. Where is, wasn't Emile Michael going, chiming in as well? What did Emile have to say?

37:50I tried to pull that up, but the underlying post got deleted. Oh, it got deleted? Okay. Well, we'll move on from it. We'll tell you about Vanta. Automate compliance, manage risk, improve trust with Vanta. Vanta helps get you compliant fast. And we don't stop there. AI automation powers everything. Meanwhile, over on X, people are saying, calling me unk. They're saying I'm a boomer for my post, which I think is fair. But again, there's more nuance. I don't know. I like there's a lot of it that I think is cool. Should we have the founder on the founder? Open invite to the founder. OK. Open invite to the founder of the brain.

38:33I am actually very interested to hear what is what the actual product is. I feel like the the narrative around around these these like, you know, rage baby stunts is always. like, oh, don't play into it. Don't play into it. I'm happy to play into it. We played into it with Clue Lee. We had Roy Lee on the show three times. And what I said the first time was what I continue to say, which is that, hey, you're going to have to build a real product. Part of my thesis is that Roy effectively ran this strategy as aggressively as you could. There was a picture of him with a stripper. Yeah, that was very rage-baity.

39:17And that was the peak. And then I think he's walked it back, and he's adapted his strategy. And so our take was like, he's going to have to build some good software at a certain point, a good product. And we tried the product, and Tyler churned. And it just wasn't adding a ton of value. It wasn't moving. Yeah, but now they're iterating on the product. And now they're iterating on the product. And so now I would imagine, during the Clearly Hay Day, I would imagine that 80 % of the team's hours were spent on marketing, and 20 % were spent on product, maybe. Now I think it's flipped. And I think that's very bullish.

39:47I think that's good. But Tyler, you had a rebuttal? Yeah, I mean, I would just say like this seems very different from like you see the gambling on your credit card statements or whatever, like from other accelerators. Like this feels very different. Like this is like the marketing. Like I think people kind of tend to group all of this stuff together where it's like what Grant Daniel was talking about yesterday. Some of these are like immoral companies that you could say. I think this feels different than that. It feels much, at least to me, it feels much more of a marketing stunt than like, this is the product we want people to be gambling while writing code.

40:23Yes, but what if the whole product is, and this is all we've seen, is a VS Code fork with minimal autocomplete. They're always a year or two or five behind Cursor and WindSurf. and yes, it has the ability to add Tinder and steak and sports betting in it. And like that, the product that they are telling us they're building is what they're building and they stay with it for five years. Like, what do you say then? Yeah, then that's not good. So basically you're just saying, you're open to a pivot. It's not even a pivot. Like it seems to me like the company is, everything, at least in my opinion, points to this being a marketing stunt.

41:12The company is named Clad Labs. It's like Clad Labs. This feels like a marketing stunt. It tells me that there actually is something different underneath than just adding. Yeah, no, no. I am super optimistic about it. I would love to see, like, I don't know, by the end of the month, we know what is actually going on here. Certainly by Demo Day. When's Demo Day? December 3rd, I believe. December 3rd, we should know what's actually going on with the company, what's the real product. So they did their stunt. They got a bunch of people hopefully to sign up, check out the website. You got your early user base.

41:51Typically in YC, you just ask your other YC batch mates to try it. Maybe you need to go broad, go viral. You did that. Now draw the rest of the owl, basically, is the prompt. Or at least say what you're planning to do. Say what you're planning to do. Let's move on to Dwarkesh. Dwarkesh. I haven't watched the full thing, but a massive interview has hit the timeline. Dwarkesh Patel, Dylan Patel, sat down with Satya Nadella, and they got an exclusive tour of Fairwater 2, the most powerful AI data center in the world. I feel like everyone's been saying they have the most powerful AI data center. It appears that Microsoft has leapfrogged Colossus 2.

42:38Well, I think maybe this is wrong, but I think Colossus 2 will be bigger. It's just currently it's not big. Like this is the current biggest. It's the current biggest. Okay, okay. Okay, great. So you had a chance to actually sit down and watch this whole interview before we started the show. Can you give me some takeaways? Where should people, are there any timestamps that we should pull up? Are there any takeaways that people should know before they go and watch it? Yeah, I wrote a bunch of notes. So maybe some of these aren't interesting. But before you take us through those notes, let me tell you about graphite.dev, code review for the age of AI.

43:10Graphite helps teams on GitHub ship higher quality software faster. GitHub, of course, is a Microsoft product. Yeah. Okay. So I think broadly, I would say it was pretty enlightening. I usually think of Satya as being very non-AGI-pilled. And I think this was a bit of an update. So there's a bunch of reasons for this. I think so. So early on, he, like, in the very, I think this was one of the first questions. He's like, what is AI? AI is basically two things. this Satya is saying this, there's cognitive enhancement. So this is your tools, or this is your autocomplete, this is your co-pilot, stuff like this.

43:45And then there's the guardian angel. And this is the very AGI build where it's kind of lording over everything. And so he actually does say these two things are very possible. And then we kind of move on. He says... Co-pilot and actual pilot. Yeah, basically. Okay, so the next thing he's talking about kind of how he thinks about pricing structures of AI broadly. So there's this kind of conflict between subscription models and usage model of pricing. And he says, I think he's broadly more kind of focused on, at least in the short term, on this subscription kind of thing, right? Which, like, he makes a big emphasis on this, right?

44:30Because you need to be very specific on how you price these things. because doing the actual serving of the models is so expensive. Okay, so then throughout the interview, he keeps emphasizing the point that Microsoft is a hyperscaler. And so that means a bunch of things. That means that they're going to keep supporting multiple models. It means that they want to prioritize the kind of long tail of high-margin users. So you can kind of compare that to Oracle. who you can think of Oracle as basically... Prioritizing one potentially low margin power user, which is OpenAI. You're giving bare metal, essentially, to one customer, OpenAI.

45:15And you're betting the whole company. And he says, like, if you're going to do that, you should just vertically integrate that company, like you're part of that company. That's what Satya says, like, in response to, like, why is Oracle basically eating your entire business? Sure. Like, in the past, like, five years, Microsoft was super ramping up their CapEx, their build and then they just basically stopped and let Google, Oracle, Amazon basically build that up. So that's kind of the main reason. There's a bunch of other stuff he's talking about. They asked him about chips. So all the big hyperscalers have their own chip play, right?

45:47There's training, there's TPUs, OpenAI is doing their own chip. And Microsoft does have their own chip, but it just kind of, it's like the actual production is like way behind everyone else. It's like, okay, why is this chip so bad or why are there so few of them? And then Satya basically brings up that Microsoft has IP to everything OpenAI has except for consumer hardware. And chips are not consumer hardware. Chips are not consumer hardware. So he's like, well, okay, how do we get the best chips, the best model-specific chips? We'll just take OpenAI and we'll just build on top of that. I wonder how much that actually transfers, though, because where does the chip development live?

46:28Where does the IP live? Like if OpenAI goes to— I mean, it's design, and they can get line time at TSMC. Then they can just produce it and sell it to any customer they want. Yes, but what if they do it within NVIDIA or something? Like what if they go to NVIDIA and they're like, hey, for the next run of NVIDIA chips, we'd like you to consider this architecture. Why would they need to do that? Because they want their models to be more performant on the chips. And so they go to Jensen and say, hey, we're one of the biggest buyers. So make the next version extremely performant for our chips. And I think they're already doing this.

47:01They're co-developing chips with NVIDIA. They're co-developing chips with other companies. And if they co-develop with Broadcom or they co-develop with AMD or maybe even Intel in the future, but that IP lives with Intel, then no, Microsoft doesn't just get it, right? Yeah. So it's like only in the event that, I mean, both of these are going to happen. Like there's going to be internal efforts and there's going to be external efforts. But it is a funny reminder that the internal efforts get copy-pasted over to Microsoft. One other note, Satya signaled that he's open to buying capacity from neoclouds, like Oracle, Nebius, Lambda, Iron, Nscale.

47:44Isn't he already doing that? To fill the, yeah, he is. But, but, uh, he, he, uh, semi-analysis has a concept called like the pause, which is like, um, basically a gap of like this, like insane period of demand that you're trying to meet that demand. Um, and Satya actually says specifically, you're right, you're rightfully calling out the pause in the interview. Interesting. Right. Yeah. I think continuing on why he basically stopped building out, the chip question is also very important because you can think of it like if you build a bunch of data centers right now, they have very specific power requirements that are directly based on the chip.

48:24If you're building based off the H100 chip, that's different than if you're building a data center based off GB200s. And it's not like fungible. You can't just trade one out for another. So that's another one of the reasons why you don't want to basically have insane buildout right now because you don't think you know that chips are going to get much better. Sure, especially with ASICs, right? Yeah, exactly. NVIDIA is constantly saying, we've got this next chip coming out. Yep, yep, yep. You want to build up your data center with those chips because otherwise they're going to depreciate and you're going to have these basically data centers in five years that are using the old gen chips or two generations back.

49:00Yeah. Anything on depreciation?

49:06Yeah, I mean, so, yeah, basically, he gives like two reasons how you can justify data centers, right? Because data center basically depreciating in like five years, or the chips, which is a big part of the cost. And so there's basically two ways to like justify the actual build out. One of them is basically you can think of like research as just being like R &D spend. So you basically just have to like, you got to do the research, like you need to do the spend basically. And the other is just like, he keeps bringing this up, everything has to be super demand-driven. So that's also why he's not, it's like, again, in comparison to Oracle, Oracle is basically, maybe you could say that they're kind of skating to where the puck is going, or trying to figure that out, and then doing a bunch of debt, et cetera.

49:51And then Microsoft, Satya is saying, no, where is demand right now? How do we fulfill that demand? Basically, exactly. If we're not fully built up, then we can lease from the neoclouds. Yeah, makes sense. I'm super excited to listen to this full episode. The Wall Street Journal also had an article about the news as well. It seems like this is all in line with Microsoft's just announcement that they have a super factory. I like it. We went from AI factories to AI super factories. We went from regular artificial intelligence to super intelligence. We're ramping up, but we're still like seven levels away from the final boss.

50:34Because after super intelligence, of course, you have giga intelligence, then ultra intelligence, then super duper intelligence, and mega factories, exa factories, terra factories. There's all these different terms. Okay, so I will say one last thing I thought was funny. Extraterrestrial intelligence factories. Yeah, very sure. Okay, one last thing. Dwarkesh asks Satya, like, does he buy basically the revenue growth of, like, when OpenAI or Anthropics says they're going to be, like, 70, 100 billion in, like, three years? He's like, well, you know, they have to justify their fundraise somehow. And then he basically doesn't say much else besides that.

51:11Wow, Satya, absolute dog. I love him. He's the best. Cooking. He's killing it. It's amazing. Chat was asking about hair routines. I think we have to comment and unfortunately disappoint because I think we both ride the, apply the same approach, which is sleep, diet, exercise, baby. Well, and no, no hair products, no hair products, water, and water, some sauna, some workout. So just keep the rest of your body healthy. And I think the hair will be healthy as well. I actually think that's how it works. Yeah. Yeah. Yeah. It's just, it's just funny. We're never, we're never going to have a shampoo or a hair product sponsor because we don't really use either of them.

51:50No. Uh, but, uh, I like creatine and people say that that's bad for your hair, but it's creating the hair loss medication, hair loss medication. People think it, people think you'd lose your hair if you're on creatine, but, uh, I certainly haven't experienced that. I have experienced that Julius helps with maintaining good hair because it's the AI data analyst that works for you. Connect your data. Stop pulling your hair out. plain English. Get insights in seconds. No coding required. Yeah, stop pulling your hair out and get on, Julius. What are the other factors in here? Microsoft spent more than$34 billion on CapEx during its first fiscal quarter and said it would increase its total infrastructure investments over the next fiscal year.

52:33It is among several tech companies pouring a combined 400 billion into AI efforts this year with demand for AI computing and companies saying they need ever more capacity. Very fun. Okay. One last thing. Another thing that I was surprised to hear is that Satya, he made a big point of saying that like there is a super intelligence lab within Microsoft. They're going to be training their own models. Yeah. Isn't that Mustafa Suleiman is running it? Yeah. And so he says maybe they'll be completely trained by Microsoft. Maybe they'll just do like fine tuning and training on uh like gpt open ai models yep um but like they are going to like actually have they're going to be training their own models they're doing kind of the full stack yeah when we were when we were about to go on stage with satya i was uh texting with doug o 'laughlin from semi-analysis and uh and he was like ask about msl or not msl uh mai mai ask about mai like what's the strategy for mai uh and we and we asked a couple people and we got a little bit, it was sort of hard to really pin down a clear strategy.

53:37I don't think that they were ready to really divulge exactly the full strategy. It'll be interesting to see what models they train. Do they just go bigger? They have the biggest factory. Could they just take this Fairwater facility and say, hey, let's go train something that's 10 times bigger than GPT 4.5? Do they believe that pre-training scaling laws hold or not? That's what I'm curious about. Or are they going to do something that's more precise? Like, will they do a pre-train for Excel? Will they do a pre-train for Word or something like that? Is there some other tactic that they're going to employ?

54:13There's a bunch of interesting things. They're clearly very GPU-rich, tons of sharp engineers, tons of interesting product surface area. Where will they actually go? Interesting times. Yeah, I mean, he talks a lot about, like, application layer stuff. He actually says, like, the wrapper model, wrapper companies are basically debunked by models getting better. so I think he thinks that Microsoft will basically take over a lot of that like application layer stuff he talked about Excel agent a lot there's like the PowerPoint stuff. That's why I've always felt like the dynamic between OpenAI and Microsoft is so interesting because OpenAI just has massive ambitions in the enterprise they want to create Sam alluded to an AI native Slack recently, Microsoft has Teams you can imagine uh open ai having like word processing yeah excel like product you know you can imagine them ultimately competing on like every single layer including at the eventually at the cloud layer what do you think the dynamic is uh between this like you know my microsoft always loves to say like hey yeah we have all the ip but it feels like they're not actually fast following like they could Like, OpenAI launched Atlas.

55:25Like, they could have launched, like, Edge Atlas. Microsoft could have launched, like, Edge Atlas, like, the next day. And that would have been like, whoa, what are we doing here? Like, that's a shot. They could have launched Sora 2. They could put Sora 2 in Excel. How about that? The BrainRot Excel. BrainRot Excel. You know, they could have done that. They didn't. And so there's clearly, like, even though they have access to all the IP, They still have a differentiated view on how the products get built. They're not just saying, yo, yes, one copy of Atlas, please. One copy of Sora 2, please. We'll just launch our own competitor app.

56:03They could have put it in LinkedIn. There is a way to integrate that. Microsoft's clearly not moving so fast on that front. They're being a little bit more methodical. Yeah, this is why I'm very excited to see what comes out of MAI. Yes. It seems like they have a lot of good people. Yeah. Like they've been doing a bunch of like talent acquisition stuff. Yes. So I'm curious if they're going to start like like right now it just seems Microsoft is still very slow. They're kind of reactive and they're like, OK, we should add, you know, another AI helper to Excel or something. But it's not like really kind of built in like low level yet.

56:38Yeah. I'd like to have Mustafa on. I'd like to talk to him about where they're training models, whether or not that seems like an interesting conversation. I'd like to ask him some hard questions. There's been a debate on the timeline. John, the X chat says, Edge has every single feature that Atlas has. It's co-pilot mode. It's better than Atlas or Comet. Okay, interesting. Maybe they don't need a copy then. They're good. But there has been a debate over how hard of questions we ask. Is it a hard question to ask Mustafa at MAI, you know, what type of model he's training or should we ask some sort of other hard question?

57:14I wanted to practice some hard questions. I think we should start asking harder questions. I have some hard questions here for you. And I think if we practice, we'll become better interviews. Are you guys ready for these? Should we just start hitting guests with like insane math questions? I wouldn't call them insane. I do have one here for Tyler. Is every even number greater than two, the sum of two primes? Answer the question. It's a simple yes or no question. Is every even number greater than two, the sum of two primes? Answer the question. Stop dodging the question. Just answer the question.

57:54It's a yes or no question. Is every even number greater than two, the sum of two primes? It's an unanswerable question, of course. It is the Goldbach equation or the Goldbach conjecture. There's also, can there exist an algorithm that decides whether any computer program will halt or run forever? It's the halting problem. It's a possible question. What will OpenAI's valuation be in 2035? A hundred trillion. A hundred trillion. Oh, you answer that quickly. We'll come back to that in 10 years. See if that was a hard question. What's the nicest thing ever done by Adolf Hitler? Tyler, answer the question.

58:31Stop dodging the question. Tyler, ask that to Karp. Karp was on a roll yesterday. Some of those clips were... Here's a question we should ask a lot of founders that come on here. Does your dad know you're bankrupt? I think if we hit them with that, people will be like, that's a hard question. That's a hard question. Here's one for Tyler. Prove God exists.

59:00What is truth? That's a hard question. Have you ever done illegal drugs, Tyler? Answer the question. Answer the question. Good answer. Who's your favorite host, me or Jordy? Answer the question. Answer the question. I'm my favorite host. Oh, taking shots. What is time? I like these hard questions. Okay, Jordan at Semi Analysis is down to hop on and correct a few things we said on the article. I'm going to drop him the, he was listening. He was listening, said Al in the chat. Where is he? Let's bring him on. I'm working on getting him the Zoom link. Okay, fantastic. Fantastic. In the meantime, let me tell you about Fall, the generative media platform for developers.

59:51The world's best generative image, video, and audio models all in one place. Develop and fine-tune models with serverless GPUs and on-demand clusters. You can get started. Let's see. What else? Henry Kravis gave a talk, the founder of KKR. The post appears to have been deleted, but I won't say who it's from, but I will read it to you anyway. The world needs another fund like it needs a hole in the head. I think it was supposed to be off the record. And I think one of our friends was maybe live tweeting it against their request. Or maybe he just corrected it and realized that he didn't want to post that.

1:00:29But there are a lot of funds, I suppose. I was doing a little deep dive on Blue Owl. I want to get to know Blue Owl more. It's a fascinating company. Merger, there's a SPAC involved. They have two co-CEOs and three co-presidents. The top ranks at Blue Owl are absolutely stacked from what I read. They are, of course, powering the AI build out with debt, private credit. They are a private credit fund up there with Aries. They've done very well. They actually have three different businesses. Only one of them is doing AI data center build outs. And even within the AI data center build out fund, they fund other stuff.

1:01:07So it's interesting to see how much risk they're taking, how much their business is really dependent on the AI buildout. It's also, there was a, they got in a little battle with Jamie Dimon, because Jamie Dimon was saying that some of these bank failures are the fault of private credit. When you see one cockroach, there's usually more. He's kind of saying like, hey, like, you know, we see these defaults. We see these large companies defaulting. maybe this is the start of something bad. Maybe there's a lot of other bad companies out there. Yeah, there was something with BlackRock had lent, I think,$100 million to a business recently.

1:01:46As of like two weeks ago, they had it marked as fully sound. And then the company almost immediately went bankrupt. So concerning. We have Jordan from Semi-Infacific. Welcome to the show. How are you doing? Hey, Jordan. Good to see you again. Welcome back. All right. What did we get wrong? What's going on? This is all, of course, Tyler's fault. He was... Break it down. Give us some notes. Yeah. No, I don't think you guys got anything wrong, but I think I just wanted to promote that we put out a companion article that comes out with the interview. Oh, fantastic. Was in there. He's being called out by some other people on Twitter for hunting around in the background while they're interviewing the Azure teams.

1:02:34wait a dylan patel you mean yeah oh we we got to go through frame by frame and find him like oh he's gonna go look because he's wait we actually break down what what does a dylan patel scoop look like when you're on site and security turns their back is he like looking for like oh they're instead of this nvidia switch they're using amds i've seen him call stuff out where it's like wait a minute this company is using a different company's technology is that what he's looking for yeah i mean i don't know if he was looking for the credo cables or anything those So those are obvious with the purple housing.

1:03:06But if you look at his last tweet, he just said, to be clear, while Scott was explaining the data center, he was super intently focused on figuring out the fiber patch panel config. We know about some providers that have gone out and bought cables, like 20 ,000, 40 ,000 cables. And then they mess up the config of the patch panel and they end up finding it too short. And they just have to send them back and get all new cables. Not a negligible amount of money. These, these cables are like at least a thousand bucks a piece. Yeah. They're really heavy. We've been talking about internally, how many cables can you bench?

1:03:45Okay. There we go. There we go. Yeah. Dylan Patel seems like a bit of the Fox who in Microsoft is the hen house and they really shouldn't have let Dylan Patel into the, into the new data center. Cause you know, he's going to find some stuff. There's going to be some scoops that like, I'm sure there's a PR team at Microsoft that's like, okay, here are the talking points. We're going to keep him here. He's not going to look behind this curtain. There's a bodybuilder over there. Don't go in that closet. And Dylan's, I'm sure, on the case, probably quickly becoming the greatest investigative journalist of our generation.

1:04:20Give us some more insight on the depreciation comments. Yeah, definitely. So there's a section of the article that goes through this. I think you guys asked me this on Monday. I didn't have a properly prepared answer because I actually didn't know what Michael Burry had been tweeting about and kind of this. about with this so maybe for background uh burry is claiming that the hyperscalers like including meta but also um azure yeah on oracle google are artificially boosting their earnings by extending the useful life of the it assets so you can see from 2020 when they would report their numbers these it assets like servers switches storage would be three to five years yeah on a life cycle And they've now extended that to five, five and a half or six all the way across the board.

1:05:14In some ways, I think this is a bit of a, you know, game of catch up. If one provider does it, everybody else has to follow suit. But we go into detail in the article about like what this actually means. And, you know, Burry's argument that this is understating the amount of or boosting their earnings. it's understating the amount of depreciation that's going on on these GPUs, is really predicated on NVIDIA's comments that the product cycle is now two to three years long. Right? Sure, sure, sure. And I think that is like... But that's on the development of a new chip. Like, from my perspective, depreciation, although there is the economic question about how much value can you get out of the tokens that you sell generated by an A100 in five years, that's a good question.

1:06:03But that's not how we think about depreciation. Like if you buy a mechanical arm to put a glass windshield on a car at an automotive factory, you're just wondering how long until that mechanical arm breaks. And I'm sure there's been immense pressure on NVIDIA to let the chips not burn out in two years. Like, you know, they probably have done a lot of work. It's a very expensive chip. It's a very expensive rack. Like, is it that crazy to assume that the 50th percentile might be five years now? No. No, and there's basically no precedent to say that a chip would fail or wear out as you're describing it in two to three years.

1:06:42Like there's the hardware OEMs, they have contracts that are standard for three to five years and they offer extended warranties for six and seven years. The big supercomputers in the world that run as a total system or even individual servers kind of get life cycled, they run for five, six, seven, some of them up to 10 years in production. Right. Not to say the years that it takes to actually turn this thing on. And these are the environments that use liquid cooling and some of the latest and greatest chips that are actually similar in comparison to the GB300. And then even if you go to the providers directly and you try to rent a V100 today, which was launched in 2017, right, seven, eight years ago, I can still rent V100s in data centers from providers like Amazon.

1:07:27I mean, there's plenty of A100s for sale right now. So there's nothing – to me, the proof of this argument would be predicated on NVIDIA releasing chips that so drastically outperformed the current generation in two to three years that all hyperscalers everywhere are so incentivized to go through another CapEx cycle. They've got to buy all new chips and rip out all the existing ones, and we're still so power constrained that they have to do all that. And that seems like a much farther leap than saying we might be able to run these chips for five or six years in the data centers themselves. I wonder, do you have a reaction?

1:08:08So Ben Thompson's been writing about the AI build out and the bubble, potentially the benefits of a bubble. And one of his like bull cases for a bubble was basically that in previous bubbles, you get a glut of IT infrastructure or even just steam engines or railroads. But his example was dark fiber. And the idea was if you overbuild, well, then you get a bunch of extra infrastructure that you can use and it's cheap. And his counter to the AI build out was that depreciated H100s just aren't as valuable as fully depreciated fiber lines. But I just don't know if that's true. I feel like if there's a ton of depreciated H100s out there in 10 years from now, I think there's still something useful that you can do with that in the same way that you can push bits across fiber lines.

1:09:03It just, it won't be super intelligence, but there will be a base load of, you know, generic knowledge retrieval or just, hey, people just want to chat. And the fact that it's so cheap now because the assets are fully depreciated means that you can just have a chatbot in interface every single conversation in every app. And it'll just be so much, it's sort of like a Jevons paradox thing. I'm not saying that the economics will stay the same. The margins will go way down, but there will be a benefit. it's not like these things just disappear after five years. They don't just break. Yeah, no, they don't just break.

1:09:38I think your comparison makes sense, but there are some nuances compared to the fiber lines. Just to say that if new GPUs come along from other providers that are so much more performant than the current ones, then at some level, it doesn't make sense to keep the power turned on for the old one. It's just the operating expense is too much. like we estimate this at 30 cents per kilowatt so if the bottom of the h100 price gets below 30 cents or somewhere close to it you're probably better ripping them out and replacing them with something new even if you know for the car analogy this is still a beater that somebody else could drive for the price of gas and insurance right yep yeah yeah i was talking to jordy about it i was saying like there is a world where to just put in really concrete terms that most business people might understand just, you know, if your job is, you know, you have a Diet Coke business, you deliver Diet Cokes, you have a gas car that delivers the Coca-Cola, and you have the opportunity to switch to an electric vehicle, you might do that because bringing down your total cost of ownership, your annual OPEX would be great.

1:10:47But sometimes companies just don't want to spend the CapEx to actually do that. And so I'm wondering if there's a world where the GPUs get traded down to a point where people are like, yeah, the OPEX is higher, but I'm still running it for this niche use case. There are still mainframes that are running. There are still on-premise cloud. There's companies that haven't fully moved to the cloud. And I'm just wondering how long some of these GPU racks might just be sitting around where someone gets it cheap and they're just like, yeah, that's the thing that filters every invoice that we get. and it just runs and it runs like it runs grok 2 on it or whatever and it's fine it's good enough yeah well i mean look i put in the article a link to azure's announcement from september kind of pleading with their users to move workloads off of v100 gpus that are eight years old so

1:11:44please don't get upset at us when we turn off this instance and you you know i don't think they're doing payroll on these things but like things gonna shut off so interesting i i mean i totally take your point working in it for years there's there's so many people that come along and when there's like some sort of last call for the sale of spare parts for these old gpus and they want to keep these systems running they they suddenly show up and want to order a bunch of them yeah and it could be for all sorts of reasons but usually it's just inertia of not wanting to move the old workloads onto the new stuff.

1:12:18We're also completely discounting the nostalgia market. I mean, people like air-cooled Porsches. I imagine that Dylan and you guys are going to be like, yeah, I need a V100 just sitting in the office. I don't like this new technology, this new GV100. You want to talk nostalgia, we've got to go back to Kepler or Pascal, way before the Volta guys. I don't know. Yeah, yeah, yeah. Way before the Volta. What about on the chip side? What was most notable about Satya's comments there? Yeah, I mean, I find it fascinating that Satya clarified they have access to all IP, including chips and including systems.

1:12:56So this is on the models that everybody jumps to. I think OpenAI has a ton of IP that's going into their chip program that Azure could effectively take and sell as part of the cloud services. if they want to partner up with Broadcom and just produce a bunch of chips that are, you know, more similar to a TPU than they are to a GPU in the future. Yeah. They've got that optionality. I wonder if there's a way to puzzle around that, though. Like when I was debating with Jordi, I was saying like, well, you know, obviously OpenAI has a team that works with NVIDIA and with Broadcom and with, you know, maybe Intel in the future, AMD now.

1:13:33is there a world where the IP lives with the other companies for the next five years or something? It seems like maybe tricky, but I'm just wondering, like, how much of the IP actually lives within OpenAI versus their partners? Well, I think it's, so Sacha made the point in the podcast that in some cases, Microsoft seeded parts of the OpenAI chip program with IP from their Maya program. Oh, sure. I think it's this whole mosaic of where the actual IP sits. But at the end of the day, there's going to be something that comes with open AI that they learn. And you can even go beyond the chips to the systems that the chips go into, to the network that connects the chips, or to the software, like the runtime and the inference stack, which is to say how they run inference efficiently on these chips.

1:14:26there's a whole stack that gets built from the actual hardware through to the API that produces tokens to the user in the chatbot and they have access to all of that that open AI would consider proprietary so I maybe to the point you guys were making earlier comparing to Excel or to Microsoft's other strategies it it seems like this is a case of open AI going really fast and executing really quickly and their moat being speed and you compare that to microsoft who has an existing moat of distribution so they've got 400 data centers in azure 70 regions free cash flow free cash flow is also a bit of a moat too yeah definitely um they've got all the makings for like you know who do you who do you think wins the race to turning on public instances of an open chip that anybody can rent, right?

1:15:25OpenAI or Microsoft Azure with a bit of a head start. Like, I think they have some good optionality there with OpenAI's program, with the relationship with NVIDIA, with Maya. But the point is that right now, they are losing the race on chips to NVIDIA, the CPU from Google, maybe AMD. They use some AMD GPUs right now. So they've got to go and execute right um whether or not do you think satya expects azure to compete with an open ai cloud in the next few years i think if open ai really develops a cloud absolutely because because satcha says in that article that they um well in in the interview that um they don't want to be beholden to one customer right i think he says like we have five big deals with five customers And that's like the bulk of Azure's compute right now.

1:16:21And I would say it's really one deal with one customer. That is the bulk of the incremental revenue in Azure right now. So, I mean, I go hands on testing this stuff and I've talked to 140 different companies about actually using Azure and they are a distant third compared to just AWS and Google from a hyperscaler perspective. And I talked to Clem from Hugging Face this morning, and he shared some data with us that backs this up. Like if you look at the downloads from Hugging Face of open source models, the downloads that originate from an Azure IP, there's more than five times less when compared to AWS.

1:17:01Wow. Right? OpenAI can use it all with their private repos and stuff. But if you look at the long tail of the market, where the long tail represents everybody but OpenAI, basically, they're getting five times less of that business right now because they're so focused on OpenAI. So I think Satya wants to do more than just OpenAI. It's just a matter of actually going and doing it. Yeah. Do we have more of a clear narrative now on what motivated the pause? like you can see it so clearly in this data center pre-leased capacity in the semi-analysis article it's so obvious what's happening from 2023 you see the blue bars just increasing increasing and then just complete flatlining everyone else is growing um it seems like this it seems like this this uh this piece this uh video with dorkesh and dylan feels a little bit like him maybe teasing like, hey, I'm getting back in the race.

1:17:57Like the headline with the Wall Street Journal is like, I got the biggest data center. That feels like I'm unpausing. But, right, you're not like, oh, I'm happy to be a leaser. And oh, by the way, I have the biggest thing. Actually, I own the biggest thing. Like that's kind of an odd dichotomy there. And I'm wondering if there's a clear narrative, was it specifically like the nature of the open AI relationship that led to the pause? or was it just concern about the overall market or the viability of the technology? Or was he like plateau pilled and had seen that GPT 4.5 wasn't getting adoption too expensive to serve?

1:18:35Like, like, do we have a narrative for like what motivated the pause yet? I don't know if it's clear. I think we have different pieces that we can piece together. the things that Satcha says in the interview point towards fungibility, diversification outside of open AI, a lack of foresight in two years ago to what the actual demand would be. They didn't feel like fishtail where they went too far on building demand and then they wanted to come back and now they didn't have enough and so now they're renting from neoclouds like Nabius, like Lambda, iREN, Oracle directly. They've got N-scale deals.

1:19:16I mean, they're clearly using neoclouds to take on that extra bit of risk. And going forward, I mean, we are, I think the result is that it's incredibly bullish for Microsoft actually serving this demand when they have the first right of refusal and OpenAI's growth continues to go crazy. And they are kind of back on track to leading the way with coding models in a way that they were, you know, maybe forecasted not to with Anthropic coming and taking a bunch of share there. They are all in on taking down Anthropic as a direct competitor with their codex models now. I think Microsoft just flinched a bit and they're back on track.

1:19:57We're pretty bullish on them growing in the future, but it's notable that Satya talks about things like diversification around open AI and around globally. The idea to me that there's a geopolitical conclusion that you can draw from his response on the question about the pause where he says, what if I want to build in India? What do I want to build in Europe? What if I want to build elsewhere? And therefore, we pause North Carolina. That seems to say that governments around the world can attract investment in data centers if they implement a bunch of harsh regulations on data privacy. And more and more companies are going to want to train models based on personally identifiable information or other types of data that's regulated and needs to stay in those countries.

1:20:44And therefore you need GPUs around the world. You know, that could be something that they just kind of learned and therefore pause North Carolina, spin up something else. I haven't seen that exactly play out on the global deployments, but it could be coming. How real is this fungibility of the fleet thing? Because the narrative around Google is that they have a TPU. They're so vertically integrated. They deep mind. It just feels like when you go to Gemini, you get something that's down to the metal. And then at Microsoft, it's like, pick your model. And then also they're subleasing from all these other clouds.

1:21:23But as I see from ClusterMax, there is a wide variety of performance metrics and results and qualitative, even the feel, of these different NeoClouds. And so is there some sort of risk to, you go to Azure and they put you on the lowest ranked ClusterMax NeoCloud and it's like, I would rather be up here on the top tier lease. I wanna be subcontracted down to the platinum tier, not the D tier. Is there any risk to that? Or is Microsoft actually able to take a NeoCloud that is in the D tier and give you the Azure level of service? Yeah, I think it really depends on at what level you're assessing the fungibility.

1:22:08So I think in some ways the tokens or the tokens produced by the model endpoints are almost completely fungible. Okay. You know, you can find certain the models in private scenarios, but generally speaking, like tokens from one provider versus tokens from another provider are effectively the exact same. As long as you're, you know, hitting the like quality metric you need from that model deployment on that GPU. Yeah, yeah. But if you look at the underlying architecture, like when Azure claims to add 100 ,000 GB300s coming online this quarter, I mean, that is not fungible to every single user on a cloud service.

1:22:42Like GB300s are rented 72 GPUs at a time and 135 kilowatt rack. Your workload needs to be ported to ARM because it's a great CPU. You need to have the performance per dollar benefits of the Blackwell GPU to justify it over the previous generation hopper. it needs to use a GPU in the first place. So I just spent a bunch of time in the article talking about the importance of, you know, in the interview, talking about the importance of databases of like, you know, non-GPU related services that actually run the web app, that actually store the data, right? Turbo puff. There's lots of Azure capacity beyond GPUs coming online too, right?

1:23:23Yeah. You know, I might call them CPU data centers now and they consume a little bit less. Yeah. I wanted your take on a couple other things. One, the new Anthropic News,$50 billion investment. Any reactions to that? Yeah, I think notable that they called out FluidStack by name and they're not calling out the underlying providers, as you said. If somebody like FluidStack can help them deliver on a bunch of what we think is TPUs deployed directly first party, you know, then there's, I mean, there's all sorts of stuff that they can do because under the hood, it's pretty clear that FluidStack is deploying these TPUs in TerraWolf data centers in Buffalo, at their Lake Mariner facility, and elsewhere in Texas, and then at Cypher Mining in Texas.

1:24:17If you look at some of the, you know, you got to pair two new press releases together, the one from the underlying provider and FluidStack, and then the one from Anthropic with FluidStack. But the point is that Anthropic is definitely ramping up just at the same level as OpenAI. Maybe not quite at the same level, but they both seem to be clearly believing that statement from Greg Brockman. Like if we had 10x more compute right now, we'd have 10x more revenue. And so therefore, their constraint is compute. Bring it online and just keep growing. That was certainly the case with the early iPhone launch.

1:24:56Like the iPhone, like Apple's earnings were extremely predictable every quarter because they were like, as many as we make, we will sell. And so they were completely supply constrained. It does feel like on the question of backlogs, Anthropic has just been much less aggressive. Like they have been somewhat saying the biggest number every once in a while, but they haven't gotten into the trillions of backlog. and it feels like open AI, Sam's almost like, oh yeah, sure, give me all the weight of Stargate, even though that's sort of a separate entity. And he could have easily fended that off and been like, well, that's Stargate, that's sort of a separate thing, that's not all on open AI, we need to clarify how this all fits together.

1:25:38But Anthropic's been a little bit quieter on the like, okay, we have this crazy RPO that's going on all over the tech industry. What was your reaction to CoreWeave's quarter, uh jim kramer was going pretty hard what did what did kramer say he he was just like it was blackmailing it was yeah it was just a funny interview he's like what's going on what's going on because they doubled they doubled revenue but they sold off is that yeah well i mean they sold us a ton in the last month yeah i think like 20 about a yeah 26 something like that but But Kramer was saying, like, why are you relying on a Bitcoin miner to, like, help fill your capacity?

1:26:18Are you sure they can deliver? And he was talking about Core Scientific, which CoreWeave attempted to acquire. It got rejected. But I know it was yesterday. Platinum on ClusterMax. That's all I need to know. Buy and hold. Get lost, Kramer. Yeah. And the CEO was saying like, yeah, and like he didn't name semi-analysis because I think the CNBC audience is maybe not familiar yet. But he was saying like, we're platinum rated. We're platinum rated. Oh, we said platinum rated? Oh, that's amazing. I love it. Yeah, I mean, it's in the earnings call. Semi-analysis is mentioned, I think, three times. But definitely, yeah, we went into this.

1:27:00We're very excited for this. This is big. Yeah, I think we went into this, Jeremy and the data center guys, Rake, Dan, they put out a note earlier in the day to the core research subscribers that the core is at risk of short-term delays in their CapEx, but not to their revenue, which is driven by the data center partner, Core Scientific. So, yeah, when you're relying on somebody to add 250 megawatts and, you know, they're only going to get you 150 by the end of the year, that changes your guidance on CapEx. And I think the, you know, this is like, yeah, look, this is the reality of the industry right now, which is that when projects are so big, measured in the hundreds of megawatts, small delays or small changes in a plan can really impact like short term financial guidance that people have in place.

1:27:55um i don't like personally i don't think this changes any of the fundamentals of like coreweaves engineering or a lot of the like experience and what customers they have signed but we saw this like a very similar reaction when when the information put an article about oracle having a delays right yeah and you know i think people are people who don't know about the data center industry or or things like this they they see things like oh yeah these these gpus take a months to come online after they get installed and they're starting to depreciate the asset, but they're not generating revenue yet.

1:28:29Like what's going on? It's like, yeah, that's, that's typical. I mean, it's completely standard. That's business. Yeah. It's stuff takes, I mean, you know, maybe us federal government is, is different, but in my previous job, we tried to hand over these supercomputers, the us federal government, it'd take like a year and a half to pass acceptance. But at the same time, if, if, if investors aren't aware of that, and then they learn that for the first time, they could be surprised. And so that could be something of what we're seeing in the gyrations in the public markets, I suppose. It makes sense.

1:28:58And there was a, I think there was a broad recognition that the Bitcoin miners have an uphill battle to figure out how to run these facilities when compared to established players like an Equinix or a digital realty or, you know, switch or something, somebody that like has been running tier three data centers for a long time. It's just different. Give us an update on the energy model that's coming. What can you tell us? Is it this year? Next year? Is it my Christmas present? How does this work? There's some previews. I think you should have a Jay on the program next time or bring Jeremy back. Yeah, I would love to have both of you.

1:29:35Talk about it. Yeah, I'm specifically, I'm very, very interested in the just American energy forecast for the next couple of years. because it feels like the entire industry is sort of hinging on like the trend changing, in my opinion. It feels like we need to bring up the level of energy production for everything else to happen. And whether we hit some sort of regulatory block or some massive scale block or a capital block or a glut of some sort or some gyration correction, there's so many things that could go wrong, but I'd love to know what the forecast actually is from the folks who have gone so much deeper.

1:30:14I'll give you the quick hit. So everybody's trying to do turbines behind the meter right now, NatGas, right? And you've got a number of different players, but if you look at Schneider, Vertiv, Bloom, one of them, Bloom, is ramping up. First of all, they're all sold out. So the model will cover like how sold out and exactly how sold out. But yeah, you want to look at Bloom ramping up production and you want to get Schneider and Vertiv on here and try and figure out exactly why they're not. Let's see, you know, what does it take to get those guys to recognize what everybody else in the industry is recognizing right now?

1:30:51And then let's dig into solar, wind, nuclear, hydro, all these alternatives that, you know, you'd like to have in there, right? Why is Elon putting power plants on the other side of the border and then piping the electricity back to his data center in Memphis because of short-term regulatory approvals. There's all of that political stuff that goes on to influence this market. I'm so excited for it. Thank you for coming on again twice in a week. We'll see what happens the rest of the week. We really appreciate you taking the time. Maybe go for a three-peat. If you're not already subscribed, what are you doing?

1:31:29Head over to semianalysis.com. Buy the most expensive subscription you can find, especially if you're a venture capitalist. You can't call yourself a VC in the age of AI if you're not paying semi-analysis the big bucks. At least a couple hundred thousand dollars per month. Per month. I demand it. Anyways, great to see you, Jordan. Thanks for coming on. We'll see you soon. Bye. You mentioned Turbo Puffer. I'll tell you about Turbo Puffer. Search everybody. Serverless vector and full-text search built from first principles and object storage fast. 10X cheaper and extremely scalable. The soundboard is doing very well.

1:32:03Someone leaked their revenue recently. Turbo Puffer. It's quite shocking. Well, we have a shocking guest. Brian Halligan from HubSpot joining us. He's the co-founder of HubSpot. Welcome to the stream, Brian. How are you doing? Sorry to keep you waiting. Welcome to the show. Never better. How are you guys doing? We're doing fantastic. What is the latest in your world? Have you been captured by the AI build out? Is this something that's keeping you up at night? Or are you watching it more as an observer? Or are you checked out entirely? Where are you on the, how much does it get your pulse up? Because me and Jordy, it's every day for us over here.

1:32:39Where are the trenches? I'm very checked in. I come at it from another angle. What I've been doing for the last year is I run an in-house CEO practice at Sequoia. I do three things. I coach CEOs, most of which are building companies on top of the AI wave, all of which. I create communities of CEOs so they can collaborate together. I have a kids' table of CEOs of kind of small company and an adults' table of CEOs in large companies. And then I'm creating content. So I've got a podcast. Launches tomorrow. Oh, let's go. Yes. About CEOs and how to be a CEO. I love it. How do you go for a startup to scale up?

1:33:24Tune in tomorrow. Fantastic. What's the format? What's the format? Is it guest-driven? been? Is it turning kind of the lessons that you talk about at the kids' table and at the big kids' table? It's a six-hour daily live stream, actually. It starts at 9 a.m. and ends at 3 p.m. Makes us look lazy. The basic thesis is the playbook I grew up running HubSpot from 2006 to 2021. A lot of those plays don't seem to work as well. Sure. And the rubrics, we thought, like, for example, I think what's how the way Justin Wong runs this company is super interesting with 60 direct reports and no one on ones and kind of public criticism.

1:34:09The way Elon runs his company, his empire is really interesting. The way Brian Chesky does it is quite interesting. And I notice all these CEOs have kind of new place and they've got a new perspective on things. And so I'm inviting them on kind of one by one to chat about how did you become a CEO? How do you like your job? What are the best practices? What is all wrong about CEOing these days? And I think it's my first guest is Parker from Rippling Tomorrow. Oh, yeah. Nice. And it's a good one because Parker gives all the tea on getting fired by David Sachs from Zenefits, all the tea on the drama with Deal on Spygate.

1:34:48And then in between all that, he is a very, very thoughtful CEO. He's not a big fan of conventional wisdom, as people know. But he's doing some really cool stuff. So that's kind of episode number one tomorrow. Yeah, I like that framing as like focusing on the actual job of the CEO. Because there's other shows that could do like, okay, let's do the product review. Let's do the product announcement. meant, but having one feed where you can go and actually understand the philosophy of a whole host of CEOs. That's fascinating. I think it's kind of interesting because I was a CEO for so long. So it's like, I'm really a journalist.

1:35:26It's like a conversation about what the hell's really going on here. How did your approach evolve over the 15 or so years that you were CEO of HubSpot? I don't think it evolved that much is the interesting part. You just one-shotted. I was basically the same guy from 2006 to 2021. And I followed like ye olde playbook of the way people run companies, like the Bill Campbell style playbook. And that's what I think is interesting. It's kind of changing, and that's what I'm trying to dig into and kind of go deep on. It's like Stanley Tucci. You know, Stanley Tucci's show when he goes to Italy and he's trying to find his roots.

1:36:05I'm searching. I'm searching. What are your thoughts on... You want to know some surprises? Yeah, please. Okay, a couple things. There's a breed of CEO that I see a lot of these days that I call a five-tool CEO. Now, I'm a big baseball guy. And there's such a thing in baseball as a five-tool player, somebody who can hit, somebody who can hit for power, somebody who can catch, somebody who can throw, and somebody who can run at an elite level. very, very rare in baseball. Who's an example of a modern five-tool player? Shohei Otani. A six-tool player is Shohei Otani. Okay. But he can actually run.

1:36:52He can hit. He can hit for power. He can throw. He can catch. He's a very talented and very unusual cat. Alex Pregman is one of those. I'm a red-sucks guy. But there's a bunch of these folks who kind of remind me of that in the software industry where they've got vision, they can code, they can design, they can recruit, and they can sell their product. And that's, I think, pretty rare in software. They're all over the place now. And they're getting funded, and I work with a bunch of them. Just a few examples, like Parker's one, for sure. Brett Taylor, for sure, from Sierra's one. Gabe Stengel from Rogo, kind of a new company's ripping, is one.

1:37:33uh mati from 11 labs i have no idea how to say his last name kareem from clay dara from delphi like on and on and on i'm super impressed with the new breed that's coming out and how talented they are who do you think the barry bonds of business is or the mark mcguire of business fantastic performance as a ceo clearly using performance enhancing drugs if someone if someone tapped me on the shoulder and said you know i don't know if you heard this but elon's actually an alien oh yeah i mean that that puts you know you gotta be surprised if someone said it turns out he was an alien would you guys be like oh no no no that's that's crazy you'd be like yeah maybe it is it is funny the performance sensing drugs thing because uh like me and geordie have joked about how it's not illegal to be on a stimulant i obviously partake in caffeine, the legal stimulants, but some folks go a little bit further.

1:38:33But the reason that it was so controversial and so illegal in baseball is because isn't baseball a sanctioned monopoly by the government? I believe that's why. And so that's why during the whole baseball scandal, the steroid scandal, they came in and they had to do like a hearing on Capitol Hill. It was a crazy, crazy moment. I don't know if you tracked that. I think the real reason is because baseball been around for 150 years people who broke records in the 1920s like babe ruth broke important records in the 1920s but you can have like a ranking of who's actually done the most home runs best batting you can't do that in basketball because they change the rules so much you can't do that football they change the rules so much so that's why they were so resistant to change the rules in baseball and that's why they were so up in arms about the steroid thing but i do think founders do the equivalent like when i was growing up i'm a beer drinker you know

1:39:28You go to a Sequoia founder dinner? Yeah. First of all, it's at 5.30, like early. And everyone has a salad and it has water. And it's back in the room coding by 7.30. That's definitely not how I grew up. Do you think beer gave you an edge back in the day? I don't think beer gives anybody an edge. I think it helped me connect with some employees along the way, for sure. particularly salespeople. What do you think about co-CEOs? Yay or nay? Okay. I think Sand Hill Road is very negative on this. Except on stewards. We have co-stewardship. Yes. We can talk about that too. But I talked about this on one of the pod episodes with Vlad Tenev.

1:40:16He was a co-CEO for like the first seven or eight years of the company. And I was very surprised to hear that. And he was very positive on it. I think you can do it. I think it's a little dangerous. You better know the person very well and be able to finish your sentence. It's like my co-founder, Dharmesh and I, we weren't co-CEOs, but we ran it like a partnership. And we thought of it like a partnership. And the employees thought of it like a partnership. If we called ourselves co-CEOs, that would have actually worked quite fine. That makes sense. Want some other things that surprised me? Yeah.

1:40:49Here's the weird thing. two of our most successful entrepreneurs of our generation are jensen huang nilon musk and they've obviously got really unique play uh playbooks like uh jensen's got his 60 direct reports and the no one wants all this stuff we just talked about elon has his has his algorithm and like credits the algorithm and a lot of his success i ask all of the ceos on the pod like who's on your mount rushmore who do you follow like who's your inspiration for me it was Steve Jobs, like my generation is Steve Jobs. None of them really say, you know, Jensen, like no one's copying their playbook.

1:41:29None of these CEOs have 60 direct reports. And not a single person has mentioned the algorithm. I think that's quite odd. Do you? Yeah, we actually polled YC founders in the last demo day. We asked probably 20 or so who was their favorite entrepreneur, like who they look up to. And I was expecting like half to say Elon and maybe, you know, a third to say Jensen, something like that. Just like vanilla market cap. It's a question. We didn't prep them. We didn't tell them we were going to ask them this. This is off the top of their head. And there were maybe like two or three Elons, but it was all over the place.

1:42:08And it was way down the stack. I mean, there were some people who were like, the founder that I look up to most is like a founder who started their company five years ago and is a unicorn now, but like certainly not like one of the greats. Like obviously doing fantastically, but like, you know, this is like a series B or series C level founder. And they're like, that's who I actually look up to. I was shocked. I don't even think a single person said Jeff Bezos. I don't think anyone said Bezos. A guy with 120 billion of 2024. Free cash flow. Yeah, it was crazy. It was much less concentrated in the Mag 7 than I thought it would be.

1:42:45And I don't know if that's a shift in how younger entrepreneurs or newer entrepreneurs are thinking about what success looks like. Like the, maybe the downstream critique of like the, the Teal monopoly thesis is that like, it's actually really hard to go start a business that is in the social networking category or in the online commerce category. It's just harder to do that. And so maybe people, some people read zero to one and they said, Hey, I got to start a company, but then other people read zero to one and says like, Hey, I better not, I better not set my sights too high because realistically, how am I going to compete in, you know, some of these monopolistic categories that have been truly dominated by the hyperscalers.

1:43:31It's going to be hard for me to go zero to one in those categories. And so I need to maybe find a smaller market. I don't know. Maybe it's a trend. One of the things that surprised me about them is they don't really have anyone that they're really following or looking up to. They're very much their own people. And they're kind of making it up as they go. And they're writing the rules kind of based on what they think makes sense. They don't have, like when I grew up, I had two people, I have two CEOs, Steve Jobs, Jerry Garcia. Those were my two heroes that I tried to base my behavior around. and they were my cortisones.

1:44:10These guys don't have cortisones. The other thing that's interesting about all of them is like, be yourself, everyone else is taken. They're all very, very, very different homo sapiens. Like Vlad Tenev is very, very, very different from Parker at Ripley. Anton from Lovable is very different than Mati from 11 Labs. They are all quite unique, which I like to see that. And they're their own people. they have a lot of agency. Are you interviewing anybody that runs their company remotely? Nobody's doing remote. Everybody is in the office. Remote is dead. Everybody talks about 996. I think there's a lot more talk about 996 than walk on 996.

1:44:57I don't think very many people are actually doing it. The CEOs are though. Every CEO I talk to is under, like I say, what's your stress level? 1 to 10. almost all of them say 10. Yeah. What do you think about this idea that a lot of companies are 996ing in a hybrid sense? That because the tools have gotten so good, I mean, a software engineer can fire off a prompt at home on their phone that will go and start doing work that gets reviewed in the morning. I'm firing off a deep research prompt for something that we are going to talk about on the show yes tomorrow and so there's this like ambient level of remote work that's happening on top of the in-person work that's happening and so i might if i were to push back it might be that uh remote is not dead we said yes and we said let's go back to the workplace let's work in the office and then let's also work remotely at home on saturdays on the nights and weekends uh because that's what it takes to win and so the 996 thing is happening but across to spaces.

1:46:07I totally buy that it's so much more because it's asynchronous now. We went remote at HubSpot pre-COVID. The culture's great, it's still great, but you lose something. You definitely lose something. I'm very supportive of the enough thing. If I were starting a company today, I wouldn't do remote. Your headquarters in San Francisco. San Francisco's a fair of a place to scale a company. Salespeople are hard to find and really expensive and not loyal, developers are even harder. And so I do a big hub in San Francisco and then another big hub, let's say in Toronto, where there's a lot of salespeople and a lot of developers and be like, we're in office and at least four days a week.

1:46:45I think that's a good call. Ironically, the people who I think that CEOs who I think are the hardest core are the two European founders I interviewed, Anton from Lovell and Mati from 11 Labs. Like they are seven days a week. They expect everybody in the office for seven days a week. Those two are the hardest core, even more so than our American brethren. Well, the pressure on them is immense because people have such, so low of expectations for Europe. So, you know, they have to deliver for their country. It's really, it's really high. It's kind of patriotic for sure. Of course, of course, they need to lead the thing.

1:47:18Talk about the stewards. One second. I was curious, is you said the CEOs you interviewed, all of them or the CEOs you talked to, all of them say their stress level is 10 out of 10. For the most part, they do. 15 years running HubSpot. When was your, like, does that resonate? Was it kind of steady state 10 out of 10 the whole time? Or did you at some point figure out a motion that allowed you to maybe hover it, you know, a seven, which is maybe like a healthy amount of stress? It was pretty close to 10 the whole time. Just white-knuckling it? It was pretty stressful. The whole thing was stressful.

1:47:58The thing with HubSpot in most companies is it looks pretty smooth from the outside, but it was like two steps forward and one freaking giant step back, and two steps forward, a giant step back, you're stepping in it. At HubSpot, we were kind of a wartime company. I like Ben Horowitz's book, and he's got a whole chapter on wartime versus peacetime CEOs. All of these CEOs are wartime all of the time. and if they don't have a war going on, they kind of create a war. I thought that was kind of interesting. It's a wartime era. The only CEO I interviewed that is not like 10 out of 10 stressed is Nikesh from Palo Alto Networks.

1:48:35He's a cool cucumber. Yeah, I wonder why. I mean, the company is like so secure. He doesn't work. A lot of these guys are wrong. I mean, yeah, like this has to correlate with strength of market position, right? Like if you are running a company where Sam Altman's out there being like, I would love to eat off of your plate. I would love to launch a competitor. And then like Satya Nudela is like, oh yeah, we're going to do that too. Like you better be operating in a 10 out of 10. Yeah, Sam competing, you know, effectively competing with Microsoft. Well, Microsoft has all his IP and he has to give him 20 % of his revenue and he's got to pay him a quarter of a trillion over the next few years.

1:49:19He might come for your nice little startup. Virtually every company, I'm just looking down my list of kids' table, adults' table, virtually all of them are extremely nervous about what OpenA is going to do. It's a long conversation at every board meeting. Of course, of course. The steward thing. I'm so interested. Why doesn't Sequoia just have one CEO? Why don't they have a super GP, like a head partner, named partner? You can just rename the fund. Kleiner was doing that for a while with Caulfield Buyers. There's so many different ways to approach it. The Blue Owl we were talking about earlier, they have two co-CEOs, and I think they have three co-presidents.

1:50:05These names kind of mean nothing, but they also have extreme weight. So what is a steward in the Sequoia Capital context? Why not CEO? Why not president? Why steward? And then what does it mean to have two now? Okay. I like the, I actually like the word steward because, you know, Don Valentine started and Michael Moritz and Jim Getz and Doug Leoni. Like it's been a long line of really strong leaders and Ruloff that they've handed it down to. And the thing that's interesting about Sequoia, so I've been in there for a little over a year, is they are not at all like even remotely resting on their laurels.

1:50:42They are paranoid. They're aggressive. They don't want to blow it. And so Stewart is actually a decent word. In this particular case, I think Ruloff was ready to move on. He wanted to move on on top. And there were just two really obvious candidates. Alfred Lin's got an incredible track record. He runs the early team. He's a good leader. Pat Grady's got an amazing track record, runs a growth team. Like, picking one of those would have been tough. They work well together. It's like, let's give it to them both. And I think Sequoia has done that kind of thing in the past. So it's not unprecedented.

1:51:16And I think it's going to work well. And they don't call themselves, they call themselves stewards, but internally they're like, just call us partners and they've made a big deal out of that. I imagine that they go around the office saying, you have to call me Steward Lynn now. You have to call me Steward Grady now. And also don't make eye contact with me. I will wear long flowing robes. No, yeah, it's just interesting, the actual aesthetics of it. I mean, we don't need to get into the actual dynamics of the changeover from Roloff to the co-stewards. It's just, I wonder if a co-CEO dynamic is sustainable, or if it is an audition process for the final boss, the final CEO.

1:52:08I mean, it hasn't worked a lot. I mean, Netflix is doing it. Spotify is doing it. But you can't think of a lot of examples. And it looks like it kind of works with Netflix. Like, that's been going on for a while, and it's doing pretty well. I mean, Spotify is really – there's just not a lot of precedent for it. So, I don't know. It wouldn't be my first choice to be a co-CEO or to fund a co-CEO. But in some cases, it can work. I think it's a little unorthodox. There is an interesting dynamic where – They call it a partnership for a reason. Yeah. You know, it's not a corporation. It's a partnership.

1:52:46And so that vibe has stretched over five decades for these folks. Yeah. I wonder the benefits of having two people in the same – in the role. Like it becomes less pressure for one person, but then you also don't have necessarily like the buck stops with one person. there's all these like odd dynamics like there's benefits and costs to to having uh the do the i think they can get a lot done and they can divide it up and get a lot more done that was the case with my co-founder and i like we divided a lot of stuff up and we got a lot of shit done together in sort of in our respective parts and i deferred to him a lot he deferred to me a lot what do they call it do doarchy where there's two kings monarchy is one key doarchy is two kings Very rare.

1:53:34How have you been processing this new wave of AI agent startups that are, in my view, oftentimes competing with traditional SaaS companies that may have only been started a few years ago? How have you been processing it? Well, I'm involved with a lot of them. I mean, they have escaped the gravitational pull of the Earth's revenue. Like, holy crap, are they going fast? And HubSpot grew fast in its day. And it's like, wouldn't even get a second look from a VC now. And it's remarkable what's going on. And I tweeted a week or two ago that, like, it's a bubble. If you get an offer to sell your company, you should do it.

1:54:21You should take money off the table. And on one shoulder, I'm like, it's definitely a bubble. The valuation is crazy, 100x and whatnot. On the other side of my shoulder, like, I just have never seen in the history of mankind growth like this. So maybe it's not. I don't know. Where do you guys come out of that? Are we in the middle of a bubble or is it just match the growth? I think we've had a, you know, we talk with a lot of investors and sometimes we get a little concerned when people come on and say they don't care at all about margins. We worry a little bit about those clips ending up in documentaries about, you know, whatever, whatever this moment in time.

1:54:54You're either in a bubble or you're in a crash, and we're definitely not in a crash right now. I would say my concern comes from companies that spring up and are raising because they're positioning their product as this AI agent, when in reality, it's hard to tell. uh why it's going to be a better solution oftentimes than something that looks like a normal SaaS product. And so I just get concerned because I know that the SaaS companies in their category have access to all the same tools and they have distribution and they can potentially use the advantage they have of having this big product surface area and a bunch of core existing product and then just compete on all these new product lines.

1:55:46and I just wonder if some of these like AI agent, these more native AI companies will struggle to actually compete with, and you've seen this, it's not like companies like Notion, for example, didn't sit around and say like, no, we're not gonna integrate LLMs into our product. Like that seems cool, but we're just a, we're a notes product and we're gonna stick to handwritten text. Like they were incredibly quick to integrate all the best models into their product. And so I just don't see the most recent generation of teams asleep at the wheel at all. Yeah, and I'm old enough to remember the last platform shift from client-server to SaaS.

1:56:29And the client-server companies were very much in denial of SaaS for a long time. I remember Oracle just being very vocal that SaaS is a bunch of BS. And obviously SaaS did very well. Some of those folks did well, and they came across. SAP did pretty well. Oracle did pretty well. Oracle did fantastic, actually. The thing that's different this time around is all the SaaS players, including HubSpot, were very early on it and are investing heavily in it and are building cool stuff. And they have a little bit of an advantage, and they have a lot of data and a lot of context, which is really useful if you're building agents.

1:57:02So I think SaaS is alive and well, and I think a lot of these SaaS companies, HubSpot, ServiceNow, Notion, so many of them are building cool stuff. Like inside of HubSpot, I think of it as like a platform with applications and agents. You can build agents on it. You can use our agents. And then it's got a co-pilot that you can refer to. It's all starting to work. People are using it. The thing that's interesting about some of these agent companies is they're working in areas where there weren't SaaS companies. Like Harvey is doing incredibly well. It wasn't really a, you know, it wasn't an area with a lot of big SaaS companies in there, or Rogo selling to investment bankers.

1:57:44Like a lot of the early huge wins on the agents are in funny areas. There's a company called Juicebox, absolutely ripping, is in recruiting areas where they're working like a big incumbent. Like they're filling in there, and some of them are co-pilot-y, sort of be working alongside of it and some are autopiloty and the autopilot ones i think will be the ones to really watch yeah it feels it feels really hairy for microsoft to go after legal ai just because that's not really they they sell mostly general purpose tools uh and so if you're building an excel plugin where you bring an ai to excel i'm worried a little bit about where that goes but if you're building something that yes it ostensibly is a word document but your entire business is off in this industry that's heavily regulated and Microsoft has only penetrated with the broad tools and has never really built a vertical solution in that category, I feel much less worried about you being attacked from the labs.

1:58:45But thank you so much for taking the time to chat with us. I'm excited for your new show, excited to check out the first episode. And the number of stewards at Sequoia is doubling. So any day now, you might get the top of the shoulder be the third fourth fifth steward they're adding more stewards every day and we're pulling for you see you guys see you cheers google ai studio create an ai powered app faster than ever gemini understands the capabilities you need and automatically wires up the right models and apis for you get started at ai.studio build our next guest is dr feifei lee from world labs she is a world-renowned computer scientist thank you so much welcome to the show taking the time to talk to us today and congratulations on the launch.

1:59:27Massive. Hi, guys. How are you? We're doing fantastically. I have so many questions about what you're building. I'd like to start with why you're building it. Specifically, what was the moment where you decided to go and start the company? Because the big institutions haven't exactly given you the cold shoulder. You've been associated with Stanford and Google and on the board of Twitter at various times. You could be doing work inside of a hyperscaler, I'm sure. Why did you want to build a company outside of the big labs? Well, because we had a vision. We had a vision we really believe in. We believe that AI as a civilizational technology will help humans and superpower humans in many aspects of the intelligence industry.

2:00:24One of them is spatial intelligence. And that is something that is, in my opinion, still in the budding stage. And our vision was ahead of most people. And this is something that my co-founders and I really feel passionate about because without this, AI or AGI would not be complete. And we want to pursue that vision and that mission. Talk about how you think about spatial intelligence. There's so many data primitives from GIS data, GPS data. There's dot clouds and all sorts of different ways, even just video footage, like what we're generating right now. We're generating spatial data to some degree.

2:01:15Where did you think the initial data sources were lacking? And then how do you think about building up the stack to kind of advance this and move it a step forward? Yeah, you're totally right. Bispatial intelligence is actually as huge or horizontal as, say, linguistic intelligence or language intelligence. So there is many aspects of it. But fundamentally, I think it's deeply perceptive. You know, it involves seeing and understanding. It involves reasoning. It involves the ability to create in the mind's eye what worlds look like. and also the fundamental ability to be able to interact with the worlds, whether it's physical or virtual, in a very profound way.

2:02:08And if you put all these together, we haven't had technology that can really do all of the above, and we're still on the journey to unlock that. But one of the things we just have done is really putting out the world's first 3D world generation model that is available to everybody. And not only is it generative, it allows users to interact and create and edit. And your question about data is actually an important one because it's hard. Unlike language, unlike videos, unlike images, you don't have too many 3D data, too much 3D data out there on the Internet. They're much more specialized. So as a group of technologists, we actually have to just work harder.

2:03:02And frankly, we use a hybrid approach in data. Some are large scale from the internet. Some are large scale from simulation. Some are real world captures. And a lot of data also comes from algorithmically processed data. That's really important because it's just a much richer problem. 3D is a much richer problem. So the, I mean, the product that we just saw, it looks a little bit like a video game. I can imagine a bunch of different ways that this gets built into something that feels like a video game. Or maybe there's this new trend of folks who are using AI tools almost as like these, they just enjoy generating songs on Suno or they enjoy just creating for the sake of creating.

2:03:54It's like you sell a guitar and then someone plays the guitar and they never commercialize their music. They just enjoy playing the guitar. And so I'm wondering if you have a view on, is it too early to tell what your customer base or what the user base will look like? Are you just hoping to get this out and then see who shows up? Maybe there's a robotics company that needs a bunch of data, and so they use your model to generate synthetic data, or there's people that just have fun with it, and they wind up paying because they enjoy the service. Yeah, I wonder if people sort of underrate world models because they look and feel today often like video games, and so they don't understand kind of the significance.

2:04:32And so, yeah, I would love to understand. You're so right. Yeah. You're so right. I think people underrate world models because we haven't had world models. So it's not people's problem. It's that the technology is really hard. It's the next frontier of AI. And like I said, this is the first time a generative 3D world model is available to everybody. And humanity has only no one world, which is the 3D physical world. But now suddenly we are in a multi-universe, multiverse situation because of this technology. So who will use it? I agree with you, actually. You know, just the sheer joy of the immersive experience for the sake of enjoyment itself is really fun.

2:05:19I was just talking to a user earlier today about the VR button. You know, we, on our product, click on the VR button. It takes you to, if you have a headset, you can just purely enjoy the immersive experiences. and, of course, add some favorite music of yours. That will be even better. But dialing all the way to professional use, there is a lot of professional use cases. If you go on our website, we have a page called Marble Labs where we show creators from VFX, from game developer, developing world, from design world, architecture, from robotic simulation, even clinical research, which I personally never thought about, are already finding use cases.

2:06:11So it's very horizontal. But I think creativity and simulation are going to be two very large areas that people will find this really useful and superpowering. okay, help me understand tool use in the world model context. So there was a moment where the large language models, the transformer-based GPT-3, GPT-4, those models were getting really good at math. And they were basically just memorizing math. And you could just ask them, what's 100 plus 50? And it would just tell you 150. And it was sort of learning math. And then at a certain point it broke down and it couldn't do really complex math just in the actual weights.

2:06:59And so the solution was to teach it to write Python and give it a Python REPL and let it write some code. And then you get the perfectly accurate math when you need that from the, not just the LLM weights, but the actual model, the experience. And so what I'm interested in is that I've seen, I've seen evidence from Jeanne and your model and world models. Generally, there's this idea of persistence and the base model getting good enough that if I go up and I paint the wall and I come back later, the paint's there, the model remembers this, maybe gravity, there's all these things that are simulated.

2:07:33But at a certain level, it feels like if I need an inventory in my game, maybe we should just use a database. And so I'm wondering where you see tool use coming in to layer on top versus where you think you're on like a scaling law where if you just keep working on the base underlying model, you will get all of that functionality for free? Yeah, great question. So I want to answer your question in two different ways because you ask a very important question technically, which is the scaling law of these models. But then you also ask the really important question of the tool use, right? On the scaling law side, 3D generative world or just world model in general is actually an earlier technology than language models.

2:08:23So we're seeing budding signs of scaling law, but this is an interplay between model architecture as well as data. So what I can say right now is we're absolutely seeing budding signs of it. But the curve hasn't gotten to the point where we see LLMs yet, which makes this technology really, really exciting. And being at the forefront of it, we really do have the best 3D generative world model is exciting. On the tool use side, this is where I find it fascinating. I think when it comes to the usage or the use cases of world model, there is a lot of professional use cases, whether we're talking about VFX for movie industry or game developers.

2:09:17And when it comes to professional use cases, I personally really believe that superpowering creators and developers is very important. That means it needs to give them the sense of control and agency. You know, if you try our model and product, there is an option just prompting a result out. and sometimes it's a slot machine but it's to be a pretty good slot machine yeah and that's great right it's amazing but um as a creator in your mind's eye there is actually so much nuance there is so much story in your mind so much um so much you want to express whether it's for a game or for a movie, for a story, for even like trying to get the robot to train the robot.

2:10:13So we actually put a lot of thought in using native AI capabilities to allow the users or the customers to engage in the editing and controllable creative loop. So if you're prompting something and you don't like the wall color here, we actually allow you to say, well, change the wall color from green to purple. Or say, add this object. Or say, I want to expand here. Or allow you to stitch together the worlds that you see. Or allow you to actually give a 3D layout so that you can put the couch where it is. So this is very important. And this is the tooling side of it, right? It gives the controllability and agency to the human collaborator.

2:11:09Yeah. You can go. By the way, my team told me I have to put this on now. There we go. Thank you. You're so sweet. Love it. You look fantastic in that. What kind of progress can we expect from World Labs on the next two, three years? like what are your kind of ambitions in terms of just general progress oh my god um two three years i i think we need to um two three years is a long time i things are moving so fast spatial intelligence is so horizontal and it's a platform i do envision that marble our product is a platform that can empower creativity simulation. And I think we're going to see a sea change in multiple use cases on the gaming side, on the VFX side, on the metaverse side, on the robotic simulation side, on design side.

2:12:15And all these are really, I wouldn't say it's almost ripe for changes. And we're seeing that level of excitement from many, many users now. What game developers or platforms do you think will be most quick to adopt world models? Because it feels like Roblox, I'm sure you've chatted with them. like Roblox should be jumping on this because it could be something that any of their users could get a lot of value from immediately speed up that creative process. But I can think of any, and some of the visuals we've shown on the screen, I'm just like, I wish that I had World Labs when I was playing first person shooter games as a kid because you could just generate infinite maps for the games that you're playing.

2:13:08You're watching this and you're like, yeah, the environments look beautiful. I just wish I had an assault rifle in this game and they were bad guys. Yes, and if you look at our page on Marble Labs, there are shooting examples already. So I don't know. I think the sky is the limit, literally. Of course, I think what's exciting in this Gen AI era is that a lot of just creators and technologists, many of them are individual developers or indie studios and teams, small teams, are really jumping in really fast. And Gen.AI models like ours really lowers the bar of entry for many of these use cases. So we're also, to be honest, we're also talking to bigger teams.

2:14:09I'm actually very pleasantly surprised by the level of enthusiasm from bigger teams as well. The team's going crazy. Having a lot of fun. They're actually playing with your website live. Is it possible that we're in a World Labs model simulation at this very moment? Sure, yeah. I'm glad this is time travel. Generated this way. Getting back to somewhat reality. The models are getting really good. Yeah. I would love to know your thoughts on market structure and how things might play out. Because I've been completely convinced of your vision for where we are on this technology, how early we are, and what's going to happen over the next few years is just the, I don't want to call them easy wins, but the logical play out the scale and you get somewhere really exciting.

2:15:05But I'm interested in the market dynamic. So in chat, in LLM, we saw sort of a somewhat of a monopoly emerge, a clear winner in consumer chat knowledge retrieval with OpenAI's ChatGPT, of course. Now other firms are obviously working on that and competing. Then in the enterprise, you have more of an oligopoly maybe emerging between Anthropic, Gemini, and OpenAI's API. There's obviously a long tail of other providers. Then you have the wrappers. And I'm wondering if there's going to be something similar that happens in world models. Is it worth thinking very hard about being the winner in consumer specifically and being not partnering with Roblox, but being the next Roblox, being the platform that wins consumer, even if it means pulling back from an API business in the short term?

2:16:05Or is it just too soon to tell and it might be kind of totally different? I'm just wondering how you're thinking about the long-term market structure for what feels like a distinct technology from LLMs. That's a great question, actually. By and large, I think it's a little bit early to tell. So right now, like I said, we just rolled out the first generally available, publicly available model. And we're going to focus on being a model company for a while because science is still early. We want to move fast and there is a lot to be unlocked. I also think that I don't think we're a model in the way that we are defining it, especially in the deep spatial intelligence way, is as consumer as a chat bot.

2:16:59Because I think, you know, like 3D is a representation. It's a medium that has its own characteristics. So I think that we're going to see products that are going to build around these models in different ways. Some are for creators, some are APIs, some are for possibly even different use cases. And I'm not saying there is no consumer. I actually think the market might surprise us, especially on the metaverse side. I think it's going to surprise you. I think it's going to be a huge consumer category. There's a lot of work that needs to be done to actually make it a place where people want to hang out for hours and hours and hours.

2:17:54But I do think it's going to happen. And I think that it's a bit of a race. But it's extremely exciting to even just be at the model layer and then experiment at the application layer. We've seen model companies get into the application layer. Yes, exactly. Of course. Of course. No, we're absolutely – I mean, model itself is an application today at the FastApp. Yeah, yeah, yeah. But we are going to focus on model. So this is the WorldApps is a deep tech company. And I think focusing on model is the right thing to do right now. Yeah. How are you thinking about fungibility between formats? It feels like there's maybe a uniquely unique value in having a world model that can export Gaussian splat, but then also export geometry, also export just an MP4 for different stages of like a VFF.

2:18:57X pipeline, for example, something that Hollywood might want to do. Do you think that you need to have like translational models or is there like one model that becomes like multimodal that can handle all the different formats, whether I want an OBJ file and I want a 3D point cloud or an Alembic, I can just get whatever I want from the same model or do I need specific models to translate between them? I don't know if you have played Marvel yet. We are actually multi-model. We are right now exporting, you know, Gaussian splats, mesh, mesh colliders, MP4s, and images and panels. But is that all from one model that's trained or is there like a translation layer between the outputs?

2:19:43Is there like one foundational truth and then like, it's like ChatGPT is not specifically trained on there's not a spanish version and an english version and a french version it just learned every language when they did the big pre-train is it the same thing or is there actual like uh or is there a uh like a translation layer that you need to do in a series of different models oh what i can say now is we are a foundation model marvel is yeah yeah okay got it cool um uh more more broadly where do you think uh ai is under hyped and where do you think gets overhyped oh that's a spicy question old model is underhyped there we go yeah talk your book i love it i agree though i completely agree i think this is a deeply underhyped category so so to be honest um you know one one thing i consider myself a scientist in my heart And I actually really don't like hyping.

2:20:42I think, you know, that's just, you probably have seen me not on the hype train in most of these discourse, right? So as a scientist, I think world model is underappreciated because it's so new. And in terms of overheated, I do think that Silicon Valley as a whole tend to mistaken clear vision with short distance. sometimes for example 2006 stanford self-driving car drove the first 140 miles in the nevada desert for mankind self-driving car it took 20 years for waymo to be barely uh on the road as a l4 and there's still it's still so limited and this is a massive amount of effort and it's also riding on the coattail of a very mature industry, especially hardware and distribution and use cases and all that.

2:21:52So this is an example to show you there are clear visions. For example, you know, robots that can do all kinds of house chores. And as a mom, I'll tell you that that would be amazing you know start with cleaning my bathroom but but that clear vision is important but the journey is going to be long it's uh you know we have issues with hardware we we still need to figure out data the world model brain of the robots also need to improve and and and all that so that's just an example and and also creativity right creativity industry which we are also engaging in. One thing I don't call it a hype, I call it a misleading sentiment, which actually really bothers me, is that we don't want to replace human creators.

2:22:52Human creativity is precious. It tells the story of our species, of our culture, of our society, of our community, of each one of us. What we want to do is to superpower and augment creators' capabilities. And sometimes the communication of technology is a little skewed towards, oh, the model can do all the work in creativity. And I personally really want to highlight human creativity as something that's so fundamental to who we are. And AI is here to help and augment. Can you help me understand the relationship between the current AI hardware build out and your technology, your model, your business?

2:23:43I would imagine that you're a beneficiary because there's all these data centers. There might be some slack capacity in the future. But at the same time, I've been hearing that a lot of the LLM foundation model companies are partnering with NVIDIA to try and steer the development of the next generation of chips to be hyper-optimized for LLM inference. And that might be okay for you, but it might actually take us away from something that's optimized for what you do. And so I'm wondering if you could just take me on a tour of what's exciting on the AI hardware build out from your perspective. Where are the risks?

2:24:24Yeah, so actually I was just in London receiving this award with both the algorithm folks as well as the hardware folks for AI. Congratulations. Yeah, so I think as you. What are AI generated? It's the Queen. No, it's not AI generated. I believe it is recording of an audience clapping, but Jordi triggered it from his soundboard. That is the Queen Elizabeth Prize for engineering from King Charles himself. Congratulations. We had it here to mention to you, but I'm glad you brought it up. Yeah, because you're talking about hardware. I was literally in the room with my friend Bill and Jensen. So as you know, the most exciting thing for modern AI, for the birth of modern AI, is the convergence of hardware, software, and data.

2:25:18So from hardware side, I think it continues to evolve. For world modeling, especially our technology relies on rendering, we do think we will keep our conversation, our side of the conversation going with the chip makers. Because some of our requirement, especially on the rendering side, as well as on the training, the model side will be somewhat different from LLMs. And, you know, I would love to, I will, and I would love to call for the chip industry to also work with us on this front. What is exciting?

2:26:04I think that's a great question, actually. So I was in Middle East visiting some of the construction of data centers. And that was pretty epic seeing how large the Stargate data centers are being constructed. And it really feels that we're in a different phase of industrial revolution. You know, we never, none of us lived in the days of the steam engines, the electricity, when the scale of industrialization in that time was just multiplying for humanity. I think we are now living in the AI industrialization era and seeing that scaling up is quite exciting. Hypothetically, last question for me.

2:26:57Hypothetically, if I gave you unlimited access to a one gigawatt data center, the best and brightest, the hot off the presses, the freshly installed one gigawatt data center, probably billions of dollars invested in this. would you do a training run? Would you run your model? Would you just scale up? Or would you say, hey, I would need more time to actually advance the way we think about training before we press go on the big run? Yeah, that was going to be my question is like, at what point does World Labs become capital constrained? And are you capital constrained today? Yeah, it's the whole bitter lesson.

2:27:45Research and new ideas. There are different parts of the business that are bitter lesson-pilled and others that aren't. And I'm wondering, like, what pieces are currently on the great scaling curves? Yeah, like I said, the scaling curve, from a model architecture point of view, we're still early. So if I have one gigawatts and that many chips, I'm more likely going to actually run some parallel models and experiments to really hole in the bet. So it doesn't mean we're still trading large models. It's just not at the level of today's LLMs yet. So are we capital constrained? we are capital constraint you gotta start hyping more we're gonna hit the gong we hate capital constraints how much have you actually raised how much can you share publicly it's known that we've raised more than 240 million dollars publicly why did you hit the gong just now because we're increasing the hype We heard your capital constrained.

2:29:03We hate that. Yes, yes. It's a good sign. It's a mutual hype. I'm wearing your hat. Yes, we appreciate that. And we appreciate you taking the time to come talk to us. Thank you so much. This was extremely exciting, extremely enlightening. And yeah, we're very excited for you and your whole team. Congratulations on the launch. And we'll talk to you soon. Yeah, thank you, guys. Thank you. Bye. See you soon. Before we bring in our next guest, let me tell you about Profound. Get your brain mentioned in ChatGPT. Reach millions of consumers who are using AI to discover new products and brands. Let me also tell you about Linear.

2:29:36Linear is a purpose-built tool for planning and building products. Meet the system for modern software development. I'm going to ask Kari to add this sound effect when you open Linear. Projects and product roadmaps. I'm loving the new stingers. Look at that. That looks fantastic. Our next guest is Scott Sanders from Forterra. He's got some big news. How are you doing, Scott? What's going on? Good to see you. Good to see you, too. First time on the show, introduce yourself. Let us know what you do, how you describe Fortara, for those who don't know. I mean, first off, long-time listener, first-time caller.

2:30:08I think it was actually the subject of the original TVPN print the internet situation. Oh, yeah? Where you were printing out physical memes, and I criticized that. You criticized it? Okay, well, you came around. I came around. Was it like we were using up too many pieces of paper? Is this an environmentalist thing? You were anti-printing or something? What was going on? No, I was just, you know, the internet exists on your computer of what's the need to physically print it off every day. Well, I printed your news. I will put it out there. We printed your news here. Yeah, we printed your news. Because we care so much.

2:30:44We wanted to physically hold. We like the Wall Street Journal. We like the Financial Times, you know. You'll come around. We should do like a TBPN, like for every episode, we plant a tree just to make sure that, you know, we're replacing. But I feel like you, I feel like we're kindred spirits. You, you work on physical things. We, we work in, in physical, in the physical realm. Of course, there is a software element to everything that both of us do, but give us the update on, on, on what you're building because it does touch the physical realm, correct? That is, that is true. We do, we do build hardware.

2:31:18Yeah. We, we announced a$238 million Series C today. I think there's a sound effect happening. There we go. Congratulations. There we go. Massive. That's fantastic. We're building distributed systems for defense starting from the ground up. We feel the warfighter, you need drones, you need maritime capabilities. There's great companies building that. But we're focused on building stuff that primarily goes on vehicles because at the end of the day, try not to walk into combat. if at all possible. And we want to be able to equip those warfighters with technology and distributed compute at the tactical edge that enables vehicles to be self-driving, active communication nodes and relays, launch things like loading munitions, put radars into place to detect aircraft or detect enemy shipping, and then put anything from cruise missiles to anti-ship ballistic missiles in the back so that we can deter our enemies from invading countries and want them to invade.

2:32:18Yeah, my mental model for this stuff is a little hazy. Maybe you can help clean it up. I feel like there's usually like one company that does just autonomy. There might be another that's doing some sort of battlefield tactical mapping, some awareness things. It sounds like you're doing multiple. How do you actually interface with the other the other touch points? I mean, who do you need to integrate with? Since some people might be familiar with maybe what Palantir builds or what Anduril's built on the operating system level. Can you give me some concrete examples of groups that you're working with or technologies or platforms that you're building on?

2:33:02What's the wheelhouse case study for 4Tera? Yeah. And on top of that, I feel like the challenge from a product roadmap standpoint, and product planning standpoint is you have to understand what are the systems that we have in place in our armed forces today and then how the battlefield is actually changing. We've heard from other people on the show that a lot of the equipment that we have might not be relevant in a world where the front line is just drone-on-drone action. Yeah, I think one of the best examples of doing this is actually with one of our partners, Raytheon, or in this case, Chaos, in that quick little snippet.

2:33:39where with Raytheon, we and Oshkosh, we integrated our edge compute platform and our self-driving software to take a traditionally manned vehicle and make it unmanned. So one, by putting the edge compute on self-driving vehicles, you're enabling other algorithms and other platforms to run on top of them. So we think about the ecosystem as a partnership approach where we're not going to make everything that a warfighter needs. In fact, no defense company, defense tech or otherwise can do that. And so we open up our architecture and our system so that a Palantir can run at the tactical edge, not on a giant server rack somewhere.

2:34:16Or in the case of the small multi-purpose equipment transports that you're looking at, we put Chaos' long-range biostatic radar system on it to detect drones. And our future integrations with those guys will enable them to utilize our compute and our autonomy later. We don't care who's doing the C2. We obviously do our own because you kind of have to. But if that's Androld's Lattice or Palantir's Gaia or smart system, we're very open to that. And we've broken up our system in a way that enables the user to sort of make the decision of what capability they want for each different application layer and not wrapping it all into one holistic platform unless that's what the end user and the customer want to do.

2:35:03What was yours or the team's reaction to Pete Hegseth's speech last Friday? We had Catherine Boyle from Andreessen on talking about how Hegseth was basically making a case that was it Cheney made back in like 2001? So these are basically reforms that have been, people have been asking for for a long time, but we haven't been able to actually deliver. Yeah, we were lucky enough to be at that event. And, you know, defense tech reform is people, is a thing people have been talking about since before defense tech was a thing. You know, the acquisition system has become extremely complicated and I think difficult to work with for many companies.

2:35:50And it's, you know, to have a Secretary of War coming out and saying like, look, this is an actual problem, right? Like the rate limiting step on getting technology to operators is very rarely whether there's the technology that exists that can solve the problem. It's can it work its way through the prototyping into production phases and getting to scale production with an end user. And that's been very hard. Palantir has been at this for 20 years. Andrel's been at this since 2017. and definitely not going to do that math in public, although it'd be quite easy to do. I'm not a math guy. He's a growth guy.

2:36:29That's amazing. Okay, speaking of growth, how is the company growing? I mean, we've texted about this. It's all SBIRs, right? Yeah, it's all Sivers all day long. Actually, no, I think we're too big for SBIRs now, which is great. we're about 550 people total and so we're not still not that big in defense tech or defense world still relatively small yeah trying to be a scrappy startup and you know work with not only the legacy primes who build incredible systems right there's no one in defense tech building 8x8 25 ,000 pound armored vehicles that can go head to head with you know major land forces across Europe and there's only a couple small startups that are starting to do the missile systems that we want to put on the back of these platforms but our approach is we open up and enable our architecture to work with nearly anyone to deliver that that capability at the tactical edge we really just feel like humans being the blunting force of combat it's a pretty bad idea we've probably 20 veterans at at for terra and a lot of us feel very strongly and very passionately that i would much rather have a robot make first contact with the enemy or be the platform that is emitting a large signature on the battlefield something firing a missile firing a howitzer moving a radar so the platform is inevitably targeted it's not a human that's getting targeted it's a machine and we're trying to change that dynamic you know one platform one program at a time well uh thank you so much for coming on the show and thank you to all the veterans uh it was veterans day yesterday right and thanks everyone for their service.

2:38:08And congratulations on all the progress and congratulations on the Series C. And next time you come on, we will not print out your news. We'll just go in blind. Yeah, we'll just go in blind. Because we don't know how to use computers. As long as the next time isn't at 10.30 at night, I'll be stoked. Wait, are you in Lisbon right now? Were you guys at? Web Summit? Josh is at Web Summit. I'm in Paris at Tectonic. I'm in with some customers. Okay, cool. Awesome. Yeah, we can coordinate better on the time. Congrats to the whole team on the progress. Yeah, awesome. See you soon. We'll talk to you soon.

2:38:42Have a good day. See ya. Numeral. Numeral.com, not numeralhq.com anymore. Numeral.com. It's sales tax and autopilot. Spend less than five minutes per month on sales tax compliance. Our next guest is Jamie Allaire from Circle. Just had earnings. Welcome to the show, Jamie. How are you doing? Good to see you. What's happening? I'm great. Good to see you guys. Welcome back. Fantastic. Welcome back. Thank you. Give us the rundown. What's the latest in your world? Yeah. Well, this morning we announced our Q3 earnings. It was a tremendous quarter. We saw USDC grow 108 % year on year, which is fantastic.

2:39:23The amount of activity happening on-chain with USDC grew about 600 % year over year to$9.6 trillion of on-chain activity with USDC. Hit the gong for that. I think that's the biggest gong ever. Maybe we've hit the gong for various countries' GDPs. I think we might have hit for a GDP once, but that's it. The only thing that's triggered. Transaction volume. That is remarkable. That's a lot. Of course, that's the nature of these networks. Yeah, a lot of other great stuff. I mean, we launched something called CPN, which is an on-chain money movement payment network, earlier this year. That is growing very fast.

2:40:05Dozens of financial institutions coming on that network. It's already gotten to an annualized payment volume of about$3.4 billion. Key, though, just from a financials perspective, we saw adjusted EBITDA grow 78 % to$166 million in the quarter. That's a 57 % adjusted EBITDA margin, which is about 730 basis point expansion. And so strong performance on that as well. And then I think really importantly, we recently launched ARK into Testnet. And ARK is a new layer one blockchain network that Circle has been building into the market. We had over 100 major companies, major banks, payment system operators like Visa and MasterCard, major AI companies, major exchanges, e-commerce companies, gig economy companies, all these working with us, collaborating us to deliver ARK.

2:41:00And we also announced this morning that we're exploring a token, a native token for ARK Network, which is an exciting development for us as well. And so we're seeing a lot of activity in this space. We had the Genius Act pass recently, and so we're also just starting to see more and more major institutions, major financial firms, payments companies getting in the game and collaborating with us and building products with us as well. Can you tell me anything about the Anthropic relationship? There's this very interesting concept of like using stable coins to power agents in the future. There's that.

2:41:41And then there's also just like Anthropic might need to pay someone internationally. Can you give me any guidance on like where we are on that? Yeah, I mean, look, a couple things on that. I think the first is that we've been working alongside others on these standards for agentic payments using stablecoins. Like MCP, which does not have stablecoins yet, but it should, maybe in V2 or something. Well, yeah, and standards like X402 and Google's agent-to-payment standard. We're collaborating on both of those. We're releasing tools for those. And we actually have a number of agent developer tools, companies that are supporting these things as well.

2:42:18And so that's actually happening. It's actually one of the key design centers for the ARK blockchain network that we're building, which is we envision a world where over time, the velocity of money increases and more and more of our economic activity is machine intermediated. And that machine intermediation is going to be in these new computing networks, these provable, trusted computing networks, which we call blockchain networks. And AI agents will become, we believe, the primary intermediary that's intermediating activity on these. And so we see that as sort of, if you look at these trends converging over the coming several years, we see these kind of compounding each other.

2:42:58And so, yeah, we were excited to have Anthropic as a collaborating partner for us with ARK. And I think we had a number of other AI-focused firms there, too. And I think it just speaks to the fact that we need a substrate. We need an execution environment that AI agents can use that are trustworthy, provable. And look, AI agents need to be able to make a$0.05 transaction for consuming some tokens or a$50 transaction. It needs to work anywhere in the world where software works. And that's what these new networks are really good at. And so there's not a payment system that's ever really been built to support that.

2:43:39And I think something like USDC and then these new networks like ARC Network and standards like X402 and others really help with that a lot. Talk about just like the strategy of going full stack and how you guys are positioned against competition at the different levels. You guys are an issuer. You're building your own network. You're going to have a native token.

2:44:05Nowadays, it feels like there's probably a new stablecoin that gets created every second of every day, right? Not just a net new of an existing stablecoin, but an entirely new stable. You guys are well positioned in a lot of ways, but I guess talk about how you view your competitive position as a variety of other networks come online and new stablecoins come online? Yeah, absolutely. I mean, I think the first to start with the core, which is our stablecoin network has really strong network effects. We have liquidity network effects globally. We have infrastructure network effects just in terms of where we're plugged in around the world.

2:44:43There's tens of thousands of products and services that integrate to our network, which really drives the developer flywheels. And so we're in a very strong position there. And we're taking that and we're expanding kind of down the stack and up the stack. And I think in our view, if you think about where we are in the evolution of this, the global economic system is enormous. And we expect more and more of the economic system to, as I said, become machine intermediated on these new kind of economic operating systems. And we're still at the front edge of that. We have a market that really grew out of speculation on cryptocurrencies, and now we're seeing more and more of actual financial contracts, actual trade flows, payments flows, treasury activities, investment products, lending products.

2:45:28All of this is actually starting to come over, but we're at the front edge of that. And so I think our view is that there are going to be some very significant scaled internet platform companies built out of this transformation over the next 5, 10, 15 years. We think that these can be as big or bigger than prior internet platform companies that were built around centers of gravity around search or social or mobile or cloud or e-commerce. We think that that's going to take place. And we think that we're in a very good position to provide really great infrastructure across those different layers.

2:46:01And I think if you look at what we've been able to do with ARK, even just getting to Testnet, the range of major firms from around the world that touch literally over$100 trillion of assets, whether those are custodians, exchanges, clearinghouses, payment systems, et cetera, we're putting forward a model that we think can work for scaled mainstream adoption. And so that's an infrastructure problem. And we're doing it in a way where we can build a distributed network of distributed stakeholder participation, distributed governance, and kind of bring that into the world. And obviously, that's what we're considering as we evaluate the design of an ARK token as well.

2:46:42And then on the other hand, we definitely see an application layer emerging on top of these platforms. And I think likewise, we're still in the early stages of the application layers being built. We are definitely taking a crack at part of that by building a payment utility, this CPN as a payment application. But we're designing it in a way where other financial institutions can use it, benefit from it, and generate revenue from it. We're not trying to compete with them as well. And so at every layer of what we do, we try and build kind of market neutrality and interoperability at the core. And that has served us well.

2:47:20I mean, one of the things we've pointed out this morning in the earnings call is, you know, we actually run, as of October, the largest cross-chain interoperability protocol in the world. CCTV, which is Circle's cross-chain transfer protocol, in October was greater than 50 % of all cross-chain traffic accounting for all assets across all blockchains. And that's really significant. We really want to be part of the solution to interoperability while we build at these different layers of the stack. We've had a number of conversations about the local regulatory environment, how that's changing recently.

2:47:57But what's going on internationally? Last time you were on the show, we talked about how part of Circle's Edge is having offices in a bunch of different countries and being set up to navigate what will ultimately be unique regulatory environments in a bunch of different countries. Well, it's been amazing, actually. Since the Genius Act, we've seen almost every major government in the world accelerating plans around building local rules around stablecoins. We've seen, for example, Hong Kong in August went effective with their new rules. We're seeing regimes working on this from major emerging markets to some of the largest developed markets like Korea, which is accelerating their efforts around stablecoin rules.

2:48:43UK accelerating their efforts. So now everyone's sort of getting behind this model of private sector, bank and non-bank, issued stable coins, clear supervision, usually supervision by the central banks, just like we're going to be overseen by the national bank regulator here in the United States. And so that kind of standard is cascading around the world. And I think also a lot of these regulators are looking at ways to support kind of reciprocity or kind of consistency as well. And so I think that's really important as we build infrastructure that we know that it'll work in all these different places around the world.

2:49:21So I would say the activity for us internationally has really, really accelerated. And you'll see a lot more as well as we go forward. But I think the U.S. is actually really providing leadership right now in this whole area. And the rest of the world is taking notice and following. It's kind of a national competitiveness issue as well. Fantastic. Well, congratulations on a successful earning season. We'd love to have you back on the show soon. Thanks, guys. Great to catch up. Have a great rest of your day. Likewise. We'll talk to you. Cheers. Yeah, bye-bye. Go bye. Before we bring in our next guest from the newsroom waiting room, let me tell you about Fin.ai, the number one AI agent for customer service, number one in performance benchmarks, number one in competitive bake-offs, number one ranking on G2.

2:50:05Our next guest is Vlad. He's building lighter. We have some big news on the funding side. How you doing, Vlad? Good to meet you. Welcome to the show. Welcome to the show.

2:50:19What's happening? How you doing? Hey, guys. Can you hear me? Yes, we can hear you. Please introduce yourself. Introduce the company for those who aren't familiar. Awesome. Yeah, good to be here. Actually, I think you had my former co-founder, Scott, on a couple weeks ago, who's not working on Cognition. Oh, Scott, of course. We've had him on a bunch of times. TBPN Hall of Fame. TBPN is made possible by Cognition, the makers of Devon, the AI software engineer. We love doing ad reads for Cognition. We're the makers of Devon. Oh, nice, nice, nice. But we're not here to promote them. We're here to promote you.

2:50:55And Leiter. Yes. Exactly. So I'm Vlad, I'm founder and CEO of Leiter, which is a decentralized exchange built on top of Ethereum. We can talk more about how we built it and why. But we launched in January and kind of had pretty strong growth this year with our mainnet. And we actually did a public launch in October 1. But yeah, and I guess before, later I worked on a lunch club with Scott. And prior to that, I guess I'm kind of like, I'm not a crypto guy that got into finance. I'm more like a finance guy who took a detour to AI and then got into DeFi. Yeah. Talk about some of the choices that you've made, like why Ethereum compared to other chains?

2:51:52Why decentralized exchange? like of the technical decisions that you made, I feel like we've talked to a lot of folks who have picked out different quadrants if we do a two by two quadrant or something like that. You've clearly landed in one particular piece, one particular spot on the market map. How would you describe the trade-offs that you've made and then why did you make those trade-offs? Awesome. Yeah, that's a great question. And we actually started working on LIDAR right around three years ago. So mid to late 22 is when we were thinking through that idea maze, as it were. So it's interesting to look back now.

2:52:36And the genesis of the idea was that we were thinking about digital assets. Digital assets have been around for 10, 15 years. They've been traded a lot. But most of them represent some kind of blockchain or decentralized protocol or rails for the future of finance. But the weird thing was that 99 % of the activity of how they're traded actually did not use the rails of the stuff that these assets are building. It's actually trading in a centralized way. And centralized exchanges, there's some good ones. I think some of them definitely have provided value to customers, but at the basic premise, a centralized exchange doesn't actually use the blockchain at its core to verify what's going on.

2:53:31In some cases, it's actually worse than TradFi. At least in TradFi, you have regulations and the legal system to fall back on if something goes wrong. The thing is that actually almost three years to the day is when FTX collapsed. And we were looking at that when we were thinking about how to build lighter. And we thought, okay, well now people will get it and volume will move away from centralized exchanges after one of the big ones failed. And that didn't happen. So what that taught us was that actually the decentralized exchanges that existed at the time, their product market fit was so bad people would rather trade somewhere where there's a one in six chance of the whole thing being a fraud than using the decentralized products.

2:54:17So then we thought, okay, well, this is something we should be part of the solution to. So for us, it's like, okay, what we want to achieve is low cost, low latency, secure, verifiable, and composable. And the first three of those are pretty self-explanatory, but But what does verifiable mean? It means that everything that happens, there's a cryptographic proof that it happened fairly and correctly. And why is that important? Well, if you ask the reverse question, what if it's not verifiable? Well, that means there's some part of the system where you're just like, trust me, bro. Which is kind of, again, defeats the problem.

2:54:56Even if there's one part like that, the whole thing could be manipulated or corrupted. So that's a problem. And then composable is also important because at the end of the day, we're not just trying to build a sandbox for people to speculate on. We want to build something that's composable with the rest of DeFi, whether it's lending, stable coins, staking, prediction markets, all this other stuff. So they can kind of all work together in smart contracts. So anyway, so we had those five goals. And going back to your question, so Ethereum was the obvious choice for the security layer. For it to be secure, we had a pretty strong view that the layer one that's by far the most secure out of things you can build on is Ethereum.

2:55:43Solana has built a great L1 as well, and others have too, but Ethereum is very secure. It's been stable for a very long time. So that was pretty clear to us. and yeah we can dive into how we achieve the other goals but as far as like why decentralize and why Ethereum I think that gets into those points I don't know if we have enough time to go super deep into more of the technical side I'm more interested to hear about how you met Ken Griffin honestly so could you tell me the story of the Ken Griffin recruitment I would love to know like how did he even know to reach out to you what happened there tell me the full story because I think that's fascinating.

2:56:27By the way, I'm also based in Miami now for the last five years. Not only did I learn a lot of stuff from Ken about finance and business, but also that South Florida is a good area. Yeah, but, you know, so I guess, you know, the thing with my time at Citadel and, you know, so I was 18 when I finished Harvard and actually even while a student. Let's give it up for finishing Harvard at 18. Congratulations. Continue. What's that account like VCs congratulating themselves? We're congratulating you. You just said a fact. You just said a fact. But, yeah, so actually, like, even at school, I did some trading and was working on kind of a side project with two friends around options.

2:57:24And so, you know, so I was doing that and that didn't quite get off the ground. But I got to know the team at Citadel through, you know, essentially like the research firm. So, you know, it's not, you know, it's not like Ken personally gave me a call or something. But I got to know the team and they wanted me to join. And then kind of towards the end of the process, Ken got involved and spent a bunch of time with me, you know, to tell me more about what their high level goals are at Citadel and their strategy and why it would be a good fit for me and kind of making sure that I'm comfortable joining them.

2:58:03And I guess without going into too much details, I mean, he made it a very attractive proposition for me and was pretty involved, you know, the first couple of weeks, especially of my work there. And, yeah, you know, he's to this day someone I still really look up to. And yeah, there's just fascinating lineage here. There's just fascinating lineage here, obviously. Ken Griffin went to Harvard, was trading in his dorm room. uh it took him a couple extra he didn't he didn't graduate at 18 so he sort of mogged him on that front but uh but then of course uh he had a deep relationship with robin hood and robin hood markets came in and invested in this round and so there's a whole bunch of like full circle moments i i did want to quickly ask if you think that the industry fully learned its lesson from ftx because i feel like there's i've seen some chatter over the last few weeks around uh various exchanges saying like no your your funds are frozen but don't don't worry we're we're working on it and it's you know maybe sending some alarm bells Right.

2:59:12So October 10th was an interesting moment for the space that actually DeFi worked well. whether it's Aave or Morpho, all those protocols. I think our competitors handled those market moves in different ways, but I think all of them in the context of DeFi worked the way they were supposed to work. Some of the centralized exchanges, I think maybe, again, the problem there is lack of transparency. You don't really know did they actually liquidate their customers correctly or not? Maybe they did, right? Maybe everything worked exactly as expected, but if it didn't, there would be no way we would even know that.

3:00:01I'm not aware of any cases of specific wrongdoing by centralized exchanges in that cycle, but again with the lack of transparency, there was no way I would know that. We do know there have been some blow-ups of smaller hedge funds like on-chain hedge funds, right? And what ultimately caused those is unclear, but the one thing I will say is I remember after FTX, there was this whole thing about like, okay, let's even centralize exchanges, let's make them more verifiable and prove reserves. Now, of course, what does that mean? Because you can prove assets, but how do you prove liabilities? But still, there was some push to do that.

3:00:43And I remember Vitalik had some ideas on how to do that. And Binance was looking into it. And then I think someone asked CZ about it. And he was just like, well, Vitalik is very technical. Some of the stuff is too complex or something like that. So I don't know if the centralized exchanges have really truly learned their lessons on that front. But did anything particularly egregious happen that I know of October 10th? I can't say that. But again, And the whole point is that it's supposed to be transparent and trustless. Well said. Well, glad you're building what you're building. Yeah. Well, thank you so much for coming on the show.

3:01:26And congratulations on the massive fundraise. Very excited to continue following the story as you continue to build this company. But have a great rest of your day. Awesome. Hope you enjoy the beautiful Miami weather. Cheers. Come back soon. Have a great rest of your day. Thank you. See you. Let me tell you about Adio. customer relationship magic. Adio is the AI native CRM that builds scales and grows your company to the next level. You can get started for free. Also, I'm going to tell you about 8sleep.com. Jordy's traveling, but I got a 91. Seven hours, 55 minutes. I'll give it to you. Let's go.

3:01:59And our next guest is Andrew DeSouza from Bordy. He's in the restroom waiting room. Welcome. What's happening? How are you? Great, great to meet. and welcome to the show. Kick it off maybe with a quick intro on yourself, background, and then let's talk about Bordy. Yeah, yeah. So I'm the founder, CEO, creator of Bordy, the AI super connector. Bordy is helping founders meet investors primarily, but connecting people for business and economic reasons. We launched a program about two weeks ago to try and get 100 founders to help close the round. We have 5 ,000 founders raising$16 billion signed up and 600 investors spending about a trillion dollars of capital.

3:02:44Some size. So, yeah, I think we've probably got about 10 % of the active fundraising happening right now among founders that are working through Bordy, which is pretty exciting. Awesome. So, yeah, I guess immediate thought is with products like this, like one of the challenges is just like balancing the network basically. And with InVenture, it's oftentimes the companies that have the most demand for their equity or the companies that are, you know, maybe performing the best oftentimes don't need intros, you know, they're in the position of like saying no. How are you thinking about kind of like maintaining quality on the platform when you're a business and you want to be as big as you can possibly be?

3:03:36and I imagine you have a strategy there. Yeah, I think that was something that I was worried about. This is the reason why we didn't start with the idea of connecting founders and investors. It just sort of happened organically that this was the activity that was taking place. I've been shocked at the caliber of founders that have come in looking for help. And I think as a CEO, as a founder, your job is to create demand for your share. It doesn't matter whether you're Elon or Roy Lee or a YC founder. That's your job. And you'll take all the help you can get. Most founders only raise capital two or three times in their life.

3:04:11And having someone that can help you navigate that, get the right introductions. I think there are probably 50 or 100 founders in the world. I think this is the reason why we talk about this sort of AI bubble right now. It's maybe 50 or 100 founders in the world that every investor is chasing, right? Because that's the hype cycle. there's probably thousands of founders who have incredible businesses we're talking to them um that have incredible businesses incredible metrics and they're just you know they just haven't they're not based in silicon valley or they're not you know straight down the middle um pattern recognition

3:04:48i'm back they may not be based at all stupid good luck good luck raising if you're not based these days. Yeah, that makes total sense. Where does this go from here? This feels like a good early wedge, but hard to build a billion-dollar business yourself just within venture. So I imagine you have a number of other pathways or use cases for the introduction engine, or I don't know how you... Yeah. Yeah, we've built Bordy as like an AI board member. So what is the most important meeting for the CEO to take at any time? Right. Is it a customer meeting? Are they hiring an executive? Are they trying to get on a podcast?

3:05:37Is it a big partnership? Is it corp dev? It happens to be that fundraising ends up being a high-urgency place where you can build a great relationship. But I think at Bordy is an economy. Every time you connect two people, there's some economic opportunity that is unlocked. There's some latent economic potential. We can financialize it. I love it. That's actually what you're doing. It's true. It's actually good. Okay, can I give feedback to Bordy? If I get an intro and I take the call and I'm like, can I be like, yo, Bordy? Never again. Never again. Because, I mean, everyone's gotten an intro from somebody and you're like, wait, why did you make that intro?

3:06:19Like I jumped on. They didn't know who I was. How am I on with the venture fund that loves firing CEOs? Don't introduce me to them. Introduce me to the friendly ones. Yeah, so how do you, because I can imagine that's like one way that you want to make sure, you want to get to the point where the entire industry is onboarding, like every single founder, every single investor. But as an investor, for an investor to like not churn, they need to be able to like give feedback and say like, hey, like this didn't meet my bar for a meeting. I don't, you know, and if you keep sending people meetings or intros that don't meet their bar, eventually they'll just say like, I'm going to, you know, log off.

3:06:59Yeah, yeah, absolutely. So, so this is, this is one of the things we take a lot of, we take feedback. Bordy has goodwill with people, right? So his idea, the same way that if anytime I make an introduction, I'm gambling goodwill. I'm going to, if I make a bad introduction, I burn some goodwill. If I make a great introduction, maybe I've gained some goodwill. And you're sort of doing that calculation at any given time. The version of Bordy, if you just sign up, you're talking to a low latency, relatively low cost on the phone, on WhatsApp. You're just talking to, you know, the standard version of Bordy.

3:07:28What we tried to do, we had this experiment. We were running this fundraising thing. We said, what if you could remove the constraints of latency and cost? What if I could spend 100 times the tokens? What if I could let Bordy run for 30 minutes trying to find the perfect introduction and position you in the perfect way? Can I get better results? And that's kind of what we're seeing. So now we've had 5 ,000, and Bordy's talked to 5 ,000 founders that are raising in 600 investors in the last two weeks. He has now picked the top 100 that he knows he can help that are going to be in demand from the investors in his network.

3:07:58And we put them into a shared Slack channel with us and Bordy and the founder. And we've now got those shared Slack channels with people. And we're like, hey, here's some investors. What do you think? Get feedback from the founders. We're doing the same thing with investors. So the investors are sort of signing up and saying, here's a few founders that are raising. We think they'd fit your thesis. and Bordy's just learning. So you're like, actually, that's not a good fit. Or I've talked to this founder, more like this, less like this. And over the course of a couple of weeks, we really dial it in on both sides.

3:08:22And we just make sure that you're spending time. And it's less about capital allocation. It's more about like, how are you allocating your time and attention onto the deals that are actually likely to materialize? As an investor, the deals are going to win. As a founder, the investors are actually going to give you a term sheet and fund. That makes sense. What are some other kind of like types of work that you expect Bordy to do in the future? Yeah, we're already starting to see a lot in sort of corporate, you know, enterprise sales and biz dev and partnerships. There are there's it starts with the corporate venture capital arms.

3:08:54So we're getting a ton of corporate venture capital funds coming in. And now we're talking to their innovation teams. And it's the same thing. It's like, hey, look, I got to I got to stay abreast of what's happening in my industry. And it's almost impossible. Right. Maybe I'll watch TVPN and see who comes on there. But like, that's almost too late. Right. It's actually like I've got to I've got to see who's starting a company. you know who's just being funded who who is going who do I want to be who do I want to be tracking and so board is starting to help with that and making those introductions hiring is another one that I think is interesting you know especially when people are passive they're not actively looking they don't want to take a bunch of recruiter calls but they'll talk to board like I might start a company but if there's a really great startup out there I might join the team and so those kind of conversations are quite interesting as well but it's like whatever you would call your board member whatever whatever introduction is going to move the needle for you that is worth the time of the CEO, that's what we want to work on.

3:09:46That makes a lot of sense. I'm looking for a board member who can take me on a track day. That's my only thing I want to get out of them. Take me to the track. Put me in a race car. It better not have a passenger seat. And then get out of the way because I'm building on my own. Thank you so much. Yeah, I mean, that's a good challenge. Maybe that's customer development. Maybe once you get into the really high tier of boardy spend, you do a track day. Well, you know, so Nico Rosberg is a big boardie user. No way. There we go. Oh, he's been on the show. Big time boardie user. That's awesome. And so he could probably make it happen for you.

3:10:18Very cool. I love it. I like it. I like it. It's one of those things that in kind of working on, when you describe it, of just like a place that a founder can go to just get help with their business, it's not it's surprisingly not something that uh not nobody's you know made that pitch on on the show to date which is surprising because that's pretty much you know feels like very valuable to a lot of the companies that are coming on yeah fantastic well thank you so much for taking the time to come talk to us congratulations on all the progress yeah great to meet you we'll try out the product and uh yeah you need to find like i guess we're raising yeah that'd be great there I'd love to give it a shot.

3:11:02Have a good one. We'll talk to you soon. Cheers, Andrew. Goodbye. Let me tell you about public.com, investing for those who take it seriously. They got multi-ass investing, industry-leading yields. They're trusted by millions. I'm also going to tell you about AdQuick, out-of-home advertising, made it easy and measurable. Say goodbye to the headaches of out-of-home advertising. Only AdQuick combines technology, out-of-home expertise, and data to enable efficient, seamless ad buying across the globe. Our next guest is Parag Agarwal from Parallel AI. Welcome to the show, Parag. How you doing? Hey, hey.

3:11:31Hey, good to see you next. What's happened? Are you at a newcomer event right now? I am. Oh, yeah, you got the step-in-route behind you. Cerebral of the Pro AI Summit. Congratulations. What did you announce on stage? What are you announcing here on TVPN? I announced on stage that we just raised CDZ. We gave$100 million from some of the best investors. Yeah. Yeah. Let's give it up for the best investors. 740 million. Kleiner Perkins Index Ventures. There we go. Got some Spark Capital participation. We love to see it. Kostla's in there. First round to rate. Wow, really? Really? Murderers row. Mamoon Kleiner is joining the board.

3:12:16That's exciting. That's exciting. Kind of looking like the Figma board. We got a good crew here. I like this. So take me through the actual company progress, the progress that unlocked this$100 million Series A. Obviously, there was some demand. There was some research progress. What was the shape of the data points that went into the ultimate deck that got this across the finish line? So we got first time two years ago. We started with this sort of find the sky idea that agents are going to use the web and they're going to use them a lot. And we need to build new infrastructure for them. Yeah.

3:12:54There weren't any agents around in that moment. None of them were using the web when we started. They were trying. Some of them were trying. Yes. And this year, everything changed. Everyone's now building an agent. And every agent wants to connect to those. So the market just came alive, I want to say six, nine months ago. The most frontier customers started doing its last years and became our design partners. This year, the number of customers, the number of queries, the diversity of work, like there's coding agents using us. There's people in sales, building agents using us. Finance, risk, compliance, insurance, healthcare, people doing scientific research agents.

3:13:43People in pharma and biotech figuring out how to use web search APIs to build agents. So like it's a completely different world this year as it relates to agents. and so agents using the web. So yeah, to make it super simple, people are using parallel web systems within maybe a SaaS provider is integrating parallel into their products so that if somebody's doing research, let's say in a CRM on a customer, you guys can help them use agents to do a web search and fill in that CRM with data. That would be one example. That is a perfect example. You see this white offline. Our products are APIs. Our users are agents.

3:14:27Our customers are people building agents, right? So if someone builds an agent to say, okay, I'm going to take your CRM and add very interesting data to it so that you can take better actions, that agent that they build will call our APIs in order to search the web, in order to pull content from the web, in order to get the highest quality information to be able to fill up that it is. React to this post from Jim Cramer. He says, when I use ChatGPT or the others, I can tell quickly that they won't pay up for good sources. We all know this. So what's the point? They need to start paying newspapers and stop using bogus sources we would never use ourselves.

3:15:11We need integrity, not volume. There's a bunch of ways you could read that. But my take was there is a data quality opportunity, especially in the enterprise, especially where willingness to pay is maybe really high. And so how are you thinking about making sure that Parallel and the deep research API has the highest quality data forever? Perhaps I'm being partly with Jim Claymore. I don't always.

3:15:44So there are two things there. How do you get the highest quality deep research? That is a lot of things. High quality data, authoritative data is a part of it. And then there's a lot of other technology we're building on top, whether it be at the ranking layer, the retrieval layer, the crawling layer, the reasoning layer, to get you higher quality deep research. I ask you to try a deep research product and compare it to OpenAI's and we win nine times out of 10. So I want everyone to try that and tell us if I'm wrong. So, go to the data question. I think there are deals being done by people. The world I don't want to live in is only Google and OpenAI's agents have access to some data.

3:16:29And most publishers don't get to make those deals. And most other agents don't get to have that access. That is the core mission of our business. We started with building products for AI customers. We've recently started figuring out what the incentives are for publishers to allow AI agents to use them, which will include payments, which will include other forms of value, which is transactions, and which will include distribution, which are the things that people care about on the web. And having run Twitter before, I understand the concerns of someone publishing on the web and business models around ads are likely to evolve.

3:17:17And that's part of the reason we end up doing this company. Because we think we have an opinion or two to bring to life here. Do large revenue multiples scare you?

3:17:30No. Good. No, as long as you know what you're doing. There we go. Confidence. I love it. Love the confidence. Gotta be confident. But yeah, I love how even since the last appearance, the vision and opportunity just seems bigger and bigger. Totally. Yeah. Makes a ton of sense. Thanks for having me. It's great to have you back at this rate. I'm sure come back on before the end of the year. Yeah, you're probably back in a couple of weeks. I'm going to ship a product in a couple of weeks and be like that. Fantastic. Head us up then. Congratulations. We'll talk to you soon, Prague. Have a good one.

3:18:09Thanks for having me. Bye. Let me tell you about Bezel. Getbezel.com. Your Bezel concierge is available now to source you any watch on the planet. Seriously, any watch. I want to do a lightning round. We got a bunch of posts. We're going to try and go through them as fast as possible. First, Brian Johnson did confirm. This is from Brian and Garel. Brian Johnson confirms he was not one-shotted by a recent Magic Mushroom trip. You said if Brian Johnson comes out of this five-gram trip and says, yeah, we're going to conquer death. we're still on then he's certainly a true believer what did he say he replied to tbpn's post he said we're going to conquer death so brian johnson is feeling good looking good back on the timeline uh not one shot i suppose one of the founders of uh chad ide says uh he quoted my post and said really great post however it doesn't apply to chad ide for five steps ahead different plane of thinking, but really great posts are reposting for visibility.

3:19:07Nice post, little bro. I want to say, I did sneak a peek at your post during the show. And I don't think you got called unk as bad as I expected. You made it sound like you were getting destroyed. You did very well. You got over a thousand likes and people are like, oh, you're taking shots at Gary Tan. No, Gary Tan chimed in. He said, good points. Big thanks. A lot of people did identify that you were rage baiting the rage baiters, which is funny. I had to use their own tricks against them. Yeah, you did. I think in this case, you used the tricks for good. And I think it's a great post and I enjoy.

3:19:49But maybe we have to have the, there's always room for follow-up. Maybe we'll talk to Gary Tan later this week. We'll get him on the show. Think about how he's thinking about the future of YC. Maybe we'll have the founder of Chad Labs, Chad IDE, whatever it is. Yeah, we DM a little. We will also be demoing the product at some point. I believe we'll be getting a preview of what the actual product is. If there's something awesome behind the scenes, we're going to tell you what that is. I just love responding and being like, yep, actually it doesn't apply to us. We're five steps ahead. Or five steps ahead.

3:20:21Different plan of thinking. I mean, that is just a complete trump card. How do you respond to that? how do you levels i think he wins uh much like wander find your happy place book a wander with inspiring views hotel great amenities dreamy beds top tier cleaning and 24 7 concierge service it's a vacation home but better what is this la lamp introducing a little lamp says uh shavir shark a sharkery uh an emotional and expressive robot to reshape our attachment to set to technology is this a cgi uh vibreel or is this a real product can we do you have any idea who put this in the show notes. I put it in the show notes.

3:20:58Okay. Do you think this is real? It looks real. Yeah, it looks real. It looks real? Okay. It looks pretty cool. I wonder, so would you buy this? What is the value of this? It's obviously very cute, very cool. It's the Pixar lamp, but it should exist, but why? What does it do? It's pretty cool. I don't know. It's weird. It just feels cool. It's a cool thing to have. Put a speaker on it. I'm like 90 % sure it's real. There was a Apple research paper that came out like, I don't know, six months ago. It was all about like expressive robots. Sure. And they had a bunch of like demos and stuff on how they could build this stuff.

3:21:32Yeah. It looks very exciting. Yeah, well, the website is Lelamp.com. That's a good one. L-E-L-A-M-P.com. You can pre-order it. Lelamp like follows around your cursor. Pretty cool website. Let's see what the checkout is. It's only 50 bucks? Wow. Okay. This is very cool. I have been delight baited. I have not been rage baited by this. I think this is a great launch. And in other news, Jim Cramer is posting about quantum computing. He says, Regetti, what an opportunity. Ka-ching, buy IBM. What does that even mean? Cramer posts like a crazy amount. I mean, he's got 2.3 million followers. He's posting like every half hour.

3:22:15The guy's a machine. He's going to be on the show in a couple weeks. We're very excited. I cannot wait. It's going to be a lot of fun. Anyway, that's our show for today. I do not tolerate Kramer slander. Yeah, no, no. I think he... He's entertaining. He's insightful. He's controversial. He's a total package. We love Kramer here. He's great. Oh, there is a review from Andre Carpathia. I missed this one. Oh, there's a couple of posts that I missed, actually. Andre Carpathia says, I took delivery of a beautiful, shiny, new HW4 Tesla Model X today. I like this hardware four. He cares about the chip inside that thing.

3:22:52He doesn't care about what the light bar on the front is or the door handles or whether the doors go up. Andre Carpathic cares about the GPU inside. So I immediately took it out for an FSD test drive a bit like I used to do almost daily for five years. Basically, I'm amazed. It drives really, really well, really smooth, confident, noticeably better than what I'm used to on HW3, my previous car, and eons ahead of the version. I remember driving up Highway 280 on my first day at Tesla nine years ago where I had to intervene every time the road mildly curved or sloped. Note, this is V13. My car hasn't been offered the latest V14 yet.

3:23:31On the highway, I felt like a passenger in some super high-tech maglev train pod. The car is locked in the center of the lane while I'm looking out from the Model X's higher vantage point and its panoramic front window listening to the incredible sound system or chatting with Grok on city streets. The car casually handled a number of tricky scenarios that I remember losing sleep over just a few years ago. It negotiated incoming cars and tight lanes. It gracefully went around construction and temporary in-lane stationary cars. It correctly timed tricky left turns with incoming traffic from both sides.

3:24:07it's a glowing review glowing review of of Tesla self-driving from the man who basically invented it and started the whole process, I don't know if he started it but he was incremental he was instrumental in the development of it so let me go out there we also missed it but OpenAI released GPT 5.1 rolling out to all users this week NIR says NIR says ChatGPT is officially in its Fiji phase now if you're wondering why the upgrade doesn't come with benchmarks. Have fun. Yeah, no benchmarks. So yes, there's two models. There's 5.1 thinking and instant. So if you go to like the actual like model router on like chatgbt.com or whatever, the only like there's GPT-5, which you can do pro, which is like the super long, almost like a deep, deep research product.

3:24:58And then the rest of them are just the 5.1 thinking. But yeah, no, no benchmarks or like any kind of really technical report. In the blog post, it was a lot about the personality of the model. She also released a separate blog post on her substack. And there was some mention of like, they're definitely like, they tried to address kind of the 4.0, like one-shotting stuff. They're like, okay, we're definitely aware that there are some problems with a very small amount of people becoming very addicted to the models. But yeah, nothing very technical out of this. It was only 1 % of our user base. It wasn't 1%.

3:25:44You're saying it was 10 million people that want to kill Rune? Use that 10%. Rune, up your security. Anyways, we actually have to go over to Base, which is a conference hosted by my dear friend, Will Cofield. So we got to head over there. We're going to have a little chat. That'll be fun. More podcasting. Another two hours of offline podcasting. You love it. Cannot wait. Cannot wait to be back tomorrow. Have a great rest of your day. Thank you for hanging out with us. Leave us five stars on Apple Podcasts and Spotify. Subscribe to the email at tbpn.com. And we'll see you tomorrow. Goodbye. Cheers.