Loading summary
Transcript758 lines
- [00:00]
Jordy
You're watching TVPN.
- [00:02]
John
Today is Thursday, November 13, 2025. We are live from the TVPN Ultradome, the Temple of technology, the fortress of.
- [00:09]
Jordy
Finance, the capital of capital.
- [00:11]
John
Ramp time is money, say both easy use, corporate cards, bill payment, accounting, a whole lot more all in one place. There is a bunch of breaking news. The big news out of OpenAI, of course, the OpenAI show continues. Sarah Fryer had some comments about ChatGPT's growth potentially slowing and it's unclear. It's not in decline, it's maybe deceleration. We'll have to dig into that. It had me thinking about debt and I was thinking about just the fact that the debt has come to tech for the first time really. And this was sort of my take and I'm a little bit. This is an area that I know the least about. And so I was doing some research, learning about a different industry since it's just so abstract to me because I've never worked in private credit or really seen that industry or even just really studied it.
- [01:06]
Jordy
You appreciate leverage. You've never been a big.
- [01:09]
John
I appreciate it.
- [01:10]
Melissa Tokmak
Yeah.
- [01:10]
John
And mostly I was just wondering, like we keep going back and forth on the debt is coming to tech narrative as like, it's very scary. Like when debt comes, only bad things happen. We live through the global financial crisis and there's a lot of jitters when debt is around. It's like, oh, you could get wiped out, you could blow up. The backstop comes in. It just feels like all of a sudden we're talking about things with a much more serious consequence than like, oh yeah, a startup raised some money and it didn't pan out and it was a zero and it wound up being a write down, but it was part of a portfolio of equities averaged out across a whole bunch of different LPs. Like, there's no, there's no even when, even when, you know, like Theranos blew up, it was only equity holders that were lost. It wasn't this higher entire industry. And so it wasn't, it didn't turn into this like systemic issue.
- [02:07]
Melissa Tokmak
Right?
- [02:07]
Jordy
Yeah.
- [02:08]
John
But now it feels like with the 1.4 trillion of backlog that OpenAI has kind of opened up across a whole bunch of different deals, there is this worry that, that maybe the level of indebtedness could be risky. The level of risk in the system, the level of investment in the system could be something that's bigger than just, oh, if you're in this one name, you're taking a big Risk. Now it's maybe like, hey, we're all taking a risk. And if we're talking about backstops, at least. And so I was trying to understand. There's this old phrase from 2006 coined by Clive Humby, classic coinage. I love a coinage. He said, data is the new oil. And back then in 2006, his point was he was working as a data scientist at Tesco, which is this British grocery store. I don't know if you know this story, but he was working at this British grocery store chain. And his point was we have all this data on a customer is in the rewards program. We see that they buy a Thanksgiving turkey before Thanksgiving. We see that they buy this type of paper towels or this type of milk or whatever. We have this data, but we don't really do anything with it. The data is not valuable. We need to refine it, much like oil into gasoline. And once we refine it into gasoline, then we can do things like targeted advertising and we can increase our customer value. So it was basically just a generic call to action for taking data science seriously. For just don't just have the data there. Understand that the data is valuable if you extract it, if you work on it. But the metaphor, people have been saying data is the new oil for I guess, two decades now. And it never really sat that well with me because unlike oil, data is not perfectly fungible. Yeah. So one, one tranche of data is not equivalent to another. Like Reddit is clearly very valuable since it kind of provided the backbone for GPT3. All the analytics data that flows out of some mobile game is basically a.
- [04:12]
Jordy
Lot of data is worthless.
- [04:13]
John
A lot of data is worthless.
- [04:15]
Jordy
All oil has at least some value, essentially.
- [04:19]
John
I mean, I guess there are different levels of crude. Right. There are different grades. And I was actually trying to play out the metaphor more and I was wondering can we get to a place where we can wring intelligence out of raw data like the oil? And the result can be low octane gasoline, kind of like midwit level slop, an AI slop. Or it can be jet fuel, like a deep research report that's actually pretty great, or some code that's really reliable and really useful. But it all depends on the processing methodology. But the more interesting data is the new oil. Take that I don't think was considered in 2006 is that maybe the tech industry is going to look like the oil and gas industry soon. I was looking up how much debt is in the oil and gas industry. It's over a trillion Dollars of debt and it's like it's fine. Yeah, exactly. Clap. It's fine. It's not this huge systemic issue. It was 2 trillion a decade ago and then it went down and then went up. And it's all just a funct of like how much oil and gas is going on, what are the new projects, how big are the projects, how much debt goes in. Like just having a lot of mortgages in America is not intrinsically risky.
- [05:28]
Jordy
Okay. The difference is that if you identify oil in the ground and you figure out how much it's going to cost you to extract it and how long you think you'll be able like basically estimating like how much, how much oil is actually available in this site, then you can lend against that pretty predictably because you know that the price of oil is going to fluctuate. But in general, as long as it's in a, in some range, it will be like a profitable operation to pull it out of the ground. And I think it's a little bit easier to lend against that than GPUs today when we're. The big debate is around depreciation schedules and will these GPUs. You know, we have a sense that a data center that has power and basically a box with a lot of power will be valuable in the future. But if you're, if a lot of the cost of a new Data center is GPUs, it's harder to gauge on what the value of those GPUs will be in, in four years than it is. Okay, it will this oil like production site still be producing oil in five years? I think that's a bit easier to answer and easier to lend against maybe.
- [06:43]
John
I mean sometimes there are tracks that only produce oil for four years and you underwrite it against a four year depreciation schedule. And as long as you get the, as long as you match the risk to the reward, the deal pencils out just fine. But I understand what you're getting at and I think that as we dig into the OpenAI news, I think we'll have more. We can synthesize some of the recent leaks and rumored statements around potentially a plateau and demand for tokens on maybe the consumer side. But it is just like a wildly different question. Like the fact that you're walking through that math is very different than what the venture capitalists in 2000 were doing. Like Ev Randall, who's coming on the show on Friday tomorrow. He always says he goes back to the Google prospectus From when they IPO'd and Google was like the most pure play, just beautiful software business. So Google from 2001 to 2004 grew from 86 million in revenue to 3.2 billion in revenue and net income over that period went from 10 million to 400 million. And that includes stock based comp. So they were still making 400 million in profit with the stock based comp. Googlers made a lot of money, they gave away a lot of stock and so it didn't look like an oil business. There was not this big capex build out. There was not this big or even this crazy R and D phase. There wasn't that much capital that went into Google before it became this monster cash flow machine. It was just infinite money glitch. It was sort of an infinite money glitch. It was this beautiful algorithm that was just discovered and it was so elegant and it just produced this monotony monopoly insane like growth rate for so long. And then of course they've been challenged and they expanded and there's million things and then eventually capex did come into the picture as they grew their cloud, cloud infrastructure, gcp, all this other stuff. But for a long time like tech just meant take a bet on a company and it's either a zero or a trillion dollars or something like that. And so it's a lot different. And I wanted to dig into like the actual structure of one of these deals because I don't, I think that tech people, I was, I was almost going to call this like why is no one talking about Blue Owl? Because people obviously on Wall street are definitely talking about Blue Owl. It's, it's a public company that stocks I think down like 30.
- [09:20]
Jordy
Poster child.
- [09:21]
John
But, but it's the data center of finance.
- [09:23]
Jordy
Private credit.
- [09:24]
John
Yes, private credit, exactly. And so I wanted to understand like how does Blue Owl actually interact with one of these data center deals? Because that's important to understand like where the risk winds up living. So I'll break one of these down but first I tell you about Restream 1 livestream, 30 plus destinations, multi stream and reach your audience wherever they are. So for Hyperion, you remember the Hyperion release, Zuck went on Threads and announced that he was going to be building a 5 gigawatt data center. It was going to be as big as Manhattan.
- [09:53]
Jordy
It's like somewhat of a Manhattan Project.
- [09:54]
John
Somewhat of a Manhattan Project, exactly. So the crazy, crazy thing about that deal. So he spins up the, he puts out the announcement post on, on, on threads, says hey we're going to build this 5 gigawatt data center campus, it's going to be online in a few years. It's going to be as big as Manhattan. And he shares some of like where it's going to be, how many racks are there going to be, square footage, stuff like that. But he's basically just announcing that, like, hey, the project's financed, we're ready to go on this. Like you would expect that when that it's a $27 billion deal, you would expect that. Okay, Meta went down, they spent $27 billion. It's worth it. They're going to. No, they got paid 3 billion. They got paid 3 billion. And the reason is because Blue Al financed it with external debt and they are basically paying Meta upfront for the right to have them as a, as a tenant, as a leaser for a very long time. So they get this, like, we have Metta as a client. Meta is always going to pay their bills. They're not, they're like, no matter what happens with the AI build out, they're going to be good for it because they have this cash machine. So they are like the best possible tenant. Not like some fly by night, oh yeah, I'm a startup, maybe I'll be around in a few years. It's like, it's matter. They're going to pay their bills. And so you have this massive data center project that's going to be paid for even if it's not producing any valuable tokens. Zuck's still going to, he's not just going to default and be like, yeah, take the company. No way he's going to pay. And so in exchange for that, they had 3 billion upfront. And so there's just each one of these deals, I think the more you dig into them, I want to have more of these people on. Mohamed El Arian at Pimco was formerly at Pimco. I know he can explain this a little bit more. I want to have more people on the, on the show to help us get up to speed on this because this feels deeply important to the current AI buildout. Boom. The tech story, it feels, it feels like an entirely new piece of the puzzle to understand where this technology is going, and I don't feel equipped to understand it at all. Barron's did have a great article about Blue Blue Owl and a very funny interaction between Blue Owl and Jamie Dimon and they're going at it and I think it's interesting to read through. So let's read through a little bit of this to give you a little bit more flavor on what's going on at Blue Owl. Because if you're just in tech, if you're just in venture, you might not know that much about them. But first, let me tell you about Privy Wallet infrastructure for every bank. Privy makes it easy to build on crypto rail, securely spin up white label wallet sign transactions, integrate on chain infrastructure all through one simple API. So in Barron's I had this. This article's from October 24th. I had it on the table. We never got to it. We're getting to it now. The title of the article is Private Asset Star. Blue Owl has been flying high. Is it too close to the sun? This feels like headline. Jordy would write. I'm very skeptical about what's going on in the AI buildout in the AI boom. But let's dig in and see how it looks.
- [12:53]
Jordy
I like the phrase private assets star. Yeah, start calling our friends private asset stars.
- [12:58]
John
For sure. For sure. So the article says, suddenly Blue Owl Capital is everywhere. This past Tuesday, the upstart alternative investment firm with an aptitude for private credit announced a financing deal for Meta Platform's $27 billion AI data center in Louisiana. That is Hyperion that I was mentioning earlier before the week before at the PAX C I S Alternative Asset Summit in Los Angeles. Blue Owl's co CEOs co CEO Mark Lipschitz called JP Morgan Chase's CEO Jamie Dimon cockroach warning about risk in private credit. An odd kind of fear mongering. So what happened there was, we talked about that, that blow up in the private credit world and I have a little bit of background on this. So where, where did he say this? So I need to actually pull up what happened with the, with the private, with the cockroach statement because it's very funny. So basically, First Brands. It's First Brands. Let me see. First Brands. First Brands. Okay. So basically private credit has been growing a ton. We've talked about this a few times. Aries is massive now. Blue Owl is really big and the whole. And there's basically been this little bit of a fight emerging between where the debt is coming from. Do you do private credit or do you go with the traditional bank route? And so Jamie Dimon, at least I'm pretty sure he's going head to head against Blue Owl in a bunch of these deals. And so there's this question of like, are they chirping at each other intentionally? And so Jamie Dimon was cautioning investors about potential risks in the credit market by invoking A proverb when you see one cockroach, there are probably more. And so he was referring to recent loan defaults such as the bankruptcy of auto parts maker First Brands and subprime lender Tricolor holdings as warning signs of broader credit issues. So Jamie noted, or Dimon noted that JP Morgan took losses on some bad loans and implied that trouble in one corner of the credit market could mean undiscovered problems elsewhere, implicitly casting doubt on the booming private credit sector. And so Mark Lipschitz fires back and he says, I guess he's saying that there might be a lot more cockroaches at J.P. morgan. And so he's actually saying like, oh yeah, maybe you should go check out their books and see if they have other bad stuff because so First Brands collapsed was an isolated case of alleged fraud actually in the syndicated loan market. And it was not in the direct lending arena where Blue Al operates. So Blue Owl had no exposure to First Brands. And yet Mark Lipschitz was still, you know, firing back at JP Morgan for kind of casting doubt off on the direct lending arena where Blue Owl plays. So there's these cockroaches, there's these cockroach statements and they kind of go back and forth on this. But the history of Blue Owl is also interesting. It's this like merger between a few different companies here and it's part of this broader boom in alternative.
- [16:14]
Jordy
Blue Owl was the primary lender for coreweave.
- [16:19]
John
Coreweave and they've also done Stargate, they've done a ton of stuff, but interestingly their data center business is I think like less than a third of their overall business. They have a lot of other stuff going on here. So George Walker, who's the CEO of the old line money management firm Neuberger Berman, he's a cousin of President Bush, he says it's extraordinary what they've done. It was just a startup and now their $26.6 billion market market cap compares to a number of large century old financial institutions. There were some. Blue Owl's backstory entails some rich behind the scenes machinations. But more significantly, it reflects the stunning trajectory of private markets which have tripled to 26 trillion in assets over the past decade. The company's also. Yes. So I mean, I do think it's important to keep the scales in mind here. The 1.4 trillion seems so big in the venture context and we think about Sam Altman as a venture backed founder, but he's now playing in a market.
- [17:27]
Jordy
That he's playing a Hyperscaler game.
- [17:29]
John
Yeah, he's playing a hyperscaler game. And so when I think about it's like 1.4 trillion. That's the same size as the oil and gas market.
- [17:38]
Jordy
Meta's Manhattan like Manhattan Project scale Data center, the 5 gigawatt data center that you talked about them doing this deal with blue owl on Meta's also just did a $30 billion bond offering which has 4 billion of 4.2% senior notes due in 2030 and then all the way up to 4 and a half billion of 5.7 senior notes due in 2065. And there was an order book of around $125 billion for the $30 billion issuance. So there's a massive amount of demand, huge demand. So this is not at least that we know of what OpenAI has not been raising this style of debt for the business. And it's unclear if there would be like a ton of demand for OpenAI's on balance sheet debt.
- [18:29]
John
Totally.
- [18:30]
Jordy
Given that it's unclear if they're going to be able to spend what they've already.
- [18:35]
John
Yeah, but underwriting a Data center with OpenAI as a client is very different than underwriting OpenAI directly. So there's some really funny quotes in this article. Blue Owl is the pretty girl at the dance right now, says Wall street trader David Williams. We're talking many billions in private credit. Ah, yes, private credit. Though Blue Owl has three lines of business. Private credit, specifically direct lending in private equity deals, is the firm's calling card and growth engine. And the straw that's stirring Wall Street's punchbowl lately. They also have this like GP business. If you want to buy a GP stake in Alternative Asset Manager, you can do that through Blue Owl. But what everyone's interested in is this private credit specifically for AI assets, at least from our perspective. I'm sure there's other people that find the other pieces of their business much more interesting. But its core direct lending business has 145 billion in AUM out of 284 billion total. So that's about half the fund. And that was conceived the firm started in 2016 as Owl Rock at the Putnam Restaurant in Greenwich, Connecticut. Of course, comfort food its best by principles. Doug Ostroover, formerly the O of GSO Capital Partners, now Blackstone Credit. Craig Packer, a former Goldman Sachs partner, and Lipschitz, a former KKR partner. The name came from the wisdom of an owl and the stability of a rock, says Lipschitz. And the website was available.
- [20:05]
Jordy
That always helps.
- [20:06]
John
So instead of relying on inventure, we all think of the GPLP private markets fund structure, right? Jordi Alrock change this. They don't do the typical GPLP split. They use what's called business development companies, BDCs. So those companies issue stock and lend money to businesses, usually those with junk credit ratings. So something like a one off data center that really only has like one client. And it's not like Apple, it's not Microsoft, it's not actually like, you know, been in business for 30 years, not the government. And so it's going to have a junk rating, it's going to be a higher interest debt instrument. And, and this has actually been a trend. Other major alt firms are also turning to BDCs which support higher yields. And so BDCs send some 90% of the interest collected on those loans to shareholders through dividends. So they've basically created the same structure as a real estate investment trust or something close to it. And so this has allowed them to scale. So two of Blue Owl's BDCs are publicly traded, others are private. They have Blue Al Capital Corp. Which yields 11.4% and Blue Owl Technology Finance which yields 9.9%. Both are down down about 14% this year, the former from January 1st. And so Goldman Sachs recently called the fears overblown about the risk of falling rates and weakening credit. And they cited Blueout as undervalued, noting that it has a stock price to fee related earnings multiple of 21.7, which is 5% below its two year low. So the stock has been beaten up, but it still has like a buy rating from Wall street firm. Blue Al has generated a stable, highly predictable stream of earnings, says Ostrover, the other co CEO. It makes no, it makes no sense that we're down more than our peers, he says. If anything, we should be down less. I love it. I love a defense. Wall street may be maybe particularly wary of direct lending as shares of both Blue Al and Aries, which specialize in that business, have fallen hard. It's also true that both stocks had previously outpaced their peers. The second leg of the Blue Al stool was created years earlier when a Lehman Brothers executive, Michael Rees, started a fund at Neuberger Berman that bought stakes and asset managers like D.E. shaw. This is what I mentioned earlier about buying GB stakes. Rees named his endeavor Dial after his children Dylan and Alexia. He just took his two kids names, pushed them together, raised its own capital from Koch Industries and invested in the likes of Silverlake and Vista Equity Partners. So he was a vehicle to allow you to buy GP stakes in Silver Lake and Vista Equity which are it not publicly traded.
- [22:46]
Tomas Puig
I believe.
- [22:47]
John
In 2020, Blue Owl merged with Owl Rock. And so the the resulting company was called was named Blue Owl. A bank working on the deal had called it Project Blue. So Blue was added to and and Rock was dropped because it was Owl Rock before. And so the hatching of Blue Owl was problematic to some companies in which Dial had invested, particularly sixth Street Partners and Galoob Capital, both of which sued claiming the new company created a competitor with information about their operations because of course they own GP stakes in the companies. And so in 2021, Blue Owl bought Oak Street. Real estate capitalist Chicago based firm, specialized in sale leasebacks and triple net lease deals. As its third business. This wing of blue L has AUM of 71 billion and is home to the Meta Infrastructure deal. So it's like a remarkably balanced stool. I'm sort of shocked. I was expecting it to be like, oh, we hear about Blue in the Meta context and that's their main business or it's a new thing and it's only 5% of their business. But in fact they have a pretty, pretty diversified offering across a few different products. And so this wing now has 71 billion and is home to the Meta Infrastructure deal and others such as Stargate data centers in Texas and New Mexico. Next, Blue Owl is working on AI deals with N Scale and Valor Equity to finance purchases from Nvidia, whose chips go for $30,000 and up, according to people familiar with the matter. Blue Al built its direct lending business by borrowing from the Silicon Valley playbook of scale first, monetize later, or by underpricing established private credit firms to gain deal flow and then raising fees later. Most of the firm's direct lending business is done as part of private equity buyouts. So Thoma, Bravo, Blackstone, Warwick Pincus, these companies come in and then Blue L does the debt side of that. These investments generally entail floating junk bonds, a business pioneered by Drexel burnham in the 1980s. But it was typically done with banks. Now it's done with these private credit firms. The private credit market has grown from 2 trillion in 2020 to 3 trillion at the start of 2025. So again I'm like, that's growth. But that's not the craziest growth I've ever seen. When I go back to global financial crisis, you hear about these like 10x run ups in these derivative markets. I don't know. No, it just feels like, it feels like we're still in the early stages of actually ramping this piece of the capital markets and marshaling that to the really crazy stuff. It feels like the crazy stuff's coming in two years. I don't know.
- [25:24]
Jordy
Yeah. And that aligns with Doug from Semianalysis Point. He was like, debt. We're still early in the debt cycle. We're early, but at the same time, market has jitters. I was working on our 2026 merch this morning.
- [25:38]
John
Yes.
- [25:38]
Jordy
And I had about an hour call, and so I missed the fact that Nvidia's down 4%, Core Weave's down 8%, and it's a little bit shaky out there. We're certainly not in white suits today.
- [25:51]
John
No, no. Well, let's go over to the timeline. Let's go over to some of that other news that we wanted to touch on today. But first, let me tell you about Cognition, the makers of Devin, the AI software engineer. Crush your backlog with your personal AI engineering team. So, Alex Heath has a scoop here in Sources. During a recent private call, OpenAI's investors asked about external signs that ChatGPT's growth is slowing. CFO Sarah Fryer.
- [26:21]
Jordy
The external signs were, I think, like App Store data. There was some data out of Europe.
- [26:26]
John
Oh, yes, yes, yes, that's right, that's right.
- [26:28]
Jordy
And it was hard to read into the European data because Europe, Europeans, they don't work after. No, I mean, it was coming off of summer. Right.
- [26:36]
John
And.
- [26:37]
Jordy
And, you know, ChatGPT is popular with.
- [26:39]
John
Students, but European summer hasn't ended yet. European summer ends, like, late December.
- [26:43]
Jordy
No, I have to push back on that because when the French television network came.
- [26:47]
John
That's true.
- [26:48]
Jordy
They were clearly done. That was about a month ago. They were clearly back from summer holidays and they wanted to learn about the AI talent work.
- [26:55]
John
Yes, that's true, that's true. No, we're obviously joking there.
- [26:59]
Jordy
Anyway, so there's.
- [27:00]
John
There's been. There's been early warning signs. Well, yeah, walk me through some others.
- [27:04]
Jordy
So I can walk through Alex's coverage says on Monday, OpenAI CFO Sarah Fryer held a private. She was really hoping to just not be in the news cycle this week. But when you're the CFO of one of the most important companies in the world, it comes with the job.
- [27:24]
John
What bag holder is leaking this?
- [27:26]
Jordy
Yeah, that's pretty private call.
- [27:27]
John
You're an investor and you're leaking bad news to sources. What are you doing?
- [27:34]
Jordy
Yeah, what are you doing?
- [27:34]
John
Did you get out or something? Like, are you have you, have you somehow facilitated some short position?
- [27:40]
Melissa Tokmak
Yeah.
- [27:40]
Jordy
Why are you jumped off the call market?
- [27:42]
John
This is very founder friendly. Whoever's doing this. Whoever's doing this is not very founder friendly. Okay. Anyway.
- [27:49]
Jordy
Anyway. Sarah Fryer held a private quarterly earnings call with the company's biggest investors. As usual, the numbers she shared were mostly up and to the right. But behind the strong top line fig figures, a quieter question hung over the call. Was ChatGPT's momentum starting to slow? During the Q and A portion of the call, sources say Fryer was asked to reconcile ChatGPT's meteoric growth and weekly users from 250 in September 2024 to over 800 million now with external signs that the app's growth has slowed in recent months. Close followers of OpenAI's business have been whispering about these signals from research firms since late summer. But this was an opportunity for company backers to hear directly from leadership on the matter. After telling the investors to take third party estimates with a grain of salt, Fryer acknowledged a chink in ChatGPT's armor. She said time spent had declined slightly in response to, quote, content restrictions the company rolled out in early August. She then referred to the loosening of those restrictions that CEO Sam Altman has said will be implemented for adults in December. So, so this is them. Sam came out and said, we're going to allow erotica on the platform. And Sarah says, and OpenAI expects the decline in time spent to reverse. And so this reminded me of a conversation we had about exactly a month ago where I said, I don't think them announcing that they're getting into erotica is a sign of strength. I don't think that's something that you do. You do just because you want to. Right. In my view, it felt like clear. I mean clearly there's user demand for it, but at the same time that felt like something that you, that they would do in order to stimulate growth while they get a bunch of other monetization online. Right, so like commerce, ads, etc.
- [29:43]
John
Yeah, no, that makes sense. Yeah. I mean the original like founding team at OpenAI was incredibly idealistic, right? Like incredibly. Like you're going to work on a non profit, on like super intelligence like AGI, like you truly are working on like what you see as one of the most important problems. What I would agree with is one of the most important problems. Then of course, like, you know, eventually the, eventually the company evolves and you and you bring in business leaders. But at the same time, like.
- [30:22]
Jordy
I.
- [30:22]
John
Do believe they Say I want to cure cancer. I believe that.
- [30:26]
Jordy
I believe that too. But the reason I reacted strongly to it was that there had been messaging around the same time of I don't want to be in a world where we have to decide between curing cancer and free education for the world.
- [30:40]
John
Yes.
- [30:41]
Jordy
And so then at that same time deciding we're going to do erotica.
- [30:45]
John
It was very weird timing. It was very weird. It was very weird timing that those two statements, like, came out one after another.
- [30:52]
Jordy
Yeah. I'm actually surprised why they're waiting until December to roll out the content.
- [30:58]
John
So in this scoop, do we have any, do we have any specific data on what exactly. What exactly is indicated in terms of ChatGPT's growth slowing? Can we actually try and define that a little bit more? Is that users. Because we're already at almost a billion users. Is it time on site? Is it monetization? I mean, deceleration? We were talking about this. OpenAI has decelerated revenue before because I think they tripled and then they went to a doubling or they were quadrupling and then they went to a tripling and so they actually decelerated in 2024 and then they re accelerated in 2025. And so I was kind of saying like, well, you know, there's a good chance that you could see deceleration in the future. It's happened before. To be accelerating forever is basically impossible. But it would be interesting to track exactly how ChatGPT's growth is slowing. There certainly feels like there's just a level of saturation. Do you have the stats?
- [32:06]
Jordy
Yeah. So similar web put out some information on month over month change in total visits to leading gen AI tools. ChatGPT is at the bottom of a list that includes Gemini, Deepseek, Perplexity, Grok, Claude, copilot and Meta AI. The key difference here is that ChatGPT is so much bigger than these other platforms that, that they could still be adding more users on a per user basis than these other tools, even if their growth is slower.
- [32:39]
John
Yeah, yeah, that makes sense. Tyler, what do you think?
- [32:41]
Tyler
I mean, they do expect their growth to slow down. So like from this is epoch AI, it was like OpenAI revenue estimates. So 2025 is 13 billion, and then they expect 2.3x, 2026, 2x in 2027 and 1.6x.
- [32:57]
John
Yeah.
- [32:58]
Tyler
So I mean, it's not like they're just saying like it's, it's going to go from 2x to 3x to 4.
- [33:02]
John
Yeah, exactly. So I wonder I wonder how much of this is just framing something that was sort of already priced in as a. As like a bad thing. Like, I feel like people were expecting deceleration. And so if she's, if she says, like, if she's on the call and she says, as expected, we're really big, we're going to be decelerating the level of new users that we're adding and.
- [33:35]
Jordy
Then I don't think she should say that.
- [33:38]
John
Yeah, maybe that would be bad framing. I don't know, it just doesn't seem that crazy.
- [33:42]
Melissa Tokmak
Please don't say that.
- [33:43]
John
I don't know. It doesn't seem that. It doesn't seem like that bad of a thing to say. Meta is not accelerating top line, top line users. They have like 3 billion users. Like no one's expecting them to accelerate top line users, maybe like randomly, 1/4. They accelerate, but not continually. And so I don't know, it just feels like an odd thing. Did you get a chance to read the Ed Zittran article? This thing?
- [34:14]
Jordy
I did, but I didn't. I felt it. Ed is such a massive OpenAI hater that I think it was hard to. And the sources were pretty unclear. It was hard to read too much into it.
- [34:27]
John
Okay, well, let me tell you about figma.com, think bigger, build faster. Figma helps design and development teams built great products together. You can get started for free. Should we head over to some timeline? What else people are saying? Cairo Smith says it's going to be very funny when LLMs plateau around 120 IQ and what we've created is just a digital guy, a God.
- [34:52]
Tyler
I mean, this doesn't make any sense. If we have like infinite digital guys, that's like literally a guy is just like a worker. Yeah, if we have infinite workers, that's like insanely bullish.
- [35:02]
Jordy
Yes, it'd still be bullish, but it's been prompt, people.
- [35:07]
John
He didn't say bearish. He said. He didn't say it's going to collapse the economy when we just get a digital guy. He said it's going to be funny. And I agree.
- [35:16]
Tyler
I guess that's true.
- [35:17]
John
It's funny.
- [35:18]
Tyler
But this is still like a very bullish take. I think people, I think you might read this as like being. He's kind of bearish.
- [35:23]
John
Yes, yes, yes. Yeah, no, I know, I know. I think you're right. If you get a digital guy, that's pretty powerful because guys can do a lot of stuff. It's valuable.
- [35:30]
Tyler
I love guys.
- [35:31]
Jordy
Yeah, you need a guy. For everything.
- [35:32]
John
You do need a guy for everything. And you will in the future.
- [35:35]
Jordy
One of the great luxuries, right? The middle class has apps.
- [35:38]
John
Yes.
- [35:38]
Jordy
The wealthy have guys.
- [35:40]
John
Yes, yes, yes. And the apps get better with the. The apps get better with AI agency, AI agents. Right. Because you have an app that acts a little bit more like a guy than an app.
- [35:58]
Jordy
Real quick, scoot in the X chat almost bought a counterfeit TVPN hat.
- [36:04]
Tomas Puig
No way.
- [36:05]
John
Watch out.
- [36:06]
Jordy
There is counterfeit TVPN store. These are not by us. They've made it look like it's by us. I'm not going to name the link, but we have not sold any merch. We will make the merch available as soon as possible. I was working on it this morning before the show, so it's coming. But do not buy any of the counterfeit merch. My big concern with that site is I don't even know if they ship it.
- [36:36]
John
Yeah, that is a big question.
- [36:38]
Jordy
And they also made like 100 products.
- [36:40]
John
They made so many products. And I did email them, and to be clear.
- [36:44]
Jordy
And our lawyers emailed them and we've submitted a bunch of.
- [36:46]
Melissa Tokmak
Take that.
- [36:47]
Cameron Winklevoss
Yeah.
- [36:47]
John
When I emailed them, I was like, hey, like, I assume, like, you're just a fan. Like, I was. I was being too nice. I was being golden retriever mode. But I did say, I was like, hey, like, I, you know, I appreciate this idea. This is very cool that you're enjoying the show. But, like, we just don't want people to get confused. We have our own plans for the store.
- [37:07]
Jordy
They're like, okay, I'll make 200.
- [37:10]
John
They just didn't respond at all. And so then we sent a takedown notice. And we will be fighting that tooth and nail. So stay safe out there.
- [37:16]
Jordy
But please don't buy it, because we have nothing to do with it.
- [37:20]
John
Tyler, did you get a chance to read Fiji Simo's latest blog post, moving Beyond One size fits All?
- [37:27]
Jordy
I hope you didn't read.
- [37:27]
Tyler
I studied.
- [37:28]
John
This was sat your ass.
- [37:30]
Tyler
We talked about this for a tiny bit yesterday.
- [37:32]
John
Yes. Yes.
- [37:33]
Tyler
This was just the 5.1 release.
- [37:36]
Jordy
Yeah.
- [37:37]
Tyler
Nothing, I would say super substantive in it. She kind of is talking about how I think with 5.1, they were going.
- [37:46]
Jordy
Like, we made our digital guide faster, better, stronger.
- [37:52]
John
Is that what it is?
- [37:53]
Tyler
The EQ of the model lot rather than iq? So that's why you see a lot less benchmarks. I think it's just hard to actually benchmark that kind of stuff.
- [38:00]
John
Yeah.
- [38:00]
Tyler
But the actual style of the model talking about kind of safety ish stuff where there's like, you know, I mean.
- [38:08]
John
It is crazy following this company so closely because in here there's a line that says with more than 800 million people using ChatGPT, we're well past the point of one size fits all. And 800 million sounds amazing. Except I feel like I heard the 800 million number like two months ago and I feel like they have been accelerating so fast you expect them to be 50, 900. Exactly. And so the fact that they're repeating the 800 number is like.
- [38:32]
Jordy
They're like, sorry, we can't add a third of the United States every month.
- [38:36]
John
I know, I know, I know. It's very, very high stakes. It's very, it's very impressive. They built, to be clear. But I just, I am really keyed that like 800, 800, because I was, I was excited they were going to hit a billion. It was going to be a big moment. And yet it feels like maybe that's a next year goal. But Near Cyan has been going back and forth on this. Near said ChatGPT is officially in its Fiji Simo phase. If you're wondering why the upgrade doesn't come with benchmarks, have fun. Run says you are confidently wrong about the internal dynamics of this. It could be better summarized as an infra cleanup. Up and near says. The source for my top tweet is Fiji's blog post from today which discusses the release and its goals. I don't really know what else to say. I don't know. Is there hunger for benchmarks anymore? I might actually take the other side of this here and say that I like that they're getting away from benchmarks. I wish they didn't do a five point. I don't want any more confusion. What is 5.1 versus 5? Just make it better and don't do a release and certainly don't tell people because what if.
- [39:52]
Jordy
Easy for you to say because you're not in love with a specific version. John.
- [39:56]
John
I'm in love with five. I'm in love with five. Rune, bring back five. I don't like 5.1. I need five specifically. Not four. Zero, not 5.1. I need five.
- [40:06]
Jordy
Five, please.
- [40:07]
Tyler
I will say when.
- [40:08]
Jordy
Just put the five in the bag. Run.
- [40:10]
John
Yeah. Come on. Bring back five. Bring back five. We need. Need to cyberbully Rune until we bring back five. Even the most minor tweak to the model is unacceptable.
- [40:20]
Tyler
You can still use 5 Pro.
- [40:22]
John
I know.
- [40:23]
Tyler
So that's, that's got to count for something.
- [40:25]
John
Only five.
- [40:25]
Tyler
Thinking, okay, but, but I was saying when GPT4, like GPT4, like, not four or anything. When that was the best model, they would do updates. They wouldn't like, say, oh, this is a new model.
- [40:36]
John
Yeah.
- [40:36]
Tyler
And people could definitely tell. Oh, sure, like, okay, they released a model, it's worse. And then everyone on Twitter would hate it. But then I think.
- [40:44]
John
So you think putting a version number actually helps fight back against that? Because people are like, oh, I get it why it's worse. You changed it. Instead of there being a surprise under.
- [40:55]
Tyler
The hood, I think it's more. It's just easier for people to tell it that it was actually a change when they're noticing something that they've been depending on. It's a little different now.
- [41:06]
John
Yeah, I just don't understand why you're surfacing it in the UI of. Of A. If I open my ChatGPT app at the top now, it says ChatGPT 5.1. Like, this is a consumer iPhone app. Today's Pulse is here talking about Blue Al Stargate Investment and ChatGPT 5.1. And I just have to wonder if like the 5.1 is like unnecessary. Like, if I open up Instagram, it does not tell me what version of the REELS algorithm I'm on. They're going to change it every. Every day. Just change the algorithm all the time. Just make it better. And yeah, if you make it bad, I'm going to churn. So don't do that. Make it better every day, forever, and just keep shipping ship every single day. I'm sure that internally there are version numbers of Google search, right? Because they push to a main GitHub branch or something or whatever they use for their monorepo. But there is version tracking for the reels out there algorithm. They just don't surface that to the user. So I don't know why they're surfacing 5.1 to users after there was like so much pushback over 4.055. All this other stuff, it seems like, I don't know, it seems like a mess. You know what doesn't seem like a mess? Vanta Automate compliance, manage risk and accelerate trust with AI. Vanta helps you get compliant fast. And we don't stop there. Our AI automation powers everything from evidence collection and continuous monitoring to security reviews and vendor risk.
- [42:33]
Jordy
There was one more note from Alex Heath's article on OpenAI that actually I think is worth sharing. He says after Meta's last earnings call, sources say, and this is confusing because the name of sources is I don't.
- [42:47]
John
Want to hear what Alex Heath has to say specifically give me some like people who are close to the matter. I don't, I don't really. I don't want to know what sources says. I want to know what sources.
- [42:56]
Jordy
What sources.
- [42:57]
John
People close to the matter.
- [42:58]
Jordy
Sources says that sources say. Okay, sources says that sources say CEO Mark Zuckerberg joined an internal employee Q and A and shared a warning about the AI bubble. First, he shared a breakdown of how different players from startups to big tech names like Meta, should think about timing their bets. He described three camps in the industry optimists who see superintelligence emerging within two to three years, moderates who expect breakthroughs by the end of the decade, and pessimists who think it'll take well into the2030s. Each outlook, he said, dictates how aggressively a company is invest. Then he expound expounded on a version of the answer he gave me recently. In our last interview, he noted that while unprofitable startups like OpenAI and Anthropic Risk bankruptcy, if they misjudge the timing of their investment, Meta has the advantage of strong cash flow. He also made the point that while big tech has historically been relatively debt debt free compared to large companies in other sectors, the AI infrastructure race is leading Meta and its peers to start using leverage in a more normal way rather than relative to their size. Like he told me in September, Zuckerberg acknowledged to employees that met his market cap could suffer if his timing is wrong and the bubble bursts. But the message was clear. We'll have the balance sheet to survive and emerge stronger than most on the other side. So anyways, Super Dario was quoting that and said the obvious end game in the next two to three years is that Microsoft acquires OpenAI, Google acquires Anthropic and Tesla acquires XAI. Only the large caps survive.
- [44:28]
John
That's a nuclear hot take. That is crazy. Crazy. How would that even? I don't know. Can Microsoft get the rest of Open AI? I mean I guess they probably have.
- [44:39]
Jordy
Depends on the price size.
- [44:41]
John
It's a $4 trillion company versus a 500 billion.
- [44:46]
Max Hodak
Yeah, I don't know.
- [44:47]
John
It doesn't seem like impossible. Let me tell you about graphite.dev code review for the age of AI. Graphite helps teams on GitHub ship higher quality software faster. Get started for I want to run through some more of these posts. Yujin Jin says in contrast, OpenAI employees stayed for two plus years sold $6.6 billion of equity last month. Many hit the $20 million cap. Morale and vibes are high, but so is the turnover rate. New OpenAI hires are often shocked by how many slack accounts get deactivated each day. This is a screenshot of an interaction between Jack Morris and Liang Chen Luke at Xai Jack says there are dozens or perhaps a couple hundred. X OpenAI, XAI, Google, DeepMind researchers, founding companies in the current climate and this was talking about the simple answer. The liquidity of anthropic options is the worst among those frontier levels.
- [45:43]
Jordy
This is talking about how a lot of people have been leaving various labs, less people have been leaving anthropic. And so Leong Chen is saying the simple answer.
- [45:53]
John
Yeah. And so Andrej Karpath he says bullseye. It's interesting how large of a fraction of people don't see the dominant first order term that drives behavior of people in companies. You can construct a powerful world model just by understanding one, just by one, understanding the system and two, assuming there is only this single term like liquidity, how much cash you have. What would be interesting is if you could is if companies started offering liquidity in the form of annuities. So imagine you have an employee who's like a rock star. They're going to sell $20 million of stock and they're going to basically be post economic. If you could instead say we're going to be paying you out like you're selling now, but you get a billion dollars a year or something. There needs to be some way to sort of cap. Is there some way to cap. Cap the actual amount? I guess 20 million was the cap, but I don't know. There's some way to deal with this. Like if you don't get employee liquidity, they'll leave for something else. They'll just go somewhere else that pays them a higher salary. If you give them too much liquidity, they'll leave and start new companies. Very, very tricky to manage the team. But that is the nature of these, these companies.
- [47:22]
Jordy
Chad buyers. That is his real name. He is a chad in the literal sense and figurative sense. He says one of my strongest beliefs is that it's going to take 20 plus years to get AI penetrated into the real economy. I filled out a piece of paper at the doctor's office last week.
- [47:41]
John
I filled out a piece of paper at the doctor's office last week too. It was crazy. And I was wondering like when will we see a fast takeoff in Docusign? Fun. I finally realized why DocuSign has so many employees because you need to go to every doctor's office in person apparently for decades to get them to use online form filling technology like General SAS really does not. Has not permeated as much of the economy as people think. A lot of people still on spreadsheets for all sorts of stuff. A lot of people still on. On paper and pencil. There is a. You know we joke about being pro ramp anti paper receipts. Of course there's a company that makes paper receipts that's worth $20 billion. $20 billion. There are fax machine companies. The fax machine industry is still over a billion dollars. Still a billion dollar industry. Crazy.
- [48:30]
Jordy
I would think it was actually more.
- [48:32]
John
Yeah, maybe that's, maybe that's too small. It's sort of hard to like calculate because a lot of these things have been like rolled up into to other companies.
- [48:39]
Jordy
And big facts. Big facts wants you.
- [48:42]
John
I think Canon is one of them. And so I don't even know if Canon breaks out their fax business anymore because they sell so many cameras and other equipment. But yeah, it's interesting. NIR says imo, in my opinion the entire AI field switched from explore to exploit two years earlier. Everyone convinced themselves no, this isn't the case. Look at our exploration and it's like watching someone go on a 50 foot walk and find a cool tree when the entire continent is still covered in fog of war. Now that the terrain seems known, it should be harder to convince yourself. I suppose this makes sense given a lot of people hint at being good as gone as soon as they have enough money. But no, not me. I've been gone for ages already. That's a very funny post I suppose. Weren't we talking about this yesterday? This idea of like where will the next innovation come from? Where will the next breakthrough come from? Will it come from any of the. Will it come from Xai?
- [49:47]
Jordy
Will it come from DeepMind?
- [49:49]
John
Yeah. How much do you need the college campus? How much do you need that environment?
- [49:54]
Jordy
Yeah. Will it come from a university?
- [49:55]
John
A university the universities seem to have not like it's very odd that the university system did not produce the transformer paper. Feels like the perfect thing to come out of a university setting.
- [50:04]
Jordy
Yeah, I mean it's really tough right now. You can stay in a university system and be a student and be taking on debt or you can go work at a lab and make have a good shot at least if you did this a few years ago, have a good shot of making $20 million in a few years and it's hard to give up that kind of opportunity.
- [50:21]
John
Yeah. This Wall Street Journal article has given more context on the AI boom. Says the AI boom is looking more and more fragile. AI st swung downward as doubt rises about sustainability and payoff. Perfect isn't good enough and any sign of weakness is a disaster. This is what's happening. It's like you double revenue and your stock trades down. It's very, very odd.
- [50:41]
Jordy
But everything core weave, who again is the only neo cloud in the platinum tier semi analysis is down 45%.
- [50:53]
John
That is remarkable. It's like they've built a like by all accounts, fantastic product. Product. Fantastic product. I mean like, I don't know, maybe Samuel got it wrong, but I don't think so. But it feels like they built something that as infrastructure delivers at the level of the hyperscalers, just like a fantastic product. And yet the market like sort of ran away with that narrative and now it's pulling back a little bit. So recent history suggests that the gloom won't last. But the shakeup serves as a strong reminder that the early years of AI pose a challenge for investors accustomed to measuring returns on a 12 month time horizon. Generative AI services require massive data centers and state of the art chips and server racks that don't come together quickly. The companies at the heart of the of AI are now talking about years, plural, of all major investments still ahead. So everything has kind of sold off a little bit. Oracle is down the most from its three month high. Nvidia is down a little bit. Google is neck and neck. They're doing great. And oddly, Apple didn't even make the chart because they're not, they're so not indexed to AI right now. The latest episode of fragility started last week when shares of some of the sector's leading lights lost ground after a broad based recovery. On Monday on news of a possible end to the government shutdown, AI stocks fell again Tuesday. Nvidia's down 4% today, lost 7% last week, slipped another 3% on Tuesday. Tuesday, while leaving it well shy of its $5 trillion market cap.
- [52:28]
Jordy
Looking at the trailing 12 months, Apple is up 21%. And Microsoft, which owns a third of OpenAI, is a huge AI beneficiary, has invested a ton in it, is only up 18%.
- [52:40]
John
Wow.
- [52:41]
Jordy
So Apple, which has been going through.
- [52:43]
John
This is the, this is the secret. Just don't invest in AI.
- [52:46]
Jordy
Do nothing.
- [52:47]
John
Just don't do it. Just skip it entirely.
- [52:50]
Max Hodak
No.
- [52:52]
John
Just do nothing.
- [52:53]
Jordy
Cook's like, wait, why would I spend 100 billion on capex. I can. He's like, I can sign up for chat for like 20 bucks.
- [53:01]
John
So hard. She's like, yeah, you know, we have this, we have Safari, we have, we have this web browser. You can go to use AI from there. Just do that on your phone. That would have been the correct thing. Instead of the getting over their skis a little bit on the branding, fortunately not on the financial side. So they've done very well. There is, of course, real reasons to worry about the sustainability of the boom. Chief among them is that there's far more AI computing infrastructure spending than there is AI revenue, a gulf widening by the day. OpenAI says planning to spend $1.4 trillion in the next eight years, but is only pulling in around 20 billion of annual revenue today. And it lacks a clear business model to reach the hundreds of billions of it needs within the next few years to keep the spending growth going. OpenAI is projecting losses will swell to 74 billion in 2028. So skittish has the mood become that CEO Sam Altman felt the need last week to defend the company on X, saying the spending was understandably causing concern.
- [54:02]
Tomas Puig
Wow.
- [54:03]
John
He says he understands your concerns, Jordy. He pointed to his plans to boost revenue with new consumer devices, robotics efforts. Efforts, an AI cloud computing service, none of which currently exist.
- [54:14]
Jordy
And this is why, when we were. The Monday after that interview when we were talking about it, saying that wasn't a strong answer. Because all those things seem like businesses that will lose a lot of money even if they're successful until they can reach some huge scale. Look at Meta's efforts in hardware. Look at, look at early days of any hyperscaler. Look at any robotics company. Right? These are not cash engines. They're cash incineration engines that could one.
- [54:46]
John
Day, if they want to get more cash generation, stop incinerating so much cash. OpenAI should. I know they're doing a lot, but they should expand into just rolling up H Vac businesses. Just buy a bunch of H Vac.
- [55:02]
Jordy
Adding agents into the workflows.
- [55:03]
John
Don't even. Agents. We don't even need to do that. Just buy a good durable business and roll it up. Plumbing, electrical, roofing, storage units, storage. There's a lot of good money in storage units. If they could get into some storage units, just buying a storage facility and.
- [55:22]
Jordy
Then just cloud storage. Storage. Storage units that look like clouds. Yeah, cloud storage.
- [55:31]
John
Lawn mowing business businesses. They could get into some. A bunch of lawn gardening businesses. There's a whole bunch of opportunities.
- [55:38]
Jordy
Landscaping.
- [55:39]
John
Yeah. I mean just buying multi family homes, just buying some multifamily single family, single family homes, getting the rent payments. They put the money in, they buy the house and then they get the rent payment and that's how they make the money. And I think that could be. It's a proven business model. Like we know it works. It works for a lot of people. A lot of people. People, they start with one single family home, they grow it, they keep.
- [56:03]
Jordy
Box of Orange is an OpenAI Property Management LLC. So they have the property management company and they actually own the properties with another. And so they can actually play both sides, play both games.
- [56:18]
John
They could get into drop shipping. Think about it.
- [56:22]
Jordy
They could set up a TVPN merch store.
- [56:23]
John
They could set up merch stores.
- [56:25]
Jordy
Counterfeit merch store.
- [56:25]
John
Sure, for sure. That's what you know. They keep talking about agentic commerce in the chat GPT app. Imagine like you go there and you're just like, I need some, I need some, need some T shirts. And it just, it just instantly sends you some T shirts. They're getting into drop shipping. They could also launch a course. They could launch a course.
- [56:43]
Jordy
How to build an AI startup.
- [56:45]
John
How to build an AI startup. 2000 bucks. Buckle up buddy. Get ready to pay two guys grand. Sam Altman's already been driving around in hypercars. You know, he has the garage. He has the garage for it.
- [56:59]
Jordy
To be a course bro.
- [57:00]
John
To be a course bro. That would be great.
- [57:02]
Jordy
Actually, he has a much better collection.
- [57:04]
John
He really does. I would pay for his course. Honestly, I would 100% pay for this.
- [57:08]
Jordy
Guy in a McLaren in a P1. And he's telling you, I will teach you to be rich. I will teach you to build an.
- [57:15]
John
AI how to do deals from Sam Altman. I would 100% pay 2 grand for that course. Course. I'm not kidding at all. Like 100%. It's worth it. That would be better than any college course ever. Be incredible. Well, I'm glad we're having a good time. Let me tell you about Julius the AI data analyst. Connect your data, ask questions in plain English and get insights in seconds. No coding requirement. Wired, Julius.
- [57:46]
Jordy
Julius has an out of home campaign right now in snl.
- [57:48]
John
It's really good.
- [57:49]
Jordy
That says ever since I was young, I, I wanted to transform data into actionable business insights.
- [57:54]
John
It's the best. It's the best.
- [57:56]
Jordy
Fantastic. Great work.
- [57:59]
John
Meridu says if you're down today, you're a certified beta bubble boy. You literally bid up sandisk. Wtf. Alternatively, you can call yourself a bad beta B I T C H People are having a lot of fun, Dario.
- [58:16]
Jordy
Sandisk is only down 15% today.
- [58:19]
John
So Sandisk by the address.
- [58:21]
Jordy
Yeah. The White House last night tweeted we are so back in all caps.
- [58:26]
John
What was that? What did that mean?
- [58:28]
Jordy
What did they mean by that? In what way were we back?
- [58:31]
John
I have no idea. But I have. I have. I have news I need to share with you that I can't share in the stream. But let me tell you about Fall a generative media platform for developers. The world's best generative image, video and audio models all in one place. Develop and find tuned models with serverless GPUs and on demand clusters.
- [58:51]
Jordy
In the chat says the OpenAI McLaren Museum. Open up a museum?
- [58:58]
John
The McLaren Museum. Just tickets just like adding to our revenue like we're going to charge 25 bucks. You can bring your kids.
- [59:06]
Jordy
I would visit, I would buy the course, I would go to the museum.
- [59:09]
John
Yes. Well, our first guest of the show is in the restream waiting room. Let's bring him into the TVP and ultradome. How you doing Spencer?
- [59:18]
Tomas Puig
Good to see you.
- [59:19]
Jeffrey Katzenberg
You.
- [59:20]
John
What's up guys? What's up?
- [59:22]
Jordy
I hope you haven't been watching the last five minutes because we're having.
- [59:25]
John
We've been all over the place.
- [59:27]
Spencer
I want you to know that I have been watching and guys, we are so back from the White House. That's got to be a reference to the government shutdown.
- [59:35]
John
Oh yeah.
- [59:36]
Jordy
Okay.
- [59:37]
John
Okay. I thought it was.
- [59:38]
Jordy
I forgot about the government.
- [59:39]
John
Lost in the markets.
- [59:40]
Jordy
I forgot about the government.
- [59:41]
John
I don't think about the shutdown.
- [59:42]
Spencer
Right.
- [59:43]
John
Okay. Well that's fantastic. For those who don't know you, introduce yourself a little bit. Give us a little backstory.
- [59:49]
Spencer
Yeah. Everybody, I'm a general partner CO2 help lead our growth fund focus on late stage transcendent technology companies. So we work with companies across of course software, AI obviously and then also some stuff in hard tech space X and oral companies like this. So we're 25 year old hedge fund launched a privates business 15 years ago ago. We're about 70 billion of AUM today across. There we go.
- [60:18]
Jordy
Where's the mallet? We're looking for the mallet. There it. It's in front of your computer, John. There we go. This is for CO2. We are thrilled to have you. I've had many fun conversations this year. First one on the air. The reason for today's appearance is none other than cursor.
- [60:39]
John
Oh yeah. What's New in Cursor world, right?
- [60:42]
Spencer
The GPT rappers are, you know, their demise was maybe greatly over exaggerated on X over the past few years. Look, we've gotten no cursor for a long time. We just think it's a really special company. Michael and his team are incredible. I'm sure maybe you guys saw Bree's piece about the company, the unique culture, how they approach hiring. Have just never seen a company with such a deep bench of really talented young folks. And then look on the growth, the growth speaks for itself, right? I mean we're seeing folks on Twitter today talking about the company being the fastest ever to a billion of revenue.
- [61:22]
Jensen
Right?
- [61:23]
Spencer
It's definitely in that air which they disclosed today being at a billion of revenue. And then if you think about the things that have transpired over the past year with Anthropic coming out with an excellent product in Claude code, Open Air responding with Codex over the summer, again an excellent product. All these things happening at once. And then you sit here and you wake up today and say, wow, Cursor's got an incredible business, incredible product velocity and incredible team. And now their own model, which they shared yesterday, is their number two most used model on the platform in Composer. So, you know, there's a lot of good companies in San Francisco. It's kind of a nice thing. Thing like in early 22 everyone was very worried that all of our 21 investments and everything was just going to kind of wipe out. There's actually a lot of good companies in San Francisco. It's a great thing.
- [62:12]
Jordy
There's a decent narrative violation, right?
- [62:15]
Spencer
There's a lot of good companies. There's a few great companies also, but there are very few of these companies that the word of the day or the word I try to always come back to is transcendental. And you kind of know it when you see it or you know it when you're around it. And that's what we think Michael and the team are doing at Cursor.
- [62:36]
Jordy
Funny, funny story. Cursor is. I'm incredibly inspired by their growth and user love and everything they've accomplished over the last few years. But I had a painful moment when they first kind of exploded onto the scene because I realized that a moment on one of the co founders had been following me since like 2020 or something. I was like, I, I had this, I've had this like, you know, lingering fear over the last few years of like, you know, getting followed by somebody not reaching out, not investing and certainly, certainly would have been smart a Few years ago to, to.
- [63:16]
Spencer
Well, listen Jordy, it's not too late. We think there's still a lot of upside, you know, so we'll see. Fingers crossed.
- [63:23]
John
One of the frameworks that I've heard kind of bandied about around what's going on. Like, a lot of people maybe want this to be more of a winner take all market than it is. And yet maybe because of Jevons paradox or just the nature of like, it's very hard to understand how much code does humanity actually want to write? Like, it's potentially a ton. And so you're in this like entire entirely new market that's a blue ocean and it's massive and it's just growing, growing, growing. And so there's opportunity. And I think you're in a unique position where you're in a number of names that do overlap somewhat and there's a lot of opportunity. I'm wondering if you think that's the right framework or is there more of like a winner take all Monopoly thesis with this particular market or if there's other historical anecdotes that you go to kind of understand how this might play out.
- [64:28]
Spencer
Yeah, I'm not sure what the best. I mean, I'll get to the anecdotes. It's hard to find the best analog and analogs are always difficult, especially for this. We're investors in OpenAI. Anthropic and Cursor.
- [64:40]
John
Yeah.
- [64:40]
Spencer
We're also investors in Glean Harvey Open Evidence. A number of companies across the kind of AI stack. Our view. Right. If you just take a step back. Back. Right. People talk about the bottoms up tam sizing on developers. To your point, people want to write a lot of code. 30 million developers in the world. Right. A fraction of a percent of the global population today. And so what's that number look like in 10 years, 20 years? What should that number be? I mean that's not really an answerable question, but probably much higher. And then if you take a step back, tops down, you've got 5 trillion of globe global IT spend and about a third of that is labor today. So hey, OpenAI is putting up historic growth numbers, Anthropics putting up historic growth numbers. Right. All that's, well kind of directionally known. We see all these companies growing really, really healthy and doing it in a way that we think is pretty sustainable.
- [65:40]
John
Did you have a reaction to Satya Nadella on Dwarkesh X? Dylan Patel? There's this interview and he kind of about it feels like he was kind of sending the signal I'm going to steamroll you if you try and build an XL agent. I'm defending the castle that is xl, but also I'm a platform for a bunch of other things.
- [66:03]
Jordy
XL co pilot.
- [66:04]
John
Yeah, exactly. And I think that I would bet on him being able to defend the castle that is accel. But I was going back and forth with Jordi about it feels a lot harder for Microsoft to put up a fight in AI for lawyers because they just don't have a practice there. And yes, lawyers use Word documents, but it just feels like a slightly different go to market slightly different product. And so I'm wondering on how you think about the edges of like where the hyperscalers or the neo hyperscalers, the foundation labs might stand Steamroll versus not because Cursor is an interesting example of a company that feels like almost steamrollable and yet hasn't been right.
- [66:51]
Spencer
Hasn't been right. Again, if you look back at the past, if I told you, hey Claude, Code is going to be this amazing product that's this good and you can use it, you think, oh my God, right? Cursor good. Like I'm short and cursor.
- [67:02]
John
Totally.
- [67:03]
Spencer
What a lot of people. So again, I just think it's. We get to sit in a unique spot where we see all these things happening and go, wow, we're just so early in of terms terms of the pie growing, the tide rising with all boats. I think I messed up that somehow. But to get back to your question around the hyperscalers and who wins where? It's hard to say exactly. I think that's kind of the big question right now. If you look back at like the beginning of the year, I think there was an open question around are, are these API businesses going to work? Is there long term market margin in the API business? Is something like Cursor just going to get run over? Is something even like anthropic in the API layer going to get run over and squeezed? I think now people are kind of going, well, these businesses look really good. They've scaled really nicely despite all these things to your point, happening at once maybe. Now the question is what's the architecture of that market look like with the hyperscalers over the next 10 years? It's hard to know. What I would say is if you take a step back and you look at the cloud AI, record revenues for the, for the three hyperscalers, I mean they've been inordinate beneficiaries of this trend, right? Open anthropic Cursor. Harvey, all these companies are great. The hyperscalers have done really, really well. And the move to cloud and the move from on prem to cloud that's being pulled forward by all this is really, really compelling for them. So like they don't need to worry about winning in our view. Right. Like lawyers in beating Harvard. Like that's, that's almost, it's just further down the economic ladder in my mind for them. But you know, we'll see the. And then the other point. Oh, you asked about Excel.
- [68:48]
Tomas Puig
Yeah.
- [68:48]
Spencer
Excel is fascinating, right? The ultimate like dominant software product over many decades. We've been looking for kind of the cursor for Excel for a while. We've tested a few products. We've got an internal team here that builds a bunch of things that we.
- [69:04]
Jordy
Guys have built your fair share of models over the years, I would hope.
- [69:09]
Spencer
We do a lot with, with data science and now increasingly with AI. We haven't quite found something that works yet, but there are a few folks building some interesting companies that we're not invested in them yet. So I'm not going to say what they are, but it's an interesting opportunity. Could I think that more likely than not, Microsoft still ends up owning a lot of.
- [69:29]
John
I mean the closed source thing is very important. Like there is no VS code in that category. People talk about this with cursor for biotech, like what's the lab notebook? And I was pushing on like I was trying to go deep with an investor who kind of came at me with a thesis of cursor for bio. And I was saying, okay, well if you're going to invest in the cursor of bio, what's the VS code of bio? What's the open source? What's the open source standard that you can find fork that every bio lab uses? And they're like, oh, well actually the labs use closed source stuff and those enterprises do not just want to give up the data. So you're going to be fighting for the walled garden and all this other stuff. So it does feel like the cursor for X model, I mean it was really hot at yc, what 2 batches ago everyone was doing cursor for X and it feels like it was wishful thinking in some markets.
- [70:24]
Spencer
Yeah, I think that's exactly right. And what's interesting is for biotech in particular. Right. And we've heard Sam make more public comments recently. Right. About their effort with the sciences and how much time they're spending there. Yeah, I just think that's where the value ends up accruing is something like an open anthropic probably throws their hat in the ring at some point. There's Isomorphic Labs coming out of Google. There's the Chai discovery business, which is a compelling company. There's a number of folks who are kind of building, building these pseudo lab companies that work with biotech. I think that's probably what ends up being more interesting because to your point, there's not like a VS code for biotech today, and there are just very few markets where there even is a VS code. Analog. Right. Part of the magic of Cursor is they identified an incredible market where the models worked really well early, which was coding. This has been the frontier of model capability ability for a while. You had the ubiquity of VS code and of the terminal and all these things, and then you just built a really nice product around it and have continued to just push the frontier of what's possible. And now you layer in their own model with Composer. So yeah, we're excited about it, but it's hard to find markets with the same anatomy and structure and kind of opportunity. To your point, people are kind of squinting and like kind of forcing ideas a little bit.
- [71:47]
John
Totally, totally, totally. Yeah.
- [71:50]
Jordy
Jordy, I'm sure, I'm sure you guys just try to back visionary founders that don't come to you for help unless they, unless they really need it. Hopefully. But how do you think private companies. How much attention should private companies in AI be be paying to the public markets right now? Because we're seeing so much growth, revenue growth in the private market markets. And at the same time, you know, we were just, you know, core weaves down 45% in the past month. There's a lot of jitters in the market right now, and it's hard not to pay attention to some of these signals even if, even if a company's own revenue growth is skyrocketing. So I'm curious how you think the best companies are going to kind of manage through the next 12 months.
- [72:38]
Spencer
Yeah, I mean, look, the fortunate thing for the best companies companies is almost all of them have a war chest in terms of capital on the balance sheet today. And I think the capital demands of these businesses is a bit exaggerated relative to what, what, what it is would be, maybe my personal view relative to like what's on X or what's in the media around these AI companies.
- [72:59]
John
Yeah.
- [73:01]
Spencer
In terms of like paying attention to public markets. Right. I think it's, I saw like Subu tweeted I think a few days ago, something around. Right. Like the market's been above its 50 day moving average for like over 100 straight days. Like something like 130 straight days or something like this. So you know that the market's been good for a while. When you see a stat like that, I think founders are aware of that and are pretty aware that it's a good time to raise capital and fortify a war chest. Right. Which we've had a really busy Q2 and Q3 and now early Q4. I think that's kind of the way to think about public markets is, hey, when's a good time to raise capital? When my, when my cost of equity is lower. But outside of that, like, I just think tunnel vision is everything. And again, just to, to bring it back to Kershaw, like this is a very focused company, very heads down company, doesn't make a lot of appearances. Right. Like just stays in their lane and focuses and I think that that's kind of the ultimate value. Right. You don't want to be spread to the in. And look, there's a lot of folks like myself running around San Francisco now looking to give founders money and tell them when the value of their equity is pretty good. Right. So we help give them that signal.
- [74:16]
John
Yeah. There's also this interesting. It's hard to measure corrections in the venture ecosystem because they're much less quantitative than just looking at. Oh, okay. You could, you know, map the.
- [74:30]
Melissa Tokmak
Yeah.
- [74:30]
Jordy
Correction often looks like, wow, that company hasn't raised in three years.
- [74:33]
John
Yeah, yeah, that. Or you know, there's some hiccup in LP fundraising or something. But there, there's there's many corrections in the public markets that I can think of that were like hiccups in the private markets. But for the most part it was everyone just kind of being like, oh, there's some crazy stuff going on. Okay. Like let's cancel a couple meetings and okay, we're back on two. Two months later. Like the COVID 19, like sell off was like a massive correction in the public markets. And for most, for most startups it was like, yeah, you needed to understand what could your company continue or were you in like hospitality or something that was going to be heavily affected. But for a lot of companies it was just like, oh, your March raise turned into a June raise over zoom and it was fine.
- [75:21]
Spencer
Yeah, I mean, I think that's spot on. Right, Right. Like there is a level where.
- [75:26]
Max Hodak
Right.
- [75:26]
Spencer
Because we do have a public hedge fund. And so what's interesting is when you have conversations like the one you guys were having prior to this.
- [75:32]
Jeffrey Katzenberg
Right.
- [75:34]
Spencer
You know, as a hedge fund, you're always behaving in the market. Right. There's always a choice to be making. Right. It's very unusual to be in 100% cash or something like this. Yeah, right. Is a private fund, is a growth fund. We can just choose to not invest for a while. Right. We can just wait a bit. And for us, like the convenient thing is, you know, we only make, I think this year we've maybe added about half a dozen new logos to the portfolio. So for us we're only making three to five core big investments a year. On the growth side right now is kind of our cadence. So it helps us kind of work through whatever that cycle is and just focus on the asset, focus on the entrepreneur, focus on what we view as the 10 year trend in store for why this company is going to be so durable and so powerful over time. But yeah, like you don't see it. Right. I think the other thing that's maybe like less discussed is we've had a bunch of companies and I've seen companies outside our portfolio too that were kind of your 2017 era SaaS, businesses that have really seen a nice bounce back and a nice acceleration over the past few years with AI really like reinvention is the wrong term. It's just kind of like some product extension in pull.
- [76:43]
Max Hodak
Yeah.
- [76:43]
Jordy
The CEOs and the management teams and the teams are just re energized and excited because there's this new capability that you can integrate across your entire platform.
- [76:53]
Spencer
Totally. And if you, if you were some, if you were some, you know, flavor of a source of record, your ability to now use that data and create economic value around it is just inordinately better. So yeah, I mean, look, the market's been up for a while. It's definitely interesting.
- [77:11]
Jeffrey Katzenberg
Yeah.
- [77:11]
Spencer
But for us on the private side, it's a little bit easier because we just get to focus on the companies.
- [77:15]
John
Yeah. How do you think people are, do you think people are reading too much into Sarah Fryer's like private phone calls these days or. I mean it feels like, you know, OpenAI has like a billion users. Deceleration might be sort of like expected. Like, you know, people don't excel, they don't expect Meta to accelerate on DAUs. Do you have any thoughts on like what's going on there?
- [77:40]
Spencer
I've got a big term because I knew this would come up and so I thought of this before. This was My one big term here.
- [77:45]
John
Okay?
- [77:45]
Spencer
Anxiety displacement, guys, we've got anxiety displacement. That's what's going on. You know, like, people talk about anger displacement.
- [77:53]
Jordy
Right?
- [77:53]
Spencer
All these things.
- [77:54]
John
Yeah.
- [77:54]
Spencer
But market's been up for a while. Like, that's a. That's a. That's like a fact.
- [77:58]
John
Yep.
- [78:00]
Spencer
Chachi BT is less than 3 years old. Sure is a product, right? Yeah, it's growing really quickly.
- [78:04]
Jordy
Right.
- [78:05]
Spencer
It's. It's. They've disclosed they're almost 10% of the world is using the product weekly.
- [78:09]
John
Right.
- [78:09]
Spencer
Like, that's pretty remarkable. Right. The market's been up a while. We had, like, I still think as a society, people haven't really processed the whole lockdown thing. We had the government shut down for, you know, a month and a half. Like, there's all this stuff going on and there's all this anxiety. And I do, I think, like, that's been very displaced because people, you know, like, OpenAI is a great company. It's a remarkable business. There's just. It's hard to think of an analog for what they've achieved in less than three years is a product. And I think Sarah and Sam and the entire team do an incredible job. So people are definitely reading too much into anything and everything sometimes. I think there's a little bit of a malicious spin on some of the stuff they've put out also, just to be very candid about that. But, yeah, I mean, it's understandable for there to be a lot of anxiety in society, in the market, and it's understandable that it gets displaced onto the company a little bit, but I don't think that's fair.
- [79:07]
John
Yeah, it was a little weird. I mean, we were reacting to this idea that there's a private call with investors. Like, what type of investor is then leaking, like, potentially bad news to the press? It seemed like an odd chain of events, but there's certainly unlimited demand for. For bearish takes about ChatGPT right now because they've. They've been on such a tear every day.
- [79:30]
Jordy
People log on and ask for what's the most bearish thing about chat GPT.
- [79:36]
John
Yep. Yeah, yeah, right, Right.
- [79:37]
Max Hodak
That's like.
- [79:37]
Jordy
That's the.
- [79:38]
Spencer
That's the tweet prompt. Like, that's your X maximalizing, like, prompt right there. Probably like that.
- [79:44]
Melissa Tokmak
Right.
- [79:44]
Tomas Puig
What.
- [79:44]
Spencer
What chart can I put together or what can I do? But look, I like. We'll see.
- [79:49]
Jordy
Right.
- [79:51]
Spencer
Worked with the company for a long time, think it's an amazing company and been really impressed by what they've done. I continue to use the product every day a ton. I think 5.1, the new model is really a nice upgrade too. I enjoy it. So, you know. Yeah, I think it's, I think they do a great job and I think over, over, rotate on it.
- [80:12]
Jordy
But we're excited for 5.67, which would be only a few upgrade cycles away.
- [80:19]
John
Rolling my eyes, that one. Can't do it. Can't do the stupid six. Seven memes.
- [80:24]
Jordy
It's over. Killed it just now.
- [80:25]
John
They're over them. I like the new coinage, though, the anxiety displacement. It's good.
- [80:32]
Jordy
Let's talk about other growth funds because that's just a fun, fun topic for every investors talk about everyone else. Do you think anybody has, have you seen any other funds kind of blink yet or hesitate or kind of slow down at all A little bit worried? Because it's just like we're seeing every day we get like half a billion dollars of funding announced on the show. But that's a lag. It has like these rounds really got done in late summer, early fall, and they're just kind of getting out there now.
- [81:12]
Spencer
Yeah. I mean, I'm sure there are some funds that have slowed down. I'm sure there are funds that are deploying more than us and less than us and those things. Right. I'd say generally, I think the vibe when you're catching up with someone over coffee or lunch is a lot more positive than over Twitter. Right. And there's a lot more of like, well, how exposed am I to the AI trend? Right. Like, there are a few growth funds that feel like I've disclosed. Right. Hey, I'm a little behind this and I'm trying to catch up here and there. So if anything, like, I've felt more anxiety about that then, oh my gosh, you know, we just put more money and pick your big premier company. And you know, I haven't slept well about it, so that's not necessarily a great signal either. Right. That's kind of like an average signal. But I'm sure there are some folks slowing down, not sure who, who they are per se because we have a lot of funds that are definitely deploying a lot more than us in terms of absolute dollars and just absolute number of deals. So for us, again, we really just try to focus on our core mission, which is, hey, what are going to be companies that matter as public enterprises 10, 15, 20 years from now and in some cases a little bit, bit sooner. Right. Yeah.
- [82:29]
John
What do you think about the capital intensive pre revenue AI companies. We had Fei Fei Li from World Labs on. I don't know if they're fully pre revenue, but it feels like a lot of the World model projects, the generative 3D worlds, Gaussian splats, what Google De Mine is working on with Gemma. Or is it genie? Genie is the model. It feels like something that I just think will be like the next Roblox. Vaguely. But there's nothing that you could do to underwrite this against revenue growth. And yet a lot of these projects are like to do the next thing, we need $200 million. How would you as a growth investor even square that? Or would you just say, hey, let's just come back when we can actually see some adoption data.
- [83:16]
Spencer
Yeah, normally we would fall into the second bucket. Like generally in terms of how we operate, it's different for everyone. I think all those rounds are ultimately just a representation of the idea size, totally of the opportunity. And then of course also of the entrepreneur and the team. We did make one investment very early that maybe fits this mandate, which would be Skilled in the robotics space, which has been a great company, has executed incredibly well and we.
- [83:45]
Jordy
Luke Metro's over there now.
- [83:46]
John
Oh yeah, that's right. Yeah. We just know someone who's on the team, but not one of the founders.
- [83:52]
Jordy
That left Anduril to join. It's skilled. Right. SK I L D I D Yep, yep.
- [83:59]
Spencer
So that was, that was one where we leaned forward a little bit earlier than we would with with most things, but and again that was on the back of just being very bullish on robotics on a 10, 15 year timescale and that team and where they sat in the stack. But generally we tend to wait a little bit to see a company be a bit more fleshed out, but it's always case specific.
- [84:20]
John
That's amazing. Well, thank you so much for coming on the show. Thanks for hanging out. Well, we will talk to you soon. Hope you have a great rest of the day and congratulations on the cursor round. We'd love to see you.
- [84:30]
Jordy
It's great to see you, Spencer.
- [84:31]
John
Have a good one.
- [84:31]
Jordy
Cheers.
- [84:31]
John
Goodbye. Up next we have Tyler and Cameron Winklevoss. But first let me tell you about Turbo Puffer search. Every byte, serverless vector and full text search built from first principles and object storage. Fast 10x cheaper and extremely scalable. The Winklevoss twins are in the Restream waiting room, I believe. If not, we have a slight delay. Let's follow up on that. Go back to the time, back to the timeline. Dario Amade predicts that we will get to 90% on swe bench verified in a year. That was one year ago. It's been one year. The best performing model, Sonnet 45 with parallel compute gets 82% on swebench verified. Close to 90% but not quite there. They put him in the truth zone. RIP to Dario. What do you think, Tyler? Bearish.
- [85:19]
Tyler
I mean I think he was pretty close. It was.
- [85:21]
John
Is pretty close. Good enough these days.
- [85:24]
Tyler
That was way more of a bullish take than most people.
- [85:26]
John
Yeah, it is very impressive. Totally. Do you think it's saturated? People are saying that.
- [85:33]
Tyler
Yeah, I think a lot of those kinds of benchmarks are generally saturated now.
- [85:36]
John
Yeah, yeah, yeah, yeah. I mean it is interesting that like simultaneously you hit the 90% on sweebench did very well there. But on the flip side you have Andrej Karpathy saying like it's slop and like you can't actually use it for like the Frontier software development that he wants to do anyway. I believe we have the Winklevoss twins in the restream waiting room. Let's bring them into the TVP in UltraDome. Tyler and Cameron Winkle Voss, welcome to the show. How are you guys doing?
- [85:59]
Jordy
Come on guys.
- [86:01]
John
Hey, background, give us the update, give us the news. I know we're running late, so let's just jump right into it. I assume everyone knows who you are.
- [86:11]
Spencer
Cool.
- [86:12]
Cameron Winklevoss
So we launched cypherpunk, a ZK cashed DAT yesterday. It trades under Cyph Cyph. We're really excited about it. The mission of the company is privacy and self sovereignty starting by accumulating zcash.
- [86:30]
Melissa Tokmak
And.
- [86:34]
Cameron Winklevoss
We hope to invest in other technologies that promote privacy and self software.
- [86:42]
John
This delay having some WI FI issues. Sorry.
- [86:47]
Jordy
Say more about the structure. This is Leap Therapeutics. You guys have rebranded it. This was an existing public company. But what more can you say on how this came together?
- [87:00]
Cameron Winklevoss
Yeah, so it's an existing biotech company and we basically took it over and invested via pipe into it and then rebranded and changed the ticker today.
- [87:11]
Jordy
Awesome. So I guess zcash has had a lot of momentum recently. At the same time a lot of the dats have. Have you know, struggled in. In more recently. What like talk through kind of like the next how. How you're thinking about, you know, making sure just kind of navigating this time when the markets are pretty choppy overall.
- [87:39]
John
Sure.
- [87:40]
Cameron Winklevoss
So I think number one, the long term we really believe in and I think that when you look at Bitcoin as a store of Value or how you store your value. Zcash is really how you move your value is very sound and we're obviously very bullish on that. But in addition, we're the largest investor investor into the dad, so we don't have like a lot of fast capital that's looking for a trade. We're just long term hodlers or in the case of zcash, zodlers. And we just, you know, we plan to hold for a very long time. And I think that's one of the differences is that other dats have, you know, had faster money and people that are moving in and out of it. And that's why we didn't fill the pipe up with we took the vast majority of it. I put a $52 million check in. So the vast majority of equity will not be trading out of the shares.
- [88:39]
John
That makes sense. Well, thank you for the update. Sorry about the wifi connection. We have some technical problems. We'd love to have you back on the show and talk more about what's going on in your world because there's so many interesting projects that you're working on. But really appreciate you taking the the time to give us a quick update on cypherpunk Technologies. And congratulations on the, on the pipe closing, the rebranding, every all the progress. So have a great rest of your day.
- [89:04]
Jordy
Yeah, thanks for joining, guys.
- [89:06]
John
Bye.
- [89:07]
Jordy
Cheers.
- [89:07]
John
Let me tell you about Google AI Studio. Create an AI powered app faster than ever. Gemini understands the capabilities you need and automatically wires them up the right models and APIs for you. Get started. AI Studio, who is building AI apartment.
- [89:21]
Jordy
Agents to make the WI fi work.
- [89:23]
John
That would be cool. It would be cool if we could deploy an AI agent. I mean, I guess that's like the next generation of the Restream waiting room. Like the Restream waiting room. Somebody talks to them in a different waiting room and checks the WI fi. That will be something. Maybe it's a 2026 project. Maybe it's a 2046 project. Who knows?
- [89:45]
Jordy
That could be the final box.
- [89:46]
John
Hopefully our next guest is dialed into the Restream waiting room. We have Max Hodak from Science. How you doing?
- [89:52]
Jordy
What's going on?
- [89:53]
John
Welcome to the show.
- [89:54]
Max Hodak
Hi, guys. Excited to be here.
- [89:56]
John
Good to meet you.
- [89:57]
Jordy
Great to have you.
- [89:58]
John
I believe we met like years ago at an Oppenheimer screening. Potentially. I don't know, very possible.
- [90:04]
Max Hodak
It sounds like something that probably happened. Yeah.
- [90:06]
John
Anyway, for those who don't know you, please kick us off with a little bit of an introduction on Yourself and, and maybe the company as well.
- [90:14]
Max Hodak
Sure. So I mean I've spent most of my life thinking about how to engineer the brain. My origin story, I think in this field started when I was in the fifth grade and I saw the Matrix and I was like, I have no idea if we're living in a simulation, but we are definitely going to build one. I'd kind of broadly characterize my ambition as to eventually disappear into the simulation, never to be found. And I as an undergrad talked my way into a lab that was doing neural recording in primates. And really that was where most of my education happened. And then in 2016 I got pulled into co founding what became Neuralink. And I was there for four and a half years and in the spring of 2021 started this company, Science. We're now about 180 people. Our main product is a retinal prosthesis. Like really the first retinal prosthesis that really works restore vision. It was in, it was actually on the COVID of Time last week, which is pretty crazy experience.
- [91:05]
Melissa Tokmak
Wow.
- [91:06]
John
So you like the Matrix? Do you like the 13th floor? Have you seen, seen that movie?
- [91:11]
Max Hodak
I have not seen the 13th floor.
- [91:12]
John
Oh, the 13th floor. It's like one of those movies that got terrible reviews. But it goes a little, I feel like it goes a little bit farther than the Matrix in terms of simulation theory and simulation hypothesis. It's a lot of fun. But anyway, we can get back to the actual story. So how far are you in building this business? Give us a general update on the shape of the business. It seems like you're in the office right now. How big is the company? What's the problem?
- [91:37]
Max Hodak
Progress overall. Yeah. So we're the, so we, we have a couple different elements of our pipeline. So the, the retinal prosthesis, we think like that is first like just an end in itself. Like if you can restore vision to the blind. It is like, it is a quest that humans have been on for thousands of years. That is an unsolved problem. And I think like we've made there's like real progress on that. There's, we've got two programs. One is a retinal chip called Prima, which is this little. It's a little chip implanted in the back of the eye under the retina. Retina that has all these light sensitive cells. Works in conjunction with a laser projection glasses worn by the patient. That finished a major clinical trial last summer that was published just recently in the New England Journal of Medicine. These patients go from Being unable to recognize faces. They can walk around because they've got a little bit of residual peripheral vision. The trial was in an age related macular degeneration, but they definitely can't read. They are really profoundly blind and disabled. And the best patients in this trial could go from reading none of an eye chart to reading the entire. I mean there's videos of these patients, like filling in crossword puzzles. It's really very cool. And then separately from that, we also have another key program in the company is a different approach to building brain computer interfaces where instead of placing wires not in the retina but in cortex, instead of placing wires or cables physically into the brain or genetically modifying the brain so that there's some things you can do with optogenetics or sonogenic genetics. Instead what we do is we grow up these heavily engineered biological cells that we hide from the immune system and then we just sit this on the surface of the brain. And so we don't place anything into the brain itself. But what these cells do is they grow in and they wire up and they form new biological connections and they can form billions of them. And so I think like the way to understand this or like I think actually fairly direct reference for how to.
- [93:26]
Jordy
Think of Sean's face right now.
- [93:28]
John
This is like.
- [93:29]
Max Hodak
So we call it a bio hybrid neural interface. Have you seen James Cameron's Avatar movies?
- [93:33]
John
Yes, yes, yes.
- [93:34]
Max Hodak
Do you know the Ponytails?
- [93:36]
John
Of course I know exactly where you're.
- [93:37]
Max Hodak
Going to plug into like their tree memory store or their horses. So I think the question is like if you wanted to build a ultra high bandwidth neural interface, like how would nature do this? I think what nature would do is it would grow a new cranial nerve that has like a USB port at the end.
- [93:53]
John
Yeah.
- [93:54]
Max Hodak
And that, I mean that, that is super cool research, but that is definitely a research project. And so the way the company is architected it is that is kind of paid for by the retinal prosthesis.
- [94:03]
John
Sure.
- [94:04]
Max Hodak
Which is a very practical near term medical device. We hope to have. We've submitted for marketing approval in Europe for that. We're going through the review process now. The FDA actually is being much slower. It is possible it won't reach American patients for a little bit longer. And that's kind of crazy that Europe is going to get it first. But that's where we are.
- [94:21]
John
That's rare occurred narrative violation, but exciting.
- [94:24]
Max Hodak
Narrative violation for sure. But so yeah, hopefully that'll be on market next summer, maybe making money. And that is big enough to pay for the rest.
- [94:31]
John
That makes sense. So why go with a retinal prosthesis instead of cutting a hole in the skull and putting electrodes directly on the brain and trying to deliver the signal that way? There's a lot of BCI firms, obviously you co founded one that seem to be approaching the problem that way. Is that just like, do you, do you have a particular view on that? Is that just farther out and this is a way to get to market faster? Or is there something fundamentally like higher bandwidth about your approach? Like what are the different technical trade offs?
- [95:13]
Max Hodak
So BCI is not a product, BCI is a field, okay? And there are many different types of BCIs for many different types of applications. Like obviously you can't go into the retina to date decode like a video game controller out of the brain simultaneously. You can't stimulate like frontal cortex to do like to do like sensory feedback. And so it really depends on the type of thing that you're trying to do. And so in vision, which we, I mean we think that a visual prosthesis is a, is a brain computer interface. We also think that cochlear implants are brain computer interfaces. You don't need to be drilling in through the skull to get to corporations cortex. So if you want to restore vision, you kind of have a choice of, you've got the retina. The output of the retina is the optic nerve that goes to a deep brain structure called the thalamus. And then the thalamus connects up to cortex and so within the retina. So let's just take a look at the options that you have here. So in the retina, normally light shines in from the front. It hits the rods and cones. The rods and cones are the light sensitive cells. There's about 150 million of those. These connect to about 100 million intermediate cells called bipolar cells. And those compress down to one get like optic nerve cells that connects to about 1.5 million cells in the thalamus and that connects up to like 200 to 500 million cells in cortex. And so no one, I mean people have been trying to stimulate vision into the brain for many decades. And until the, the clinical trial that we just finished, no one had ever gotten form vision, like structured vision that the brain could intuitively assemble into a whole. Like you could get patterns that if you like looked at it carefully, it's like, oh, there's like a line here, there's a line here. Here it's connected. That must be an A here's a line, that's an N. But in the prima trial, they could read off words at a time, and that had never really happened before. And a big difference is that we are stimulating that first layer of cells, the bipolar cells in the retina. And we know that if you just go one layer deeper, from the 100 million bipolar cells to the 1.5 million optic nerve cells, if you stimulate those optic nerve cells, you don't get this. You just get these diffuse flashes of light that you can't really attend to. And the brain does not intuitively design assemble together because the brain has already compressed the signal. And you have to then figure out what is that transform, how did the brain compress this? Or how did even just the retina compress this? And so the thalamus has the same issues as trying to stimulate the optic nerve, except it's under 8cm of brain tissue under the skull. And then once you're up in cortex, you're dealing with hundreds of millions of cells that are distributed over large areas that you just can't stimulate selectively. And so if you do this, you absolutely will get flashes of light. But converting that to form vision that you can intuitively read is a totally different problem. And then even if that worked, even if that worked perfectly, one is an outpatient surgery going through the soft tissue of the retina. The other is a four or five hour brain surgery, drilling through the skull. I think one of those is. You're going to kind of win sales.
- [98:10]
John
Yeah. Wait, so you said outpatient? Walk me through. Through comparing the level of intensity of the surgery to something like Lasik.
- [98:19]
Max Hodak
So it's a little more than Lasik, but it's not a ton more than Lasik.
- [98:24]
John
What about getting your wisdom teeth out? I'm just coughing.
- [98:27]
Max Hodak
I mean, you could do this. So in the trial, many of them ended up being done under general anesthesia. But to be honest, in these cases, general anesthesia is really as much like a commitment mechanism for the surgeon and the patient that it is like there's any medical reason to do it. I mean, you just like, you can't change your mind halfway through.
- [98:43]
John
Oh, okay.
- [98:44]
Max Hodak
Commitment doesn't. Well, you. But from a, like, experience perspective. So you could do this with. You can make an injection next to the eye. The eye goes dark and numb for a couple hours, and the. And then you can go in through the soft tissue of the eye. You leave the chip, There's a little injector. The surgeon presses a lever. It leaves the chip under the eye. They Come out. They're done. And so it's a really very simple.
- [99:06]
John
Yeah. And then, sorry, just to be ultra clear, like the chip is in the eye, then how am I communicating with the chip? Is it wirelessly? Is there a device that's on the other side of my head? You said glasses maybe are interfacing with that.
- [99:20]
Max Hodak
Yeah. So if you look at this chip under a microscope, it has all these little hex cells on it, and every one of these hex cells, like the Science Prima chip, it's essentially a solar panel. And it works in conjunction with. There's glasses that are worn by the patient that has a camera looking out at the world. Although you could really get the video feed from anywhere. And then there's a laser that projects the image onto the back of the eye in the. In infrared.
- [99:45]
John
Okay.
- [99:45]
Max Hodak
And because you can't see infrared, you can't. Like, if you have residual peripheral vision or any. If you're not totally blind, this doesn't interfere with that. You can still have that. But the infrared laser, where it strikes the implant, it works like an overhead projector. Like if wherever there's white that is projected, that is like, that's energy. And wherever there isn't energy, it's dark. And wherever the laser is absorbed by the. The implant, it stimulates the cells directly above that pixel. And so this is pretty cool because the implant is powered by the information that's projected onto it as a solar panel. This means that there's no implanted battery, there's no cable. This tiny little, fully wireless, 2 millimeter chip is the whole thing. And also because the eye moves relative to the projector, the projector is shining in from the front of the eye. And then the eye moves and the image changes. This means that the brain can easily fuse it together with their existing vision. And so there's some pretty cool stuff here. If you show one of these patients a solid green bar all the way across their visual field, they'll see a contiguous bar, even though the implant only fills a small area of the total area of blindness. And they'll say it's green, and then it turns white, because we can only get black and white right now. And then it turns green again. But the brain fuses all of this together. And so even though the implant only has 400 electrodes, as the eye moves around, you don't experience the image that falls on the eye. All falls on the retina like a camera. The thing you experience is the brain activity of the world. Modeling the brain. And so as the eye is moving around, it's updating the world model and that's the thing that you see. And so even though it's like 400 electrodes, you can't think of it like 400 pixels on a screen. You think of it as just getting the information to the world model and the eyes moving around and the brain's cross referencing all that. So it actually does significantly better than you'd think from being 400 electrodes.
- [101:32]
John
That's fascinating. How do you, how do you think about the analogizing around the artificial intelligence community of what you know of the brain? So a lot of people in computer science or AI might say with, with LLMs, we've built this piece of the brain with the hard drive, we built the long term memory. Has, has what your research, have you mapped any of that onto the current state of artificial intelligence? Has it proven insightful to help for you to understand what's going on in AI?
- [102:10]
Max Hodak
This is funny because I mean at the very beginning of Neuralink and OpenAI, we were in the same building in San Francisco and we'd have these discussions about like, oh, who's going to learn more? Is AI going to learn from neuroscience or is neuroscience going to learn from AI? And I think it has been revealed that, that like I asked, I caught up with one of those guys a while ago and I was like, oh, like in retrospect, what do you think AI learned from neuroscience? He thought about it for a second. He's like the concept of a neuron.
- [102:33]
John
Yeah, that's it.
- [102:34]
Max Hodak
That's basically it.
- [102:36]
John
That's like the very, it's literally just in the name. Neuroscience is neuron and that's it. And then nothing else because yeah, like the whole brain structure concept.
- [102:44]
Max Hodak
Yeah.
- [102:44]
John
Anyway, continue.
- [102:46]
Max Hodak
But going the other way, I mean AI has been so useful. I mean there is a really interesting convergence going on. This is kind of called like NeuroAI, where neuroscience is learning, learning a lot from artificial intelligence in ways that I don't think any of us really would have anticipated.
- [102:58]
John
Okay, explain.
- [103:00]
Max Hodak
And have you come across the Platonic representation hypothesis? So there's an empirical finding that different neural networks trained with different architectures, different objectives and different concrete data sets, but for the same type of thing like images or language or audio, they produce these similar internal representations. And what I mean by rep, like sometimes you'll hear people say like, oh, these models are like stochastic parrots or they're just like, there's glorified auto completes. Like these people are safe to ignore. Like the mathematical objects that you see appear inside these models are super interesting and look a lot like the representations that you see in the brain. And so that is hinting at like there's some like, deeper fact about the universe that we're figuring out here that basically if you have a lot of compute power and you kind of run it in these ways, then you see these like, these data, these like mathematical objects kind of emerge. And what evolution did and figured out in the brain is like, looks a lot like stuff that you kind of see in these transformers and these other AI models. And there's definitely some interesting unification happening there. I mean it's not, it's still like it, it's a little more than speaking totally metaphorically, but it is still like, I'd say like instructive rather than literal. But that's getting like every month there's like some new cool thing that comes out around this. I will say that I don't know. My view is that the transformer is like a reasonably good model of what cortex is doing, but there's a lot more of the brain and so there's other parts that aren't fully cap that we're going to need something else. But it's not like the transformer is wrong. I think it's probably part of the story.
- [104:43]
John
Yeah, that makes a lot of sense. I wonder what you think about just the idea that it takes millions of years to train a model to drive a car and it takes a 16 year old a couple weekends to do it, or the amount of energy that I will consume in one day of reasoning generating my own reasoning tokens is way less than what it takes to run the data center. There seems to be some sort of like exchange ratio that's like we're off by a couple orders of magnitude. And maybe that's an algorithmic question, I don't really know. Do you have any thoughts on that?
- [105:21]
Max Hodak
Yeah, I mean evolution has done, I mean it is really, it has minimized, it's been very good at minimizing energy and optimizing some other stuff. But I mean that I think is like the advantage of the biological brains. Like every. And now and then I see pitches from companies that they're saying, well, AI is really energy intensive. So what we should do is we should grow up cultures of biological neurons and train them to do intelligence tasks. Because you can kind of do this. You can grow up neurons on electrode arrays and you can condition them to learn things by stimulating them in different ways. Like when I was in college, I grew up a counter strike aimbot. There's actually not that hard to do. And, and I, and this is a thing that I think. Nerd snipes many people in this field at some point, but I, I don't think that that's like the way to go. I think that's there's like just really structural advantages in silicon. Like if you compare and contrast these two approaches.
- [106:16]
John
Yeah.
- [106:16]
Max Hodak
Like in the, in the deep learning models, like you can see all the weights, you can introspect them, you can like stop the model, you can change one, you can replay it like samples out, you can copy to disk, you can like send it over the network. Whereas with the, the Bilo biological living, these organoid brains, you can't see the weights, you can't copy it to disk. It's like what it learns is the time integrated experience it's always had. And at some point four or five months in it will randomly get infected and die and you'll have to start over.
- [106:44]
John
Yeah.
- [106:45]
Max Hodak
And so I think there's just like structural. It's like the thing that the biology does is it's energy efficient. But my response to this is like generate more power.
- [106:52]
John
Yeah, I like the idea there's no.
- [106:53]
Max Hodak
Such thing as a low energy wealthy society. Generate more power.
- [106:56]
John
That's. Yeah. I like the idea of the solution to AI is just like have kids or something and you're like you've reversed it all the way down to like if I want artificial intelligence, I can do it biologically. Just have kids.
- [107:07]
Jordy
Are you and the team getting much leverage out of models today? Is it accelerating your progress or is it really just about having like deep domain expertise and being more obsessed with the problem than anyone else and hiring the smartest people in the world?
- [107:26]
Max Hodak
Interestingly, I think companies like this are really limited by infrastructure. And so like one of the things that I got from my prior boss was I totally received the gospel of vertical integration. And the. And we're really not, it's not like we're held back by like genius scientific insights. For the most part we're limited by like, oh like well we need to get a new, like a new material and are like microfab deposition tool. But this requires hooking up some gas plumbing which requires getting some specialist vendor to come out and weld it to the machine. Or you're out of animal housing and building. That is an architectural design process and then permitting and then construction. You're really limited by infrastructure more than your genius scientific insights. And so even if you had This I think, when I think about being in the takeoff era and what is the impact of progress in AI, we're still limited by how that can impact the real world in really meaningful ways for these assets. Adams things. But I'd say there's two places that AI has had a bigger impact. The first is ironically comprehending regulatory standards and generating regulatory documents.
- [108:27]
John
I was about to say, if permitting is the Boutleneck, can you have a permitting agent that just goes and spams the permits until you get exactly what you need?
- [108:36]
Max Hodak
Yeah, I mean they're not quite. You end up doing a lot of editing, but you definitely just don't want to spam the permits. This makes the regulators mad. But the, but I mean the, the filing in Europe for. To ask for approval for our retinal prosthesis, which we submitted last summer. Yeah, it was. I forget exactly how many. It was like tens of thousands of Pages was a 65 gigabyte PDF. Yeah. And that depends on like hundreds and hundreds of standards that we're just expected to know all the details of.
- [109:01]
Jordy
Yeah.
- [109:02]
Max Hodak
And so being able to talk to these things is super useful. Like chat with this, this, these data sets is super useful rather than like having. Yeah, like having all big data team of regulatory experts who are all kind of. To do this through meetings. And then the other place that we've had big success with AI internally is in our protein engineering program. So there's conventionally many of the problems that we're interested in you'd have to solve with better electronics or better physical devices somehow. And we're now at a point where often when we find a problem, we ask, can we make a protein as a solve this? So a couple months ago, we published a paper on a new type of optogenetic protein, which is. These are proteins that can make a neuron light sensitive that is not normally light sensitive. So we could control it with light. And these typically require very bright laser light in order to work. And we were able to use AI models to find one that is so sensitive that it is responsive to not just daylight, but indoor optimism office lighting. And so that allows us to substantially reduce the power consumption so that we can. Because often the brain implants were often limited by thermals, like how much energy you can consume depends on how much is limited by how much you can heat the brain. So if you have more sensitive opsins, then you can have your LEDs be dimmer and have more of them. Then you can think about going from thousands to hundreds of thousands. But then also that might turn actually into our next generation retinol product. That will have to go through clinical trials. That'll be a process. But it's possible that in five or seven years you won't even need the chip at all. You'll just get an injection and then we can just make the bipolar cells themselves light sensitive and then that won't even need the glasses potentially. And that really comes out of the big breakthrough on being able to solve these problems with proteins is an AI enabled thing.
- [110:52]
Jordy
Chat says we should have you ask how to explain this to preschoolers. Some labs, various groups have talked about the revenue opportunity and just automating science. And it's usually very general where they're just like we're going to come up with a bunch of ideas and then hypothetically they give the ideas to various people to execute on and they get some type of royalty on it. How much opportunity do you think there are for a more general foundation model company to just come up with a bunch of ideas and then actually capture real value from it? It feels like in some ways the pharmaceutical industry maybe doesn't have a shortage of ideas. They have a shortage of infrastructure and funding in order to test enough ideas to actually get viable solutions.
- [111:46]
Max Hodak
Yeah, I think that a real bottleneck here is just the translation to human subjects research and then to the market is really difficult. We have have like the like a thing I sometimes joke about is like we're in a golden age of mouse oncology. Like if you're a mouse with a cancer, like we've got some great things for you.
- [112:06]
Jordy
For all the mice out there.
- [112:07]
John
That's great to hear from the mice. That's fantastic.
- [112:10]
Max Hodak
We owe to them for all that they've done for cancer. But like that's the trade off. And so if you can. And the, and I get it, it's, it's not, you can't just say like, oh well the FDA is the, is the problem. Like the problem, like the problem is that human subjects research is like life or death. It's like, it's a, it's like no joke. And like I've had the experience of like a patient goes into a new surgery for the first time or they, or you're going to direct them with something and you're going to wait. And like that is a very stressful experience. Like you want them like and that's like that's stressful experience for people like me. I'm not even the one getting it right.
- [112:43]
John
Yeah, yeah.
- [112:44]
Max Hodak
And so the, there's trade offs. There That I think are very deep in like our, like the way that we in our civilization value human life, which I think is right. And so you have this knob of how much risk do you think of taking in human subjects, research and how fast do you get new things? And we know that the toolbox of science is very, very powerful because when you look at what's possible in the animal models, you have these amazing things that are possible, but then getting that into humans is like. And it's not just to say like, oh well if we, like, like oh, we should just deregulate all of this. Like there's, it's more complicated than that. Although I do think that there's a little bit of that. But at the same time, I mean certainly this, there's an intelligence effect. Like I think that there's, it's going to be inevitable that these things, I mean the fact that you can fold all the proteins, at least in static forms, is a big deal. I would absolutely not bet against improvements in AI leading to improvements in medicine and healthcare. And I think there actually just to go one step further, I think then the thing that we're gonna have to reckon with is healthcare. So like if, if you thought like, like 20 years ago TVs and phones and computers were way, way, way more expensive and now they're much cheaper, but we spend more on them total. Because this is like a technological growth industry. Like normally If Nvidia sells 20% more GPUs next year and their revenue goes up and their earnings go up and the stock price goes up and like everybody's stoked and about this. But as time goes on and there's more things to spend money on in healthcare that produce better outcomes for longer lives, there's like spending should increase. But because we pay for this, these kind of insurance schemes which are kind of pseudo fixed buckets of money. Like if, if there were real breakthroughs in healthcare that allowed people to live much longer and have much better outcomes, but they cost money and you could spend like 10 times as much on healthcare. Like that would be a catastrophe. Like you do not want to spend. Like right now our system would not handle spending 10 times as much on healthcare. Yeah, but that is kind of directly at odds with it being a technological growth industry.
- [114:47]
John
Yeah. What about the BCI industry? I think it's interesting. We've seen like this boom in quantum computing and there's public companies do all this stuff, but it feels like, well.
- [114:58]
Jordy
So specifically I feel like the interesting question with bcis is like I want to understand your timeline.
- [115:03]
John
Yeah, people aren't.
- [115:04]
Jordy
It feels like there's these therapeutic use cases where somebody's blind and they're willing to take some level of risk in order to, to see or at least see something. And then there's like the utility timeline and entertainment timeline where I just have, you know, instead of wearing glasses, I just have a screen that's just embedded in my retina and I don't need to wear glasses and it's always on and I can turn it off. And I feel like that's where, like it feels like that's where you're going is sort of a general purpose technology on a long enough time horizon. But maybe I'm off.
- [115:43]
Max Hodak
Yeah, I mean, I think, I mean it would certainly be cool if when your eyes are open you've got the world of atoms and when your eyes are closed you've got the world of bits. The near term is all these are medical devices, especially on the cortical side. I think these are very serious brain surgeries that healthy 40 year olds are not. Not going to be like if your hands work. A keyboard is a really great brain computer interface. And there's a lot of research that goes into the design of the Xbox controller or the design of the VR inputs because if you can convey the signals from your brain to the muscles, you can just capture that and that works really well. But at the same time, I think many people eventually become patients. These bodies are great until they're not. And as they start to fail, we should have better options. And so I think that bci, I think that there's a kind of a meme that it is an AI adjacent story, but I actually think it's like a longevity adjacent story. I see the only organ I really care about is the brain. As far as I'm concerned, kind of the rest of the body exists to keep the brain alive and healthy and move it around and cause it to do things. And I'm going to be fairly disappointed if I'm murdered by my pancreas or my heart. And so I think you can get to this point where you say, well, instead of you've got all these hard problems. Yeah, we've made progress on cancer and cardiovascular disease and metabolic disease. Disease. But again, when you think about health insurance, you can't insure something with a 100% loss rate. And that is what we were talking about in medicine. And I think you can get to this point where the brain is the thing that is really special and makes you, you and when I look at a person, I see an agent and a robot. And biology makes great robots, but the thing we really care about is the agent. And if you can kind of deal with the agent directly and the rest you can swappable parts, then instead of needing to cure cancer, we might just be able to avoid this entire entirely. And that I kind of see as one of the central elements of at least how I think about the promise of BCIs. And so I'd say a thing that I think could really be possible is like by 2035, it won't be widespread. I think it'll take longer than that for sure. But by 2035, can you offer patient number one the choice of dying of pancreatic cancer being inserted into the matrix?
- [117:53]
John
Yeah. What's the strongest steel 2035 you can offer for the anti transhumanism arguments? Because I mean, I've been steeped in this exact Silicon Valley lore for basically my entire life. I've heard it articulated a bunch of different ways in media, video games, the Matrix. And yet I've also seen a lot of pushback. There's a lot of. Yeah, there's a lot of pushback against the transhumanism stuff. What's been the strongest argument that's that you've had trouble debunking, if any.
- [118:28]
Max Hodak
Yeah, I mean, I don't really like the label transhumanism. It just has like all these connotations. I also don't like the word, the term. I think it is important that these things, like people don't want to be different, they want to be themselves. I think it's important to talk, to realize that things we're talking about should make you kind of just as much you as you've ever been and just like with the best quality of life you've ever, ever had. And I think a concern with AI is that it could be massively undermining our agency. It could lead to a massive loss of agency for humanity and the emergence of something else. Whereas I think that these BCI technologies are really about increasing human agency. It's really, it's a very like. Yeah, yeah. That kind of exists to facilitate you. And it should not. And it should be. It's not other than human. It's like it, it is a very pro. Human technology.
- [119:26]
John
Yeah.
- [119:27]
Max Hodak
I mean, as much as any kind of healthcare. Right. Like we try like you, we do heart transplants, we get, we're working on artificial hearts. You have dialysis, you have, we treat cancer. We like think that these Things are worth trying to improve on. And it's only transhumanist in the way that like any healthcare is that you don't accept that just like, like the state of the art in the middle Ages, if you were like playing in the wrong forest and got a scrape on a branch, you could suddenly die of a bacterial infection two days later. Like, we developed antibiotics. Like, this is only transhumanist in the way that antibiotics are. It's this continuation of that story.
- [119:57]
John
That's a great. That's a great argument. I love it. What? Yeah, Jordy, you have anything else? I was going to ask about the merge.
- [120:07]
Jordy
Yeah, I'd be interested in. Interested to get your personal definition of the merge. Sam Altman wrote in 2016 that it could be a scenario where people become best friends with the chatbot. Maybe the merge is something like AGI, where we just keep moving the goalposts.
- [120:25]
John
Yeah. Do you think it's inevitable or do you think there's like some sort of fork or where do you think the unexplored discussion service area is around the consultation concept of the merge?
- [120:35]
Max Hodak
I mean, the way that we interact with AI as. I mean, there's many different ways that this could go. I think like the, the failure mode of like Terminator seems less likely. But I definitely agree with Sam that like an underrated failure mode is people just start like delegating their decision making to it because the models just make better decisions. Like there's. I was reading an article that one of the US military, like combatant commanders is now like running personnel decisions by ChatGPT. And if you're just like realizing like, hey, like these things make as good decisions as I make, then you. They kind of come like pre merged. Like, you don't need a device for that necessarily because we are just kind of like, like causing it to happen in the world, like whatever it wanted. And there's like some dark versions of that. Like you can imagine, like an extinction scenario is that we just really become dependent. We're like, man, these things have great judgment. They really know what to do. We should ask it for advice and listen to it. And persuades a bunch of humans into suicide. I think you got to keep an eye on those rates. They won't be zero. And I think it's unreasonable to expect them to be zero because you're talking about hundreds of millions of people. But do those trends, what do those look like over time?
- [121:43]
John
Yes, yes. Keep an eye on the rates. That's the great summation of what's Going on, the baseline's not zero, but if it starts climbing up, you got to watch it.
- [121:52]
Max Hodak
Exactly. Just like self driving cars, humans are not perfect. Don't hold them to these unachievable standards. But like, keep an eye on the rates.
- [122:00]
John
Keep an eye on the rates. I completely agree. That's a great take. Thank you so much for coming by the show. Congratulations on the progress. Jordy, you have anything else?
- [122:09]
Jordy
Yeah, this was incredible. I wish we had more time.
- [122:10]
John
Yeah.
- [122:11]
Jordy
But come back on whenever you want.
- [122:13]
John
Yeah. I really appreciate this.
- [122:14]
Max Hodak
Thanks for having me on. Good to see you guys. This was super fun.
- [122:16]
John
Yeah, we'll talk to you soon.
- [122:17]
Jordy
Cheers.
- [122:18]
John
Bye. Let me tell you about profound. Get your brand mentioned in ChatGPT. Reach millions of consumers who are using AI to discover Discover new products and brands. You get a demo rofound. Our next guest is Andrew Dudham from Hims and Hers. Two companies? No, just one company. Hims and Hers. Welcome to the show. How are you doing?
- [122:36]
Melissa Tokmak
I'm great, guys. How you doing?
- [122:38]
John
Thanks so much for taking the time to jump on the show. For those who might not be familiar, can you just kick us off with a little bit of an introduction on the state of the business right now? It's a public company. People know telehealth.
- [122:50]
Jordy
Yeah, I thought you were gonna say for instance, anybody that hasn't heard of Hims, because I'd like to find one.
- [122:56]
Melissa Tokmak
You guys have some great hair. So I feel like maybe we've got some customers on the call. So Hims and hers have been around for about eight years. It's actually our eighth birthday this month.
- [123:06]
John
Congratulations.
- [123:07]
Melissa Tokmak
We went public in 2021 and the vision is to help people feel great through the power of better health. So we connect you on your phone, you pick up your iPhone, you click an app and you immediately get connected to one of a thousand or $2,000 doctors who registered in your state and then can immediately engage in everything from cardiovascular disease risk at home, testing for testosterone, menopause, weight loss, prescription treatments, all the way down to stigmatized things like sexual health and mental health and depression, anxiety. So we treat millions of patients on the platform, 10 to 15,000 patients per day. So while we are in many ways a new and innovative telehealth company, as people call us, we're actually probably one of the largest healthcare systems in the US today, just given the pure volume of patients that we treat. And it's all digital. So no matter what zip code you're in, you have access to world class consistent care, standardized care, which I think is one of the most important parts of all this digital healthcare revolution, which is no matter where you are, you get great care, as if you're 10 minutes away from Stanford and go see a dermatologist right out of outside your door. So this year we're on track, you know, two and something. Billion in revenue, profitable and growing super fast, which is exciting. Launching a whole bunch of new categories.
- [124:25]
Jordy
Hit it gong, hit gong, hit for you. Let's talk about labs.
- [124:33]
Melissa Tokmak
So labs is an important one. We launched today two packages where you, for a few hundred bucks, can get over 120 buyers. Biomarkers. This is the most advanced diagnostics, frankly, that are out there in the market. And I spend a ton of time figuring out with, with concierge doctors and others what is kind of at the leading edge where you can get twice a year annual testing and then have all those biomarkers come into the platform and have thousands of doctors essentially overlook it, coupled with our AI to be able to prescribe and treat very specific treatment plans for you. So it's not just a data dump of here's, you know, 120 biomarkers, good luck, throw it in chat should be, but. But it's actually real providers that are then helping you make tactical personalized prescription next steps. So we have about a million square foot of pharmacy compounding in the U.S. so we actually make a lot of these medications ourselves. So if you're a person that comes in like me, low vitamin D, you know, testosterone's not as optimized as I want to be. So it's going to be a zinc and ashwagandha or magnesium supplement along with core pharmaceuticals, let's say, for heart disease, like a cell stem statin, we'll be able to actually put that together for you in a form factor that you love, you know, customized into a beautiful personalized treatment and then delivered to your door for something like 30 or 40 bucks a month. So it's really this vision of a completely unified system that gets to know you, optimizes you, and then actually can verticalize the making of treatments for you so that we can see you get better and help you kind of get preventative. Given we're in a country here where almost everybody dies of preventable disease. Right. It's like heart disease, heart attacks, diabetes. It's stuff that takes like 10 to 20 years to develop. We should be able to get ahead of this stuff and actually start people. You help people change that course of direction.
- [126:19]
Jordy
How do you, how do you think the labs market Evolves. There's a ton of players in this space. This is obviously not your core business, although it feeds into the core business. I think there's players that are trying to use Labs as like a wedge to then go and compete with you guys. You guys are kind of doing the opposite. But do you think that margins on Labs like effectively go to zero over time?
- [126:43]
Melissa Tokmak
It's going to go to zero? Yeah. I think anybody who's in the labs business today whose core business is labs needs a new business. You've got Quest and LabCorp and Great Partners. They've got thousands of locations, 10, $15 billion public companies each of them. There's no way that those margins will be constrained when you have a platform like ours where it's in our, it's in our best interest and it's in our consumers best interest to verbalize this infrastructure over time and give it away for free. Right? For, for you guys or anybody who's a Hims or hers customer, it's to our benefit that we do a full panel and, and get you that at cost or even as legion. Right? Like we can, we'll, we'll pay you to take it eventually as you actually verticalize this infrastructure. Because the data you then get back from it allows us to give you very tangible next steps and manage your care over, you know, 5, 10, 15 years, which is really where the long term relationship and value creation comes from. So, you know, for us, you know, we're excited about these price points. It's 1.99for a base package. It's 499for the advanced package. You know, my hope in three years from now, and my CFO would be mad if I said this, but it's like my hope is this. 30 bucks, right? And it just gets added on to your Hims and Hers member membership. And if you have a treatment with Hims and hers at all, you get this stuff for free, right? Because it just makes the platform and it makes your precision care that we can deliver so much better. Right? Say you're a hair treatment and you're taking our oral chew for hair loss, which is like finasteride and minoxidil. And then we figure out that you're pre diabetic. Our ability to compound that with a low dose of metformin or some type of other microdosing GLP1 or GIP that just transforms the actual clinical outcome for you. And it all started with a very simple relationship, which is, I'm worried about my hair. I want to start taking advantage of.
- [128:36]
John
That, I want to know your take on Theranose. I have this hot take that it was never a good idea at all because you can just take a full vial of blood and even if it worked, it wouldn't be that big of a product because if I go to the doctor and they take this much blood instead of this much blood, it's just not that big of a deal. Do you think if Theranos had worked that would be a successful product? Would that have reshaped? Was there ever a chance they would have been a great thing?
- [129:06]
Melissa Tokmak
Yeah. The only reason that the idea for Theranos is interesting was if you can do at home diagnostics as opposed to in office phlebotomy. So that burden of leaving your house, going to a doctor, doctor's office. The number one fear in the country is needles, right? So you sitting there and somebody, you know, putting a needle in your arm to take out that tube of blood. Whether or not it's a tiny tube or a big tube, it actually doesn't matter. You know, you had to go and park and take time off work.
- [129:34]
John
I see that as a terrible problem in our society that people are afraid of. Noodles should not be fearful. I don't like that at all. I feel like we should solve the fear problem before we solve the needle problem.
- [129:46]
Melissa Tokmak
I think if you can do any, and this is something we're working on and we've talked about this, I think in the next year or two you'll have at home diagnostic devices that you can put on your arm. Click a button, you feel nothing. It's kind of like a cgm, right? And you can actually probably eventually look at the interstitial fluid. So you're not even doing injections deep into the blood, but you're actually just doing kind of like top capillary fluids to be able to interpret what some of these base biomarkers markers are. If you can pull that off and it costs 10 bucks to do that and you just click a button, take it off and mail it back to hims, that I think really transforms access because so many more people I think will do that versus the whole, you know, kid caboodle, like going and scheduling and showing up in office.
- [130:29]
John
Yeah, yeah, that makes a lot of sense.
- [130:31]
Jordy
How, what, what's your like, framework for how the GLP1 market shapes out?
- [130:40]
Melissa Tokmak
Yeah, it's a crazy market. It's like one of the most important categories. Probably in the next few years, I think you're going to have more and more treatments on the market. Very Quickly, you've got Netcera deal with Pfizer and Novo that have been taking place back and forth. You've got Kylera that just raised a huge growth fund, that's got a GLP GIP that is at par or possibly superior to Terzonek Appetite. You've got Viking coming out probably in 29 or 30. You then have Wegovian Ozempic going generic sometime in 2030 or you know, 31, which is going to bring the price down to, you know, it costs, you know, it could probably cost $10 to make one of those vials. Right. So I think in the next few years the landscape is going to be incredibly competitive. I think the prices are going to come down dramatically. This is something we're very excited by. When we first started putting these treatments out, the prices were $1500 a month and now they're down to about 150. I would guess in three years from now you can be buying, you know, one of the best treatments for 50 bucks a month. And at that point it's going to be just cash pay. Customers are going to be able to use hsa, fsa, not have to deal with the, you know, the pain in the ass of insurance. And then the massive market will expand pretty dramatically.
- [131:54]
John
Yeah. You said you have like what, thousands of square feet of compounding? Something like that. Tens of thousands.
- [132:00]
Melissa Tokmak
Yeah, yeah, we've got a million. A million square feet.
- [132:02]
John
Million square feet. I feel like that's been controversial in the past. There's been battles between you and maybe the fda. Maybe the.
- [132:12]
Jordy
Just a little controversial.
- [132:13]
John
Yeah, I mean we read about it in the Wall Street Journal. Is there, are you staying the path on compounding? Is that where you want to be in a decade? Is compounding here to stay or is there a world where you wind up partnering with the legacy pharma companies?
- [132:28]
Melissa Tokmak
I think it's both. Right. I think we will inevitably partner with a lot of these companies. You know, we have the largest distribution platform for therapeutics in the US for consumers. Right. So if you have a next generation treatment, we should be talking because if we want to get your medicine to a lot of people, you know, there's just no faster route to do so. So I think that's a very logical and natural thing to happen. And we, we've been doing things like this. We recently invested in Grail, which is one of the, I think the most advanced early multi cancer detections blood test, simple blood tests can detect over 100 cancers as early as stage one. I had every single Person in my family take this test. Every year, more things like that are going to come to market, and I think we'll come to the platform. But for compounding in specific, there's really clear guidelines with regards to FDA allowance of compounding. There's different types of pharmacies. They can do different types of things for very specific reasons. Reasons provider personalization, form factor, side effect management, etc. We're a public company. You know, our chief medical officer was the chief medical officer of Walgreens. Deb Otor, who's on our board, who runs the risk committee, was the woman who wrote a lot of the compounding legislation at the FDA for a decade. So I think we play very clearly by the rules. I think this is one of the first times in history that we're able to give this level of personalization to a lot of of people. Historically, this level of personalization only was possible for people that, you know, were spending $50,000 a year for coner doctors and they could have treatments made for them. And I think it's our ambition to figure out how do you get that for the 1% to everybody. And so I think there's friction without question in that system, but the regulatory framework is very clear. And I think we're just staying the course. And I think over time we'll be able to figure out ways to make the other parties in the ecosystem feel like they're getting well compensated as well.
- [134:22]
John
Yeah. Makes no sense.
- [134:26]
Jordy
How would you describe your management style? The company's up massively from the IPO. I feel like I started tracking HIMS during the D2C era. Right. You guys were the poster child of D2C. This meteoric route rise.
- [134:44]
Tomas Puig
And then.
- [134:45]
Jordy
And then obviously, you know, the, the public markets have been wild. Yeah. I'm curious, like, how, how you navigate, you know, the, the market internally, you know, even with the team.
- [134:57]
Melissa Tokmak
Right.
- [134:58]
Jordy
It feels like you guys are very built for it. But what's your approach?
- [135:02]
Melissa Tokmak
Yeah, you know, when we. I remember waking up one morning and looking at my wife and the stock was $2.87, like we were trading at, I think like 0.4 times next year's revenue or something like that. It was crazy. So I think the team has a stomach of steel. You know, whether it's stock, $70 or $2.87, I think, you know, most of the executive team will retire with this company, including myself, because I think there's this incredible opportunity to figure out how you redefine what is a health care membership that Every single one of us wants to buy. Like it's got the cutting edge, it has the most, you know, the best diagnostics, it's the most affordable. Affordable. It's an Amazon prime version of healthcare that you can afford, right. It's like 500 bucks a year. And for 500 bucks a year, why deal with insurance in this country when everybody's insured, but everybody has a high deductible plan with a $2,000 deductible. Right. So while everyone is insured, nobody can actually get the benefits of insurance because they can't actually afford to use all the cash to then get the $1 of insurance. So I think our management team is, it's not a very glitz and glamour team. Right. Like you're not seeing us out on TV 24 7. It's a really heads down team. It's a gritty team. It's a team that loves to build. We have a ton of fun together. And I think it's a team that genuinely focuses on what can be accomplished in a decade and not can be accomplished in the next quarter or two. Which is why I think a lot of the investments we're making have a much longer time horizon. But as a founder that runs the company that has the high boat stock, we have that privilege to be able to invest over that long period of time.
- [136:42]
Jordy
That makes sense. What kind of horror stories have you heard about Chinese peptides? There's a meme going on on X of people talking about it. Obviously some of the pricing coming out, people buying directly there is pretty wild. What are the risks that people should be be aware of?
- [137:02]
Melissa Tokmak
Yeah, you've got to make sure, you know, the pharmacy suppliers are working with are good quality pharmacy suppliers. There's something called a certificate of analysis that you know, high quality FDA oversight facilities in the US can deliver third party tested and validation. You can actually ask for this certificate. We give customers their certificate for any of our compounded products which actually shows the independent laboratories saying this is exactly, exactly what you know, we say it is. So I think you want to have organizations or pharmacies that have that type of coverage, that type of third party testing, you know, good manufacturing practices and ideally have some type of oversight from, you know, state boards or fda. I think there's a lot of stuff you can buy online that's you know, truly not for human use. And they say that as a way to kind of loophole around. But it like truly is not for human, a noose. And you know, I would encourage people to Try to stay away from that kind of stuff.
- [138:00]
Jordy
Random question. I'm curious to get your take on it. Something like 1% of US GDP is dialysis doesn't necessarily seem like a problem for hims to solve. But from everything you know about, you know, working in and around this industry, how do you think, do you think it's a solvable problem or is it?
- [138:23]
Melissa Tokmak
Yeah, yeah. You know, my grandfather was on dialysis for like this five or seven years before he died. And it's a, it's like the worst thing in the world, like watching somebody on that, right? You have to show up to the facility every week. If you don't, you literally, your body deteriorates and you die. Dialysis is a result of you having totally failed, taking action on you being sick for 20 plus years, like a very long, long time. Similar to cardiovascular disease, right? Like you show up in the ER at 60 and you have a heart attack. You know, that took 20 years to build up. Like there's just no other way. It happens. And so something like what we launched today I actually think is the first step of the solve because you get for a couple hundred bucks something like a lipoprotein little a test which tells you your predisposition for advanced cardiovascular disease. Which means even if your cholesterol numbers are amazing, they're not amazing enough, right? You have to have like the best gold standard cholesterol numbers, like you're an 11 year old kid to avoid a heart attack because you have this predisposition to risk. Same thing with diabetes, right? Diabetes, A1C, fasting insulin, all of the are fasting glucose. These all rise over time. So if we're 30 or 35 and have these numbers, you can just chart what it's going to look like if you 50. And so, you know, micro steps in health, in food, in movement, in, you know, metformin at the right time, or GLPs or preventative statins or PSK9 inhibitors. Like all of these tools exist. And for the most part these tools are not extremely expensive. Right? So the real burden in the US healthcare system isn't actually lack of innovation, it's lack of education and access. So how do you get more people tested faster? How do you make it easy and less scary for them to take that first step and then keep them motivated along the way? Because you know, these treatments don't make you feel amazing the next day, but they'll save your life 20 years from now. And I think that's where a brand like ours spends a shitload of time is you, you actually have to love hims and hers because staying on this medicine is, is good for your health and, and it's hard because you drop that habit and then you drop that benefit.
- [140:37]
Jordy
Makes a lot of sense. Well, thank you so much for joining. It's great, great to finally meet. Certainly we're, we're an inspiration during again that DSC era. It was like just a crazy, crazy moment in, in time and it's awesome to see you guys execute in the public markets.
- [140:55]
John
Yeah, congrats.
- [140:55]
Melissa Tokmak
Thank you guys for having me on.
- [140:57]
John
We'll talk to you soon. Have a good one. Before we move on, let me tell you about linear. LINEAR is a pretty purpose built tool for planning and building products. Meet the system for modern software development. Streamline issues, projects and product roadmaps. We have our next guest in the studio, Melissa Tokamak. Tokmak from Netic. We also the market is selling off like crazy. People are asking us to cover it. But we will talk to Melissa first. How are you doing?
- [141:20]
Melissa Tokmak
Hello.
- [141:21]
John
Good to see you. Have a seat right there. We do need to cover. I'm glad we have you on the show because the market is, is in turmoil but your business is maybe less indexed to the market, Is that correct? That is correct. Explain, explain, you know how explain the business, the news, take us through it and then we'll get your take on.
- [141:42]
Melissa Tokmak
All sorts of things.
- [141:43]
Jensen
Well, I mean we serve the essential services businesses that are really backbone of our American economy by nature.
- [141:50]
John
Countercyclical.
- [141:50]
Jordy
Let's give it up for the backbone.
- [141:51]
John
The backbone of the American economy. Me. Thank you.
- [141:54]
Jensen
Okay, so these businesses are, these enterprises are in industries like H vac, plumbing, electric, solar, consumer health, energy across the boards. So whether the market is going down or up that somebody, a business or a consumer always needs these businesses. Actually today I'm coming from San Francisco. As you know, all of San Francisco needs a lot of these services today because the weather is a common complete disaster. It has been raining. No, complete rain. Everything is shut down. Yeah, electricity goes off all the time. And I'm sure a lot in the Bay Area will be needing these services all day today. So what we do is provide really these enterprises with an AI revenue engine.
- [142:40]
Jordy
Right.
- [142:40]
Jensen
So in these times of need, they can actually handle all of the demand with AI agents and serve the community, communities. But also when the demand is soft, they can predict the need and turn these relationships into multi time recurring relationships.
- [142:57]
John
Okay.
- [142:59]
Jordy
So I feel like every time I call a plumber or a roofer or an Electrician People like that. They don't like the delay and the lag of getting like there's always like crazy lag in terms of getting back and just being like, yes, I can see, I can come by at this time.
- [143:15]
Melissa Tokmak
Time.
- [143:15]
Jordy
And I feel like the lag is because they're super busy and if you can help them respond more quickly, they can generate a lot more.
- [143:24]
Jensen
Exactly. Actually many of them are very large businesses and the lag is because everybody needs it. When you need it, odds are millions of other people need it at the same time. And I'm sure like this has been talked here in this program too. There is a national shortage in skilled labor in these industries too. So it's actually, I mean Jensen talks about it a lot and it is the type of labor we need in the country. And until then there.
- [143:49]
John
Is that what Jensen was talking about? I feel like he said he like, you know, we don't have enough plumbers. But I felt for some reason he was talking about like plumbing in data centers. Yeah, it's the same.
- [143:57]
Jordy
Right.
- [143:57]
Jensen
So H Vac technicians or plumbers, what the data centers need cooling.
- [144:02]
Melissa Tokmak
Right.
- [144:02]
Jensen
There's not that many or to be able to stand it up. But if you think about takes a lot of years to get actually trained for these jobs.
- [144:12]
John
Sure, sure.
- [144:12]
Jensen
Probably more than I went to school for at Stanford or you went to school for actually it's very important. And then afterwards you have to do it in real life, the training as well. So with data centers and consumers and businesses across the world, the need for these businesses a lot. And it's not only about H vac plumbing, electric. When you say essential services. Right. It's also across energy and solar or. Or consumer health like the Bay Club. Right. That we have been serving. So it's really the things that everybody in their daily lives need and has to interact with.
- [144:47]
John
Yeah. Walk through the typical stack of H Vac repairmen. I imagine somebody could just have a Yelp page and a Venmo account. Then some of them have a full website with a booking system and almost like their own little mini error. Probably not something they built themselves. Maybe they have a Shopify site or something and then at a certain point they get a roll up happens, they get bigger, they get more industrial, they get more mid market. And then I imagine that there's some sort of central, you know, point of record. Are you plugging into that, are you trying to replace that? Are you trying to be the first, the first choice for setting up all the touch points?
- [145:27]
Jensen
Yeah. So we primarily touch work with what you described as mid market and large enterprise. A lot of these businesses are owned by private equity and or still owner operated. They have built it from zero to hundreds of millions of dollars in revenue. So they will actually. I love that you went through that stack. They might have a more sophisticated tech stack than here. Right. So these businesses run on ebitda, right. So it's extremely important that they think about their efficiency, efficiency, the customer relationship and how do they serve their communities. So primarily it will focus on what type of data they have and where do they keep it, various software solutions and of course third party aggregators where they might do advertising. So we will partner with different solutions that they might have to make benefit of any of the data they can. But primarily what we're doing is bring frontier AI to these industries so that when their demand is very high instead of losing that they can handle it all at the same time. And when the demand is soft, they can generate net new demand. Right. As you know, these industries, just like we talked about, San Francisco today is very volatile due to seasonality or some other external effects that they can't really control. Right. So that's why it's very important how you need to be efficient. When the demand is coming, you have to capture it all. And when it is not, you're thinking.
- [146:55]
Jordy
About, yeah, it's like a retail. You can't sell during Black Friday, during Q4, you're cooked. Like you're probably not profitable.
- [147:01]
Jensen
Exactly, exactly. Or even how to predict the next thing. Like one of our customers. There were tornadoes in Missouri a few months back and before the tornado literally happened using Netic, right. Reached out to potentially affected areas and talked about generators. How would you feel if your electricity went in the studio? You would not like that because your business would be down. Right. And but then afterwards I tried to.
- [147:25]
Jordy
Get a generator when the power was.
- [147:26]
Jensen
Out last time it was a denied buy.
- [147:28]
Jordy
It doesn't work well, it doesn't work.
- [147:29]
John
Once the, once the crisis stops because.
- [147:31]
Jordy
The electrician is like yeah Buddy, I got 200 people calling you the same thing.
- [147:36]
Jensen
I will, I will help you out. That's the best in there.
- [147:39]
Jordy
I'm covered now.
- [147:40]
Jensen
But also, also when the disaster actually happened in 90 minutes maybe they got thousands of calls, right. So how do you help? But it was great for them because they could answer every call with Netic and they had already maybe rescheduled the non essential jobs that they had so that they could help the community in that moment. So that type of proactivity and focusing on the future and like revenue generating interactions for the businesses is very important for these businesses.
- [148:12]
Jordy
Talk about traction.
- [148:13]
John
Yeah, sure.
- [148:13]
Jensen
Yes.
- [148:14]
Jordy
You raised a new round.
- [148:15]
Jensen
I did.
- [148:16]
Jordy
Today we announced very, very low dilution. Let's let her hit the gong as hard as you. As hard as you can.
- [148:25]
John
$23 million Series B led by Founders Fund. Hit that gong.
- [148:30]
Melissa Tokmak
Wow.
- [148:31]
John
With authority. At a 45 hour, 450 million dollar cap, which is a strong 4x step up in valuation.
- [148:38]
Jordy
Very low dilution. Very low dilution, yeah.
- [148:41]
John
Why?
- [148:42]
Jensen
Well, I think from the beginning I really.
- [148:45]
John
You hear this good omen. Still going when it still goes.
- [148:47]
Jensen
Same strength we put into our business, right? No, I mean that is actually the answer from the beginning. What was important to me is that we build a business with strong fundamentals. Right. It is not about the hype, it's not about the valuation. For me, the most important thing, even in investments, we work with the best and we're lucky and grateful. That same people triple down in a row in netic. And when you have a business with strong fundamentals and strong margins and scaling efficiently, you don't want to raise more than you need. And if you do that, by the way, that would be a detractor, actually, in the type of talent that we're hiring today, they do want to work and nimble teams, they do want to run through walls. They want to be here because they want to build the future of sbc.
- [149:33]
Jordy
I want to run through a wall.
- [149:35]
Jensen
Yes.
- [149:35]
Jordy
You should get one of your customers to build us a wall. We'll run through.
- [149:40]
Jensen
Exactly.
- [149:41]
John
I have a question about models. We were talking to Brian Chesky and he's starting to add features to Airbnb that allow the booking of a chef. It's a very different, different model for someone who's traveling. They want a private chef or something like that. But he mentioned that he's having a lot of success with open source models. Obviously he's operating at massive scale and so every dollar counts. Are you seeing luck with open source models or are you sticking to the closed source models?
- [150:14]
Jensen
Yeah, you actually can't stick to one the way to. Because if you think about it, maybe the difference there is we serve essential services industries. I'm like a utility to these companies. I can't go down, I can't be wrong. So the way that we build our own ML orchestration. Right. The way that you build and think about models and what's really helping with each task that you need to do in that orchestration is very important. So you have to think about what's best fit in terms of address verification in this case versus understanding what does Jordy need when he's calling me. Right. Or messaging me. So that would be need finding. So the whole point is how do you think of these modules and what do they need to get done and what is the model that can give you the best answer? Right. And you have to obviously build a lot on your internal evals to be tracking that continuously and make any changes as you need. Right. So today definitely we do use quite a bit of more closed source models. But depending on your evals, if something is not keeping up with what we need and the new functionality we add, we would always test and think about any other model.
- [151:30]
John
What about in 10 years, you think.
- [151:32]
Jordy
In 10 years before that? Are you using voice models at all?
- [151:37]
Jensen
Yeah, we don't do voice to voice. We do have voice actually to today we support our customers from Voice Text online, like any type of channel you can think we are. We're the single inbox. Right. For anything that they're really getting. So voice, yes, we support, but voice to voice is actually not there yet. Speech to speech. So we do really orchestrate all of that in terms of speech to text reasoning models and then text to speech afterwards to really give that accuracy. Reliability. Reliability as well as the flow, the feel. Right. And hopefully maybe in 10 years. You're asking we will have to think that. Not in 10 years, maybe even a few. Where's voice to voice speech to speeches going? So according to that, we have to continue.
- [152:19]
Jordy
What's the sales cycle like?
- [152:22]
Jensen
To be frank with you, I think we really sell based on roi. So we show our customers.
- [152:30]
Tomas Puig
Right.
- [152:30]
Jensen
And they can talk to our existing customers at any point. Anyone interested in interested private equity firms that we support, Large companies that we support.
- [152:37]
Tyler
Yeah.
- [152:37]
Jordy
Private equity firm's been one of the main channels in here where they're just like, hey, we have a roll up of a bunch of different underlying businesses. Let's roll.
- [152:46]
John
We have a number of rollups and we want the same software.
- [152:48]
Jordy
We have a roll up of rollups.
- [152:49]
John
I mean that's what private equity firms are these days today.
- [152:51]
Max Hodak
Yeah, they are roll ups of rollups.
- [152:53]
Jensen
That is correct. It has been one of the good channels, but also it can be direct to large companies itself. I think the reason we love working with private equity because in today's world they understand the importance of AI and they're looking for an AI partner. Yeah, right.
- [153:08]
Jordy
It's not really about we need an AI strategy. What can we buy?
- [153:11]
Jensen
Yeah. Actually, you'd be surprised. Like, I think a lot of tech companies may be looking for something or a strategy to put on a board deck. A lot of these businesses I work with, I will say they're absolutely incredible and better entrepreneurs than entrepreneurs in Silicon Valley sometimes. I see. And they are focused on real value.
- [153:29]
Jordy
Experience, free cash flow.
- [153:30]
Jensen
Yeah, it's. I mean you can't hide behind raises and valuations. Right. Like all you see is that, does this help my revenue?
- [153:37]
John
Yeah, yeah.
- [153:37]
Jensen
Very easy. And for me too, I can say, hey, I will not give you any random words, AI this, AI this, blah blah, blah. No, let me show you how is this going to affect your revenue? Let's talk about that. So when you do it that way, I think the cycles are pretty, pretty fast.
- [153:56]
John
Yeah.
- [153:57]
Jordy
Do you think people are scared to compete with you? You seem like a pretty formidable opponent.
- [154:04]
Jensen
I don't know. I hope not because I.
- [154:07]
Jordy
You enjoy crushing.
- [154:10]
Jensen
I do enjoy crushing.
- [154:12]
Max Hodak
Crushing.
- [154:13]
John
How are you seeing not only me.
- [154:15]
Jensen
My team too, and this is just for them from here. You all that are watching right now, you're all be. And there's no one else in this world that I would rather work with other than you incredible people coming from scale, hrt, databricks, mit, Stanford. But I don't wanna even talk about the accolades. Who cares? They all, most of them have deployed AI applications in the real world, in production and they run through walls every day to deliver for the real world. Even though they could be anywhere, they could go in any company.
- [154:50]
Jordy
Well, for the series C, we will get a wall built and you can run through everyone.
- [154:55]
Jensen
Exactly. I will be doing that.
- [154:58]
John
Yeah. Thank you so much for coming on the show.
- [155:00]
Jordy
Amazing progress.
- [155:00]
John
Congratulations.
- [155:01]
Jensen
Thank you for having me.
- [155:02]
John
We'll talk to you soon. We have one more guest. Stay with us. The market is crashing, everything is in turmoil, but the business continues. The show goes on.
- [155:12]
Jordy
The White House needs to announce a mortgages second round.
- [155:17]
John
Let's move to 100 year mortgages. That will be the solution to all this. No. The only thing that can save us right now is numeral.com. sales tax and autopilot spend less than five minutes per month on sales tax compliance. Also 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. And our next guests are here. Let's bring them in to the TBPN Ultradome. We have Jeffrey Katzenberg and Tomas. Welcome to the show folks. Thank you so much for taking the Time to come down to the TVPN Ultradome. Good to see you again. Congratulations on the news. Let's get some introductions first. Who are you? What organization are you with? Let's kick it off there.
- [156:02]
Tomas Puig
I'm Tomas puig. I'm the CEO and founder of Alembic.
- [156:05]
John
Thank you.
- [156:06]
Jeffrey Katzenberg
Jeffrey Katzenberg, GP at WonderCo.
- [156:10]
John
Third time on the show, second time on the show. Something like that. But first time on the show here. Congratulations. And what's the news today?
- [156:16]
Tomas Puig
Oh, well, the news is, is that we actually just raised $145 million.
- [156:21]
John
Fantastic. Why don't you go hit that bomb? Please enjoy this.
- [156:24]
Jordy
Enter the Dragon.
- [156:25]
Melissa Tokmak
Yes.
- [156:26]
John
Yes. Yes. It's a. Whoa.
- [156:31]
Jordy
We haven't seen that before.
- [156:32]
John
It was. That was a. Okay.
- [156:35]
Jordy
That was a very aggressive ring.
- [156:36]
John
I like it. I like it. It's still ringing now. It has a nice sound. So take us through the business. How are you pitching it these days?
- [156:43]
Tomas Puig
So what's really interesting about is we do a little bit of a different thing in the AI space. We do causal AI, and really what our executives that we work with in these Fortune 500 Global 2000s want to know is they want to every chain reaction and lever that moves the metrics that they care about about in the business at any time. And so where we started originally was on the marketing and sales side. And so they would be like, hey, I spent $100 million on a stadium name. All these unknowables, these large content pieces, all this brand. We all know it worked at the time, but nobody could actually prove down to the dollar, down to the actual effect of what it would be. And so when we built the company originally, we were, you want to know how to shine a light inside this black box and actually be able to get the real dollar value so that you can start telling stories. Because there's been so much of this improvement in, say, the programmatic side of the house buying these ads. Well, what ended up happening is once we built that, we found out it actually worked really, really well. And we got a lot of incredible clients. And then they started to ask us to do other things, like being like, oh, oh, hey. Now can you see how all that affects foot traffic? Now can you see how all that affects my ordering systems? And as we started putting this causal model out further and further, we realized it actually worked on an incredible amount of stuff. And so then our clients started asking us to be like, well, can you do our FP and A for finance? Can you do our supply chain? And so that's really kind of where the business came from. But the core of it is that, like, when you have low information or low amounts of stuff, you want to be able to actually affect the metrics you care about.
- [158:20]
John
Take us through some of the case studies. Who have you worked with? What's the most concrete example of you talked about buying a stadium. Have you literally found whether or not crypto.com arena penciled out? Have you, have you looked at these? You know, I grew up at the Staples Center. I, I, I miss Staples. It's been renamed. But, but I want to, the red.
- [158:40]
Tomas Puig
Stapler that, like, I want to at.
- [158:42]
John
Least know that, that it was worth it financially. Is it, was it worth it financially? Can you tell me that?
- [158:47]
Tomas Puig
Yeah. So we did a, let's check Staples.
- [158:50]
Jordy
Stocks and say yes.
- [158:52]
John
Should Staples have stuck with the sponsorship?
- [158:55]
Tomas Puig
See, the problem is it doesn't count for execution risk.
- [158:57]
John
Yeah. Yes.
- [158:59]
Tomas Puig
So when we, there's always unknowables.
- [159:01]
John
But, but what can you now?
- [159:02]
Tomas Puig
No. So we did a great case study with Delta Airlines that we actually presented with their CMO at the Nvidia conference, at the GTC conference where they were sponsoring the Olympics. And one of the things about these large sponsorships is there's two big aspects people don't talk about. One is every time you've ever worked in the business side of the house, they go, hey, you need enough historical data to be able to do something. The second is it takes a lot of time to get a response. Well, when they were doing the Olympics, they were doing the promotions. One of the most interesting things we found out is we analyzed all this ad work that's coming out. You have only two weeks. You're holding up tens of millions of dollars on the P and L line while you're doing this.
- [159:42]
Tyler
Right.
- [159:42]
Tomas Puig
These are not cheap options. And when we pulled the study, we actually found out that the best piece of content that functioned for them, the most profitable, was not actually the content. That was the ads. The ads did okay.
- [159:55]
Max Hodak
Right.
- [159:55]
Tomas Puig
30 seconds, 60 second spots. But what ended up happening is they actually, if you watch the Olympics, they had the Delta medal presentation ceremony where like every time an American athlete would win a medal, they would take it, put it on, you know, their shoulders, and that would be, you know, a really emotional moment. Well, it may seem kind of obvious in hindsight, but when you have the Eiffel Tower in the background, really emotional moments, you sell a lot of tickets to Paris. But the key is if you can know that within a few days, you can either double down Right. You can sponsor the next Olympics, you can do all these things, and then you can actually act on it. Now, the problem also with when you do big, huge campaigns like that is you walk into the lounge for Delta Airlines and Team USA is the WI Fi passcode. It's not like one simple campaign, right? You pivot an entire company to a messaging set. But we were able to tell them within a couple days, down to the dollar, a material amount of cash that they were able to pull back, and then they could actually show that to, you know, executives.
- [160:51]
Jordy
So all this oldest problem in advertising, right? Of like, I know my advertising, like, that's a line.
- [160:57]
Jeffrey Katzenberg
I'm 100% sure that 50% of my brand marketing is working. If I only knew which 50%, which has been true in brand marketing forever going back to the 1980s.
- [161:08]
Jordy
Yeah, I'm sure in entertainment, especially, because there's less. Like, the attribution is really challenging. We did a bunch of marketing, and then people went to movie theaters, and then some people streamed it, and it's hard to actually track it all the way through.
- [161:19]
Jeffrey Katzenberg
But what the brilliance of what Tomas and Olympic have done is that using. Well, first of all, and he should explain it because he's the brainiac here. The new math that came out of contact tracing from COVID which actually has impacted many different businesses in terms of research and causality and attribution.
- [161:43]
Jordy
We got advertising learnings out trace, by the way.
- [161:47]
Jeffrey Katzenberg
It's absolutely.
- [161:50]
Jordy
It's a biggie. Very American to like, you know, entrepreneurs says, oh, okay, turn it into an advertiser.
- [161:57]
Jeffrey Katzenberg
But when you think about it, is, is that either for that, for COVID testing, or for any of these attributions, what they are able to do, which was not possible even only three or four years ago, which is to be able to ingest billions of rows of data from all sources and using machine learning and AI to actually digest that and make sense of that and actually then be able to see directional in time. He'll explain neural networks and how this actually works, which, as I said, is way above my pay grade here. But I can tell you which is to your. To your question here, which is when we went to the top brand marketers, the first reaction was, no way. You know, this is like, you know, yeah, it's not possible. And this is just like telling us, you know, you know, pot of gold at the end of the rainbow or, you know, dumbo flies or, I don't know, whatever you want. And they went, not, not possible. And that's when these guys would come in and do these pocs. Disney, Mars, Accenture, Delta, and you know, in the world, once is an accident, twice is a coincidence, three times you go, okay, well, this thing actually delivers. As fanciful as the notion may be, it's real.
- [163:26]
John
Yeah. Can you tell me a little bit of the history of how, how like product placement works in Hollywood? Because that feels like the classic example of difficult to measure. But you've been putting the car in this one.
- [163:41]
Jeffrey Katzenberg
But it's when you start to actually think about this. There's a trillion dollars a year globally spent on brand marketing. And brand marketing is everything from crypto on the Staples Center.
- [163:55]
Melissa Tokmak
Sure, sure.
- [164:00]
Jeffrey Katzenberg
To an interstitial to imagine in the Disney parks all the things that you're, you know, that they're able to offer to their brand partners to putting a logo on a car. What do you, what, you know, what's the value of having Oracle?
- [164:15]
John
Yeah.
- [164:16]
Jeffrey Katzenberg
On the race car on a race car in this. Or a patch on a base. On a baseball player or a basketball.
- [164:22]
John
Player, you know, so is it. Has it always just been intuitive or has it been more relationship driven?
- [164:27]
Jeffrey Katzenberg
No, it literally is intuitive. It is. I mean, he can tell you there's an old math. Mmm. He can again, Tomas can really take you through, explain it. And that literally dates back to the 1980s. And one, it's directional, not specific. And two, the lag between the time it happens and you're able to gather the data on it was months. So it doesn't. The value of it was really questionable.
- [164:57]
John
I guess my question for you is it's very clear that with the progress in AI, there's the ability to find insights in data. Clearly we see this across everything. My question is there's so little ground truth that how can your clients ab. Test your solution against something else. If I go to, to a coding agent and I ask it to generate some code, I can run the code at the end and I can say, well, yeah, the code worked, or I can read the report. But if I go to you and I say, how much was this sponsorship worth? And you tell me it's $4 million, and I say, okay, maybe it was four, that could have just been a guess. So how do you justify the results that you spit out? Since we can't run a controlled trial.
- [165:45]
Tomas Puig
So validation is one of the number one things we get asked constantly. I'm sure when we give out papers, it's really funny. We have one paper where it's like a two page brochure and then the next 15 pages is how we validate.
- [165:57]
John
Yeah, exactly.
- [165:58]
Tomas Puig
And one of the things that people don't talk about is it's actually very possible to test these things. So what you can actually do is you can do what we call backfit testing. You could set the machine backwards in time and then you can test against things that, you know, you can withhold information. You can do all sorts of things to actually see whether, you know, you feed the machine up to like last year.
- [166:17]
John
Yeah.
- [166:17]
Tomas Puig
You know, what all the results are going to be this year. So when you predict what something's going to be, you can see how close you got. And so I think that when we're testing things, we use a variety of both synthetic data testing methods, because we're not, you have to keep in mind, we're not a transformer model or an LLM.
- [166:33]
Melissa Tokmak
Sure.
- [166:33]
Tomas Puig
It's an entirely new methodology. And, and so when we grab this stuff, we have both synthetic data we can build. So there's systems like, for the nerds, like Tigerbyte that does causal synthetic data, or there's real data like EF, MRI data that we actually know that there's a cause and effect that we can pull from physical world or physical bodies that you could test the algorithms against as well. And then from the business side of the house, there's lots of ways to do it. Now, one of the things I like to correct on this is it's like we often say the mantra of our company is we're about being directionally correct, not specifically wrong. And so when you need to make decisions, a lot of time when we do these simulations, we're like, here's the top 20% of things you need to absolutely keep doing because they knock it out of the park. Here's the 20% of things that are just literally hurting you.
- [167:14]
John
Okay.
- [167:15]
Tomas Puig
And when you have zero information, like you're in a pitch black room and you have no idea where the exits are, would you rather take a really, really good educated guess about where the exit is or would you rather wander around the dark?
- [167:27]
John
Yeah.
- [167:27]
Tomas Puig
And so I think that one of the things about business intelligence is when you have zero information, the value of the information you can get is that much more important.
- [167:35]
John
Yeah.
- [167:35]
Tomas Puig
And so we do very.
- [167:36]
John
I'm so confused about the back testing thing. Like, if I like, how would you go back and assess the value of like Budweiser sponsoring Super Bowl 40 or something? It's like you can't run the counterfactual.
- [167:48]
Tomas Puig
Well, you can see counterfactual Simulations are assumptions. So think about like this. A lot of people talk about probabilistic graph. Right? Graph analysis, probabilistic graphs, whether they're causal, whether they're anything else is to a certain extent, I mean, you know, pedantic. LLMs are almost a probabilistic graph.
- [168:05]
Melissa Tokmak
Sure.
- [168:05]
Tomas Puig
Right. Undirected till you query it.
- [168:07]
John
A neural network and.
- [168:08]
Tomas Puig
Exactly. And so when we do this type of analysis, the thing about it is that we actually have seen lots of instances of the same exact thing. It's about the specificity. So we just did a calculation lyric yesterday that we looked at 1 trillion connections across six months for a company. Like that's the type of scale of analysis we're doing. We haven't seen. When you haven't. When you say the counterfactual, when you're saying what does it look like when you have a base state? We've seen the operating version of a company over years and so we know what the base state looks like when there's no influence from that. The key is you have to have enough data. So in the old world when we're doing all this analysis, we to used to say, you hear about the term overfitting.
- [168:48]
John
Right.
- [168:48]
Tomas Puig
Everybody's worried that is the model just biased? Well, in the old world you'd say I want to reduce the number of features, reduce the dimensionality. To prevent overfitting. Right. To prevent over prediction. Well, in the modern world with like computational statistics nowadays, or AI as we call it, you want more features, you get more accurate the more data you have. That's counterintuitive for how people think about these type of things. So my answer to you is we have to have immense amounts of data.
- [169:15]
Jordy
Right.
- [169:15]
Tomas Puig
And so we ingest a lot of it and then that gives us enough vision, Right. That we can see what a state would be and what it would not be. And then we also provide our customers with confidence and everything else. So there are times where we really, really know, right?
- [169:28]
John
Yeah.
- [169:28]
Tomas Puig
And we go, this we have 100% bet on. We see enough examples of this and sometimes we go, you're asking for a call on this. And yeah, we've got a pretty good guess, but you should still keep it.
- [169:37]
Jeffrey Katzenberg
The interesting thing is, you know, he's going fishing to see where, you know, the best fish are.
- [169:46]
John
Sure.
- [169:47]
Jeffrey Katzenberg
And interestingly, we did one of these POCs for a very, very big branded company and they were looking for the positive impact of event driven.
- [170:01]
John
Sure.
- [170:04]
Jeffrey Katzenberg
Brand marketing they were doing when they went out and sucked in all of this. Data to do this assessment for them. In addition to finding what were the sort of positive impacts of this which were modest. What they actually caught in the net was unknown to them. A promotion being done literally in Canada by a little subsidiary that was just like a regional commercial which somehow or another bled over into the States. So it was being run in Canada by a subsidiary there. Bled over into the States and had a tremendous negative adverse impact on brand.
- [170:49]
Jordy
Yeah, those Canadians.
- [170:53]
John
Yeah. The classic example is like social media manager intern of your Canadian, you know, offshoot is doing something that goes viral in America. And everyone.
- [171:02]
Tomas Puig
This one was even funnier. They bought out a national, the national sports National Hockey League final spot.
- [171:07]
John
Okay.
- [171:08]
Tomas Puig
And so that's gonna be huge in Canada.
- [171:10]
John
Yeah.
- [171:10]
Tomas Puig
So they. But that broadcast across the entirety of North America.
- [171:13]
John
Okay. So everybody saw it. And so like suddenly this little like.
- [171:15]
Jeffrey Katzenberg
Thing there drippy ugly hamburger and they hand just went and I'm sure maybe natively they thought it was funny.
- [171:24]
John
Okay, interesting. And so you're looking at like when, when sales data is happening relative to when the campaign goes on. And then you tease out from there.
- [171:32]
Tomas Puig
Yeah. Think about it like, you know, one of the big inspiration for the company was Renaissance Technologies.
- [171:35]
John
Sure, sure.
- [171:36]
Tomas Puig
So the high frequency trading firms and everything they can do a pretty good job about knowing when things affect each other or not. But you have to have an immense amount of incredibly high speed data sets.
- [171:45]
John
Talk about Accenture. You're partnering with them. Are they, are they just an investor or are they also to going go to market channel?
- [171:52]
Tomas Puig
Well they actually started as one of these tests we did because they were curious.
- [171:57]
Jeffrey Katzenberg
They were a customer.
- [171:57]
John
Customer.
- [171:58]
Tomas Puig
Yeah. And the interesting.
- [171:59]
Jeffrey Katzenberg
Which is the best way when you think about it, John. So they started as a customer, then said wait a minute. Yeah, we have a multi, multi billion dollar business around marketing go to market.
- [172:14]
John
It seems like a lot of people go to Accenture for these questions. Julius Bain and BCG McKinsey.
- [172:20]
Jeffrey Katzenberg
And so so then it went to hey, can we help take you to market?
- [172:24]
John
Yeah.
- [172:25]
Jeffrey Katzenberg
Which they've been fantastic at.
- [172:27]
John
I can imagine.
- [172:28]
Jeffrey Katzenberg
And then that led to when this, when Thomas doing this most recent round them stepping is the biggest venture investor they've ever done.
- [172:37]
John
Wow, that's amazing.
- [172:39]
Jordy
So big companies, Fortune 500 have been hiring Accenture and other consulting firms and research firms to help them understand what is driving results positive and negative in their business for a long time. What's the timeline to productizing what you're doing To a degree that a much smaller company, let's say a Company with like a million dollar a year advertising budget can actually start to get value out of this. That's interesting.
- [173:05]
Tomas Puig
Oh, so this is actually one of my favorite questions because it has to do with blah long term vision of the company. One of the problems you get when you have mid sized or small size firms and they're near and dear to my heart, is that they literally have never done things. So when you have a really large corporation, right, they've been in a podcast and whether they want to be too or not, somebody's mentioned them, right, they have all of these data sets across everything. But when you're a smaller company, right, you say a million dollar, your business or something like that, you haven't done everything. So there's no actual priors, there's no data, we don't know how they react. But eventually when we see enough of the universe, right, just like you hear about world models, just like you hear about anything else, we'll actually know what the causal universe looks like. What is the actual most likely outcome of when somebody does something? So in the future, in say a couple of years, we'll actually be able to build synthetic data sets that you can send us any query and we can respond to you what the most likely outcome would be. And where this gets really, really important is I think that especially when you're doing private businesses, the world of LLMs and everything nowadays, I think they're amazing but they're quickly converging, right. There's not going to be that much difference for a client between like ChatGPT and Claude. The problem is that if you're using that for business intelligence and business decisions, what are you going to do when your competitor gets the exact same answer and strategy you do do? That's a real problem. And so we believe that we will take the best private data sets in the world, do stuff for just them, get the, get our overall learnings in other places and then we can actually provide people with strategies that are unique to them. Yeah, right. Augment the other sets of intelligence.
- [174:45]
John
But yeah, there are certain brands that will get a better return on investment from being in the super bowl than others. If you are a no name brand and you, you just put up a 32nd spot in the super bowl, people are going to be like your web.
- [174:56]
Tomas Puig
Server crash, like all sorts of things.
- [174:58]
John
Yeah, yeah, your web server, you might not be ready for it, but also people might just be like I'm not ready to learn about that as opposed to I actually going to am in the market for beer. Right now. And thanks for showing me those Clydesdales.
- [175:08]
Tomas Puig
That makes sense. But I mean, to kind of like talk to your question like that is an absolute dream that I have, right, of being able to level the playing field across that, but also provide, because.
- [175:17]
Jordy
That'S one of the advantages of these massive businesses. They can spend $20 million to figure out what's really working and then do a lot more of that, whereas a small firm is like kind of doing the vibes based analysis. Started my career helping companies like advertise on YouTube channels and with podcasts and you know, they might run a $200,000 campaign and there's some like direct attribution that they get either from a landing page or a code. But then they're like, wait, our conversion rate is just going up on the site generally? Is that being driven by changes that we made at the site level, changes that we made to the offer? Is it just overall lift from podcast advertising or is it some other strategy entirely? And so the more you can bring real business intelligence to small companies, the more they'll be able to actually compete against the big guys.
- [176:05]
Tomas Puig
And that makes me happy because everybody should have the level playing field and whoever has the best strategy and product should do well.
- [176:10]
Melissa Tokmak
Right.
- [176:11]
Tomas Puig
I think that it's important to note as we kind of talk about this thing, we spent years, years building the signal processor for this thing. We had to figure out how to bring in all this unstructured and semi structured data and be able to basically do datadog for unstructured data first before we could even try the causal thing. And so we have years of working on that. And that ingestion pipeline, that skill there is what allows us to do what you're doing. Talking about you can't just be like, I'm going to slap a model on top of it. Right. You actually have to be able to have a sensor that can actually understand every data feed.
- [176:44]
Jordy
Have you found a company yet that Jeffrey can't get an intro to or directly connect you to the CEO? Like this guy?
- [176:51]
John
Imagine this.
- [176:52]
Jordy
Oh, you want to meet this guy? Yeah, I'll give him a call right now.
- [176:55]
Tomas Puig
You ever want to not take a bet on something? That's one of the things I would not take a bet on.
- [176:59]
Jordy
Right. Jeffrey will find them, hunt you down.
- [177:04]
John
Well, congratulations on the progress. Thank you so much for coming by the Ultra Dome. And yeah, good luck with the next, with the next phase, putting the capital to work. It's going to be an exciting time and I'm excited to hear more of these case studies as they roll out.
- [177:17]
Tomas Puig
Yeah, I really appreciate both of you.
- [177:20]
Jordy
Let us know when you're ready for your first podcast. Customer. We got. Yeah, yeah, we got to figure analysis that we want to do. I mean, we just do the. We do the vibe space.
- [177:27]
John
We did a billboard. We did a billboard in New York.
- [177:32]
Tomas Puig
We should just do it for fun anyways. Like, there's nothing more that I like than, like, looking at data sets.
- [177:36]
John
Yeah, yeah, we ran. We ran a billboard campaign in. In Manhattan. We ran two. Exactly two billboards. We have no idea whether or not it worked. It seemed to work because people shared it a lot on social media.
- [177:46]
Jeffrey Katzenberg
For a million dollars, he could tell.
- [177:47]
Melissa Tokmak
You.
- [177:50]
John
Bargain that time for you all.
- [177:53]
Tomas Puig
We could do some drinks.
- [177:53]
John
It'll be fun. Well, thanks so much for coming on the show. Congratulations on all the progress and we will go back to our regular scheduled programming and I will tell you about Adeo Customer Relationship Magic, the AI native CRM that builds scales and grows your company to the next level. Buco Capital bloke is also blackpilling on the timeline. It is a bloodbath in the markets. Nvidia is down 4%.
- [178:19]
Jordy
Microsoft's Michael Burry Rage quit.
- [178:21]
John
Michael Burry Rage quit. We need to talk about Michael Burry. I'm sure, sure there's something in the stack. Let's go through some quick updates as we run through the show. Oh, this is cool. This is a white pill. Google the mind. SEMA2, our most capable AI agent for virtual 3D worlds. Tyler, what's the deal with SEMA2?
- [178:39]
Tyler
Yeah, this is really cool.
- [178:41]
Jordy
So this is a.
- [178:42]
Tyler
It's like a general model that can basically play, like, any video game.
- [178:45]
John
Sure.
- [178:46]
Tyler
Which is different because, like, you've seen a lot of, like, even early OpenAI, there was like, different.
- [178:51]
John
This is amazing for me because I don't have any time to play any video games anymore since I have kids. And if with this agent, I could just tell it to go play the game and then I could go have.
- [179:02]
Jordy
Fun and then describe the fun you had to fully experience it.
- [179:07]
John
Email it to me, and I'll have Nick read the email and summarize that to me in a text message. That would be my experience of the video game. No, seriously, what I actually want this for is I hate how modern video games that take 20 minutes to set up. Have you. Have you ever experienced this? It's like going through the tutorial. Remember when I. Yeah, when I made you play Halo. When I made you play Halo, it took like 10 minutes. For you to actually get into the game and then the actual game took you five minutes to play. And so I like, there are a lot of times when I hear a new game, I'm like, I only have 20 minutes in my weekend to play this game. I want to jump straight into the action. I don't want any of the opening unskippable. I don't want any of the tutorial. Learn how to jump, learn how to crouch. I already know how to move. I know what the stick does. Don't tell me that this is going to solve that for me. Hopefully.
- [179:53]
Jordy
Yeah.
- [179:53]
Tyler
But this is actually, I think this is like one of the most interesting papers this year.
- [179:56]
John
Yes.
- [179:59]
Tyler
A big part of it is it's general. So you can put in the new game and then it does like self play essentially.
- [180:04]
John
Sure.
- [180:05]
Tyler
Which is like really important because that's like it's teaching itself. This is basically the first agentic model that can see a new completely new environment and then do self play where it gets better.
- [180:17]
John
I wonder how wild. I'd love to know how diverse the inputs are. Does it expect an Xbox controller's worth of inputs? Does it expect a keyboard's worth of inputs? Because that's more inputs that it would need to learn. That's fascinating. And then I wonder what happens if you start marrying it to generative world models in the future as those become games. Like you have this weird agent on simulated world. You're simulating both of these. That's very interesting.
- [180:43]
Tyler
A big part of it is they connected it to Genie 3.
- [180:46]
Jordy
They did, yeah.
- [180:47]
Tyler
No way it worked. So you have a generative basically video game. It's like generating frame by frame and it has a little keyboard and stuff. And then you have the generative agent that learns what the world is and then actually plays it. You see this like flywheel kind of starting very well.
- [181:02]
John
Whether you're going long or short. Short. Go to public.com investing for those that take it seriously. They got multi asset investing industry leading yields and they're trusted by millions.
- [181:09]
Jordy
They also acquired Alto. Ira Crypto ira. You're now going to be able to hold digital assets in your retirement accounts. All on public. They acquired Alto for 65 million that got announced this morning. So great pickup from the public team. Major white pill Mira Moradi startup Thinking Machine Labs is in early talks. We love early talks.
- [181:34]
John
Love early talks.
- [181:35]
Jordy
Of our favorite talks to raise a new round of funding at a valuation of roughly 50 billion more than 4x their valuation.
- [181:42]
John
Everything's cooking a few months Stop Blackpilling.
- [181:46]
Jordy
In other news, Anthropic disrupted a highly sophisticated AI led espionage campaign. The attack targeted large tech companies, financial institutions, chemical manufacturing companies and government agencies. We assess with high confidence that the threat actor was a Chinese state sponsored group. I guess they were using Claude. I think is.
- [182:09]
Tyler
Yes, I think they were using Claude code actually.
- [182:12]
John
Weird.
- [182:13]
Jordy
They said they were vibe coding espionage.
- [182:15]
Tyler
Yeah, it was pretty funny. I read through some of the blog posts and it was like some of the interactions of the hackers and it was like this is what they were saying to the model. They were like okay, good job Claude. But I think this part is wrong. You can see like the actual transcript. Very bullish for anthropic.
- [182:32]
John
Well go to 8 sleep.com get a pod 55 year warranty 39 risk free trial free returns, free shipping. Michael Burry appears to be shutting down Scion Asset Management. He said dear investors with a heavy heart I will liquidate the funds and return capital but for a small audit slash tax holdback by year's end. My estimation of value insecurities is not not now and has not been for some time in sync with the markets. With heartfelt thanks but also with apologies I wish you well in your future investments. I do suggest investors contact my associate pm.
- [183:06]
Jordy
Did he really, did he really quit right before the market started correcting? Is this one of Those like you know 90%, 90% quit right before. 90% of gamblers quit right before apparently this they finally call the top correctly.
- [183:21]
John
Yeah, it does seem odd. I mean if he, if he, if there is a crash and he was going to be short but he pulls out before like getting that short thesis to work and realizing the results of that, it really changes his legacy. It changes the meaning of that meme. It changes the meaning of the Michael Burry image in my opinion. But we'll see. I mean he might be out of step for years and we might look back on this and say that it was great that he got out and it was great that he didn't short the greatest market.
- [183:58]
Jordy
His memory will live on through meme images from the big short and through.
- [184:04]
John
Whatever he gets on his wrist. Go to getbezel.com your bezel concierge is available now to source you any watch on the planet. Seriously. And watch.
- [184:11]
Jordy
In other news, Paramount, Comcast and Netflix are preparing bids for Warner Bros. Discovery. They will have until November 20th to submit non binding first round bids. Yeah, Warner Warner Discovery is holding the auction process in the hopes of having it completed by the end of the year.
- [184:31]
John
Have you seen how expensive streaming is getting. This is on the COVID of the business. In fact, in the Wall Street Journal, the price has gone up pretty much everything. Netflix has gone from something like 5 bucks to 25 at the top end. Everyone's raising the price and now they're creating bundles of streaming properties.
- [184:50]
Jordy
Only going to be $2,000 a month.
- [184:52]
John
They're all going to have. Yeah, 2,000 bucks a month. They all have ads cost.
- [184:55]
Jordy
They all have different logins.
- [184:57]
John
All different logins. And you can barely rebundle them even if you try.
- [185:01]
Jordy
I wish that we could get Jeffrey in here to talk about this Warner Bros. Deal, but you're probably too close to the metal on this one to to provide to be able to really comment on it.
- [185:12]
John
Well, we'll tell you about wander instead. Find your happy place Book a wonder with inspiring reviews, hotel graded minis, dreamy beds, top tier cleaning and 247 concierge service. It's a vacation home, but better.
- [185:20]
Jordy
In more news, apparently vine is being rebooted under the name Divine. With funding from Twitter's former CEO Jack Dorsey. The app plans to feature more than 10,000 previously archived Vines and does not allow AI generated content.
- [185:37]
John
That's remarkable. There have been so many vine revival attempts. Elon was talking about bringing it back at this point. One more time I believe the founder of vine was talking about bringing it back and did a number of different projects. There was a project called V2 right at some point it would be fun. I was a huge fan of vine when it came out. I really enjoyed it as a new creative medium. It was very very interesting. Very very fun. Let me tell you about adquick.com out of home Advertising made 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. What else?
- [186:15]
Jordy
Other News Strategy has gone below 11 NAV for the first time ever.
- [186:21]
Tomas Puig
Whoa.
- [186:21]
Jordy
That's crazy that sailors BTC holdings are worse are worth less than their than their total debt. How is that? How is that even possible? That seems very concerning. Anyways, their debt has long maturities 2027 to 2032. They're not margin loans but eventually they'll be forced to sell if they can't make interest payments. They have a something like a $700 million worth of interest payments due next year and they have I think under 50 million of cash on hand at the moment. So they'll either need to raise more.
- [186:58]
John
Or start selling and just like immense pressure from the other products in the market. I feel like if you want access to Bitcoin, which has been Michael Saylor's flagship asset, you were able to initially mine it for free, which was weird. Then buy it on Coinbase with a somewhat clunky process. Now it's pretty simple. Now you can buy Bitcoin on a credit card, you can get it in almost every app, you can get it in an etf, you can get it in your retirement fund. There are a whole bunch of different ways to get exposure. That particular strategy is not the only one in town. Anything else Jordi, or should we wind down for the day and say goodbye to everyone? Leave us five stars on Apple Podcasts.
- [187:45]
Jordy
Everybody go DM the White House now on X and just request the hundred year mortgage. We need some type of bullish announcement, otherwise tomorrow will likely be even worse.
- [187:58]
John
Yes, well if you do, DMX has changed the way DMs work. Now there's unified DMs and encrypted chats all in one place. That is the last piece of news of the day. Well, thank you for tuning in. Leave us five stars on Apple Podcasts and Spotify and we will see you tomorrow.
- [188:14]
Jordy
Can't wait.
- [188:14]
John
AM Pacific.
- [188:15]
Melissa Tokmak
Sure.
- [188:16]
John
Goodbye.
- [188:16]
Jordy
Cheers.