Ep13. Silicon Valley’s Political 180, META AI, COVID Postmortem | BG2 w/ Bill Gurley & Brad Gerstner

28:03So what did we learn today? Because there were some debate as to whether 405 was going to be open source or not, Bill. What did we learn today and why is this so important? Well, I think there's two things we could talk about. So you tell me which one you want to talk about. First, there's the strategy behind why would a company like a Facebook meta choose this methodology? And then there's the question of what's happening with the large language model competition. Which one did you want to take? Well, let's talk about just Meta's decision to do this. Yeah. And the commitment to open source and why it's so important.

28:38Yeah. So Mark, first of all, everyone should go read what Mark wrote. It's not ultra long. It's been four or five pages, but he lays out a lot of it. You know, and he talks about when Meta has run into other closed system companies, it's created a lot of frustration and limitation to what he perceives is innovation. And he mentioned the Apple world that he's kind of forced to live in. Right. Something happened over the past. I'd say 15 years where some of the smartest companies in Silicon Valley have developed a new strategic play where they use open source as a defensive weapon rather than as an offensive weapon.

29:21That's my terminate on clumsy. One of the first to do it was Google with Android. And it's really hard for people to put themselves in the mindset when that iPhone shipped and when Apple had only let AT &T have it, and it was on AT &T's term. And Google convinced the rest of the world, all the handset manufacturers, the carriers, to support this model by telling them it was going to be open. Right. I think they backed up on that a little bit, but that worked like that got them to get behind this thing. And I guess if you look at the China market, the openness that AT &T or Verizon and Samsung thought they were getting into with Android did play out.

30:07Like there's no Google or Apple in the China market. But then others started to do this too. Google did it again with Kubernetes, which is a piece of technology that lets you move your workloads very easily between clouds. They were worried Amazon was running away with AWS. So they took a piece of technology that was inside their company, Kubernetes, and made it free. And they got the Linux Foundation to help organize, which the Linux Foundation is very good at. And they got IBM, they got, they got all these, hundreds of players behind Kubernetes. And eventually it became so successful that Amazon had to give up and support it.

30:48And that has allowed more fluid transfer of workloads and it's leveled to playing field a bit. Facebook created this thing called Open Compute Project. And they basically said, if you want to sell computers into our data center, we're going to write a list of specifications that those things have to meet in order to go in there. Well, one of them is basically no proprietary technology. So it's almost like a patent deflection move that says we're only going to support open source things here. And by the very definition, if it qualifies to come in, it doesn't have any proprietary technology, which has been, and by the way, people say, oh, well, who would agree to that?

31:36Well, Dell sells into there, Cisco. So like all these people end up kind of just agreeing and selling in. And so that is a remarkable move by Facebook to lower their cost to their infrastructure and to keep it lower and to keep themselves out of position where they can be held up. There's been a recent move in the map world to create an open source map project because they're worried about Google having too much power. So this has become a kind of a new go -to move, but you need a lot of resources to do this. You can't like us, obviously there are startups that get behind open sources, but if you're going to play this defensive move, you need to be a big player.

32:15You usually need to get the help of others. Here, Metta, I think, just became concerned, and you and I have talked about this, but I think some of the players in the AI were brought this upon themselves by claiming that they were going to change the world in such dramatic ways you awaken the giants, right? If you cause everybody's investor call to be, what are you gonna do about AI? You're gonna wake up. All the giants, right? And so in this case, Metta didn't want anyone to end up in an advantage position to them because they were so much better in AI, or they didn't want to become dependent on a third party for an AI tool that they needed and couldn't have.

33:06I thought one of the most interesting parts of the letter and Mark's pod on it today is, he said it's become personal. First, he points to Linux and Unix and it is just being a great unlock by unleashing the power of the long tail. You know I'm a huge huge fan of this. The long tail of developers. But one of the things he said, he's like, this is personal to me, right? Because he said with Apple, where we saw ecosystem control by a single company that extracted most of the rents out of the mobile ecosystem. He said it was bad enough that we had to pay a big tax, right? Because that tax hurts all the developers in the ecosystem.

33:49He's like, but you could tolerate the tax. He said, but it was absolutely soul crushing. He used that phrase, which I thought was interesting. When you develop work really hard on a product improvement and then you have a company that says, no, you can't release that into your product for whatever reason. Right, right, right. And so it seemed to me that this has become extraordinarily personal, right? Here you have a company that produces tens of billions of dollars of profit in the back room. It can fund all of these models for as far as the eye can see and they've made a decision to make it open.

34:25I think that has broad economic implications for the entire ecosystem, right? If you're a closed model company, the question first is, okay, I'm going to have to compete and keep up with this frontier level competition. That's hard enough. It takes a lot of resources. But number two, Mark's going to give it away, you know, Metta's going to give it away for free. Now you have to develop a business model that can compete with free. Yep. And so this reminds me, and we'll put a link to it in the notes, but years ago I wrote a blog post about Android and I said, it doesn't matter. This is so, so similar to what you just said.

35:04It doesn't matter if they can make money on Android. It turns out they figured out how to. Sure. But it creates a moat that's so wide around the castle. I think I drew a picture I talked about like, okay, yes, you have the water moat, but now they're going to char the earth for five square miles outside around the water. Yeah, you're not getting to our goddamn castle. And so it has, and you know, not every company can afford a multi -billion dollar moat extender, but it makes a lot of sense for these companies to do it. And there's no one today that is listening to Mark's podcast playing with the new models.

35:46By the way, on Grock, they're like nutty fast. And saying, oh, like, that guy's stupid. Like, no, it's saying that. Like, it looks like a genius today. He also said today that they've already laid out the compute cluster for llama four. So this was llama 3 .1. It was a large 405B model, also distilled into an 80 and an 8B or 70 and an 8B version. But they said they've already laid out the compute cluster for llama four. They've already laid out the data architecture for it. They've run the research on it. So it seems to me and our team's best guess is that llama four is going to come in the back half of the year, right?

36:28Maybe in November, December, timeframe. They probably don't want to be more than a couple of months behind GPT -5. So they probably think you're going to get a new release right after 40 from OpenAI that they can then try to quickly leap frog with llama four. But what was interesting, he said, not only have we laid out llama four, but he said it's kind of fun and fascinating. We've laid out five, six, and seven. Okay, and he said the models are all going to be bigger. They're all going to cost a lot more money, billions of dollars. And he kind of alluded to what Sacha said a few weeks ago that you and I heard, which he said, this isn't going to be a straight line.

37:07He reminded folks of the internet bubble and said, you may have to go through a bubble here. So if you're a venture capitalist bill and you're hearing this, okay? First, you got to go compete with free. Second, you're going to have to spend billions for a long time and these get, the incumbents are going to compete. Third, that we may in fact have to cross this chasm, some bubble -like chasm. I think what it does is it's also a shot across the bow of those folks who might want to start close model companies to go compete with that. No doubt. And one other thing I would highlight for the listening audience, like every startup, whether you were an AI startup or whether you're a startup that's been around a while that wants to enhance their product with AI, this should make you tickled pink.

38:02Like you should be so happy that he made these decisions and went, because as you said, he went further in terms of openness today than the part of what was before. More permissions and the hugging face team was clapping and celebrating on Twitter as well for this reason. And so for, as Mark had been like to call it a little tech, this is phenomenal. Correct, like phenomenal. It may not be for OpenAI, which I don't know, no Brad Gabbard, but I saw a funny tweet today where he said, is OpenAI the new Netscape, which is somewhat provocative. But Mark created that question. And I would say, I'm gonna stop and let you go, but like last thing, when I just don't think that Sam Altman realizes when he talks about things in such a grand fashion, how it makes people like Zuckerberg feel or if Zuckerberg has to take a question, is AI gonna make you irrelevant?

39:07Like right, it gets under skin a little bit. I do think about it this way. You know, if you're OpenAI, you gotta go raise your next four or five billion dollars, right? Remember, meta had allocated $20 billion a year to reality labs. And there was a report last week that they may, you know, tighten their belt on reality labs to the tune of 20%. Okay, so that's at four or five billion dollars of savings on a research project that they have in reality labs that they can reallocate to AI. Now, what I would say on the OpenAI front, I actually think that they're doing extraordinarily well. And here's the interesting thing about OpenAI.

39:51In the face of all this, their consumer, you know, Zuckerberg said, I wanna have the number one consumer application in AI by the fall. But if you look at chat GPT 40, since it was released, web visits have gone from 1 .8 billion to 2 .6 billion visits per month. DAUs have gone from 60 million to 100 million. And so, you know, as I say to my team, there's only one way, you know, to build a business. Either you gotta get a consumer to pay you or you gotta get an enterprise to pay you. And the only other company that's really shown traction with consumers paying, you know, at that scale is OpenAI.

40:32So, the question is, can they build a consumer product that is durably better than the alternative, right? That Meta's gonna put out there. And Mark mentioned that there are hundreds of millions of people using Meta AI, but they're not really paying a subscription and using a standalone product, right? It's embedded in all the other products. A lot of the time you're using it, you probably don't even know you're using it yet. But I do think that Meta has a history of grinding out good products. And I suspect it will get there. But no doubt about it. This is gonna put a lot of pressure on the closed source model companies.

41:11When you, and I mean, I, like those numbers, you're referring to as thinker or just at a high level, they're not necessarily the paid constituency. No, no, those are total users. Yeah, and I playing around with credit card data, I think OpenAI is about two billion, and theoropics about 100 million in perplexities at like 25 million. Yeah, just I'm guessing that's directly right. No, no, no, I think that's directly right. On the 20 dollar thing, but at 12 months, the retention, even in OpenAI is like 35%. So I doubt that anyone that is looking at that, even at two billion, feels great about that as a durable business model.

41:55And you add to it, Mark saying, Mark's not gonna charge. It would feel really great if it didn't cost you $3 billion every year to build a new model in order to compete. I mean, that like ultimately the cogs in these businesses, right, that you have to, because these are super quickly depreciating assets, right? You have to, you have to keep reinventing that model every single. By the way, I pay for all of them, I play, because I love playing with them, and they're so much fun and so fantastic. I would say I think one of the retention problems is simply all the substitutes that are available. And it matters, and then it speeds gonna matter.

42:33There were these old stories of Google made itself like 30 milliseconds faster. Like there were more searches. Like, and that, play with the Lama Grock demo. Like, it's a different experience. I don't know how much that will drive. Yes. So anyway, I just think there's a lot of alternatives. If Mark, if we're right about Mark's intention to, he took it personally, he's playing defensively, he's gonna make his version free. Yeah, there's no doubt about that. And he might run it for four years with no ads. Right. And I think this fast inference, and it's Grock, it's Cerebrus's fast inference, you're fireworks and other benchmark investment, I think, you know, as inference optimized on top of Nvidia.

43:24Right. You're now getting to the point with these inference engines that they're going to be, you know, it's faster than you can read, certainly. You and I had a conversation about this. That doesn't really matter because all of these interactions, you know, particularly when we think about multi -chained agent interactions, right. You may have a hundred different interactions, computer to computer interactions, before you ever see an answer. And so speed really matters because computers can talk to each other at, you know, the fastest speed you can possibly, you know, kind of contemplate. And so reducing that latency unlocks a lot of innovation, you know, that will be good.

44:05And I think there's a lot of that coming. You know, I was really impressed today with the alliances that Meta had built. You know, Mark pointed that out, but really from AWS to Databricks, to Snowflake, to Accenture, You know, the rocks of the world, fireworks, etc. Everybody, you know, launched with optimized versions of this, they've been playing with it. So, you know, in the case of Databricks or Snowflake, they can basically take these optimized models immediately to their companies. They can do fine tuning, they can do synthetic data generation, whatever their companies might want to do. So part of the way the open source works is by running the field, right?

44:53Getting everybody into the pool. No, yeah. While you were saying that, I went back to the Kubernetes Wikipedia page. So on launch, the principal competitors were VMware, Mesosphere, Docker, Azure. And then, you know, a few months later, AWS came in, but yes, if you're a participant in the market who is looking for your own solution to be advantaged and there's someone who wants to make a piece of it free and open rather than closed and paid, right? That's awesome for you. And so you write it, but it comes as a tractor that brings the other parties to the table to play. Right. And yeah, that's part of why it's so powerful in this defensive way, what I call a defensive strategic play, you are able to bring the other people in.

45:47Right. It's really fun. And once again, I think it's good for entrepreneurs. I think it's good for society. Right. Like 15, 20 year patents on big farmer drugs is not good for society. This is the opposite. This is technology getting freer, cheaper, more available. You know, and I'll give you credit for this. A lot of conversations I've had in Washington, your work on regulatory capture has actually made its way into a lot of Congress people's minds and offices, right? They're on the lookout for regulatory capture. Awesome. Right. And so I think it's pretty wild. Zuck went from a few years ago, you know, having a bullseye on him, so that's what it's for DC.

46:30And now, you know, he's somebody that people point to as a real asset against, against closed models. Yeah. And I hadn't thought about this until you just said that. So thank you for triggering this in my brain. But maybe there's a bit of a halo you get for all business. For sure. Yeah. For sure. So you look like, you know, you're the good guy. And not the bad guy. So, you know, by the way, by the way, I do think it's worth mentioning on the AWS thing. Just from talking to entrepreneurs in the AI space, Microsoft Microsoft certainly is advanced by being or Azure is advanced by being the hosting system that has access to open AI.

47:12Oh, for sure. And Amazon doesn't. Yeah. So not surprised. Yes. That did AWS was in the align. Yeah. Yeah. It makes it on long day. Makes sense. They want to they need a response to that. Yeah. And, you know, we could talk about this for a long time. Let's continue on. We have a few more topics we want to talk about. We'll make these a little bit more lightning round. But you had me listen to a podcast. You know, this week excellent podcast with Dr. J. Bhattacharya and Rick Rubin. You know, J. Is a Stanford kind of economic epidemiologist was one of the senior authors on the Great Barrington Declaration in April of 2020.

47:56And that was really this the seminal paper that showed that the COVID death rate was far lower than people thought and that the virus had penetrated, traded far deeper into the population by April of 2020 than people thought, which then had broad implications on whether or not we should lock down. You know, we're now four years later. Tell us why, you know, you were telling all your friends to listen to this pod. Yeah. You clearly are still agitated about the fact that there's been no post mortem here. No. Why is this so important for society that, you know, that we listen to this, that we come to terms with it, that we think about this?

48:37Yeah, I would encourage anybody and everybody to listen to this. It's on Rick's podcast called Techno. I can't pronounce it. Techno. Yes. Well, give up. Well, Rick obviously is one of the most interesting Americans on the planet. An incredible record producer. He wrote this great book on kind of design and how he thinks about creation that came out last year. And his podcast has this breadth of guests that's just phenomenal. So I listen to it quite a bit and I'm a huge Rick Rubin fan, even though I can't pronounce the name of his podcast. And, yeah, so he had Jay on and they talked for two hours kind of about.

49:22And they covered, I think, maybe like a good two, three, four year period of COVID and everything Jay went through. And, you know, you were talking earlier about like the media just kind of blocking and not doing research. I think if everything from the origin of COVID to the responsiveness to COVID, for whatever reason we were in this weird place and some of this came out in the Twitter files where if you said anything that wasn't 100 % consistent with what Fauci and Collins were saying, it was labeled misinformation conspiracy theory. And I did a bunch of research at the time, but I looked and found this incredible New York Times piece about the tower collapse in Miami.

50:10And maybe we'll put it in the show. Yes, because it's just so amazing. So someone spent, you know, six months, probably thousands of hours of investigative journalism to figure out exactly what happened. Yeah. And they had these infographics. They had all, it's like amazing. And you look at the origin of COVID, we have, I don't know, 20 million people dead, probably the worst catastrophe since World War II. And no one was looking. Like no one was looking. No one was doing the work that these New York Times journalists did for this collapse of this building. Yet the consequences were a thousand XM, like 10 ,000 X bigger.

50:51And I don't know if it was Trump derangement syndrome. I don't know if, you know, I don't know why. I mean, and the Twitter files once again had a lot of this where people tried to talk up. And so to me, one thing should be obvious to everyone. We don't want to go through this again. Right. Yet you look at how people talk. And if you hadn't listened to Jay, I would, I would be, I would say, if this happens again, are we any smarter? Are we going to make any better decisions? His concern is that we're not. Yeah. Right. And so I would argue that the reason to do the postmortem is that, you know, you had three prominent senior authors on this paper, 60 ,000 signatories.

51:36And it basically was early in the pandemic and it odds with the conventional wisdom. Okay. The conventional wisdom was that you needed to lock down key people separated that it wasn't that deeply penetrated, you know, of disease. And we could prevent it from spreading. He was basically making the case because they, they, he had instant data. He had, he also had incidents that locked down, to have massive consequences, which they did. Right. Right. But he wasn't given a voice. And Francis Collins called him a fringe epidemiologist. And this is a celebrated. In fact, he's, he's, his, his area studies economic epidemiology.

52:21Yeah. And so he's actually thinking about how these things interact with our global economic system. Right. He's like the perfect person to listen to. Exactly. And he got attacked. And I think any time. And I might, I might include this in, in how some of the people responded to Ben and Mark. And, and I remember I triggered on Kevin Scott, when people questioning LLM scaling called him trolls. Like any time your reaction is not to react to the argument, but to throw this label, this negative label out of the other side. My brain says, you don't have the goods. Right. Right. And that's what he said.

53:00And what was interesting here is, you know, it was all ad hominem attacks. They wanted to be, you know, it was like excommunicating them from science, you know, because you had the head of the NIH and you had Fauci who were causing calling them fringe players. They didn't deal with any of the facts, right. The science was that they had the sewage data from Santa Clara County and from L .A. County. They knew the penetration and they of course knew the death rate, you know, that they could calculate. They knew among younger people that there was almost zero deaths. And among older people, you needed to take a different sort of precaution.

53:41Now, Jay, what makes him interesting is, you know, he actually had two relatives in India who died from COVID. It wasn't like this guy didn't care about COVID. It wasn't like he was, you know, some cookie is one of the most celebrated scientists who actually looked at data, did the research, put together, you know, a very important paper. Now, this is a person who was celebrated by the NIH. He was on NIH review panels. He had written over all of them. Right. All of a sudden, they didn't like 100 different papers, et cetera. And the reason, you know, so when you said, I need to listen to this, I'm thinking to myself, that's been four years ago.

54:19But he says something at the end of the podcast where he said the problem is we're no better prepared today than we were then. And the exact same rush to judgment, right, lock everything down, we're prone to do it again the next time because we haven't done the proper post mortem here. And the proper post mortem is that we need to do the science and the research. And one thing, obviously, I hope everyone goes and listens to it. One thing that he uncovers, they've gone back through and looked at excess deaths at a number of different countries and lockdowns achieve nothing. And, and I think he makes a solid argument.

54:59I think most people believe there were consequences to lockdowns. Definitely with kids lost generations of kids from school. And they're still very suicide rates in a lost education. And all of my kids were in high school and it was horrific for them not to have the experience that so many of us cherished from those moments in time. It actually, I get a little upset thinking about it. But there's more than that really because we need to go back to and look at everything. There were, you know, there were, there, there's a lot of questions about the origin of COVID. If you look at what's, what's being discussed with in Congress with, and what they're uncovering it, it's really, really needs more, more work to be done.

55:40And I think Catherine Bach and Alina Chand, there's people that you should follow on Twitter. They're uncovering that finally. There were these incentive systems pushed through hospitals where like you could charge 30 % more if a patient had COVID. Like, does anyone gone back through and see if that was really a good idea to that? Or did people take advantage of it? And did they over selectly? It probably did. There were, there were questions about how that I brought up during my regulatory capture speech about the different types of tests that could be done in their cost and which was more accurate.

56:13And all that stuff while we don't have a pandemic, this would be a great time to go in depth. And I hope that whoever is our next president will put together a panel. And I hope Jay's in charge of it. I have no idea if Jay has that kind of time. It's like to go review everything that happens so that we can be better prepared next. Well said. Well said. Let's talk about whiz. This deal Google was rumored to be buying whiz for 23 billion bucks. There's a second hope high profile deal. The first was HubSpot that seems to, you know, have kind of blown up that Google was supposedly going to do. Asaf the CEO of whiz came out and said not to worry, you know, we're going to get to a billion dollars in revenues and then we'll go public.

57:00You know, any anything to take away here? I mean, my read on it just at a quick level is, you know, if this were to go public a pray trade at 13 or 14 times forward, that would be like 13 or 14 billion, not 23 billion. Are these things blowing up because people all of a sudden get cold feed about antitrust? Do we think that a Trump administration is going to loosen up M &A anything to see here? Well, I mean, one thing I would say is that these are, if you're lucky enough to be involved in a situation where one of these deals comes down, it's hard to know what to do. You know, and I mean, it's not nearly this number, but I remember when Instagram, you know, got their offer of, I feel it was 1 .1, 1 .2 billion and they had like 20, 20 employees.

57:52And we had funded the 14th photo sharing site and Matt comes in and said, should we do this? And like, it's easy to want to be on the yes in these situations because it's like, wow, this might, and I can remember a very famous situations where people walked away from these deals and never got close to that number again. And so the consequences are high. It's like a really intense poker moment where you're making a gut call. And so, and there's also the famous, you know, Google didn't sell to Yahoo for a billion and then became one of the most powerful companies of all time. Microsoft Facebook for 15 turn down 23 or whatever the number was.

58:37It's a higher numbers makes me a good, higher voice. Right. Right. Because as we've often discussed, like, it gets exponentially thin at the top. The number of people that make it to a billion is a fraction of the number that make it to 100 million. Yes. And the number that make it to 10 billion is a fraction of like, like, it's the air stand as you climb the mountains. Yes. Who does to them if they're just that confident in the business? There's some rumors today that it may be that the crowd strike situation makes them feel that they may have more running room. I would, I have a hard time believing this one was a concern about the antitrust because Google doesn't have a huge security business.

59:21And there's actually concern about consolidation around Palo Alto and Microsoft and Cisco in the industry are in crowd strikes. So, so I would think they'd be supportive of it. And if you thought Trump were coming in, I think it'd be less reason even to worry about it. I mean, I don't see that. Yeah, that's probably the one thing I would press on is, you know, the betting markets, I think are now 6040 with Kamala and the race of Trump winning the election. You know, we've been in this period in Silicon Valley for four years where companies just stopped. They may have sent their M &A teams to the beach for all I know.

1:00:04There just has not been a lot of activity certainly among the hyperscalers relative to what had existed before. It seems to me they did three things instead. They bought back their own stock. They all started issuing dividends and they bought Nvidia chips. Right? They, you know, they tighten their belts on people. They weren't spending more money on people. All the companies became a lot more profitable and there just was a real lack of M &A, which is a problem for Silicon Valley. Right. And I would say like having been in these situations before, usually the partner that's involved in the company, the one that's on the board or that led the investment, they're usually pretty, they tend to be overly confident.

1:00:44Yes. And that's often balanced by the other partners back at home. Right. So when you circle around the table and have these discussions, they're like, really? Are you sure you don't? Right. And if we've been in an environment where liquidity's been scarce, the odds that they were sitting there pushing, hey, hey, maybe we should do this. Maybe we should do this. I would think would be high. We saw this unusual Sequoia event where they actually recycled, striped out of their own dollars. Yes. Which speaks to the fact that the LPs feel a need for a cap. Having to come up with alternative liquidity in a world where M &A can't get done.

1:01:21And M &A was not a market bench list, but it could have been. Yes. Yeah. But it could have been as another, you know, another problem that we face in Silicon Valley, part of the vibrancy of this ecosystem relies on the, you know, not just the IPO market, but also the M &A market. Yeah. And so, you know, I think there's always been a peer pressure to appear particularly confident and to play the long ball game amongst venture capitalists. And they almost fall, they almost fall over each other trying to express this bravado. So kudos to index and Sequoia and others that have have turned down. Great lock.

1:02:05I'm just looking at the list, Andrew, and thrive that said no, I already in the team, obviously. But, but boy, you know, we talk about valuations being discounted future expectations. Nothing smells like that when you walk away from 23 billion. Yeah. You know, I think it would be less in the public markets today, but they've clearly built a terrific business. No doubt, I've heard nothing but good things. I don't, I don't. And by the way, one thing that could be very positive that comes out of this is if they turn and run at the IPO markets very quickly. Yes. Because I would love for the list of I think we have a bit of constipation here.

1:02:44And we talked about this at the CO2 event. I think people have gotten overly conservative about what they need to be a public company. So I hope they run as fast as humanly possible. We've had talk about these, you know, potential AI, I .P .O .s coming. I would I'd love for you and I to on this call to be talking about I .P .O .s more than we are. Yeah. You know, I'm I just got off a board call with company where we're mutual investors in that's this going to come public in September, October. And, you know, it's exciting to be talking about companies coming to the public. No down, no down. And I think I think it's best days and and it's maximum innovation is still in front of it.

1:03:29Right. And I was just reminding them like all of that can occur post going public. Right. And so I think for Wiz as well. So today, you know, Tesla reported tonight, they miss their numbers a little bit on on on margins, the stocks down seven or eight percent. But interestingly enough, Elon said on the call, if you don't believe that Tesla is going to solve autonomy, sell our stock. Okay. So he's like, this is not just about the number of cars we sold in the car quarter gross margin, like the reason you're in this stock is because of autonomy. And so I want to talk a little bit, you know, you and I spent a whole pod talking basically about 12 .3.

1:04:12But, you know, now they are on to 12 .5. He mentioned 12 .5 and 12 .6 on the call today. You know, what we've been able to discern is these models are much bigger. Right. They're running like 5x larger models on the edge on the car. So rather than a billion parameter model, maybe something like four or five billion parameter model. Yeah. The rates of improvement continue to accelerate. And so here's how it manifests itself, Bill. You know, in we went and did some testing of our own on this in 12 .3. You know, you could take your hands off the wheel for like 40 seconds before it told you to reengage by 12 .4.

1:04:53You could take your hands off the wheel for up to four minutes. Right. A lot longer period of time. And it has eye tracking. And so you could look away a little bit before, you know, it was telling you to take control of the wheel. On 12 .5 and 12 .6, we think they get to the point where there are 100 % hands free. So it doesn't constantly ding you to grab the wheel. And you can really start looking away from the road for longer periods of time. Now, of course, you won't have to do that. But I'm just saying it will do it before kicking in its warning systems because they're confident in the efficacy of the model.

1:05:32In order to bet, you know, Elon said because the 12 .5 and 12 .6 are so good, they're going to look for approvals in China and Europe around, you know, around. I was wondering when that was coming because the way most of the world, they have to apply for licenses for the car to run independently. And so I knew that would be a step before they were live. Interestingly, one of our analysts who covers, but they said Europe and they didn't. And not US. Well, in the US, they are ready can run full self driving in states like Texas, but not without a driver that they'll apply. No, no, the drivers in the car drivers, drivers in the car.

1:06:14So one of our analysts went to China to test all the L2 and L4. You read you've read something. You read the headlines of like deploying 10 ,000 cars and Wuhan overnight, et cetera. And here's what she, you know, came back to she came back and reported. She said, you know, they're still catching up to FSD 11. Now remember FSD 11 was so they're like three years behind on full self driving. Those were still the deterministic models before the big breakthrough of 12 .3. She said, but here's what's crazy, although they're far inferior models that Chinese consumers seem very receptive to being hands off the wheel, eyes off the road.

1:06:59And everybody's at this Huawei, Xiao, Xiao Ping, Neo, all these different companies. And so she was in them. And she said she had some really lousy experiences in the cars, but she was blown away at the Chinese consumers willingness to adopt the technology. Well, keep in mind, I mean, you've heard a similar story about like leapfrogging technology. So in Africa, there, you know, there's no landline, it's straight to mobile. And so everything, you know, changes in a different way. Car ownership in China was relatively low, like five or 10%. So most people are either bicyclists or or commuters. And so that would be one reason why your expectations would be in a very different place than if you were American who loves your car.

1:07:47Right. Right. I mean, we, this is deep, this is for another pod, but we spent, I don't know, 100 years, 70 to 100 years, maximizing America's structure and layout and topology for that damn car. Yeah. Yeah. One of the things you and I debated back and forth on this pod is just the impact that AI is going to have and the timeline against it's against which it's going to have the impact. And you've called into question, right, the the magnitude of the impact perhaps from the LLM, but we've been in lockstep that this is having a major impact on full self driving, autonomy and, you know, likely on Robo taxi.

1:08:33And so it does feel to me like this is one of those places where, you know, now if you're in San Francisco, it's not surprising to see two or three way mose around you. I think you're going to see a lot of people driving on the streets in their Teslas and then not too distant future, right. They're going to be reading a book. They're going to be looking away. Their hands are not going to be on the wheels. And, and, you know, and obviously in the robot taxes reported to pull the steering wheel out of the car altogether. By the way, on that last day, I'm going to put one more podcast in the list that I was that found super interesting, which was Misha Laskin, who is an entrepreneur that runs an AI company here in Silicon Valley.

1:09:16He went on the Sequoia podcast, but he was at he was at Google deep mine and he talks about AlphaGo. And he has some different perspectives on AlphaGo and and, but when, when problems are finite, and there's this great book about infinite, finite games, like chess, the ability for these models to compete and innovate is so much higher than when you have open -ended problems. And one of the things about self -driving is I think you define it in a way where it's a finite game, which is no wreck. Like, no wreck becomes the finite game. And so they're able to these models are able to go a lot farther when there's a finite game that can be played.

1:10:01And anyway, his, his talk is super interesting. And it makes me think about the areas where AI is going to accelerate faster, because you can do those kind of things. The open -ended AGI thing is actually, in my mind, is going to take the longest, because it's the hardest problem. Yeah, I think there are a bunch of discrete things you can define as finite games. You know, customer service being an example of that. But conversion on a website, for sure, potentially be done that way. So what the heck happened with Crowdstrike this week, right? Not only did it lose 30 % of its value, all my friends were complaining that they couldn't fly, you know, get from point A to point B, airlines ground to a halt.

1:10:45What do you think this reveals to us about kind of the nature of systems that are being built in the cloud today? And what are the things we ought to be looking out for or concerned about as a result? A whole bunch of things went through my mind. One, the one that actually, I think, J. Cal brought it up first, but that is just so, like, such a strong argument. Why didn't you stage gate how this is rolled out? Like, how could you let it have this big and impact? Like, if you did increasingly large groups every two hours, you wouldn't have never had this impact on the world. The deployment engineer that needed to go on vacation.

1:11:26I don't know, like someone like so crowd. And I think that's on Crowdstrike. So the and there were other people. There's a Twitter account that's really good at kind of analyzing crisis PR who went through his stuff. I think they could have talked about things like that. Why that failed rather than like they gave a technical description of what happened to say, oh, we uncovered it. But that doesn't really help that didn't say why this isn't going to happen again because there was a policy process failure, you know, on top of this. The other thing that that if you read Matthew Prince who wrote a large analysis of this and keep in mind, Matthew runs a different company cloud flare in the security space.

1:12:11And he's very worried about a very specific thing here. I just went and read a whole bunch so I didn't know this at a time. But apparently Microsoft will argue the reason that these security companies have access to the kernels because they were forced to let them have access because of an EU ruling. That related to how big and powerful Microsoft is now there were no non Microsoft boxes that failed. And if and people will say, oh, well, you know, Apple doesn't allow you to have access to the car. And so what what Matthew's worried about is that Microsoft's able to use this to reverse the EU initiative and then push these players out and a lot of people don't know this but Microsoft's now one of the top four security companies in the world and may want to do that.

1:13:02And so that's a new kind of dimension of competition that I think people need to pay attention to. I'm also struck like I run there was a funny article that said Southwest was spared by this because they weren't on Windows 95 yet or they weren't on Windows like some they were on really, really old windows. And I'm and I just don't understand why anyone that has a terminal that matters is is running Windows at all. Right. Like maybe it's a testament to just the power of the network effect of Windows and all this stuff. But, but you would think there'd be a hard and Linux thing that would be so much less susceptible to this kind of risk.

1:13:51You know, we have these esoteric conversations all the time we're investing these startup companies. But you know, the world has become software and you know, we take for granted every day the amount of software that's running around us. It's getting us from point A to point B soon to be driving our cars for us, et cetera. And you know, with it, I think, you know, comes significant responsibility. And again, circling back to where we started the conversation, you start taking down airlines that really makes people in Washington unhappy. I believe that. And so part of the reason I think that, you know, the nature of the relationship between DC and Silicon Valley is changing is forever changing is also as simple as this.

1:14:42Our lives today are so, you know, twisted around all of these technologies, whether it's the phone we carry in our pocket or the software that's getting us from point A to point B to the AI we're going to be using in the future. And so I think that this symbiotic relationship, this, you know, figuring out how to make sure that we're managing the relationship between Washington and Silicon Valley is the most productive way possible. For this country is, is, is going to be super important no longer acceptable for us just to stick our head in the sand and say we're 2851 miles away. You know, engagement, I think is going to be the new course.

1:15:29And in that regard, I think policies are going to matter like I don't think any party can take Silicon Valley. One of the articles said that people were forced to implement crowd strike because of regulatory reasons. So they had it implemented because someone had told them they had to. So that's another interesting dimension to this whole thing. Hey, before we wrap, I have two things I need to do super quick. One, I'm going to try and pronounce Tetra Grammitan, which is the name of Rick Rubin. Perfect. And apparently it's an important Hebrew symbol from way back before Christ. So I want to do him justice there.

1:16:10And in second, I would like to call out with respect, Mr. Daniel Eck at Spotify, who just printed amazing numbers and took his stock to an all time high. He's been grinding away at Spotify for a very long time. It's a great product that I love to use super innovative. A lot of people thought he couldn't make positive cash flow because of the way that the music label deals are structured. But he's he's after it. He did. He had a fun social media post before he did his earnings call, which is. I thought it was clever. I think I think I love how he does this. And he's just he's so thoughtful. He's so wonderful.

1:16:49He's so available. He gives back to other entrepreneurs. He cares about, especially about Europe where he came from and the company and I anyway, I wanted to. Well, the stock's up 12 % today. It's up 100 % over the course of last year, 120 % over the course of like six and billion. Yeah, so he's built an amazing business, but more importantly, again, getting back to what we just talked about. He makes our lives better every day. For all those who use the product, our lives are just happier every day. So I could just do him on a great quarter. All right, great to see you. Let's wrap.

1:17:32As a reminder, everybody, just our opinions, not investment advice.