Sharp Tech with Ben Thompson - Tesla and the Road to True Autonomy, Strategy that Starts with the Dream, Regulation and Market Questions

Episode Date: October 17, 2024

Understanding Tesla’s approach to an autonomous driving future, why some observers think Tesla is ahead of Waymo today, and questions about market structure and regulation concerns as the future of transportation takes shape. At the end: An additional note on politics as a zero sum game, and a few thoughts on the rest of the F1 season.

Hello and welcome back to another episode of Sharp Tech. I'm Andrew Sharp. And on the other line, Ben Thompson, Ben, how you doing? Horrible. I'm just really suffering over here. No one likes to hear about fantasy sports. Is there any more narcissistic sort of possible podcast topic than someone talking about their fantasy team? My problem is not my team, though.

It's my league. we could not find a time where everyone could do it. So we're doing it over text, which everyone thinks is amazing and so convenient, except for me, because I'm the commissioner, and I have to keep yelling at people to make their picks across like four times zones. For the last 90 minutes or so, you know?

And I just all, I'm just completely frazzled right now. So this is, I mean, God bless this podcast. I need to be a professional. No one needs to hear about my personal problems. I need to buckle down,

focus, deliver the people what they paid for. Oh. Well, that's what we're going to do, but I did enjoy the last 10 or 15 minutes. I got to participate vicariously in your fantasy draft on our little video call here. And it made me excited for my fantasy draft. Basketball season is right around the corner. Great things are happening all around us.

It's ironic we've been talking about Elon Musk who is like managing, you know, Twitter and SpaceX and Tesla. And I can't manage a fantasy draft. So, I mean, I guess a little bit of respect is too. Yes. Well, you know, before we talk Tesla, there was. one thing I wanted to mention, just one aspect of the SpaceX booster achievement that I forgot to highlight. We were talking about how incredible that was back on Monday's show. Just rewatching the video the past couple of days. The thing I kept coming back to was hearing all the engineers

in the background losing their mind as it happened and hearing the announcers go crazy. It sounded like one of the announcers might be crying. Yeah. Yeah. It was just a very powerful scene. And so I was kicking myself for not mentioning that aspect of the experience on Monday, because that more than anything else is what resonated with me. Just nice to see a bunch of humans all rowing in one direction and working toward human progress there. And we're not talking about SpaceX on this episode because you wrote about Tesla this week in an article that synthesized some of the conversations we had on Monday's episode. One of my favorite Sertechery article in a while. So keeping with that theme, we've got a few more emails about Tesla and the autonomous

future. And we'll start with this note from Justin. And this is a little bit of a longer note. So just hang with me. Justin says, Ben, perhaps on the next Sharp Tech, it would be interesting to discuss the value of Elon companies commercializing technology quickly. For example, not only is Tesla betting that more data will lead to level five autonomy, but their customers are paying them for the privilege of providing that data. Most do so passively simply by purchasing a Tesla, which then runs the full self-driving models in shadow mode, meaning that Tesla can see where the human makes a driving choice that is different from the self-driving model and the AI developers can respond accordingly. A minority of these customers also provide data by actively using full

self-driving, having either purchased it or via the monthly subscription. As a result, I would argue that there is another way to think about whether Tesla or Waymo is ahead. On the metrics you listed, sure, Waymo appears to be ahead. However, in terms of overall value created, I would argue that Tesla is far ahead. And then in his email, Justin runs some numbers to estimate that Tesla is currently getting data on roughly 500 million self-driven miles per year, while Waymo is getting data on 26 million miles per year. So your mileage may vary, pardon the pun, on whether you trust those calculations. But he continues and says, as one additional data point, I would say the Tesla miles are more useful. I use full self-driving for everything from my 15-minute drive to work to my three-hour drives

out to the end of Long Island through horrific traffic. It's not perfect. And I intervene periodically. but it has changed the way I drive such that on the rare occasion, when I'm forced to go back to driving a car without full self-driving, it feels archaic and genuinely painful. I readily admit to being an early adopter of technology, willing to suffer the flaws of bleeding-edge tools before they are accepted by the masses. But full self-driving also passes the spouse test. My wife adopts technology on a more typical timeline and very much did not enjoy riding in or using full self-driving in its earlier iterations. But she recently admitted she also truly missed it when it was taken away on a recent

drive because of her failure to pay attention to the road. I've heard similar feedback. Is that a gentle letting down of the spouse or was that a little harsh? Well, no, I mean, I think that's fair. She self-admitted, right? Well, honestly, I included this because I relate more to the spouse and it is eye-opening and that it is also become addictive to her. But he finishes and says,

what do you think? This was way too long, but hopefully I've provided some food for thought here. So what do you think, Ben? Does that spark any thoughts for you? For sure, for sure. I would say number one, seeing the most recent full self-driving it again,

don't want to overindex on just a few rides. But I, you know, when I last bought a car in the U.S. that I use mostly in the summer, I just wanted something that was convenient. I needed to get it quickly. But one of my core requirements was it had to have adaptive cruise control because I've rented enough cars and used that enough and you're doing long highway driving that to me it is

just painful to drive a car that does not have adaptive cruise control. And I think I would probably completely relate to the full self-driving. Now I'm just like pretty irritated. I didn't get a Tesla, even though I was on this podcast previously saying, look, I live in the Midwest. I take really long drives. You know, actually, you know, now I'm all talk because now I was just thinking about the 12 hour drive I took to Canada last year.

Totally normal, you know, because we're, you know, I'm Midwesterner at heart. Human that would suck in a, yeah. I mean, we spent, we spent less time on filling up with gas to go there and back than like one charging session for a Tesla. So I mean, there's that sort of,

that bit still holds. But yeah, It's nice. Like I like driving, but there's a lot of driving that is not enjoyable driving. And it would definitely be, be sort of nice to have. So I think when Justin says that, like once you get used to it, it's really hard to give it up. I think that's probably true. And as a sort of side note, maybe this gets into the, I think one of the most more interesting things Tesla did over the last couple of years was they've opened up their sort of their charging network.

And so other companies now will pay them and they will basically accumulate licensing fees. So that's really attractive, right? They have this additional revenue stream going on. It's going to be the standard sort of the standard is going to be open, but they have the most infrastructure as far as the stations out there. And so that's going to be an additional revenue stream. But at the same time, it's like, well, that was a huge reason to get a Tesla is because, you know, you have access to their supercharger network.

And now if that goes away, why not just buy something else? And I think that's probably evidence of the high degree of confidence they had about self-driving that, look, no one's going to go anywhere because they want sort of our self-driving. driving technology. Yeah. Now. Well, and on the self-driving point and the note about his wife, it reminds me of having cameras and sensors in the car to help with parking.

And we also, we talked a little bit about back when we were talking about Boeing and Airbus and the different approaches to autopilot software and how certain pilots just become more reliant on the software and never learn how to fly manually. And over time, that's how I've become with parking. And just it's so much more annoying to get into a car that doesn't have the cameras showing you as you back up and the sensors. If I don't have and look, I consider myself a great parallel parker. But if I don't have that technology, it's just really, really annoying. And hearing about Justin's wife made me realize eventually that's going to be true for all of us with full self-driving technology.

You might have some, you might have to some self-interspection coming up. I'm going to come over and I'm going to paint over your camera, see how well, you parallel park then. Well, look, I'm still good at it, but I don't like it. And all I'm saying is I now recognize that eventually I'm going to feel that way about driving more generally, you know, maybe 10, 15 years from now. That'll be my experience as a driver. Yeah, for for sure. Now, to be clear, when I was with in Tesla's with that experience, there was at least two times that I can recall where the driver had to take over. And, you know, so it's not even remotely close to being ready.

To be clear, it's not what is in the field right now is not competitive with Waymo. Then again, the Waymo's are geographically constrained. And that is part of the level four designation is it's fully operational. And by the way, I did misstate about Waymo remote drivers taking over. And it's funny that I kind of, when I was going through and typing out the levels, I realized, wait, there's kind of a disconnect here because level four, means it can always operate on its own. That is actually correct.

Waymo sort of reached out. It's like, well, just to be clear, the Waymo can request assistance, but it has to always be able to operate on its own. Okay. And sometimes, though, that means it just stops, right? I think that's probably why it's not taking passengers on freeways yet. It is in testing on freeways. I think primarily in Phoenix.

But what do you do in a weird situation? Like when you're going, you know, 80 miles per hour or 75 or whatever it might be, that is definitely a bit hairier. but from what's out there it is it is sort of better but there was a couple things and I mentioned I appreciate you're admitting this is sort of a synthesis of stuff we've talked about so apologies if I'm sort of repeating myself but I think there's two points that I did want to sort of distill in this article yeah number one is it's just a fundamentally different goal to be talking about true autonomy versus self-driving cars.

And actually, now that I realized in my article, I almost should have made that even clear. Waymo is working towards self-driving cars. Tesla, at least according to their presentation, which is funny because everyone was mad about the presentation and its lack of specifics. But in many respects, I felt like the presentation actually said more if you sort of knew what to look for.

And one of these was, it was almost like a retreat from past presentations. They have gotten a ton and two specifics. but now there's a bit like, wait, where are we actually trying to go? And that goal is true autonomy. And true autonomy means everyone is, like self-driving cars is just part of the infrastructure that we have. And I think this solves a lot of sort of things.

We were in a group chat and talking about like, well, how is it, how is a Tesla going to get to like when it breaks down and how it's going to pull into a repair shop and all these sorts of things when there's no wheels or drivers? And well, the answer is if everything is like this, infrastructure is going to spring up to support it, right? It's like this fundamental change will change the environment around it to accommodate it. And this is incredibly audacious, right? Here's an analogy.

I was writing about Intel and AMD this week, right? They're having a partnership. It's like, and it's like, it's kind of an obvious thing to do because they need to evolve the X-86 sort of ISA. Right. They're threatened by, they're jointly threatened by arm. and then AI is obviously sort of looming over everything. So of course they're partnering together. But anyone who's been in tech for a long time, the idea of Intel AMD and AMD,

even agreeing on a meeting time is like incredible, right? These are the most arguably the most bitter rivals in the history of tech. Like Microsoft versus Apple is nothing compared to Intel versus AMD. And like they were in court for basically 20 years straight. And lots of dirty tactics, lots of dirty tricks, mostly by Intel towards AMD, to be totally honest. But, but, but, you know, in Intel, say, look, we invented all this stuff. They tried to copy us.

Like, like, of course we're doing dirty tricks, right? But, but they're sort of partnering together. And the reason is for a very long time, and Intel realized this back with a 486, where there was a big push then to shift to a different set of construction sets were called Risk versus Sisk. Risk is sort of theoretically more efficient, all these sorts of things. And, uh, there was a split. with an Intel, which way to go. Pat Galsinger, the current CEO, was actually on the side of,

no, we should keep with X86 because it doesn't matter if it's somewhat less efficient. The software advantage is massive. For to start out afresh, even if someone comes along with a different instruction set that is better, they have to rebuild all the software. By the time they get all the software built, we're going to, our manufacturing is going to be leaping ahead. And our speed advantage from manufacturing will overwhelm any sort of like marginal advantages they have in their sort of architecture. And that was the right call. And then, and Intel, funnily enough, made a mistake. They forgot that lesson. So you fast forward when 64 bit computing is coming along. And Intel's like, okay, here's a chance to start afresh. We're going to make a new ISA, a new

kind of chip around 64 bit. And by the way, AMD is not going to be able to copy us now. They don't have any licenses. They don't have all this sort of stuff that is downstream from IBM forcing Intel to second source or have a second source for chips. They had to share everything with AMD way back in the 80s. That's the, that's the genesis of all of this. And so Intel's like, oh, this is our chance to start afresh. AmD comes along. So Intel has this big thing to build this new sort of instruction sets. And they, they and HP were sort of the main two entities working on it. But there was other folks involved. The company that was not involved was AMD. And like, it was very clear what was happening. Intel was going to shut AMD out once and

all. So A&B comes out with their own sort of extension, which is, oh, we have a 64-bit instruction set, which by the way, has a compatibility mode, which means it's backwards compatible with 32-bit. And everyone's like, yes, that's better. Like, we don't want to rewrite all the software in the world just so it works on your chip. It's better if we can just use what we already have and transition into this new world. And it's funny, that was actually the key transformation of that relationship because Intel suddenly had to license that from AMD because AMD was right and that is what the industry adopted. And now they had the grounds for a cross licensing agreement where they both needed each other had a little bit more leverage. Yeah.

That's right. AMD needed X86 and Intel needed the 64 bit sort of extension that AMD created. Okay. But talk more about the distinction between self-driving and autonomy. No, no. I'm doing the weave as it's known. Yes. Yes.

Great. So what happened, though? And so that was like in both cases, the right decision was to optimize for backwards compatibility because the gains from a clean break just weren't worth it. Then the iPhone comes along. The key thing about the iPhone relative to Windows phone, Windows phone was trying to be an extension of your computer. It literally had a start button on it, right? It was like little X's in the corner to sort of close apps. Microsoft was like sort of porting the desktop onto the phone.

And you can understand why they did that. That's what big companies do. We have all these sort of advantages. We're going to put them into place. And Apple's like, no, it's a new device. It's a new interface, direct manipulation. We have to start from scratch. And yes, we're going to take the kernel like the core operating system from macOS.

That's the core sort of iOS. But everything above that has to be fresh. Oh, and by the way, this is going to be completely powered by batteries. So the number one most important thing is not. outright performance, it's efficiency. And so we're going to use arm. This is a more efficient chip and it has a better roadmap in terms of efficiency. We're going to go in that direction.

And the key thing there is that started a cycle of building all sorts of software and getting that was already an x86 and now it had to all work on arm. Arm at that point was mostly powering. It did power some phones, but it powered a bunch of like just not important devices. that was the key thing that elevated arm to, in the long run, being a competitor for Intel and being a competitor for AMD. And because it kick started this 15 year period of building, rebuilding all this level software. All the work that Pat Galsinger in his first stint at Intel said, no one has time to do that. Because why would they?

The gains aren't worth it. Well, if your most important thing is efficiency, then the gains were worth it. It was worth it to do all the work. And when all the work sort of happen, suddenly Arm is now competitive with computers. And now it's competitive with servers. And suddenly Intel and AMD have to become friends because they are threatened by Arm because of this sort of clean break. Clean breaks happen when the benefits are so overwhelming that infrastructure builds up around it to support the new paradigm. Right.

Now, to tie this back to Tesla. what Tesla is proposing. And just to be clear, this article was laying out how I think Elon Musk and Tesla are thinking about it. It was not, I don't know enough to say whether it's going to work or not. Right. But what I think they were proposing and why Musk spent this presentation mostly talking about this future autonomous world and didn't spend and rented out the whole Warner Brothers sort of studio a lot to sort of build a demonstration of this and didn't talk about business model or all this sort of stuff. is basically he's saying, look, building something that works in the current built environment is a waste of time and ultimately a dead end. Sure, it'd be nice, like a better Uber, but what would it mean to completely transform everything?

And what he's proposing is autonomy presents the opportunity for a smartphone-like moment, where the benefits are so overwhelming that it's actually will in the long run be worth rebuilding the built infrastructure to support this new vision. And that's the goal. That's the goal. I mean, there are aspects of the presentation that are really powerful in that regard. And you call them out in your article where Tesla and Musk are on stage and one of the slides shows, you know, Dodger Stadium surrounded by parks instead of parking lots or LAX surrounded by parks instead of parking lots. I mean, that's the autonomous future that we're all sort of yearning for. whereas Waymo, when you talk about their more confined vision, expound on why there are more limits on that side of the spectrum. Well, just to be clear, let me be super clear about this.

Waymo wants this too. I think if you talk to anyone at Waymo, they want this fully autonomous future. That's where they want to get to. And again, just to be clear, they're definitely much better right now. Their cars can actually self-drive. No, and I'm glad you're being clear because that's what was confusing when you drew the distinction between Waymo as the self-driving company and autonomous on the Tesla side. Right.

I think the question, though, for Waymo is with the approach they're taking and building up in the current environment. And right now, Waymo, like everyone is, of course, moving to full AI driving. But, you know, Waymo has a lot of sort of heuristics and guidance along the way and lots of mapping and the LiDAR. And they know every inch of these streets. And it's good enough. If there's construction, they can go around it. That's not on the map. Like, they're not completely map dependent.

But they are sort of a much more sort of brute force understanding the world. And that works better for what is available today. The AI is not good enough to drive cars today. It can help. It's getting more and more, but it's not good enough today. And so Waymo adding on all these sort of layers and all these use cases and all these sensors is definitely sort of better for figuring out what's out in the world today. The question, though, is, and this was sort of the analogy I wanted to make to sort of to SpaceX. SpaceX famously, they're focused on going to Mars, right?

And so if you're going to Mars and you're not worried about like anything else, you're going to go for all these audacious sort of things because you're just focused on how do we get the most stuff into space and the most efficient way possible so we can actually get it to Mars, right? We're going to have to have the largest rocket ever built. That rocket, to do it in a cost effective way, we need to be able to use it multiple times. But we can't have legs because legs are extra weight. And the bigger the rocket is, the bigger the legs have to be.

How can we get rid of the legs? Oh, we can catch it, right? You end up with chopsticks catching a rocket. That's downstream from why and you go to Mars. Right. And along the way, there's going to be lots of cool stuff, right? I thought one of the most interesting things that Elon tweeted this week was about this new satellite for Starlink that is going to drastically increase bandwidth and latency.

I was talking to Craig Moffitt and one of his hesitations about Starlink is, look, there's just not that much bandwidth. Well, what happens when these big new satellites launch, right? And that's, but that's a good example of if you always sort of start from the present and think forward about how to build, you end up in constraints and you end up in a path dependency that might not get you to the actual ultimate outcome. And that's like what happened with the European Space Agency, right? We're just sort of building better watchers, building better watchers.

No one at the European Space Agency is thinking about going to Mars. But when you think about going to Mars, you suddenly, you just end up going in a completely different direction than everybody else and lots of new stuff. all along the way. Now, to be clear, lots of times you have big dreams and it makes you do stupid stuff, right? Right. You could argue that's the case with Musk many of the times. But in this particular case, it clearly worked with SpaceX. By having a different goal, it drove them to make different decisions that seemed the chance of possibility of likelihood was very low. But if it worked out, suddenly there's a completely new path opened. And this is my question. Does this also apply to

self-driving versus autonomy. If you're like Google and you're trying, if you're like Waymo, you're trying to get your cars working in today's built environment and you're doing lots of things on top of that. And because you just want to get to function, does that in the long run actually transform things? Or do you end up in maybe a local maxima in sort of a dead end

where your cars are too expensive relative to truly transform things? You have all these. It's hard to scale into new markets. I mean, it's all relying on humans right now. Right. And there's a bit where, Like the AI, at some point, AI will be better than human programming. But as long as they're doing both, when do they have the confidence to sort of switch over to like pure AI, right?

You're kind of always, like, we already have this level. It's not as good. Like, imagine they start expanding the geographies and then the AI is not as good as they want to shrink the geographies. Are they going to be able to do that? Right. Like, you can, you can end up. And again, I'm not saying this is going to happen. I'm just sort of thinking through this sort of framework.

And so you have must talking about pure autonomy. How do we get to peer autonomy? Clearly the only way to get that is AI doing it all. Number one, you can't have human backups. They have to truly be able to do everything. And in many respects, that is clearly going to take much longer. But if you go directly there, does it make it more likely that you actually get there? You don't end up on a side quest, right? You actually end up on sort of the main thing. That's the North Star, not serving people in San Francisco and Phoenix. You also, you referred to the bitter lesson, which was a great little essay by Rich Sutton. Am I correct in summarizing the bitter lesson as the lesson that technological solutions that rely on humans to direct technology will eventually lose out to solutions that rely solely on technology to analyze and solve problems for itself? Yeah, really. So the bitter lesson is about AI sort of specifically, which was, you know, and to you. use some of, this is lifted from his essay. It's a short essay. It's in the notes.

Everyone should read it, even if you agree or disagree. It is actually one of the more important essays. Right. I would say there are some profound ideas in there. And again, it's like five paragraphs long. Right. And so you go back to something like chess, right? AI playing chess. And everyone tried to program every sort of thing that you might do.

And the solution that actually beat the best players in the world was just they listed, they basically programmed every possible. end game in chess and then it searched for the right next best move. Right. Let the computer figure it out. And it was it was almost demoralizing in a way, right? Because it turned out the solution. Yeah, that's why it's the bitter lesson. The solution was just to list all the data and let the computer search for it.

No human could do that. What it is, it's doing what computers can only do. And this is we talk about this, right? Remember our talk about advertising. Humans think put content up, put an article there. No, a feed works. It's something you can only do on the internet. In this case, only a computer can search all possible solutions to a chess game and provide the answer.

You fast forward to Go. The search thing doesn't work anymore. There's actually too many. Even that's too much for computers. Maybe in the fullness of time there will be a computer that can search every possible game for Go. But it's like astronomically large, like more than the number of atoms in the universe or something. That's not true. But it's very, very hot.

So the way Go was solved was basically you put the parameters of the game into the AI. and let the AI play itself a gazillion times. And the AI just basically played every possible permutation of Go playing against itself. And they just learned how to play. It's like deep learning. It's self-taught. And then it played a human and it beat a human. And actually one of the interesting things is I think to a much greater extent with Go than with chess, humans learn from the AI playing Go.

Because they do stuff humans never did before, right? Chess, it's more constrained. Everyone's kind of tried everything. But go because the problem space is so large. Like AI helped introduce completely new concepts and strategies to humans. And now even when humans play each other, it's a more interesting game because this entire latent space of moves was explored by the AI sort of on its own. Again, though, the lesson was trying to put in rules and heuristics didn't work. Right.

Just letting the computer do it, you can figure it out on its own, was better. Same thing with sort of vision and audio recognition. again, from examples from his article, trying to put any sort of rules. And the key thing is, every time the humans put rules in, it works better, right? Because, of course, you're solving all these use cases. Right. It works better in the short term, which is what's happening with Waymo. Then you hit a wall, right?

And what happens is just laying the computer figure it out does better. And so the theory here and the reason the argument why Tesla is actually ahead, which is not, you know, it is a, it's not, I don't think the majority view, but as a strong minority does think that. is because Tesla, number one, that the best approach to self-driving is basically the Go approach. Just let the computer learn how to drive on its own. Don't teach it how to drive. It will learn. But how do you learn? Yes, you can do sort of simulations and you can drive a car around.

It can observe those sorts of things. But you need, it's just the bitter lesson says, oh, actually just count for compute to catch up. Now, that was written before Transformers. The bitter lesson was. I think the other lesson for Transformers is you need the data. Like, overwhelming amounts of compute and overwhelming amounts of data solve all these problems. This is Justin's point. Tesla has cars, millions of cars on the road.

All those cars, because they're all connected to Tesla, Tesla can send out. And according to them, all those cars are self-driving. Like, they're running self-driving. Now, the self-driving might not actually be controlling the car, but it's running as if it were. data, yeah. And so of course the people who are doing self-driving are the best, because they go in and they interrupt it. And then that's like a red flag. Oh, someone interrupted it.

That's like when you're doing the RLHF, that's like reinforcement warning. That's super direct. I bet it has a big impact on sort of the model sort of over time. But even the ones that are out there, just it's in the real world and counting real things. And I think it's not an accident. In a 20-minute presentation, there was two minutes spent of videos of people doing real stupid things on the roadway, which you're not going to get in a simulation, right? I mean, it was so funny.

He's like, he timed it perfectly. He's like things you never see. And it showed the guy with like one leg rolling a wheelchair backwards, like across like intersection. So yeah, probably got into your situation. A lot of situations that an individual human will never encounter. But collectively, you know, Tesla is able to sort of amass this knowledge base that should prove useful somewhere along the line.

Yeah. And so again, will this all work? Maybe Waymo just, it's good enough. And they get there first. Maybe Waymo gets regulations written that you have to have LIDAR. That'd be a big sort of a heavy blow. I really, I would strongly encourage any regulators listening. The right gauge is measurement, like disengagement, human intervention, whatever it might be, which again, Waymo is way ahead on. We should not index on sort of equipment. Now, the equipment thing is a big thing.

You know, lots of people are very passionate. You have to have LIDAR. Like, humans are bad at depth. Yes, they've been emailing us over the last several days.

So the LIDAR contingent is vocal. Yeah. I think that the Tesla argument is, in Outeridge Carpathy basically said as much. The information is in the pixels. It's just really hard to figure it out, right? But technically speaking, they are seeing the real,

sort of the real world. And just the AI just has to get better to sort of figure that out. again, is that will that work is a completely open question. For all the emailers, I understand the argument. I'm not disagreeing with you. I'm also not agreeing with you. I just don't know.

I'm saying I don't know. But I think the point is if you do want to get to this autonomous world, and this is the other broad point, it doesn't just need to lead to a transformation of the built environment. You get to a transformation of the built environment when it's cheap. Cheapness matters. Like it matters as far as like whole scale transformation. The reason the internet changed the world is because it was free, right?

Like if everything, it'd always cost something, it would be a neat little adjunct. It wouldn't be transformational like it is. And so I think that is, you know, the Tesla approach, if it works, actually does lead to full autonomy because it's pure AI, number one, and number two, it's cheap. and does the Waymo approach get there? Well, maybe ladders get super cheap. Maybe we get sort of silicon or chip ladders that don't have as many sort of moving components and it actually gets super cheap. That totally might be the case. I'm not dismissing Waymo at all.

I just wanted to explain why the Tesla approach is not nuts and also why I think there is a consistent thread in Musk's approach, which is always towards scale and cheapness. And even with Tesla manufacturing, right? Why are all the Tesla's the same? Because they're going for scale. Like, why are they trying to figure out how not to paint cars? Because it's cheaper. Why are there no buttons or whatever in the cars? Because scale, cheapness, right?

Like, if you do it by computer, you save so much money. And of course, it's presented as, oh, it's so sleek and futuristic. There's a real cost aspect to that, right? That drives sort of the way they approach things. And again, I'm not saying it's going to work. But I think that it's that consistency. That's what I wanted to highlight. No, no, but it was, it was really interesting.

And we also have a couple more questions. But just to be clear, when we envision the utopia of the future where LAX is surrounded by parks, the key point there is that vision is only attainable and plausible if you can build an autonomous future on a cost structure that actually makes sense for that sort of. of scale. And I think a lot of the wider people would say, look, if you're making LIDARs in the millions, it's going to get cheap, right? And there's going to be technological breakthroughs that's going to make it cheap. And that very well may be the case. Maybe it's almost a race between LIDAR getting cheap and sort of Tesla, you know, it's the sort of hardware versus software sort of bit, which one's going to sort of happen first. But you would also need to phase out the humans in that scenario,

the humans that are going around mapping cities and helping Waymo scale, at least in these individual markets at this point. Well, the other thing is, it's just like, again, in theory, Waymo can dual track. Like, we're going to have human programmed heuristics and AI. And we're going to shift to AI when the time is right. It just the history shows that that's actually a lot harder than it is. Yeah. Like that's the bitter lesson, right? The bitter lesson, everyone likes to think, oh, yeah, the bitter lesson, good article.

I'll keep that in mind and make sure to switch to AI when AI is ready. I'll never make that mistake. That's right. And so, yeah, I mean, maybe I think it's definitely a good chance that the bitter lesson applies to self-driving. And if so, it's better to start with AI than not. The only problem with Elon, of course, is he's talking about next year again. Like, like, can he just stop with the dates? Even Carpathy was saying, you know, end of the decade, which feels more.

more realistic, you know? Right. And Carpathie is obviously, he was the head of Tesla stuff driving for five years. So this whole strategy is probably his, right? And, you know, he was like, oh, I don't know what's gone on with the team. Dude, you were in head of it for five years. I think you know what's going out of the team, right? And he is unsurprisingly bullish about Tesla.

He thinks they're ahead. You know, so take with the appropriate grain of salt. But he is saying 10 years. Right. Why does, like. He's not saying 2025, 2026. But it's all part of the ride, you know? Well, one note we got from Travis related to cost structure.

He says, hey guys, I appreciate the recent discussions about Waymo versus Tesla business models and autonomous driving. What is missing seems to be the elephant in the room for all AI companies. Compute cost. Sure, Tesla's Robotaxy may be $30,000. But Tesla is spending billions of dollars on Nvidia GPUs and Dojo, whatever that is. in order to make the $30,000 robotaxy possible. $30,000 then is only the marginal cost. Tesla's Cappex over the last 12 months was $9.8 billion.

For context during the Model 3 ramp in 2017, Tesla's Cappex was $4.1 billion. Tesla probably spent about $5 billion building full self-driving data centers over the past year, and that will go up. Any thoughts on that? So implicit in this argument of, about scale and cheapness. When I say cheapness, I mean marginal cost cheapness.

Like what's the cost of one additional unit? And the whole bit about compute is one additional unit is free, you know, barring power and all that sort of thing, right? And so like the, the space, like the, the starship is a great example. They spent billions and billions of dollars developing that. They have a long way to go to make any, make their money back as far as that goes. But the marginal cost of a kilogram going into space is, going to be drastically lower.

What Elon does is he takes this fundamental financial equation of tech, right? This starts, the reason why the whole Silicon Valley ecosystem exists is it literally started with Silicon. That's why it's called Silicon Valley. And the fundamental nature of chips is they cost a ton of money to develop. But once you're making them, every additional chip is basically free. It's just sand. Like, that's your input.

Now, it's overstating it a bit. You have to actually process it and get sell credentials and all, all these sorts of things. But TSM, they're spending $30 billion or whatever it might be or $20 billion on a fab. That's an unbelievable amount of money. TSM is spending so much money. And then the chips come out. And basically from TSM's perspective, like a wafer that goes in is a few hundred dollars. And from that wafer, they get 100 chips, right?

So their costs per chip are tiny. All their costs are depreciation of those fixed costs. And so that is how tech works. massive, massive, massive upfront costs and then basically zero marginal cost off the back end. This model is how venture capital came about because you need to get started. You need astronomical amounts of money, but your payoff is infinite because you can just make an infinite number of things, right? So Silicon Valley starts, venture capital is a part of the Silicon story.

Then software comes along. Software is the exact same model, but even more extreme because now it actually literally is zero marginal cost. It's just like sort of copying the whole thing. But again, huge amounts of investment up front to sort of make it work. And so software becomes huge. And then you have, venture capital becomes this massive sort of industry. What Elon is doing is he's taking that and applying it to the physical world. Like, what does it mean to take a silicon mindset, to take a software mindset, and apply it to cars? What does it mean to take it and apply it to space? What would it mean to be the Apple of cars is basically the Tesla proposition?

Actually, Apple is a useful example here because the brilliance of Apple is they make these incredibly expensive objects, but they do it at scale, right? It's the scalability that's amazing. And again, it's not a perfect example. iPhones have actual costs, right? Teslas have actual costs. Rockets have actual costs. But it is taking to the absolute extreme. To what extent people talk about Tesla's being computers on.

wheels. And that is correct to the extent they're eliminating all the controls and trying to put everything. And even once they're self-driving, it's really the case. But it's taking a computer mindset and applying it to physical goods. That's the consistency in must businesses. And that that is Apple as well. And so to Travis's point, yes, absolutely, they are spending a lot and they're going to spend even more. And that's why it's like, that's why you need big dreams because the risks are astronomical. Like, like, if you build and spend all this money and it doesn't work, you are out of business. And, uh, and that's why regular businesses don't do it. The risks are too great. Well, speaking of that Silicon Valley mindset, related to that and the general conversation

about Waymo and Tesla, this could be stupid because maybe there's an obvious answer that I'm missing. But we're talking about the car market here. So projecting forward. the car market has been extremely competitive for basically its entire existence. Why is it that a bunch of our emailers and a lot of people I've seen online are handicapping the race to autonomy as if this is going to be sort of a winner take all market like other tech markets? I mean, how do you think about that question? That comes to the transformative bit. So I think in the in the self-driving, if I'm going to distinguish between self-driving and autonomy, in a self-driving world,

it's probably not winner-take-all. And just to back up, generally speaking, physical goods are not winner-take-all because you're just fundamentally constrained. Like, that's the difference about the aggregators, right? The reason why I've never said that iPhones are an aggregation platform,

even though the app store kind of is, is just you're fundamentally constrained in serving the whole market because you actually have to make physical devices, right? And so that is a very difficult thing to do. And so, yeah, by and large, the more friction there is, the more marginal cost there are, the less likely there is to be winner take all sort of dynamics. And so if you, now, there could be a world where imagine if we had roads with like compute

built in so they could communicate with the cars and all that sort of thing. Then you have like a network effect and that would be winner take all. I don't. And so that's one possible outcome, but I don't think we're going to get there. Who's actually rebuilding the sort of roads, right? You need something that can, it's kind of like this middle way. Because I talk about backwards compatibility. You know, you, you, you, what's a middle way where you have new stuff will be built organically, but you don't have to like destroy stuff that exists, right?

Like the road things always like destroy what exists to a certain extent. If we had that world, that would be definitely winner take all. If you have the Waymo approach and you're just trying to get self-driving cars amongst us, then there's probably going to be. lots of approaches that work. We have regulatory standards about how good you have to be. And anyone that clears that bar can sell into the market. And, you know, and then we'll, you know, there are network effects in things like ride sharing, right?

We saw that with Uber. It was a slight one, but even a slight network effect was enough to basically win the market. And maybe it ends up being that that's enough actually for Uber to win the market. And Waymo ends up being providing software for Uber. And then we have all these Chinese OEMs or whoever it might be building sort of cars that that actually implement the software, and it looks like the PC ecosystem

or something along those lines. That is a very viable sort of solution. I think the reason why Tesla is the other opportunity to, if not winner take all, then to be winner take most is because if they're right, they're going to be so much cheaper. They're going to be cheaper and so much cheaper and so much further ahead of anybody who might try to compete with them.

That's right. And so if you're cheaper and better, you will just win. because you're the best product in the market, right? The iPhone was the best product. It was never the cheapest. And yet they still got 50% market share. And so, you know, and so I think that is, okay, Waybo will probably figure this out and

then they'll have a nice market. Tesla is taking a way more extreme approach. But if they get there, they're already making millions of cars right now, right? Like they, they're already at scale to a certain extent. Well, and you talked about with AI, I mean, data is becoming as valuable. as compute going forward and Tesla's going to have the most data if AI ends up being integral to winning in this market. Right. For now, I mean, maybe it just takes Tesla to wrong. Google gets enough cars out there on the road. They get enough data. And then that's right. And so again, this is why it's, this is, I keep saying this is the most interesting, one of the most interesting questions in tech right now. The implication of a being the most interesting question is no one knows for sure. No one knows the answer. I mean, I've, I've really enjoy it just.

sort of spitballing and wondering about what the future could look like over the last couple episodes here because it's a nice departure from, you know, some of the typical debates that dominate tech. And there is no clear answer at the moment here. That's part of what makes it fun. It does. And it speaks to like what an exciting time it suddenly is, right? We have rockets. We have self-driving cars. We have AI. We have new devices. AI. Like tech's back, baby. It's awesome. So on that note, I have to bring the party down. Peter, who's writing from the EU, says, Andrew and Ben, we have been able to do drone delivery for 10 years, but regulation makes it virtually impossible to implement. If it takes 10 plus years to come up with clear regulation for delivering a pizza or delivering Andrew's retro basketball shirt with a drone, imagine how long it will take to come up with regulation for letting 4,000 pound vehicles going 60 miles an hour. roam freely among us. If I had to take the over under

for whether regulation will allow for fully autonomous vehicles to go wherever they want in the next 15 years, I would definitely take the over. It seems like regulation is being treated as a passing thought. And as the meme goes, my European mind can't comprehend this.

I read that mainly because I love that Peter is a European who used the European mind. I love it, Peter. That was so good. Oh, terrific. It's obviously a huge aspect of the conversation, though, and that's going to be its own hurdle. So we would be in dereliction of duty to not at least mention it. This is a great example of the strength of the American system.

Like we're way more federal, not in that things are distributed, right? And so you can go to San Francisco right now and you can get in a waybo, right? It is out there. And you can use Tesla self-driving, level two self-driving, but it's like strength. rattles the line between level two and level three. And you can do that right now. And you can go to Phoenix and do it right now. Now, an important bit is, you notice all these places are very sunny, right?

Good weather. The weather issue is going to be a huge challenge for all this sort of stuff. And we may end up where it's allowed in parts of the country and it's not allowed in others. And, you know, one of the things is an issue about Tesla is there's like eight cameras that go into self-driving. One of those cameras is the one in the car that looks at the driver. and make sure they are watching the road. And so that is an important feature. That's how Justice's wife got nailed back at the beginning of this show, you know?

That's right. That's right. And so, yeah, I think, honestly, this is just a strength of the American system. Like, we do have the capacity still for stuff to be experimented in different places. You can see if it works or if it doesn't. And then it sort of will progress. And the win, it's such a win. Self-driving car is one of the biggest wins imaginable.

But like, this is where Musk was exactly right. You're giving people back hours of time a week. Like, everyone is going to want this. It is going to sort of be a big deal. People will not tolerate regulators holding this up, at least in the U.S. And so I, I am actually fairly optimistic. What I'm worried about, I don't want regulation that specifies product design, right? Specify an outcome, you know, how many disengagement or whatever it might be.

That's right. And so, but I think I'm fairly optimistic about that. I think California is actually doing a decent job in sort of this regard. Cruise was not good enough. They got taken off the roads. But it was right they were allowed to try. If you have to be perfect before you do it, you're going to be the European Space Agency. You're just going to totally miss.

You have to have grace. You have to have room in there. And yes, we're dealing with 4,000 pound vehicles and it's dangerous. But humans are dangerous too. So many people die. from cars. And so that's the big sort of ace in the hole is Tesla's doing real self-driving today would probably already be safer than humans, to be totally honest. Right.

And so as long as we keep that in mind, we don't get stuck in the status quo. I'm optimistic. I hope you're right. I too have faith in the American system. And I, over the last couple of weeks, have become a lot more comfortable imagining this autonomous driving future. but I also, as an American, know that the American system moves slowly. Sometimes that's part of the design. That's a good thing because it moves slowly to implement new regulations.

Like the whole, I remain convinced. Well, yeah, but like car companies, unions, those are powerful lobbies that will be able to gum up the process if they want to. Well, that's why Elon must be invited by the White House to like National Electric Day or whatever it was. and the sort of small domino that has went to lots of other Elon Musk discussions that we are avoiding on this podcast. I don't know. They're taking a mixed approach to a lot of this innovation. Yeah, but you know what? God bless the fact that it's really hard to pass laws in America.

Like honestly, like this is what men with the Uber thing. It really irritates people that Uber operate in this gray zone and they would go into cities that didn't have clear laws because it was a new concept that no one ever thought about before. and they just start operating. And then they would try to pass a law. And now Uber had a constituency. They had all the riders and all the drivers that would lobby for favorable laws to Uber. And this really got a lot of people mad. It's just like, why can't they go in and get the law pass first and then go in nicely?

Because that's not how it works. Right. Right. If you pass a law first, like the European AI Act, if you pass a law first, you're going to make wrong choices. you're going to be biased to the status quo. Because the whole point is you don't know what the future is going to be. Europe now, enjoy your Apple intelligence, enjoy your meta-a-I, enjoy the future open-AI models. Oh, wait, you don't have any of them because you tried to make the law first.

And yes, it's easy to look at things that go wrong and say, well, I wish we had a law ahead of time. No, you don't wish that. You don't wish that. And Europe is experiencing that right now. If you read the actual European regulations, Apple cannot launch Apple intelligence in the EU. Like, you passed the law. Sorry. Yeah. Maybe don't pass laws first for stuff you don't understand.

Well, look, the only other factor here that I want to mention, which you alluded to earlier, Karsten says, it seems like snow, ice, and dirt block most, if not all car sensors today. Does that mean no level five self-driving cars in the winter in middle and northern Europe, for instance? That is one of the most interesting second and third order questions that I've been wondering about when we imagine what this looks like in terms. 10 years. Like, is this like a southern technology and a West Coast technology, but not a very galling.

Very galling. Yes. I think this is a huge concern for sure. And this, by the way, LiDAR is lasers, right? It has the same problems in many respects cameras do. It doesn't work well in bad weather. This is where radar is, this is why I think Teslas will end up with radar just for the sort of bad weather problem. But, you know, radar has, it's own issues and challenges. It is much cheaper. So I think it does still fit in this sort of distributed world. But yeah, the bad weather problem is I think we're not even close to sort of figuring that out. Yeah. And it's kind of fascinating to imagine just sort of a stratified transportation future where parts of the country do one thing and then other parts you can have a much more efficient

autonomous reality. We shall see. Speaking of the American system, Jonathan Sussie, said regarding one digression in Monday's podcast, politics is not supposed to be a zero-sum game. I mean, I do get why you might think that after the last several decades, at least if you look at the U.S. federal government, but that's because we've been mostly doing it wrong. As Otto Van Bismarck famously said, politics is the art of the possible, the attainable, the art of the next best. I won't belabor the point, but a recognition of the importance of negotiated settlements and the ability to enforce terms are essentially prerequisites for economic growth. Did you have any thoughts there, Ben? What I was referring to is an election is a zero-sum game.

One side wins and one side loses. And because elections in a democracy are sort of the core of it and the people's primary means of participating, that sort of extends out into the way people think about it. And things get more and more polarized and more. You're just sorted. because it is literally a sorting process. So, you know, great, inspiring quotes don't change that reality, unfortunately. I think he's correct that politics has become more and more of a zero-sum game over the last several decades, but that really isn't the way it's supposed to work. And we're in sort of a doom loop in the United States.

And hopefully somewhere along the line that cycle breaks. But I appreciated the note because I read it. And I was like, yeah, he's absolutely right. It wasn't supposed to be, but, you know, half the country is just pissed off and aggrieved for four years, depending on what happens to an election. But that seems to be where we've netted out, at least for the time being. And I think one of the challenges, certainly, this is like an internet sort of thing. Like more transparency is not always a good thing, right? Like, you know, when you see the nuts and bolts about like the push for total efficiency, total efficiency, total efficiency in politics is total polarization.

Like that's just sort of, you know, the old world of like pork barrel spending and trading things here and back and Southern Democrats and like, although the weird sort of like things that existed in U.S. political life previously made compromise a lot easier or like, you know, not made compromise easier, but sort of shielded what was actually happening in a way that made it easier these negotiated settlements. But when you're, when everything is open, uh, that's a lot more difficult. And I think it's it's one of those things where almost everything in life has a downside, even if it sounds really great. Who doesn't want more transparency? And if you can't figure out why something you want, if you can't think through what is the potential downside of this thing, you probably don't really understand what fully, whatever you are pushing for. And it's a useful thought exercise. What's the what's the thing you feel most strongly about, you're most certain about? Can you actually think through what is a scenario or what facts would have to exist for me to be wrong, right?

How many people have truly thought through the downside of transparency, right? Now, again, I'm still for transparency, but once you, and if you want to achieve the Von Bismarck sort of vision, this is how you do it, right? You actually think through this thing that I feel so absolute about under what scenario am I incorrect? And in that you start to get actual empathy for your opposition. And what is it they might be thinking about or considering? It's pretty tiresome. I talk about this in analysis, right? Don't assume people who do dumb things are stupid.

Assume they're smart. And then you can actually think, what would inspire them to do that? And sometimes, by the way, you do that and you realize stupid things are actually not stupid. right? And sometimes it's like, oh, you know, company culture or incentives or whatever it might be. And by the way, sometimes they're stupid. I was going to say sometimes it's like a Google Gemini situation and you're like, man, you guys should have had a common sense meeting before rolling ahead with that one. But, but yeah, transparency, just to clarify, you're saying essentially the transparency has made it harder to work together with the other side among our leadership. because they then have to answer to their constituencies. Is that the argument? Well, here's a better example. I think the pork barrel one is actually a great example, right?

There used to be all this sort of like unattributed money. You could do it. You could go build a bridge. Like, whatever. Like, that's how you got people on board. And then there was a big push. I think it was sort of like the Gingrich platform in like the 1990s. I can't remember exactly.

The doom loop started with Gingrich in the mid-90s. But part of it is a very structural thing, which is we're going to earmarks. That's what it is. We're going to eliminate earmarks. Um, guess what? What you do is when you need to pass some quote unquote common sense legislation that requires compromise is you peel off people by saying, guess what? We're going to build a bridge in your district that is probably not necessary. We'll have your name on it. And that'll be an earmark to this bill. And people come in with the best of intentions. We're going to trim government waste. Guess what? Sometimes government waste is how stuff gets done. And so you eliminate earmarks and suddenly. So this goes into negotiation, right? The way to have a successful negotiation is if you're negotiating over one thing, you're doomed because it's literally just a power struggle over who can get the most things.

So if you're in a negotiation with someone, you want to introduce more variables, right? So you're negotiating with a landlord, right? Like, well, what if I extend my lease, right? Or what if I sort of change this thing or do it? The more variables you can bring in, suddenly it's like, okay, this person, because what happens in you have multiple variables. You have multiple variables. You stack them in order of priority.

Most people have different stacks. And so you can say, look, this one thing is the most important to me. This other person cares about that thing, but it's the third most important to them. They have their priority number one is my third most priority. So I know exactly what I can give on and they can give on. So like in this case, you take out earmarks, you're reducing variables. Yeah. And it makes it harder to get anything done.

And this applies to a whole host of things. I mean, it applies to fantasy basketball trades, for one. Try to help each other. That's what fantasy basketball is all about. I shrunk the lineups, increased the bench this year because there was not enough trades last year. I need trades. I need some negotiations. They need some battena.

Everyone has gotten too smart. And then the trades are broadcast to the entire league. So nobody wants to get embarrassed. And so there's a lot of negotiation, but not many trades are actually consummated in fantasy basketball these days. But in any event, another good podcast. We didn't get to a couple open AI questions. Let's get to get to this last question because it's a little bit of a tease.

Okay. So we were asked by multiple people why meta hasn't started an app store or whether they would consider starting an app store. So Dan says, hey, guys, half-baked idea. But is meta not well positioned to launch an alternative to the Google Play store during this three-year period? They already have a store on their own Android for VR and have some developer relationships, albeit for a different category of apps. It seems like it would serve them well from an ads perspective, and they've always wanted to own the phone. With Orion on the distant horizon, nice bit of alliteration there, isn't this the perfect timing for them? Do you have thoughts for the five or six different people who have asked us about this possibility?

I do not have thoughts. Sorry, we're at the end of the podcast. I'm just going to point you to I did an interview with Hugo Bar again who I talked to after the Apple Vision Pro he previously worked on the Oculus team used to be an executive at Xiaomi worked on Android he thoroughly convinced

me that meta should not build a phone at all they shouldn't partner anything just skip the whole phone sort of thing so that is a teaser I actually think it was a great interview podcast so go check that out and we can maybe we can talk more about you said listen to

that, then resend your questions and we'll go from there. Yes, we will double back on the topic. And a good plug. I look forward to checking out the Hugo Bar interview. But I'm trying to get a Ben Golver's level, just the king of plugging. No one can get on golfers plug. But yeah, but yes. I had a perfect segue to get to take us out here. We've been talking about driving for the last couple of shows. David will end on this driving question. Will Lando Norris catch Max for stopping? curious to hear both your takes on Sharp Tech F1 corner before the end of the season. And before F1 returns in Austin this weekend, do you have any takes on the next month in F1? No, he's not going to catch Max because he cracks under pressure.

He just doesn't have it. Let's be honest. Everybody who watches F1, let's all be honest about Lando Norris. He doesn't quite have it. He's a B plus trying to compete with an A plus Max. for stopping who it should be noted red bull has been in crisis for the last like three months here it keeps getting worse it's amazing right uh no the the the fact he's a very very good driver like so when you grade it's sort of challenging but the he he qualifies this why he qualifies he

qualifies his teammate you know he he in the right circumstances he can dominate a race but you saw it in singapore he almost crashed two two times he cracks that's the issue there there's a vector of sort of grit and resilience that he lacks. And you see it in his comments that really drive me up the wall, always making like snide comments, like when Lewis is dominating or when Max is dominating, oh, there's a there's just a, there's weirdly melodramatic in all of his. I don't want to psychoanaly him, but there's like there's some sort of streak of I feel it feels like sort of insecurity and like imposter syndrome it feels like. And the problem is that when the pressure is the absolute highest and you just all you need to do is just execute that's that's when you crack and and uh the pressure

he's going to be facing the highest pressure he's ever faced in his life by far over the the following few races it the only reason why this is even a question of course he should win the title that his car is so much better than everybody else is it's better on every single track he qualifies better than his teammate ergo he should be pull every race and he should dominate every race and yet me sitting here, have zero confidence he will do that because I don't think he's mentally tough enough. His car is better. It's not that much better than the rest of the field. It's like 20% better. Now that they can't cheat anymore. I don't need to go there. Well, and look, I hope he can push Max because Max is the best driver in the world. He has won the last three world driver

championships. Red Bull has been the best team, but they've been an absolute mess. And I would love it if Lando could win a couple of the next races to put some pressure on Max. And Red Bull does not have the best car anymore. And I want to see how Max for Stopping responds to that sort of pressure. I have faith that he would respond. And he's done a really good job keeping Red Bull near the front over the last couple of months anyway. The same thing was a miracle. Like it's not just that their car is bad. They're particularly bad at that track. And somehow he pulls out a second place.

Like to me, that was the key race of the season. Well, and look, the deal with sports is you want to see the best athletes in the world tested. And Max has been the best driver in the world the last couple of years, but we haven't seen many tests. So I hope that Lando can at least give him some semblance of a test over the next couple of months. But I'm not really holding my breath on that one. I'm not a Lando Norris believer. Sorry to David if he's a McLaren fan, but everybody can get excited.

We got F1 for the next three weeks after just an interminable month-long break for F-1. Sorry, I need you people to be working harder for my entertainment, please. I know. Look, we all got back to work in September. I don't know what F-1's problem was. But in any event, we'll continue working next week. Monday mailback coming back.

Continue emailing us. Email at sharptech. FM. Like I said, I got a couple OpenAI questions. Maybe we'll return to the meta app question. But then, until then, have a great couple of days and I will talk to you next week.

Yep, talk to you later.