This Week in Startups - The Self-Driving Unicorn You Need to Meet & MicroStrategy's Bitcoin Bet | E2052

Episode Date: November 27, 2024

Todays show: Jason and Alex welcomed Alex Kendall to the show. He's the founder and CEO of Wayve, an AI company working on self-driving. You might recall that Wayve raised $1 billion earlier this year, but TWIST wanted to know how the company intends to commercialize -- and how soon. From there, it was back to the bitcoin mines. Figuratively, as the hosts dug into MicroStrategies' bitcoin buying bet. Precisely why it is trading at such a NAV premium remains occluded to the show, but we did chase down its debt terms. Next week: More TWiST500! * Timestamps: (0:00) Jason and Alex kick off the show (2:32) Introduction of guest Alex Kendall Co-Founder & CEO of Wayve (3:00) Alex Kendall on Wayve's autonomous driving approach and AI technology (8:29) LiDAR vs. camera-first solutions debate (10:19) Cloud Devs - Visit https://www.clouddevs.com/twist for an unbeatable offer on hiring elite Latam talent today. (12:23) Economic viability and strategy of autonomous vehicles (17:49) Timeline for Wave's technology deployment and future market scenarios (19:58) Vanta - Get $1000 off your SOC 2 at https://www.vanta.com/twist (22:34) Autonomy subscription models and AI driving behavior challenges (30:01) Kyte - Download the Kyte app today and use code JASON to save 10% on your first rental. (33:00) AI driving culture and Twist 500 update (37:05) MicroStrategy's Bitcoin strategy and market concerns (41:33) Bitcoin convertible debt analysis (49:03) Deciphering complex business models and historical examples (51:24) Evaluating new financial terms and investor advice (52:29) Final thoughts and invitation to Michael Saylor * Subscribe to the TWiST500 newsletter: https://ticker.thisweekinstartups.com Check out the TWIST500: https://www.twist500.com Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp * Mentioned on the show: Check out Wayve: https://wayve.ai/ Check out MicroStrategy: https://www.microstrategy.com/ TWiST Episode #1251 with Do Kwon: https://www.youtube.com/watch?v=PLZL8iPKsHQ * Follow Alex Kendall: X: https://x.com/alexgkendall LinkedIn: https://www.linkedin.com/in/alexgkendall/ * Follow Alex: X: https://x.com/alex LinkedIn: ⁠https://www.linkedin.com/in/alexwilhelm Follow Jason: X: https://twitter.com/Jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Thank you to our partners: (10:19) Cloud Devs - Visit https://www.clouddevs.com/twist for an unbeatable offer on hiring elite Latam talent today. (19:58) Vanta - Get $1000 off your SOC 2 at https://www.vanta.com/twist (30:01) Kyte - Download the Kyte app today and use code JASON to save 10% on your first rental. * Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow TWiST: Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin Instagram: https://www.instagram.com/thisweekinstartups TikTok: https://www.tiktok.com/@thisweekinstartups Substack: https://twistartups.substack.com * Subscribe to the Founder University Podcast: https://www.youtube.com/@founderuniversity1916

Whenever people make up new terms for revenue or expenses, any of this stuff, it's going to be a pink flag or a red flag. Pink flag is what I waive on the way to a red flag. Red flag means stop, don't participate in this. Pink flag is, do we ask some questions? If you're a micro strategy holder, my best advice to you, if it was a family member, I said, hey, I put $100,000 into this, I double my money. My best advice to a family member, and this is not financial advice for anybody else. It's me for my brother. Sure. My mom.

I'd say, sell it. You doubled your money. Sell it. And then go buy Bitcoin directly if you love Bitcoin because now you're going to get four Bitcoin for every one that you bought last month if you in fact doubled your money. And it's, you know, this whole thing, not your keys, not your coins. You probably want to own it yourself, have your own coins or maybe be in something that's super trusted. I'll have Michael Saylor here on this week in startups.

Happy to have him on. He can explain it. It's not personal with Michael Saylor. Just so we're clear. This week in startups is brought to you by CloudDev's. Building the best remote team is tough, but you don't have to do it alone. Visit clouddeves.com slash Twist for an unbeatable offer on hiring elite Latam talent today. Vanta.

Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a SOC2 report fast. Twist listeners can get $1,000 off for limited time at vanta.com slash twist. and Kite. Looking to get out of the city, rent a car with Kite. Kite delivers rental cars to your door, no counter, no lines, no hassle. Download the Kite app today and use code Jason to save 10% on your first rental. Hey, everybody, welcome back to this weekend startups.

I'm Jason Kalkanis. He's Alex Wilhelm. You know him from his tech crunch days. You know me as an angel investor and from a little podcast called All In. This is the podcast that started it all. This is where Chimov and Sacks used to hang out before All In this week in startups. on this week in startups. Well, 14 years ago, it started as a show where I would interview a founder. And then it was two founders a week. And then, I mean, it got up to six days a week and it took over my life. It's been great. I got Alex here now. We're going three days a week, Monday, Wednesday,

Friday, at noon, Texas time, 1 p.m. East Coast. We're going to talk to whoever the main character is at the moment in time. Maybe that's Michael Sala right now. Maybe it's Trump some weeks or J.D. Vance, a politician. Sometimes it's Elon. Sometimes it's Sundar. We're We talk about whoever the main character is. We talk about startups and we talk about all the technology friends. We have a big docket today. Let's talk to our first guest and just get started. We're going to talk about micro strategy after our first guest.

Okay, let's get started. Alex, who's our first guest on this week and startups? So very happy to report that I have sourced yet another Alex for the show. In this case, it's Alex Hindle. He is from the Wave team, W-A-Y-V-E. One of the most interesting companies, I think, in the realm of autonomous. driving slightly more under the radar here in the States than Waymo, Jason. We talked about Waymo quite a lot.

Yes. People might recall that Wave raised a $1.005 billion series C back in May, led by SoftBank, and then later on, Uber invested in the company as an extension of that series C. So they're backed by SoftBank, InVDivDia, Microsoft, and of course Jason's favorite company in the world, Uber. So Jason, we want to have them on. They're part of the Twist 500, and they have a very unique approach to autonomous driving. So welcome to the show, Alex 2.0.

Alex Kindle. Hey, hey, guys. Hey, Jason. Hey, Alex. So tell us about your approach to autonomy. We talk about a lot here on the program, the autonomy end game. We all know autonomy can work, but there's a final series of bosses, the final boss, being regulators and safety, maybe open roads. You tell me what you consider the final bosses, but where is Wave in terms of its contemporaries? And let's just pick two contemporaries that people talk about the most, Waymo doing. 150,000 rides in three or four cities. And of course, Tesla, which has the largest number of FSD vehicles, but they're supervised FSD at the moment. What does Wave fit between those two competitors or industry leaders? I love that. The final boss level, it's got to be AI. I mean, self-driving is an AI problem. And I think that's fundamentally the bet we made seven years ago at Wave. It's not the perception. It's not the infrastructure. But it's the decision-making and complex multi-agent environments, being able to predict how others are going to behave and, and more importantly, generalize the situations you've never seen before. You're never going to see the

same thing twice on the road and so much diversity of weather, of different behaviors, of different situations you can encounter. So the key thing you ought to realize is this is an AI problem. AI, you've spoken tons on the show about large language models, but I think that's just the tip of the iceberg with AI. Bringing it into the physical world and what we call it building embodied AI. I think this is what's going to enable self-driving to really scale, getting us off the drug of LiDAR, of high-definition maps. So what we've built a wave is one end-to-end neural network, a single AI model capable of driving in a very generalized sense in different countries, different vehicle types, including on roads it's never been to before, all with a camera-first

solution, with just equivalent of a single GPU. And the really exciting thing is that that many consumer vehicles that are being produced today have that equipment on their vehicles. And so we think this is the opportunity to bring this technology to mass market and do so through into-end machine learning. Alex, really briefly, is that GPU that you mentioned that's on cars? Does that run the whole neural network locally or is there a off-device component to the computational requirements that you're talking about? No, everything's running on board the vehicle.

So a single couple hundred tops GPU. Wow. So there was a paradigm shift at some point where we went from writing lines of code and trying to process the world in real time and say, is there a stop sign? Is there a pedestrian? What is happening in the real world? And then, if this, then that. Hard coding of the world. And then these neural networks came out and we just said, hey, here's a bunch of videos of humans driving correctly, do what they did on the streets of London, particularly chaotic place with streets that were designed, you know, hundreds and hundreds of years ago, as opposed to, say, a grid system in Arizona where it was just plopped in place, you know, in the last 50 years or so, or maybe in the 50s, maybe 70 years ago. So maybe you could just explain, did you start in the hard coding era and then move to the language model era, or have you always been in this sort of new paradigm? No, we've been all in on an end-to-end approach since, you know, my days of research at the University of Cambridge since starting the company. We've been all in. And I know it's, I think, gathered a lot more market traction recently. But seven years ago when we started, this was a really contrarian idea. Everyone in the ministry was telling me, look, this would never work. This would never be safe. But what we've seen is just the most remarkable outcomes are possible. I'll give you a couple of anecdotes. The first one is we can now train these systems to not just,

understand how to drive a car with sort of a vision action model, but we can actually also ground them in language. So these models can not only drive and understand the world around them, but also engage in language, explain their actions or even take instructions or feedback through natural language. Could I tell a car then, Alex? Could I say, hey, you know, you're going a little bit too fast for my taste, slow it down, and then it could digest that and then have that impact how it actually drives? Yeah. Or, you know, we put an example on our website of one of our car is overtaking a double-decker bus in London, and it says, hey, I'm pulling out, overtake this bus, because it's pulled over for a bus stop, and I'm going to tuck back into the lane

afterwards. It gives you, I think, a much more of a chauffeur-like experience rather than simply an invisible railway track driving experience. So when we look at this model, there was a big bet made by Tesla and Elon that was, along the lines of, we don't need LIDAR, human eyes are taking input in and they've been driving for 100 years. These cameras have higher fidelity. There's more of them and they're looking in 360 degrees. So therefore, cameras alone should be able to solve this problem. We're not going to need LiDAR. Waymo said, no, we want as much sensor as possible. We want as much maps as possible. It sounds like you sit on the non-LIDAR group. Maybe you could make the argument for and against LiDAR as people in the industry see it and where you think this will eventually wind up.

Look, Jason, we're actually sensor agnostic. I think ultimately long term, a camera first solution will be most scalable. It's got the right information to make decisions. But bringing in a redundant sensing modality can be a pragmatic way to get a system to market sooner. I think the key thing, though, is that you need to use sensors that have a good supply chain, cost effective, and give you surround redundant vision to be able to get an L3 or an L4 solution to market. But cars today that are being produced by major manufacturers have surround cameras, surround radar, and a single forward-facing light. I'm not talking about the robotaxies driving in Beijing or San Francisco, but mass market, you know, high volume vehicles, maybe not mass market, but at least the luxury end of consumer vehicles are coming out with those kind of sensors.

So, of course, they're a great platform to bring to autonomy. They're still a far cry from some of the robotaxies that you see in San Francisco that have enormous amounts of sensing. But that kind of sensing is a great place to get started. And I think ultimately camera radar or camera only solutions will be able to provide very intelligent level of driving, of autonomous driving as well. How many cars do you have on the road right now? And how long? And have you been taking rides or doing rides?

We've been developing with our fleet in London for, yeah, the past seven years. We recently just expanded to the US and I keep an eye out for what's coming next. The really cool thing was I was out a couple of weeks ago. We had maybe a couple of weeks of experience driving in the US. In the UK, we of course learn highway, urban driving, but in the UK we don't do things like four-way stop signs. There's no ability to turn right at a red light, and of course we drive on the left side of the road. Are you looking for an unbeatable deal on senior talent? Well, if so, you need to look

outside of your borders. According to G2, Cloud Devs is the best place to hire top Latin America, a market with millions of developers. Cloud Devs is the largest. pool of Latin American talent with over 10,000 senior engineers who are rigorously vetted for communication and problem-solving skills. And now that hire by cloud devs can even fill non-technical roles. Very interesting, from accounting to graphic design, with elite Latam talent. Giants like Coinbase are choosing cloud devs because its talent is matched with your time zone. And it's up to 80% cheaper than a local hire with the same Silicon Valley standards. So here's your call to action.

Because they're such big fans of the show, CloudDev's is offering Twist listeners an exclusive pay what's fair deal. Instead of their usual fee, only pay what you feel the talent is worth for a limited time only and exclusively for our listeners. Just go to clouddevs.com slash twist to pay what's fair for your next five-star hire. And what was so cool to see is after a couple of weeks of driving experience in the south of San Francisco Bay,

I came out and we took the car up to San Francisco. We'd never been there before. There's no HD map, no prior training from San Francisco, and we turned the system on. And what was amazing is we saw a drive effortlessly with the urban driving experience of London, the American culture of South San Francisco Bay. And in San Francisco, we saw our car do right turn red, go for four-way stop signs, and, of course, drive on the right side of the road. I mean, this is an amazing example of generalization,

because it's showing the AI take different concepts and apply them in new ways. to scenarios it hasn't been to before. I think this is the thing that will unlock bringing autonomy to the world. Okay, so let's just be rude here. The way you describe the system that you've built sounds awesome. It is generally applicable to environments that it has learned from before. You can take it at other places.

Don't need HD maps. Sensor agnostic. Has Waymo been really, really aggressively barking up the wrong three? Because I've been blown away with Waymo's ability to expand paid rides across the US, Alex. But it sounds like you're, going to surpass them? So are they just making mistakes? I mean, I think there's a lot to respect for the experience that Waymos put together every time I write in it, it's, you know, it is a remarkable

product experience to be how to go driverless. I think what I paint the picture is, is a different approach that has an opportunity to bring this technology scale, which is getting this deployed through driver assistance. The advantage of this brings is you can get global scale integration into into vehicles, and as I mentioned, these vehicles have the right equipment on them today to be able to run this level of autonomy. And, you know, this is at scale of millions of vehicles around the world. And this is through driver assistance that could be a hands-off eyes-on solution or hands-off eyes-off, L2 or L3. I think this is going to get to scale faster, but more importantly, bring more diverse data that should enable an AI solution to be able to

actually get to an L4, the future that, you know, I think we've been talking about for some time. The key thing, though, is what is going to bring this to a safe and economically viable solution? And I'd really challenge autonomy strategies on, is it going to be economically viable? And I think this path will lead us there. Let's talk about economic viability. The cars by Waymo are 150K, 100K, people are claiming that's the approximate cost range. I'm sure you have some competitive intelligence. Is it in that same ballpark in your estimation or your guest estimation?

I'm going to be bold and say the cost is zero because we just work with the equipment that's already in some consumer vehicles today. No, no, I'm saying what do you think Waymo's cars cost right now? Oh, yeah, I'll let you speculate on that one, but I think you're in the right ballpark. Would you guess 100, 150K is what people say most often, so you think that's probably accurate given what you know about sensors and the cost of the car itself, a Jaguar. I think for a robo taxi that's built on LIDAR with a tonnebomble computer, that might be a reasonable assumption, particularly when you're at low volume, thousands, you know, less than 10,000 vehicles. And so they are the leader, most number of rides, but a high cost.

Elon claims he's going to make these cars for 30K, that seems reasonable. Does that seem reasonable to you that these cars could be made for 30K-ish, eventually, eventually being next couple of years? The vision that I think we're heading towards is that, every single vehicle, if you're producing a vehicle that's not capable of hands off, eyes off driving, point to point where you can, you know, jump in your car, put it, ask it, where you tell it where you want to go, maybe give it the style or the type of driving you like it to do and then just sit back, read a book. Now, initially, maybe there are a few scenarios where it pulls over and asks you

to take it forward through an intersection or something like that. But over time, these data-driven systems should just get better and better until it's reliably doing all of the drive. a $30,000 car should have it in the next two, three years. And the more expensive cars already have it, those sensor packages and compute. That's right. They need the AI, but they have the hardware, and I think that's going to go mass market, because really you're talking about a, you know, $2,000 bill of materials. It's a couple hundred tops computer that might be $500 or less. Surround cameras, surround radar, maybe a redundant computer, a second computer, a redundant computer. that package, that's $2,000 or less.

Is your plan to be an OEM to all these different car makers who are struggling to get this technology in it and trying to figure out what their future is? Or is your model to go head-to-head with Robotaxi from Tesla and head-to-head with Waymo and just compete directly? Or do you see yourself as a provider of cars that have been adopted, that have the technology in them, and then provide those to the Uber-Lift DoorDash networks? What is your strategy? Yeah, we want to build the most intelligent, safe and entrusted AI model.

And I think to do that, you're going to need scale. So we want to work with the leading fleets and manufacturers around the world. I guess imagine different manufacturers will have different sensor configurations, different compute. The nice thing about the foundation model that we're building is we can distill it down into different variants. It's very flexible in that manner. But I'd love to see our AI power L2 to L5 driving.

So whether it's driver assistance or some safety features, through to hands off, eyes off driving, through to growing up into driverless as well. With this commercial consumer vehicles, the model that we want to run is seeing AI deployed in as diverse applications as possible. So you're a partnership. Yeah, you see yourself in partnership with car manufacturers, and you see yourself in partnership with fleets and taxi services like Uber and Lipp, correct? Yeah, we've integrated into cars, into vans.

we've course do we drive every day, but we've also done some grocery delivery trials. I think the more diverse days you get is just going to make the AI safer. And the key thing is this AI needs to be adaptable to the different platforms that other manufacturers and fleets are building. When can I experience it? When will it be out in a test fleet that I can access here in the States or will it show up first in like the Toyota website when I can click a button that says add wave, you know, AV 2.0 self-driving technology to this order?

That'll be the first time the public can experience the product is, as you say, just part of a consumer vehicle you can buy. Give me a shout. We can go for a ride in the Bay Area or come to London. We can go for a drive. As for timelines when this tech's coming to market, I mean, you know how much the self-driving industry has been hyped. I want to be quite careful, not publicly making a claim like that, but it's coming soon. You know, the hardware's there. We've got the clear opportunity regulations coming together.

We're working as hard as we can to get this AI shipped. It's going to be awesome. do you see the market, the chess board playing out? Do you think it's going to be Waymo, Uber, and yourself, and Lyft versus Tesla? Do you think Tesla is going to provide this to other manufacturers? Do you think, you know, other manufacturers like, say, a Toyota or something, will want to run their robo fleets? I mean, this seems like a game of Thrones right now in terms of the different approaches to winning. Uber itself has, I think, nine or 10 autonomy partners, and they've put money into your company. So obviously, they believe they're going to be the aggregator, and they'll work with many

partners. I think Tesla's made it clear that they're not going to work with the aggregators, and they want to be the aggregator and provide this. So if we're sitting here in five or 10 years, what are some likely scenarios for consumers? Yeah, I love this game of 40 Chess, and it's such a privilege after a pretty long deep tech journey to get to where we are. But I think there's a lot of analogies you could draw on, like, look, you can think about the cognitive AI, the language not all space. You could think about iPhones and Android's. I think in general, what I point out to is a couple of things. The first is it's really important, I think, you work on the hardest thing first. Otherwise, when it comes to scaling this kind of technology, you're at risk of hitting

some glass ceilings. And secondly, when it comes to adopting automation or robotics, I think time and time again we've learned is that if you require large infrastructure change or support, it's going to be much harder to roll out these systems compared to if they work seamlessly with how society works in an existing way. Founders, do you want to sell to bigger customers? I know you do. You got to get that ACV trending up. And you want to push your churn down, right? Sounds good. But to sell to those big buyers, you need to clear all of these compliance checks. You know that. That means you got to have things like SOC2 sorted out. What's SOC2?

It's a standard and ensures that companies keep their customer data safe. And if you aren't SOC2 compliant, you can kiss those big deals goodbye. You're not going to land the lighthouse customers. You're not going to be able to operate at the highest end of the market. But Vanta makes it really easy for you. To get and renew your SOC2 compliance, on average, Vanta customers are compliant in just two to four weeks. It can take months without Vanta. and they automate compliance for GDPR, HIPAA, and more.

So you can sell to bearer customers in whatever markup your startup is going after. Vanta is going to save you hundreds of hours of work and up to 85% on compliance costs. Stop slowing your sales, team down, and use Vanta. Get $1,000 off at vanta.com slash twist. That's vanta.com slash twist for $1,000 off your sock, too. So I fundamentally think, believe that autonomy of scale is going to be driven by a system that's trusted, that can drive in busy urban environments and merge with traffic, interact, and all of that kind of intelligent behavior that will come from an AI system. And then, of course, work with the kind

of hardware requirements and lack of an HD map that we've been talking about. So I'm bullish on this approach. It was great to see Tesla move that way in the last couple of years. I think there's a lot of space to run. What we want to do is enable, if you think about the 90 million vehicles that are produced each year, the Tesla might be producing a couple of million of those. The rest of them, we would love to enable those. One point of the year, I think, is the number, yeah, Alex? Yeah. So there's about two percent of the market, maybe more. Yeah, exactly. What about the other 98 percent? I mean, unfortunately, it looks like we're going to see more segregation between China and the U.S., China and the West, let's say, but for all of those vehicles being produced in the West,

we would love to let those fleets and manufacturers gain access to the incredible benefits of autonomy. What will the business model be? You'll just charge the consumer $49 a month for it or it'll be abstracted into the cost of the vehicle. How do you see your business eventually with these, as an OEM, I guess, to these car producers? How do you see your business evolving? What's the model?

Would you give them a global license and say, hey, up to a million vehicles can have this for 10 million a year, give us $10 per vehicle produced per month. How do you see your business about? As with anything, I mean, I can, you know, give a vision today, but this is going to evolve as we get going. I think ultimately robotics is an exchange of value for time. And so I think long term, a subscription of an autonomy service that gives you time back or frees up for time is where we should land. It incentivizes parties in the right way to improve the system over time. But what we're saying about, you know, getting things in without causing disruption, right now people are used to in the automotive space buying with an upfront piece price or fee. And so I think, you know, that's maybe where the industry starts, particularly at L2.

But when you get to L3, where it's a true exchange of value for time, yeah, I think that's where we'll land some kind of subscription model to consumer. And that will be shared between the fleet or the manufacturer and, of course, ourselves providing the AI that makes that possible. I'd pay so much for that. I mean, I would pay Jason 500 bucks a month for a fully self-driving car, I think. Just straight up. 6,000 a year plus the cost of your car plus insurance, but your insurance might go down to a fraction of what it is if it is running this kind of software. I mean, if you could recapture it actually do your email or actually do a podcast from the road. It's coming. We could record this from the road and you think about that amount of time short. And you might pay five bucks an hour. for it. And if you were to drive 10 hours a week using FSD, that's $200 a month, which I think is kind of where Elon wound up on Tesla, right? It's $99 a month or $49 a month. I can't remember because I just buy it for 10 grand when I buy a new vehicle. I think it was 5 or 10 grand. So you just buy it one time, but I think they've been doing a subscription. Yes, Alex? Yeah, and take those numbers and multiply

them by, you know, some significant fortune of those 90 million cars produced each year. you know, scale really matters here, not just producing autonomy for an affluent area like the Bay Area, but getting to scale, getting this out everywhere, it's an enormous opportunity. And I'd argue that's just the tip of the iceberg with embodied AI. When you start to think about what's possible at taking this beyond automotive to other kinds of robotics, this is going to be an enormous market. All right, listen, this is incredible. Congratulations on your early success.

It's great that we have this capitalistic environment, Alex, where y'all are in a dog fight to save human lives and to recapture all this wasted time and create a surplus, really, for everybody. So we wish you great success on this and great competition ultimately benefits the consumers and society. So, you know, I'm really excited that you're in the fight and I'm excited to see if I order a Corvette, if I can order a Wave in it. I really am coveting this new CA Corvette, Alex. I can't, I'm a functional, I'm a functional idiot when it comes to driving now because I've been using my Tesla FSD since they accepted it and was called autopilot.

And the idea of trying to stay exactly in the road or doing the, you know, adaptive cruise control, it's just so much better and so perfect now. I get car sick if I'm driving and other people in the car are going to get car sick. When you compare it to how smooth, you know, these systems are, they're getting really smooth. They do seem to be jittery, though, on certain, interactions like a traffic circle, a left turn, they seem to jitter a little bit too much. They're not decisive enough. As we wrap here, is there a solution to that lack of decisiveness

we see in these system? Yeah, well, you know how I mentioned working on the hardest thing first? I mean, starting in London and forced us to tackle that immediately because you leave our office and immediately you're in roundabouts, lots of merging scenarios. You've got Jaywalkers everywhere and if you're jerky, it's going to be a very, very uncomfortable ride. I mean, the key solution is this just going to a self-supervised large-scale system that can learn emergent behavior that's like if you think about if you have to handcode all of those factors that require you to behave of things that might cut in risks and other factors it's just going to cause frequent slowing down emergency stopping whereas letting the system watch those kind of dynamic

scenes and learn emergent behavior that our car pulls out and nudges into traffic pushes it way through crowds and also courteously lets others come in front of us where appropriate. That kind of stuff you see everywhere in London. How do you, I mean, let's face it, when humans drive, they cut corners and they break rules, consistently. We all know that. We all know that people roll through stop signs.

We all know people, you know, yellow means go a little faster when it in fact, he stops. So how do you balance what the training data is going to learn from humans, which is to on the margins break the rules or to bend the rules, let's say, let's say bend the rules, versus what the law says, which has come to a complete stop, nudge up to the intersection. I mean, there must be some balance here. And I wonder if it requires going to regulators and saying, hey, this is not how humans drive. This is not how they take a traffic circle, you know, the buy the books method. I think it's such an important point because if you think about, say, cognitive AI, these large

language models, it's a bit of a wild west around some of these. safety questions and rules, but the advantage is that we have in embodied AI is there is a pretty clear rule set. And I think you need to be able to separate out the safety rules. And then within the, what is like a safe envelope to make just driving decisions, there is a large design space of behaviours that you can have there. And so what I think we're motivated to solve in self-driving and where I think we will quite possibly see the solution to AI safety is a system that can provably stay within the bounds of those rules, but within those bounds provide a sense of collaboration and customization that passenger or an own occupant can help prompt the AI to

drive in a certain style. But those rules can be fairly well set. And then that driving culture that I was describing, it's very different from London and the US. And AI has been able to learn how that does differ and drive naturally. If you don't do that, you see a road rage all around you, people, you know, pissed off if you're not merging in the right way or, you know, accepting the right driving culture. In New York, you don't give an inch. In L.A., you kind of, you know, shake your fist at each other and, you know, but you kind of let people merge on the margins.

And then in Texas, people are like, no, no, you go. Go ahead. Go ahead. It's very interesting, even in the United States, how different it is, the three major cities, four major cities I've lived in are distinctly different. and it's largely because in Texas people have guns in their cars and everybody's nice to each other because you don't want to get in a road trade instance in Texas, I understand,

because it's going to go a certain direction. Everybody's very polite. In New York, nobody's got guns. Everybody just kind of yells and screams at each other, and that's as far as these interactions generally go. So are we going to be in a situation where eventually the AI models drive like New Yorkers or Texans? Are you going to be like,

I'm going to drive like a New Yorker in Texas, and it gives you a warning, hey, maybe you need to be a little more polite, a kid from Brooklyn or the Boston driving mode, Jason, because everyone... For sure. I mean, people from Boston are the worst. The worst. Terrible. Well, we're building some fun demos where you could drive for a couple of minutes yourself

and then ask the car to drive in your style or you could prompt it and ask a drive in a certain way. The key thing is it just knows the bounds of what's safe and within that. It should be fully customizable. If you travel a lot, you got to deal with renting cars. It's just a fact. It can be slow, confusing, and most of the time, the experience is stuck in the past. It takes an hour to rent a car, two hours to rent a car.

And I recently became aware of a really cool startup called Kite, K-Y-T-E. With Kite, instead of going to a dingy desk tucked into the bowels of some airport parking garage, a car is delivered right to you. That means you can get a new professionally maintained car brought to your door without dealing with the lines. And I use this. Many of you know, I moved to Texas, I'm in Austin, and I was in the Bay area. I was going to be doing a lot of meetings. I was going to be traveling around. I needed a car. All my cars were in Austin now. So I popped out the kite app. Bing, Bing, Bing. I had my car. And I avoid waiting around. And I had my own car. So I could go to all these disparate meetings I had across the Bay area. You can manage your entire trip in Kites app. Meaning you can avoid waiting around while someone else enters your data into a computer by hand. One slow, kish, don't get a time. It's like that scene from the movie with the sloth at the DMV.

Anyway, Kite doesn't do extra fees. It operates in large cities around the United States, and I've used it in both L.A. and San Francisco and love it. Here's your call to action. Download the kite app or book online at KYTE.com. That's KYT.E.com. Go ahead and download the app so you have it ready to go. Use the code Jason at checkout. You get 10% off your first rental, which could be significant. Don't forget to use the code Jason. Download the app, KYT.E.com right now. Yeah, it's going to be interesting to say, you know, I want to be zippy. And they do have like, I think to Tesla's innovation, they do have a sort of more aggressive mode in their self-driving preferences where you can change lanes more often or not at all,

which the whole point of self-driving I like is not changing lanes. I like the idea of not changing lanes because changing lanes means I got to pay more attention. I got to be more vigilant, you know, as opposed to Alex and Alex, Alex squared, that you could just, you know, kind of chill more if you're in the right-hand lane and it's just or in the middle lane and you're just cruising at, you know, whatever, five miles above the speed limit. Yeah. Where do you guys want to go autonomous driving? When we had Bill Gates over, he asked to go for some fish and chips here in London.

Ah. We could do a podcast in a car. Totally. We'll do it next time. We're there. Great job. And, oh, yeah, there it is. Bill Gates, a test driving.

Did he invest or is he just a fan? Has he invested yet? No, Microsoft is a great partner of ours. But Bill was a fan. Awesome. He's been a great supporter of the company. Well, it's fantastic to see so many players out there. Yes.

To a great competition makes for great products. So we wish you the best. And we'll check in with you in six months. And hopefully you'll have some cars on the road in the U.S. available in the Uber Network or Lyft or DoreNet. Hopefully you'll be delivering some barita. Great to catch out with Alex. And we'll talk to soon. Cheers, Jason.

Cheers, Alex. Bye. Oh, wow. It's great to see Alex. how much progress we're making here, huh? I'm so excited. Everybody is so dialed in. So what's interesting about that entire interview is that it reminds me of a company

called DeepGram and they were doing neural networks, you know, way before the open AI boom. And if I recall, I was talking to the founder and he said, look, what we did was we just took all of language and then we had a computer just learn how to talk or how to translate. And I was like, well, that's pretty cool. And this, to me, with my lack of deep technical knowledge about AI, does feel like a similar approach, and I'm very excited by it. I just, I want someone to make this work in Providence.

I don't care if it's Tesla or Wave or Waymo. I just want someone to bring it to where I've lived. That's what I want. Yeah, for sure. It'll be great to have this on the road and more options available. It's nice to see some competition, you know, in the space and it's not just Waymo and Tesla that you'll have upstarts. And I love the idea of people providing, you know, this technology to other platforms, Alex. Like, that's actually going to be one of the big wins is that you're not going to be able to buy a car that doesn't have this in it. And if there are models, multiple in the models pursuing this, you know, you'll be able to look at their track records and compare them in the same way, you know, when I buy a car, I think about my daughters and my wife,

and my bulldogs and what's going to be the safest. So I'm like, Model X, suburban, model Y. I want the safest cars for them to be in. And so I think what's going to happen is people are going to start looking at the safety records. How many miles did this drive? How many interventions did it have? How many accidents did it have?

How many fatalities, God forbid, did they have? And that's the way I'm going to judge these. Oh, I dead on. And also, I just think by the time that, you know, my thought is when my kids are 16, what will be the choice? And I think by then it's going to be so clear. as to feel irresponsible to let them drive, let alone a stick shift.

A comment from the audience. Sebastian Asprella says, interesting, no guns, for example, in Holland, and everyone is polite and courteous between drivers.

So, Jason, there is disputes about your gun, your gun driver philosophy. This is what I was told. Maybe somebody was being cheeky with me, but they kind of gave me the, you know,

Texas discussion. Like, people don't do home invasions. People don't get in road rage incidents. because people, generally speaking, have a firearm on them. And so, like, the idea of going into a restaurant and robbing a restaurant at gunpoint, there's going to be more guns in the restaurant than on the perps coming in to rob the restaurant in all likelihood.

Yeah. Yeah. I mean, it's like the okay corral, basically. But you don't know it because you don't see them. People just have guns on them. And, you know, I see signs. It's very weird, you know, moving to Texas. the signs I would see in New York were, if you have a gun, you go to jail for a year.

Minimum. And there's no discussion. That's what they try to, you know, really explain to New Yorkers. And then in Texas, it's kind of like, in our establishment, we don't allow firearms. Please do not bring, keep your firearm in your car or whatever. It's almost like they're asking you nicely. We prefer you don't have a firearm in this restaurant. Please, just, you know. But it's not like there's a metal detector at the door. No. So I'm kind of like, why is that sign even there?

That sign kind of makes me believe that people are going to bring their guns in anyway. I wonder if it's a liability thing. So, you have mentioned several times on the show that you had hoped that the Twist 500 will get filled out a little bit faster. We've been juggling a lot of different stuff. Yeah. Babies, in fact.

Babies, I mean, you know, the moving to company to Austin, adding shows. We've been doing a lot, a lot of fun. But Maddie and I, who is on the research team and also the twist team, we're going to get to 250 by the end of the year. Oh, love it. Great. So that is our pledge to you.

So we will get that ball re-rolling. So expect more interviews, guys. If you like Twist 500, twist 500.com, check it out. But yes, Jason, Microstrategie. I've been in the documents today. I've been digging through. I can answer your question.

This is what I like with our pick and roll here, our two-man game. I have been mixing it up on Twitter and hearing a lot of people saying, this is a Ponzi scheme. And the way they're describing Mr. MSTR, Microstratage, which is run by Michael Saylor, who is an effervescent, enthusiastic entrepreneur. This company has existed since the late 90s.

He got in a bunch of trouble with the SEC previously. It's been up and down. The software business, as we've talked about, is de minimis and declining, but the Bitcoin business is vibrant and being invested in massively. Yes. If you bought shares in Mr. At the peak last week, I think it was Thursday, he was trading at an all-time high of $540 a share.

When I woke up this morning and I just glanced at it was maybe $380, or 390 or something. It was... 371.30, as we speak. Oh, okay. So it's 370. So you've lost... I don't know if you made this trade. You've lost $150 a share, maybe, $100 a share or something, depending on when you got in. So, you know, this clearly was overheated. And I am not the only one who is saying, is there something here that the markets should be concerned about? Now, one of the great things about markets, Alex, is they are self-correcting. If there is an opportunity,

to short a company. There are a group of people who like to do that. If there is an asymmetrical outcome here where people believe that you should pay two and a half times for micro strategy and then the value of their Bitcoin because Michael Sellers got some financial wizardry at work with these convertible notes slash bonds, you know, financial instruments where he's going to acquire more and he's going to corner the market on Bitcoin because he now is the number one holder in Bitcoin after Satoshi and above the United States government, which has been seizing Bitcoin that are used illegally, and Trump says he'll never sell them.

So there's two aspects of this that I just wanted to put on the table right now. One is I am long Bitcoin. I've always been long Bitcoin. I bought a bunch under 100. My wife bought a bunch at 100. Mine got hacked back in the day when you had marketplaces for these. Hers did not. And we've made a lot of money on Bitcoin.

So I love Bitcoin. I think it's real. The fact that governments have abandoned and the fact that it hasn't been hacked is incredible to me. And I believe it is here to stay. Micro strategies now, the second piece of this is people like Vinnie Lingam and other folks who are in the Bitcoin community are very concerned at what Michael Sayler is doing. And they think he's going to damage the Bitcoin ecosystem and the, dare I say, brand and the trust in Bitcoin because of these volatile, in their minds acquisition strategies he has. But in the documents, what have you discovered and you could, of course, comment and what I'm saying, I don't have any beef

with Michael Sailer. I don't have a horse in this race, even though we own Bitcoin, we're never selling it. We think it's a great part of distributed, you know, diverse portfolio.

So we like having it. It's not personal with Michael Seller, unless people think I have it in for him or I'm jealous of him. I'm happy for any entrepreneur to be successful. Obviously, I've been successful myself in my career.

Absolutely. I just am always concerned when I see something disconnect from reality of markets, and Alex is an expert on markets as well, and we like to talk about it here. It's not personal. I'm not blocking him from being on all-in. If the other guys want to have him on all-in, I'm happy to have him there. It's not my decision. It's a group decision. So, anyway, there's everything you can say. I know he's been asking to come on all-in, you know, which was number seven in the rankings last week. Yeah. So, for a popular podcast. So I just want to be careful also platforming somebody and getting more of

the audience to buy shares in something when I think there are red flags everywhere. And Other people are saying red flag, red flag, red flag, I'll stop there. One of the reasons why I think the story's worth coming back to today is the religious intensity of people's complaints about you asking questions about the company. Because to me, if I sat down and said, Zoom, the software that we used to record the show quite often, the lovely enterprise software company, if I was like red flags, everybody, I'm really worried about this. Everyone would go like, well, okay, fair enough.

And if you can think that. Sure. But when it comes to anything that touches Bitcoin, you end up. dealing with these religious fanatics. And I've been trying to find a comment from our live chats. We're doing this life today. Someone was like, oh, the guy who thinks he's an expert about micro strategies is now expert on real estate. And I'm like, you came into our chat to complain about Jason's tweets to that degree. What are you doing? I mean, I own a little bit of real estate and I was on the board of Blockable for many years and invested millions of dollars. And so I do

have a slight education on manufacturing of homes and mobile homes. And I've interviewed a bunch of people on it. I don't claim to be an expert on either Bitcoin micro strategy or home construction, but I do know how to ask a good question or two. And that's what we're here for. It's, by the way, you should hope that people are constantly asking questions about companies that report their numbers publicly because that is the immune system of the market. The market will self-correct over time, but we can shorten those loops by asking a lot of questions. So, Jayes, you wanted to know what are the debt terms? What is this convertible debt that this company is putting out?

So I went into the documents. These are the $3 billion worth of notes that are due in 2009. They are convertible debt. And you asked, are they senior? They are senior unsecured obligations. They do not bear any regular interest. And the principal does not agree. Okay.

Why are they valuable? The notes are convertible into class A common stock at a ratio of about 1.5 shares per $1,000 of principal in the notes. So that means that they will convert to stock at about a $672 per share rate, which is a 55% premium at the time of issue. Which is almost double now because stocks come down. Stock has come down, yes. But usually in a convertible, you can also get your money back with a little bit of interest. And it depends on who gets to choose that. Sometimes the company would get to choose that sometimes, and this is where the devil's in the detail, the people loaning the

money would get to choose that. So that's where, and I don't know if they have to disclose that or if they have disclosed it yet, but this is where it could get very interesting, because the $3 billion, those people could just say, hey, I'll take my $3 billion back plus 8% a year, some interest on it. That could be buried in the documents we don't know about. If it is, you know, they're going to have a measure in five years. You know, it could be three and a half. billion dollars or whatever it is, that has to go back to those individuals in cash. Yes. That cash has to come from somewhere.

So either they have cash or they sell shares in the company or they shell the underlying Bitcoin. If Bitcoin were to be lower, this thing could, as opposed to going up to $540 and almost $100,000 of Bitcoin, it could go the other way, I think. And that's one of the problems with owning a large percentage being the number two holder in Bitcoin. If they have to liquidate, it can have a massive price impact. So I can answer that to some degree.

So first of all, starting in June 1st of 2008, quote, holders of the debt that MicroSatogy has sold have the right to require the company to repurchase for cash all or any portion of their notes. So a year before they actually reached the maturity point, the people who loan the money can say, we would like our cash back, please. Thank you. With interest or not with interest? repurchase price equal to 100% of the principal plus any accrued and unpaid special interest. I tried to chase down the definition of special interest throughout the...

That's in the document somewhere. So they must have defined what special interest is. I'm going to guess special interest is like plus one LIBOR or something or plus three LIBOR or something. Who knows? Because there's no underlying coupon rate, it doesn't seem to be connected to what we normally look for here, Jason. So I think basically they can get their cash back.

and the interest is relatively de minimis compared to the principal amount. But what that means is if the people who hold this convertible debt do not think that in 2009 when the notes mature that they're not going to make their money that they were hoping to, they can just get their cash back. And so that's when this feels Ponzi-ish because Michael Saylor is essentially saying, look, here's a call option on our equity down the road in the form of a convertible note today. And if you wouldn't make your money, great. but you can also make us give you cash back.

And if they need to repay, then they might need to liquidate some of their Bitcoin leading to a cascade because if the Bitcoin value goes down because they sell it, then their share price will go down further. And then all their convertible debt could become essentially a thing that could be paying cash. So I hope that made some sense. I spent a lot of times today reading SEC filings. Well, I think it is important to understand the mechanics of this. I think the SEC is going to require them to share a lot of this, maybe more than they've already

shared. And then consumers will become more hip to what's going on here. And I think somebody at a higher pay grade than us, who works in accounting, who's a CPA, could literally build a model here. They know how many Bitcoins they have. It's $200,000 or something, I think. Maybe it's $300,000, whatever the number of Bitcoin is. Yep. And then there's a price at which, and I don't know what their cash reserves are, because he doesn't like to keep cash. His whole concept is the Treasury is in Bitcoin. Yes. So I don't know how much cash he has, and then you'd have to predict how much cash is he going to have there. And I think he's talking about doing this every year.

So I wonder how much more of these converts he's going to do, because I think the concept was to do a convert every year or do bigger converts and buy even more Bitcoin. So now you have the criticism I've heard of this is he's going to have an incentive to do a bigger convertible. that convertible then is going to go buy more Bitcoin, clear out holders. People have been long-term holders who've wanted to move large positions. So if you wanted to move a billion dollars in Bitcoin or $5 billion in Bitcoin, like now you've got a whale in the system who wants to buy it. And they are a large portion of the buy side that would drive the price up. Price goes up.

Stock goes up. Ability to do another offering goes up. And maybe this can keep going. for a long time, or with the people who are providing this convertible debt, say, you know what? Interesting offering. We want tighter terms. We're going to need clawbacks. We're going to need a 2x liquidation preference. We're going to need some more structure here. And if maybe they'll put in there, if Bitcoin goes down below this number, you've got to liquidate and pay off 25% of this. If Bitcoin hits this number, you've got to liquidated, pay off 25% of it. So maybe the next note, depending on how

Bitcoin goes, will be more structured and have more covenants, as they say in the business. In other words, things that if they get triggered would cause actions to occur, i.e. a default. Yes. There are a lot of default terms that exist, but they seem to be mostly around change in company control, which is a pretty standard thing, for example. And there's some anti-dilution preferences in there as well. The thing that I have not been able to understand and chase down and really get to the root of though, Jason, is why Micro Strategy trades at a multiple to the value of its underlying Bitcoin. Everyone agrees the software business and MicroStrategy is pretty much moot.

And we all know the value of the Bitcoins they hold. And by the way, last time they told us, it was 386,700 Bitcoin. Just chase that down for you. But why do they have this massive premium on the net asset value of their underlying Bitcoin? And I have not found any answer. I read your tweets. I read the responses. and people were mad at you for asking why and then just saying, do your own research,

which didn't help at all. So have you found anything here? Here's the thing. When people tell you put in the hours, do the research, and when Michael Saylor says, I can't explain it in 300 characters, I have to come on the all-in podcast to explain it, or this week in startups to explain it, I get a little nervous. Yeah. Because almost all of the businesses, and I've invested in 400 startups and, you know, I've been a

journalist for, you know, over a decade covering them, done 2,000 episodes of this podcast. So I've been around the block, let's say, uh, in terms of assessing businesses. Anytime it's too difficult to understand in my experience. Yes. Now, I'm not saying this is micro-stratin, but when people tell me, do your own research, have fun being poor, you know, it's too complex for you to understand. The last person who told me was too complex for me to understand was Doe Kwan. He was picked up in Montenegro, I think.

Do Kwan was, um, the stable coin. That was the algorithmic stable coin. Help me out here, Jason. What was it called? Doe Kwan was doing Terra. He was the CEO of Terra. Yeah, I think he's in jail. We can link in the show notes to his episode here.

He tried to explain to me. I gave him two or three times to explain to me how this all worked. And he was arrested. I'm not saying, Michael Sellers going to get arrested. What I will say is there is a pattern. Yes. If you were to ask, I don't know, Adam Newman to explain.

explain how we workers were. I think he said, hey, community EBDA. He came up with his own term, community EBITA. Yeah. B2C yield, I think is a term that Michael Salary used this morning. BTC yield. I didn't know that there was a yield associated with Bitcoin. I don't understand what he means by that term. Like, is it... It's not yield. So community adjusted EBTA for folks who might have forgot was the famous metric that essentially we were convicted because adjusted EBITA was too strict to make its numbers look good. So they had to make it. Just any of it is too strict, you got a problem. You've got a big little problem.

Now, Red flagged. BTC yield is something that I flagged, and I said it was notable in our notes on Monday, which was my sarcastic way of saying WTF. So BTC yield is a term that does not reference yield. What it does is draw a ratio, as far as I understand, between the total number of Bitcoin that Microstrategy owns and each block of 1,000 shares outstanding. And so it's essentially how many Bitcoin per thousand shares of microcontradictory?

that are out there. Okay. That's not yield. That's a ratio. It's the ratio of how much Bitcoin you get per share. Whenever people make up new terms for revenue or expenses, any of this stuff, it's going to be a pink flag or a red flag. Pink flag is what I waive on the way to a red flag. Red flag means stop, don't participate in this. Pink flag is, do we ask some questions? If you're a micro strategy holder, my best advice to you. If it was a family member, it said, hey, I put 100,000 into this, I double my money. My best advice to a family member, and this is not financial advice for anybody else. It's been for my brother. Sure. My mom. I'd say, sell it. You doubled your money. Sell it. And then go buy Bitcoin directly if you love Bitcoin, because now you're going to get

four Bitcoin for every one that you bought last month if you in fact doubled your money. And it's, you know, this whole thing, not your keys, not your coins. You know, you probably want to own it yourself, have your own coins, or maybe be in something that's super trusted like Coinbase or Robin Hood. And I'm long-term shareholding in Robin Hood. But, you know, here we are. Another great episode. I'll have Michael Saylor here on this week in startups. If the other guys want to have them on all in, happy to have him on. He can explain it. It's not personal with Michael Saylor. Just so we're clear. I'd love to have him on the show. It'd be a great conversation. But just to be clear, we couldn't figure out the root cause of why they traded such a premium

to the NAV. No one's managed to explain to us. No financial coverage has explaining it to us. yet to read a single good explanation of why other than stonks go up. And that's not a good long-term investing position in my view. But yes, we got to wrap. Winner be back. What's our, what's our return from Thanksgiving plan? I think we're Monday. We're Monday, Monday, Monday. So we'll see everybody on Monday and talk to soon. Bye.