This Week in Startups - AI Demos: Llama 3 Innovations, vibecheck, Globe Engineer, Wisdolia and Tesla’s FSD | E1941

Starting point is 00:00:00 Chat GPT just made available to all its users is memory. And what memory is, it starts to learn about you. In the memory settings of chat GPT, which you should have available now, it starts to remember. And in this particular user's cases showing that it's traveling to Canada for a vacation, has a two-year-old daughter named Lena, daughter loves to. So the language model is storing this as bullet points in your settings. Seeing this become available globally within chat GPT. This is incredibly dangerous, by the way.

Starting point is 00:00:34 If this got hacked. Okay. Yeah. This week in startups is brought to you by Squarespace. Turn your idea into a new website. Go to squarespace.com slash twist for a free trial. When you're ready to launch, use offer code Twist to save 10% off your first purchase of a website or domain. 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 sock to report fast. Twist listeners can get $1,000 off for a limited time at vanta.com slash twist. And Wistia. All in one video platform for business with tools that help you create, manage, and measure the impact of your videos. Try Wistia for free.

Starting point is 00:01:26 at wistia.com slash startups. All right, everybody, welcome back to this week in startups. Madra Tuesdays have begun. AI Tuesdays. Sunny, it's time for demos. There's a little bit of news. But, you know, people are here for the demos.

Starting point is 00:01:41 We've got some news too. We've got some news too. We've got demos and news. Let's get started. Let's start with Lama 3. Because what's happened, you know, last time we talked about it, we had a good episode.

Starting point is 00:01:50 I think a lot of people tuned in. Yeah, what's happened since Lama 3, and I'm going to just pull in, like an example. And this will be pretty quick. And this is why, you know, sort of what Zuck was talking about was so important with respect to the ecosystem is what you can see here. This is a good example by Min.

Starting point is 00:02:10 And he's kind of gathered 10 ways that people have taken Lama and basically pushing the limits of its capabilities. Okay. And to recap, Lama 3 is Facebook's open source language model. They released it last week. They're working on number four, and it's already shot up in the hugging face rankings to like the top two language models. People are really excited about it because it's smaller, faster, better. So the first thing we're going to talk about here is a user Sabu Shuban have basically taken Lama 3 and you can make it such that you can deploy it locally on your computer.

Starting point is 00:02:47 Okay. So this is a regular old laptop and you can deploy it on that server as opposed to putting it on an H-100. is like what somebody would normally do or put it on a clock chip like you're working on. Correct. Okay. But now I would suspect if you're running it locally, it's incredibly slow and grinds to a halt. It's slower and it's good for like a maybe personal use cases or you want to use it without the internet or if you're kind of coding on a plane and you want a coding assistant without basically, you know, while you're doing something. So there are some limited use cases where this is, it's usable.

Starting point is 00:03:23 It's probably going to operate, you know, at the. the speed of like 10 to 15 tokens a second, which is reasonable for some use cases. Got it. So, but it obviously would not, if you were offline, be able to go to the open web. Okay. So this is important. Why is this important? Okay, we understand the fact you can run it on a laptop.

Starting point is 00:03:41 Why is this important? Because would anybody actually do this? I think what people are showing here is that this is a good way that it may start becoming integrated within applications that don't want to go out to the internet and don't want to pay. You know, when it's out on the internet, you have to pay someone for the inference, right? When you're running it locally on your machine, we go back to the good old days of J-Cal box software. Ah, got it.

Starting point is 00:04:05 Yeah, when I was running my entrepreneurial efforts in my youth in Brooklyn, about seven miles from where I am right now in Lauren and Han. Okay, so that makes sense. I can think of one application here. If I have this on my local computer and I wanted to index everything on my computer, all my photos maybe all my documents whatever my mp3 files whatever i had locally most people store everything in the cloud but you might have a copy people take dropbox or iCloud or outlook at microsoft drive and they and they syndicate here so yeah you can run this on your local computer like a local search engine of your computer got it makes total sense we're going to go through

Starting point is 00:04:45 these quickly someone took lama 3 8b which is a smaller model yeah and extended the context length to 160K. Explain context length, give an example. Just so people know. The default model is 8,000 tokens, right? So remember, we just use this proxy. Think of a token as a word. It's not quite, but it's a great proxy.

Starting point is 00:05:07 So the default model only comes with a context length of 8K. So 8,000 tokens. Which would be, for example, a long newspaper or magazine article, a book. Correct. You know, that's a five-hour audio book or something, might be 50, 60,000 words. A long book might be 100,000 words. Yeah.

Starting point is 00:05:26 So just to give people context here, this small model basically would take like a Sunday magazine or a New York, New Yorker article, a long article, and be able to put that in the context window. Correct. And the context window needs to be, have enough space for both input and output. Okay. And so you were trying to summarize a long book, you wanted to basically make sure it has space to summarize it as well.

Starting point is 00:05:48 But what we've seen here is people have extended. extended the training and they used 200 million tokens, additional tokens, to basically extend the context length to 160K. It's fun. This is sort of the things that, you know, Zach wanted to happen. You know, this one's very similar to, you know, things we've said, but like people have made it of co-pilot within VS code. You know, today the leading co-pilot, Microsoft had their earnings last week and they talked

Starting point is 00:06:15 about, you know, they have millions and millions of paying co-pilot users. Well, now you can use Lama 3 with your, you know, basically, you can, you can, you use it locally or you can use it remotely, and basically you can have it as a co-piloted VS code. Another user here has enhanced it for function calling, which is really useful. This is tool selection. So if you want to, you know, you want to use the model in a workflow where it needs to make a decision to go out to the internet to do something, this user here has extended it so that it can do function calling and it knows which tool to use depending on the question.

Starting point is 00:06:48 And there's like an example here. It's using like a currency converter to do. currency conversions. Here, someone has squeezed it to make it run on a single four-gigabyte GPU. These are all those things that Zuck was talking about, which could be interesting innovations. Some could be useful. Some could spark other ideas. And here, basically, what the user has done is they've done, like, a divide-and-conquer approach and, like, how they're doing the layer-wise inference. Here, a user, you know, interesting and left using rock, basically built a video summary tool. So what you can do here is you can basically drop in a link to a video and it'll create like an instant summary for you without you having to watch it, which is great.

Starting point is 00:07:33 You know, sometimes we don't have times to watch the one hour and 40 plus minute Olin podcast, Intent, J-Cal. And so. It was a long one this week, yes, for sure. I'll just do one or two more here. This is just showing the world of the realm of possibilities. Here, someone has combined lava, and we've demoed that before. That's a vision model with Lama 3. So using the capabilities of Lama 3 with the vision capabilities of Lava.

Starting point is 00:07:57 Who is lava? That's an open source project, yeah? That's an open source project. That is basically it really meant for research. You have to go get a commercial license. You can't use it for commercial without getting a license from them. And you can see in the example where they've combined the reasoning capabilities, remember of Lama 3, $8 billion, which was as good as Lama 270 billion, with a vision model

Starting point is 00:08:22 up front. And so this is really exciting as well. Yeah. Here, I'll do this as the last one, and then we can kind of move on. Well, this one might impact, this one might impact some bets, I think. This week in startups.com slash bets to see sign and I betting like lunatics on AI predictions. Yeah. So this one, basically someone, again, created a variant of Lama 3, Abacus, and basically they've made it available within their iPhone app. So this is running locally on your iPhone app for the small model, which is awesome. Wow. So this is something we thought was going to happen at some point. And now you can think, well, the context window could be all of your contacts on your phone, all of your eye messages on

Starting point is 00:09:08 your phone, things that exist on your phone. Maybe it can record your, maybe you could build an app that records your screen and then adds everything that you record on your screen puts it in LLM. That's local on your phone. And then if your phone gets dumped or like, you know, it's encrypted, then nobody would know. So there's this concern that I want to give my information to Open AI or Microsoft or some of the third party. And that's no bueno for me.

Starting point is 00:09:34 But imagine, and I had the guy, I forgot his name, will come to me in a moment, who's doing the recording everything on your desktop. and I challenged him on the pod, like, are you recording signal on people's desktop? It's like, we're recording your whole desktop. I'm like, well, the expectation would signal is that those aren't recorded. So you're kind of like building a tool without ethics

Starting point is 00:09:54 to record things that aren't supposed to be recorded, right? And then you're putting it into a model that you're storing and it gets hacked. Now you're going to be in trouble here. So this would, if it's your local copy, optimized for iPhone, let's say, or Android, stored, encrypted, all of your data for all time. recording everything you do.

Starting point is 00:10:13 Could record every phone call or, you know, the metadata of the call in and out, every conversation you have in there. You could ask it. Who haven't I talked to in the last week that I should talk? He's saying, hey, you haven't called your mom. You haven't texted with your mom. It could start to understand everything on your phone. I think this is very powerful when we start seeing these language models running locally.

Starting point is 00:10:32 Hey, startups. You ever notice how successful businesses are constantly evolving? Well, that's because they add fresh features to get their customers excited. And that's why I am all in on. Squarespace and you should be two. Squarespace is adding new features and revolutionizing their platform all the time. I've told you this before. This includes powering all their tools with AI, you guessed it. And one of my favorites, this is incredible is Blueprint AI. Yes, it's their guided design system for building a new website. It's fast, it's custom, and it's built specifically for

Starting point is 00:11:01 your business needs. They put over a billion design combinations at your fingertips so you get an online presence that's unique to your brand. We used it to build our own founding university website. The process was so simple. We were prompted through a series of questions, and then Blueprint AI took control, tailoring everything to our preferences from website structure to the color scheme, very important for me, and from design elements to font pairings. I love fonts, by the way. It's just one of the things I'm into. Boom. Within minutes, we had a sleek new website ready to go live. So check out Squarespace.com slash twists for a free trial and make sure you try this Blueprint AI tool. It's amazing. And when you're ready to launch, go to Squarespace.com slash twist to get 10% off. Your first website or domain

Starting point is 00:11:41 purchase. That's Squarespace.com slash twist. Since you brought that up, J-Cal, and I do have to give you credit, world's greatest moderator. You just have this uncanny ability because we didn't even plan this. No, we never planned it. It wasn't on today's list, but chat GPT just made available to all its users is memory. What does it mean? Memory. Okay, so they basically put this out in beta in February. And what memory is, it starts to learn about you. Okay. And so you'll see in this, example, what they do here, and I'm just going to, you know, for the folks, I'll narrate it, but as I'm doing that here, is what the user does is, I live with a golden retriever named

Starting point is 00:12:21 Ellie and Maine, Coon, and cat named Teddy. What this does is in the memory settings of chat, GPT, which you should have available now, it starts to remember. And in this particular user's cases showing that it's traveling to Canada for a vacation, has a two-year-old daughter named Lena, daughter loves to Ellie. Oh, so the language model. model is storing this as bullet points in your settings. So you don't have to tell it because I remember there was another feature not called memory. It had a bad name, but I would put in there.

Starting point is 00:12:52 I like sources and I like things presented in tables. Exactly. I forgot what they call that, but this is much better because it's happening in real time. So I wonder if we did ask it, hey, put that in a table. Hey, cite your sources. If it would put into its memory, you like sources. You like you like. You like source, well, yeah.

Starting point is 00:13:12 And, you know, it's doing it sort of automatically and also like you can force it to put something in there. But I'm just bringing this up because you were talking about how you think people are going to use this on local models. But we're seeing this become available globally within chat, GPT. And it is incredibly dangerous, by the way. If this got helped. Okay. Yeah. This is very dangerous.

Starting point is 00:13:34 And this is where I think. Breakdown why you think it's changed. You know, Google has your search data. Instagram and TikTok have their algorithms watching what you look at, what you like, etc. That could be very revealing. People have complained before like, oh my God, I'm suffering from depression or I have cancer. The health stuff comes to prominence very quickly. Oh, my God, they have these searches I did, et cetera. That's why people should use things like the Brave browser, which has a built in VPN. Shout out to my friends at Brave. And, you know, incognito mode was the big

Starting point is 00:14:04 joke like they called it, I think, porn mode, but not private mode is I think what the Google engineers jokingly called it because it doesn't do anything. And Google got in trouble, right? Because you saw this recently, they were sending themselves information on your incognito mode. Anyway, this is why this is very important. I want to finish this example for the listener, Jake Howard. So the user basically told it the name of its golden retriever and cat. And basically that went in the memory here.

Starting point is 00:14:34 That's what it's showing. And then the user in a future example, it says, create a photo of my pets writing a surfboard. And you'll see Chat GPT create an image. And it says, here is the image of your golden retriever and Maine Cooncat writing a surfboard. I hope you like how they look. And you've got to remember that in the prompt, all the user said was, create a photo of my pets writing a surfboard, please. And it used its memory to enhance the response. response to the end user.

Starting point is 00:15:04 So, yeah, I mean, this is going to be very powerful and very dangerous. So if it knows when, you know, your parent died, you got a tragedy in your life, when you got a cancer diagnosis, whatever it is, these things are really not things you want a company like Open AI or Google or Microsoft to have their hands on and to have this whole psychographic of you, which Microsoft has had. You could spoof all these things very easily. when you sign up for these services, create yourself a burner account,

Starting point is 00:15:33 tell them the opposite gender, add or remove 20 years from your age, and use a burner email address so that these companies don't track you. That's like number one. Use a VPN. Number two, I suggest NordVPN or the Belt and Brave VPN.

Starting point is 00:15:50 Both of those are pretty solid. Both of those also have sponsored the program because I love their products. Just so you know, like the way the ads work on this show, I tell them what products I have. Like, they email me talking about the products on the show. The people who are marketing, like, oh, my God, JCal loves Audible. JCal loves, you know, LinkedIn jobs.

Starting point is 00:16:09 JCal loves, you know, Squarespace. Great. Well, have him read those ads because it's not an ad. It's like an endorsement. Like, I love Notion and Coda and Squarespace. All those things you hear me talk about in the ads are actually, I'm using it every day. So we have our choice. We can skim the cream with the best partners and Brave and Nord VPN.

Starting point is 00:16:27 Two really great products. shout out to Nord and Brave. You have to use those things. Do you use a VPN and Brave or no? No, I use Chrome and I use VPNs only when I'm traveling. Big mistake. You got to have your VPNs on all the time, shields up, and you can't use Chrome because you're giving every single thing you do to Google.

Starting point is 00:16:51 Yeah. And Google has shown you probably like any corporation, whatever they say, whatever you think, you should not trust them. they can dump your search history. And look, I probably need to think about this more. But like, I have different Chrome profiles. That one for work and I want, you know, personal one. And my personal one, I like that it's looking at all that because I use the Google search app.

Starting point is 00:17:11 And, you know, news feed is that because it's kind of looking at everything I'm searching. So you've made a decision that you don't mind it. Yeah. I suggest you. It creates a better experience for me. Okay. Yeah. I mean, that's, and that is the give take.

Starting point is 00:17:24 If you give of yourself, and that's what we're seeing here, and you give. and you get more personalization. I do think, you know, and I'm not being a conspiracy theorist here, people should use Brave as their default, which shields up as their default, and then back it up with, you know, Chrome or something.

Starting point is 00:17:41 That's my best advice, and always, by default, have your VPN on, only because you could get intercepted and on Wi-Fi. And you'll forget, like, you'll forget. And, like, you know,

Starting point is 00:17:50 when I'm at an airport, I'm at Starbucks, whatever, I just don't want to not have NordVPN on. And NordVPN, by the way, has something called double VPN, which is nuts. They do like a VPN to a VPN. So like, ooh, a double layer of the VPN.

Starting point is 00:18:04 It's incredible. And then when you're overseas, I like to use Sirius XM, Hulu, NBA. Yeah, then you can get that. Yeah. And, you know, I'd say, you know, those services sometimes play cat and mouse with the VPNs to try to ban them from using it. Yeah. I would say NordVPN has the best track record for me in my experience.

Starting point is 00:18:21 I'm going to try that. You know, my current go-to is I don't use public Wi-Fi as I use my hot spot on my phone. That does help a lot. So I never go on public Wi-Fi's. And if I'm traveling, I use VPNs for the exact same reason you're talking about so I can get to U.S.-based services. And by the way, these things went from being like 30, 40 bucks a month to be like five or 10 bucks a month. So you don't really have an excuse if it's $100 a year. Like, just get it, folks.

Starting point is 00:18:45 If you hear my voice. And I'm not saying that because they're partners. You can pick whichever VPN you like. I'm just telling you the one that is the most rock solid. And I have it on my phone, my laptop, and my desktop. Boom. Rock and roll. Should we do Tesla really quick?

Starting point is 00:18:56 you wanted to talk about FSD? We only had a small amount of time to talk about it on All In. And, you know, we were asking to get predictions or whatever. So we were like, oh, you guys don't understand. You guys don't understand how FSD works. I'm like, I'm using FSD pretty great. And they're like, you don't understand it's now it's a neural network. It's not the C code.

Starting point is 00:19:14 I'm like, I do because when I click the button, it says we're no longer using C plus lines of code. We're using the neural network. Yeah. You know, I've met the guy who runs Towsla. You know, like, I do understand that, folks. And so when I said, I think it's going to take 10 years for 5 to 10% of the rides to be automated, I was talking, number one, on a global basis because jurisdiction matters.

Starting point is 00:19:38 And number two, just the number of drivers out there available and the competition, how expensive robotaxis are incredibly expensive. Teslas are incredibly cheap because they're already out there. But net, net, net, there's regulation. There's the technology working in all markets, all climates, all conditions, all roads, all edge cases. And so I thought, I don't know, in five years, if of the 10 billion rides Uber's doing, and I think there's probably that, I think Uber might have less than 1% of all rides globally in the world.

Starting point is 00:20:13 Like, I'm not talking about paid rides on just trips. Just trips. I think there's something like a trillion trips or something. So if they have 10 billion, they have 1%, right? And 100 billion rides would be 10%. So let's just say, for argumentous sake, I don't know if this is true, but let's just say Uber's 1% of all global rides. And then who knows how many, how quickly this will roll out by jurisdiction on a global basis. I suspect foreign jurisdictions will do it first.

Starting point is 00:20:42 You know, if you're in UAE, you know, you need to get the government's permission. And, you know, they could just have these up and running in Dubai tomorrow because it's a top-down government who moves quickly. Somewhere like the United States in Europe, you'd have regulators and local. and protests. I mean, that's how democracies work. So it's going to be much slower. So I took the position 1% of rides for a year for 5 to 10 years. It seems like back of the envelope that 1% of rides, by the way, is 10 billion rides a year.

Starting point is 00:21:10 So that means 10 billion rides, you know, in 2026. I don't think that will happen, by the way. That's a lot of rides. But let's just say in 20, 27, it went to 10 billion, then 20 billion, then 30 billion, then 40 billion. What's happening is it's not taking away from Uber. or a lift in my mind, it's taking away from car ownership and public transportation, as well as the dogfight between the ride sharing services. And people are like, oh my God, you're crazy, you know, next week, where this year Tesla's going to flip a switch and they're

Starting point is 00:21:38 going to own all of ride sharing. And I'm like, that's not possible because it's not legal in these jurisdictions. Listen, a strong sales team can make all the difference for a B2 startup. But if you're going to hire sharks, you need to let them hunt and you can't slow them down with compliance hurdles like SOC2. What is SOC2? Well, any company that stores customer data in the cloud needs to be SOC2 compliant. If you don't have your sock too tight, your sales team can't close major deals. It's that simple. But thankfully, Vanta makes it really easy to get and renew your SOC2 compliance. On average, Vanta customers are compliant in just two to four weeks. Without Vanta, it takes three to five months. Vanta can save you hundreds of hours of work and up to 85% on compliance costs. And Vanta does more than just

Starting point is 00:22:21 sock two. They also automate up to 90% compliance. for GDPR, HIPAA, and more. So here's your call to action. 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. You sold your last company, two companies ago, to Ford. You have a lot of experience with self-driving and with car technology,

Starting point is 00:22:44 and then definitive was sold to GROCS. We've got a lot of things in language models. With your unique perspective, you've heard my position, obviously Tesla bulls are out there saying, they're going to win it all. I get it. If you own the stock, you might talk it up. And obviously, you never bet against Elon.

Starting point is 00:22:59 But I think he'll just get as many rides as Uber over some period of time. But I don't think it's actually taking away from Uber Lyft. I think it's just taking away from people driving any service, driving themselves. And that's what's going to go away. As car ownership is going to go away to car rentals. And on 8, August 8th, they're launching and they're going to show a car without steering wheel pedals from Tesla, which then means they fire up the production line.

Starting point is 00:23:28 And we've seen the production line and how difficult that is. Elon talks about it all time. Cybertruck took a long time. I think they're up to 1,000 cyber trucks a week. I remember when Model 3 hit 5,000 a week. Like, this is really hard. He always says, like, designing a car is easy.

Starting point is 00:23:43 Producing it is hard. So who knows how long it takes to get the robo taxi up? Let's say he starts having that come off the line next year. That would be like a record, but, Let's say he's motivated, and I wouldn't underestimate. He gets there in 20, 26. They start coming off the line 26. Okay, a thousand a week, five thousand a week, ten thousand a week.

Starting point is 00:24:01 It's a dreadnought with his factory. So 10,000 a week would be half a million a year. Okay, half a million a year, a million a year, two million a year. You know, he's going to get, he'll get there. But it'll be one, he'll get one percent of rides a year every year, which is 10 billion rides. What's your take, Sonny? Am I wrong?

Starting point is 00:24:20 So, are the Tesla Bulls wrong? What's reality here? because Waymo is in two cities, three cities, right? L.A., San Francisco, Arizona, I believe. Yeah. There's a lot of other people in the space. What do you think is going to happen here, handicapped for us? I think there's like the following major considerations that like just can't like maybe like

Starting point is 00:24:37 overlooked. I think we kind of need to break it up. We are very far away from your driveway to your final destination in any service. The amount of cornercase. that it exists from every single person's driveway to every single business is very, very hard. Okay. And why are you saying your driveway? What I mean, why is that important?

Starting point is 00:25:05 You're saying your driveway, you're not saying, without sharing too much information, you know. I've been to your place, right? Yeah. And your driveway is hard for a human. Sure. It's a long winding driveway. It kind of goes downhill a bit.

Starting point is 00:25:19 So in order for it to get, you know, you got to get past the gates. right and so in order for it to come down to you know your front door and then take you out like it's it's it's going to take them great actually is going to be challenging but let's assume people are in apartments in cities and they're just going down to their street so okay what i'm talking about is i think we're further away from the convene full convenience of like what a human can do there now where i think we're going to and i'm going to give the other side of it where we're very close to is what they have in Vegas now. Have you tried the tunnels in Vegas?

Starting point is 00:25:57 I haven't been to that yet. Is it self-driving or is there a person in it? It's self-driving in the tunnels, right? And so basically you can- But there's no co-pilot there or a host. I thought they had a host in the car. No, no, no, there's a host, but I'm saying it's, but they put it in autopilot mode.

Starting point is 00:26:10 They put it in autopilot. It's supervised. So there's a safety driver. Yeah. But what you can see is that's a replacement for that Vegas monorail. That they're very close to. So I'm just kind of giving you the, the bookends here.

Starting point is 00:26:23 Okay. The hardest thing, getting to driveways in the suburbs, and I would assume you put snow, sleet, fog into that category as well. All of those edge cases, that's the most difficult. And the easiest is you've got a tunnel with no pedestrians guaranteed,

Starting point is 00:26:40 straight shot, boom, easy. You have a big area where you get in, it drives you through some dedicated lanes. It comes out, it drops you up over there. So that's 100% done, and the other one is not even close. And you're not even close. Yes.

Starting point is 00:26:53 And so where I think we land is bits and pieces now. So I think the bookends we agree on in the middle now. Let me break that down for you. I do think what we're going to have sooner than later is you see this off the 280, you know, like those car share parking lots. Yeah. So I think what we're going to have is car share to car share next. Ah, very interesting.

Starting point is 00:27:20 Yeah. So I go to the 280. I park in one of those parking lots. For people who don't know, there's a car share lot on the 280, which is the giant freeway. Every time you get off, there's just parking lots. You park at that parking lot so that you could meet somebody and you can leave your car there for free. It's a giant parking lot. Okay.

Starting point is 00:27:37 So I drive there. I get out. There's a Tesla or an Uber or a Waymo waiting for me. I get in. It drives me to Los Angeles. It drives me to Stanford or whatever or it drives me to the next one. I get off there and then I do one. It takes you to kind of well-known spots like that. So it won't take you like right to your final destination, but like some pre-programmed, some pre-programmed area on the Stanford campus it can get to.

Starting point is 00:28:01 Yes. This specific place. So it's going to take me to university. Yes. And this block, boom, it's going to take me to the Stanford campus. That's a very interesting idea. And that would be doable today, no problem, right? Yeah. And that's what I'm trying to say. That's kind of closer to the tunnel, right? Where if you're willing to drive from your house to the 280 and do that. It can take you to the Chase Center. And the Chase Center can read you back or SFO or take your pick. And so I'm confident that that's achieved as well. Okay. You believe that's achieved. No brainer. Exactly.

Starting point is 00:28:33 So I think that's achieved. And so what we're going to see is like sort of this progression of folks. I think we're going to see progressing from like those examples to then fixed within the city as well where it's like, hey, you know, I'll get you, you know, close. enough to your destination. The cities have already started making changes. Like, you know, if you notice in ASF, like, Market streets closed to regular traffic, but like taxis can drive on it. Yeah, only that, yeah. You can't pick up or drop off unless you're a bus. So you have to get off of market to do your drop off a pickup. Yes, exactly. And then just go around the corner or whatever. So they're trying to keep the flow going, which they did exactly as San Francisco

Starting point is 00:29:14 cleared out. So now Market Street's doubly empty. It went from being a, uh, chaos to being empty. Okay, great. Hey, everyone. Quick shout out to our friends at Wistia. Wistia is the ultimate tool every marketer needs to supercharge a video strategy. It's a powerhouse for generating leads. And I want to tell you about one really specific feature that Wistia has. When you're playing a video, let's say it's a tutorial video and it's teaching people how to use your SaaS product or even your consumer product.

Starting point is 00:29:41 Well, a pop-up can appear in the video that says, for more information or to book a meeting, put your email in right now. This is transformative. You can't do this with other video players. That's why you use Wistia. You get to capture people's email addresses right in the video. You have to see it. It is going to increase conversions for your webinars, for your book demos by at least 10x. And Wistia is going to seamlessly integrate into all your tools, Marquetto, HubSpot, Perot. This means all those fresh leads are going to go automatically into your marketing platform. And you can take those leads and score them and then get them into your nurture flows where you're nurturing those leads and turning them

Starting point is 00:30:18 into happy customers. Keep moving them through your sales funnel effortlessly with Wistia. They even let you add custom CTAs, calls to actions to your videos to drive prospects to the next stage of your funnel. So if you're looking to elevate your lead gen, streamline lead nurturing, and enhance user engagement with advanced analytics, Wistia is the perfect tool for you. Wistia is everything a marketer needs to create host analyze and market video. Try Wistia for free by heading over to Wistia. dot com slash startups. W-I-S-T-I-A-com slash startups.

Starting point is 00:30:49 Give me your prediction as to, because people think I'm talking at my book, oh my God, Uber, you know, I got some big stake in it or whatever. You know, I do have a stake in it, but I mean,

Starting point is 00:30:58 I'm not really talking about my book. I'm trying to, I'm trying to call balls and strikes here. So call balls and strikes for me. Let's talk about the ride sharing companies. Uber has a deal with Waymo to put them in there. Yeah. Let's take,

Starting point is 00:31:09 let's move out from just Tesla, for example, because there is a big debate. Actually, I want to go two ways. One, LiDAR versus not LiDAR, Tesla versus Waymo. And then I want to go to the networks like Uber's network, adding Waymo, adding crews, or those people competing. So which one do you want to start with? Let's start with LiDAR versus Vision.

Starting point is 00:31:30 Okay. So explain to people what this debate is and if people have won it. Because Elon said years ago, humans are capable of driving cars in all kinds of conditions. Therefore, cameras, which can see further and can see 360, should be able to do the same job. And that Waymo said, hey, LIDAR is necessary because you want to build this huge model. And there's all these edge cases, you know, a plastic bag floats across or famously the white truck that went across the highway tragically.

Starting point is 00:32:00 And the person was watching Harry Potter. And wasn't, you know, might be shame on that person. And, you know, I don't want to speak ill of the dead. But really can't be watching a movie with, you know, they tell you to stay vigilant. So take us through the LIDAR. Yeah. What I would say is you can achieve. you know, full autonomy without LIDAR.

Starting point is 00:32:18 But what I will say, so you're in the Elon camp. It is possible. But what I will say is just like planes, you can get into, you know, one of our friends' original planes, which can fly you, you know, from L.A. to, you know, S.F. Yep. But the newer plane has like collision avoidance systems and all kinds of other stuff. And so, look, both things can accomplish the same task. one just has a lot more safety features.

Starting point is 00:32:45 And what I will say is, it's not that they can't complete the same task, but the one with the collision avoidance system and all kinds of advanced avionics that don't exist in the two-seater plane is the one I would want to be in. Is the one you want to be it. So if we're talking about safety, you feel LIDAR will be some magnitude safer? Yes, because it's... Tell us, 10x, 2X? What do you think if you had to guess?

Starting point is 00:33:11 I'd say when we look at the final data, it will probably be, yeah, somewhere between two and 10x better. Like not like a hundred X better. Significantly better. Yeah. Now that won't matter if there's no fatality. So do you think people will die in a self-driving car in the next year? A Waymo, a cruise, a Tesla, whatever, in a robotaxie. We haven't had a robotaxy death. We had the tragic Uber one where the woman was playing candy,

Starting point is 00:33:41 rush, she got in big trouble, and the safety driver wasn't paying attention. You know, that was just human error. So we take that one out. Taking aside human error, will we see a robotaxy death, whoever's robotxy it is, not picking any company in the next couple of years as these things are deployed? Or do you think it's largely, you know, and it was the fault of the robotaxy, not somebody teabones it, not, you know, a deer jumps across the 280 tragedy. This does happen. And a deer smashes into, the windshield and kill somebody, putting those kind of things, do you think there'll be a fault of Robotaxy death in the first five years of this being deployed in the United States?

Starting point is 00:34:23 Here's what I'll say. I think we will not see a fatality in a LIDAR-based system for in the next two years. Okay. So you don't think a Lider will have a death? I don't think so. But does that mean you're saying you think a non-Lyter system could have a death? I think so. Yeah. Still, wow. Okay. Unpacked that.

Starting point is 00:34:46 Why? Why? Why do you think that? You know, LiDR, the systems that have LIDAR also have cameras. And so for all those things that you can't see with your eyes, the Lider is like an additional set of sensors, right? It's like the plane that has, you know, auto collision avoidance, right? Got. Okay.

Starting point is 00:35:06 Yeah. So the question is, are we 99.9.9 safe in a non-LIDAR and 99.999s or whatever. And we'll see. You know, one of the things I think that we're going to realize is in constrained environments, we will have interventions. We will have the cars pulling over. We will have them causing traffic. We saw that with Cruz and Waymo already. But what we won't have is fatalities. And the people who will die, it will be the fault of the humans and human failure, i.e. somebody jumping in front of one of these things or some biker, you know, trying to set a record going down some hill in the mission, you know, blowing through a red light or a stop sign because they're one on their bicycle hit some Strava record or some nonsense. Okay, very interesting.

Starting point is 00:35:54 Now, let's take the second one. What do you think on a business basis is going to happen here? I said, I think there's going to be 10 people who figure out self-driving in different jurisdictions and different modalities, all in the same time frame-ish, which I would say, you know, and let's call it the same 36 months. And what that means is since we have a, it's going to take a while to roll these out, even if you were the one who, you know, figured it out the 36 month, and let's say Tesla and Waymo figure it out in the first, you know, six month in that period, the rollout is so long that there's going to be an opportunity for, of those 10 players,

Starting point is 00:36:30 you know, let's say three or four of them to compete. and then Uber and Lyft have the option of the vendors who just want to sell their vehicles into a network of buying those vehicles. And there are many at scale folks, you know, whether it's Volvo or Toyota, you know more about this than I do, who will have these available and they don't want to run their own network. They want to sell into the existing networks. Is my thesis of that falls and strikes and accurate? Because what I'm trying to do is be just completely intellectually honest here, as, as an Uber shareholder and a friend of Elon's and Tesla and a fan of Tesla. I own Ford Tesla.

Starting point is 00:37:07 Here's what I learned through the Ford experience is that the most likely winner here is the one that has like the closest to the Uber model. And what does that particularly mean is that the massive capital expense is taken on by the owner of the vehicles. Okay. Explain that. If you're trying to build a network from scratch and you're trying to build a network from scratch and you're trying to own all the vehicles, right?

Starting point is 00:37:34 Those numbers become very big, very quickly. You're spending billions and billions of dollars on the vehicles, owning those vehicles, maintaining those vehicles and doing all of that. The beauty of like sort of the Uber model is you can put rules on it for maintenance and everything else, but all that's taken care of by the owner of the vehicle. And so if I were to bet, given everything that I saw the last few years, it would be the Uber model where people bring their asset into that is the one that's going to win. Which is what Elon said with Robotaxi when I guess it was how many years ago was that?

Starting point is 00:38:11 It was before COVID he did a preview of this. Yeah. I know, was that six years ago or seven years ago? Can I share something we learned during the, you know, some of those days in Ford, which was many people that drive in the ride healing are not permanent drivers. and there's a lot of really interesting incentives, and I just want to kind of commend Uber for doing this, but you can go and get a car through a car rental agency,

Starting point is 00:38:39 like Hertz in a bunch of other places, which will allow you to drive, and you really only have to do like 20 rides a week, and basically the rest of the time you can use a car yourself. Wow. Oh, that's very interesting. I wasn't aware of that. I do hear these kind of stories. So I'm, let's say, lower middle class, you know, I'm a blue collar person, student.

Starting point is 00:39:04 I do 20 rides a week. I make 10, 15 bucks a ride. I get, you know, whatever that is, 800 bucks a month, thousand bucks a month. I don't know how much you're making on the average ride, but the average is $37. Yeah, you got to pay hurts like $200 bucks. You've got to cover your gas and other things. You pay that $200, $400, whatever. You're in the black very quick.

Starting point is 00:39:24 Yeah. From some number of rides, right? If it's $37 an hour, let's say $35 an hour, let's say $30 without the gas, $30 an hour. And if it costs you, I don't know, $400 for the thing, you've got to work 10 hours a month to have a car for the rest of the time. I never thought about it that way. Yeah. You know, there's another analogy is people use this service get around and some other ones and they rent their car out. Yes.

Starting point is 00:39:47 Three days a month for $300 to folks, appear to peer. So you've rented out four days a month for $100 a day, that's $400. And if your car payment was $400 or your maintenance and insurance, everything was $400 a month, whatever it is. I'm talking about you have like a used Toyota for you or something. You're in the black. It's pretty great. Yeah. So that's where I think like what gets underestimated.

Starting point is 00:40:10 So that's where I kind of like the Uber model where they have the network. And then people will bring their car to that network and put it in the network. And it doesn't have, you know, 10 hours a day and all that. It could do 20 or 30 rides for you a month. then it covers your bill, and then the rest of the time you can use a car for whatever you want to do with it. Yeah, so this leans to me to believe that Tesla is going to have a significant advantage over everybody because they have whatever number of millions of cars on the road, adding to it

Starting point is 00:40:39 millions of cars a year. But 10% of those decide to put in the network, which seems like a pretty good number to start. Yeah. 10% of a couple million cars, three, four, five hundred thousand cars, very significant, put them into the network. They start doing, you know, five rides a day. We started our conversation as the capabilities increase where we go from the fixed route to carpool to carpool to, you know, full on over the next couple of years, you can decide how you want it used.

Starting point is 00:41:09 Now, talk about the other players. Waymo has decided they want to put their vehicles into the Uber network. You'd be able to pick just like Uber does, yellow cabs, the livery tabbs in the UK, Waymo. If they split the revenue from the Waymo's and they have the other ones, they just have the largest network in the shortest wait times, right? So that protects them in your mind as a network business? Or do they need to own one of these companies? Or merge, just Ford?

Starting point is 00:41:36 So we look at Uber as a company, right? Worth well over $100 billion. Do they need to merge with, and they own 25% of Aurora or whatever? Do they need to bring that back into the company? Do they need to own cars in their network? Do they need to do a deal with Waymo, maybe Waymo, merges into Uber becomes one thing. I honestly think what would you do on a strategic basis?

Starting point is 00:41:59 I think the automotive OEMs are out of the self-driving business. You think they're out of it? Yeah, other than Tesla. Who do you think the top three players will be? If you were to, we're sitting here five, ten years from now, who's going to be the top three or four players? I would say it will be Tesla, Waymo, and someone who has licensed technology

Starting point is 00:42:23 from either Tesla or Waymo. There'll be three players. Did you see the open source one in China? They're open sourcing it. I think Bill Garley shared on our group chat. Yeah, yeah. The open source one. So we sit here and talk about open source.

Starting point is 00:42:34 Yeah, we wouldn't shut up about open source. If you think open source wins the day here and everybody can use that software on a range of cars? I disagree with the open source one because this thing requires like continuous improvement and resources. And this is like where open sources is just not enough, right? Got it. I think. So it could be an open source program where somebody makes the commercial version of it,

Starting point is 00:42:57 the WordPress.com to the WordPress.org like we talked about. Yeah. But I think like Elon said this really well. It's like if you're not spending $10 billion a year on the resources for AI, the open source is just, you know, software at that point, right? So that's why my bet would be Tesla and Waymo in it directly. And then the third player will be someone who licenses their technology. And where you think that leaves the Uber's and Lyfts?

Starting point is 00:43:22 of the world over the next five, ten years. I think cars end up in their network. Yeah, that's what I think. You see, people have looked at this as like Tesla, Waymo versus Uber and Lyft, and the way I look at it is that group of people versus owning a car or taking your own car. And they're so, the overwhelming majority of rides, 99% of the rides in the world are done by the owner of the car driving themselves. and then there's a whole much of other rides,

Starting point is 00:43:53 I'm not counting, that are people taking public transit, right? Yep. Those two pools are so giant that they're just going to, every year, 1% of those rides, 10 billion rides will just flow into these other options.

Starting point is 00:44:09 So those will keep growing. So people are looking at it like, I think people have the wrong, they're setting up the wrong rivals. It's not Uber, Lyft, Waymo, Tesla versus, each other and they're all fighting it out. It's that group every year taking 10 billion rides from people driving themselves. And then what percentage of those rides they get, that is material.

Starting point is 00:44:32 So it's kind of like. And there's, what people discount, there's also more and more people on the planet every year slash there is, you know, more more people. More more people. More economic activity. More rights to be done, right? So it's just like it's like a growing market anyways. And so also going to induce more. So, you know, if you, you know, there are people who don't own cars, they take public transit. They make decisions on where to go to eat or movies or whatever. And I'm talking on a global basis here. So I'm not being classes here.

Starting point is 00:45:00 I'm just talking about just objective reality. So, you know, you go to another country. There might be a group of people who only eat in their local village or city or town. They don't have the option to go to the town, two towns over and have somebody drive them there for 20 minutes and eat the restaurant over there or go shopping over there. Now, if they can take, you know, a, a robotaxi for a small amount of money, they say, oh, you know what? I want to try some restaurants over there. I want to spend the day over there. I want to do a gay trip, you know, there.

Starting point is 00:45:30 Or I want a walk over there. That's not even a non-American. Like, you know, we would meet all these people that were taking lifts to work every day, right? Because, you know, they had a job across town, right? And it, you know, saved them time and, you know. Point-to-point was quicker than public transit, yeah. Exactly. All right. Wow, that's a great discussion. Okay. Let's do a demo or two here. We're 40 minutes just on news. And I think we had to do it, though. And so I hope this is a great opera, you know, for everybody. All right.

Starting point is 00:45:58 I'm going to, I'm going to do two or three demos really quick. I think court had a product. He wanted to show, too. Oh, he did. Okay. Yeah. Court, I was talking to court about something I saw. Court, you want to jump in here?

Starting point is 00:46:11 Hey, guys. How you doing? Hey, everybody. See producer court? He's in his bunker. Wow. Are you a preper court? Nobody's ever seen producer court here.

Starting point is 00:46:19 Everybody knew producer Nick, but here's producer Corey. Say hi. How's it going, everybody? Are you in a bunker? You're in your bunker? I'm in a basement. Are you a prepper? What's a prepper?

Starting point is 00:46:30 Like you have like a hundred days of food and you're waiting for like the end of the world, like fallout or something? I don't even have one day of food. Okay, great, perfect. I'm the opposite end of the spectrum as a prepper. Okay, good. Okay, you're a risker. So what was this thing we were talking about?

Starting point is 00:46:45 We had talked about like to back AI death calculator that's been getting some buzz. You hear about this? Did you hear about this? No, break it. No. No. Here we go.

Starting point is 00:46:54 Look at this. Producer, showing up with a demo. So this is the Lifetovac. dot Io, which I saw some news articles saying, be careful using this. A, they were saying it's strangely accurate.

Starting point is 00:47:05 And they even say something ridiculous, like 78% accurate, but I don't know how you'd come up with that statistic without a lot of years of testing. And this is what they do. They ask you to fill out stuff like your age. You're 47, court? I am 47.

Starting point is 00:47:17 Oh, wow. You got such a young energy. I had no idea. Yeah, you know, you're not allowed to ask when you hire people, some deep people their age. So I don't know. Yeah. It turns out 47.

Starting point is 00:47:26 I didn't realize that. I was stuck in. Maybe if you had done that, I wouldn't have snuck in. No, you know what? I got to tell you. I just judge people on their attitude and their aptitude and their execution. You know, I don't really care about itch. So 47 now, you don't smoke.

Starting point is 00:47:41 You don't exercise regularly. No major health conditions. So, yeah, any major health concerns? No, diet I put unhealthy because I'm stressing. I work too much. Much alcohol consumption. No, I don't do that. Stress level high, hours of sleep, six, positive, family history, positive.

Starting point is 00:47:57 In that next impact, I put neutral access to health care. I'm here in Canada, so it's adequate. So, socioeconomic status, I just stuck middle. So, I mean, there are a lot of statistics that you could try different things if you didn't know how to put yourself in. So where are we? Education level, fine, occupational hazards, okay, whatever. And then recreational activity. So let's press calculate.

Starting point is 00:48:15 I'm going to live 56.5 years. What? All right. So you got, you're saying you got nine years left. I think that has to do with your occupational hazards. I would say it's zero. Oh, yeah, you're right. Environmental factor impact would be, I guess, very positive too.

Starting point is 00:48:33 I got two great kids. Shout out to the kids. Let's calculate that and see if I live a little long. Not really. I don't know what's going on here. So this thing has no credibility. Or it's completely accurate and I've got to start prioritizing some of my things for the next eight years. Yeah, maybe I've got to get your affairs in order.

Starting point is 00:48:48 And so this is using AI. to do your life expectancy. It makes no sense. It has no credibility. Okay. Death calculator. I give it an F. And this thing is stupid and it's a dark concept. I'm going to give it a ridiculous F. Yes. You give it an F. Okay. So there's your first court. What do you give it? Sundeeb? I don't even want to raise it. It's too dark. It's too dark. All right. Good job, producer. I like the contribution. Okay. Well done. Dismissed. By the way, I gave it to him. I told him check this thing out. I will find out if this is legit. Oh, okay, okay.

Starting point is 00:49:21 I thought he went into the week. I just want to give some shine to producer cord who's doing an exceptional job since producer Nick left. I'll give him one of our demos, so Jake Feltz. Well, I think it would be interesting if, to be honest, if it took your watch, your weight, how much screen time you're doing, all the factors. And then also had you talk to it and, you know, answer questions. And then it might actually, just from a video of you, be able to tell from your eyes, your blood, your body type, And you put in your Renovo, AI is going to be able to tell you your life expectancy.

Starting point is 00:49:53 Oh, yeah, for sure. Very sure. Because doctors do this already. If you had your oxygen, you know, and everything, your stress level was actually from your heart rate and your sleep monitor was from your sleep monitor. So I do think the death calculator as dark as that is, is a really viable real product. I wonder if there's a startup working on it. I think a startup that did LiveLonger.A.I. Wouldn't that be a killer startup to try to look at all your data and give you suggestions in real time?

Starting point is 00:50:19 about how to live longer? I think we know the answer, Jake. How you eat better, workout, you know, go see your fucking doctor. Yeah, I mean, but, you know, I think the thing is, I don't think you need an AI to tell you how to be healthier. But it would be, I mean, it might find some edge cases like, hey, you know, you're under too much stress, you're not sleeping and not drink enough

Starting point is 00:50:40 water or your heart rate's a little weird. I don't know, maybe it finds some edge cases. I mean, the Apple Watch does that stuff for sure. You know, it's starting to give you some alerts, right? Like, it tells people to have. You know your heart rate if you're like doing like, you know, sometimes we're playing poker and I'm like, oh man, like I could see it like shoot up for me and then you can give me that shot. That's how I know to go all in. That's when I know you're bluffing. No, no, you, you always go like, oh, Sunny go all in here. I like, I, you know what I'm like saying. Just go all in. Okay, give me a demo here. Come on. Press me. Okay. Let's, we're going to do, we're going to do a couple. There's some awesome ones here. Okay. So this is by the founder of Signal. So it's funny. So my feeling is and I already did one before and we can do one live. So. Let me show you.

Starting point is 00:51:19 It's called vibecheck. Dot Market. And what it does is, you know, so this is just kind of building on, you know, last week we had the vertical search engine. You know, we've been working on this, obviously, you know,

Starting point is 00:51:30 building on our Mahalo experiences, J.C. Yes. And so on this particular one. And, you know, I did like a special machine. And so I'll do one live with you here. It's been overloaded.

Starting point is 00:51:41 So give me the best, what do you want the best of J.Cow? The best headphones. Oh, headphones. Okay. Yeah, you know, these are my focal bathies. Shout out to my friends at Headphones.com. It's not a sponsorship, but they sent me a pair of these because the founders are a big fan of mine.

Starting point is 00:51:58 And I had, they heard me say like, oh, you know, com.com, get a great domain name. You know, a decade ago, they followed that playbook of inside.com, com.com, et cetera. And they got headphones.com. And then they had all this credibility. And then people started buying the headphones there. And there's this company focal that makes these batis, B8. T-H-Y-S and these have a DAC built into them. And the digital audio conversion.

Starting point is 00:52:24 And they're 800 bucks, so they're not cheap. And I've had all the different versions. So these are like a step-up. But they have a built-in DAC, a digital audio converter. And you plug it in through a USB. And then I listen to FLAC files now where this Q-B, which is a high-res streaming service, as opposed to Apple Music. This one has all flaks.

Starting point is 00:52:44 And, man, and the speakers are really high quality. and they angle towards your ear drums. And yeah, Q-Cobus. Q-O-D-U-Z is the app I'm using with the flax. Okay. So anyway, if the servers are overloaded on this one, I'll try it in the background. Oh, wait, it's using all of the,

Starting point is 00:53:02 it's using all the subredits to do this? Yes, exactly. Do they have permission? Because remember Steve Hoffman, friend of the pod was like, you got to have permission? What's the vibes here? Are they, they have an API access? Well, they have something.

Starting point is 00:53:13 So, look, it's saying server overloaded, so we can't do your headphones. try it one more time. You had one for the espresso machine. It was really good. I did the and what's cool is it asks a few questions as well once it analyzes. It's like, you know, in the headphones, it would have been over the year. And so it does it. And basically, it gets you a top pick out of it, you know, your alternative and your budget pick. So it's like, it's like a wire cutter. Yeah. I love it. You know, AI wire cutter. Yeah. And so the question is, should it be allowed to use wire cutter? And this is the perfect example, not to constantly go back to

Starting point is 00:53:47 rights and clearing rights. What this says to me is this demonstrates why what chat cheap T4 and open eye did and stealing all that content from the New York Times is so, so unethical. Okay. And because you can see here, we all want this product. Yes. We want this product more than we want wire cutter. Wirecutter has the opportunity to create the derivative products from their IP.

Starting point is 00:54:13 Yes. This has proven my exact point. and the New York Times is going to get, I predict, an injunction against chat GPT4. Your point is that because people are not licensing the data, this is what's, this chaos is occurring. That they have stolen the opportunity to create this product from their IP. The way IP law works is. But this is coming from Reddit. Right.

Starting point is 00:54:36 And Reddit, it's their opportunity, since that's their content, it's their data. They own the data. Now, you might say, oh, the people who posted it on it. No, the people who posted it gave a license for all time to Reddit. Therefore, Reddit, and they did that in exchange for Reddit being free and providing the community and all stuff. So Reddit or Quora own the rights to build these type of products, the New York Times wire cutter under, they have the rights to this opportunity. But is the pricing here? What's your budget for it?

Starting point is 00:55:07 Oh, yeah, 500 to a. I should go 2000 plus. Right, that's fine. Okay. You need noise canceling or no? Sure. open back, close back, no preference. No preference.

Starting point is 00:55:19 And so hopefully this completes now. And it's like highlighting all the subredits is getting the answer from. Yeah. Okay. And okay. So by the way, that's Sony. I have those. Those are fantastic.

Starting point is 00:55:29 They're like 300 bucks. So Sanhizer's, that's another great brand. Yeah. That's correct. Those barodynamics are also incredible. And people do love those. So those are all fantastic and they're all dialed in perfectly. That was a perfect answer based on my.

Starting point is 00:55:45 extreme knowledge of this space. For 500 to 1,000. Do it again and then pick over 2000. I just want to see if it picks the focals. This is like instant wire cutter. I love it. It's like, you know, I don't have to end. Well, and here's a thing.

Starting point is 00:55:59 And this is where the stealing and the IP gets even more complicated. We don't know. Maybe these guys went and cut a deal with with. I'm not talking about that. I'm saying on these subredits because it's user generated content, you can be sure that people said wire cutter pick this as a. the budget pick. Wirecutter pick this as this.

Starting point is 00:56:17 And they cut and paste and they put all that data there. I see. I see. And they put, you know, the New York Times says this. Oh,

Starting point is 00:56:22 and Consumer Report says this. So, you know, where does that land? If someone says that on Reddit, where do you put that now in terms of content? Whose rights is that? Yeah.

Starting point is 00:56:34 So if I go to Reddit and I go, oh, I read on Wirecutter that the Sony, whatever is the best. So if the person is cutting and pasting it, then that person has committed a copyright violation. And, um,

Starting point is 00:56:44 It belongs to the New York Times. If that person wrote their own thing and says, I disagree with this as the budget pick, but wire could it pick at the budget pick, consumer reports pick this as their budget pick, PC magazine, Times Hardware, pick this and gadget pick this, Vox pick that, verge pick this. And they do their own like little. That's okay. Because they're making a new product, which is like they're going matter and comparing

Starting point is 00:57:06 all these things to each other. And now somebody will say like, oh, well, that's what the AI is doing. But the AI is an automated commercial thing. non-commercial, an individual doing a non-commercial message on a message board is not interfering with the commerce and the revenue opportunity, and they're not doing it at scale for every single category. This is doing every single category. So the scale of it also is part of the difference here. I give this an A. Perhaps I'm not going to give it an A plus because I haven't used it enough, but out of the gate, it's an A. Because this, from the one example you gave,

Starting point is 00:57:40 It nailed it. If we were to take that and compare it to wirecutter, Tom's hardware, PC Mags, Verge, Engadget, the best one, whatever, it would be probably very accurate. And I think that's what's happening here. Also, it could be indexing the links that are on the Reddit page. So if I was building this, I would say, hey, you know, if they link from, you know, headphones slash headphones, which I'm part of that subreddit now, if that linked actually to, I don't know, the Engadget or the Tom's hardware review, I would then go ingest that to make it work and make a better product. So they're probably doing that as well. So anyway, excellent job.

Starting point is 00:58:18 It's A. Okay. I do want to know if they had the rights from Reddit to do this. Steve Hoffman can answer that question because they have a licensing business. And this is what you should do. Get in touch and, you know, what's the market cap of Reddit right now? I've got to maybe have to put a J-trade in here because I think this will be built into Reddit soon. And I think this could be like the beginning of a great J-trade.

Starting point is 00:58:37 It should be worth five times revenue. Reddit Market Cup right now, 7.5 billion. That's a lot. Really? Yeah. On 800 million in revenue in 2023, I believe. I don't know what they did in Q1 or whatever, but let's say they did 250 in Q1 and they're out of billion.

Starting point is 00:58:54 That's seven and a half times revenue. It should really be four times revenue. So, yeah, that's a bit of a challenge. Yeah, it's overvalued right now. It should be worth $4 billion, $5 billion. But maybe people are making the J-trade and their jatrating.com based upon the thesis I'm putting out here, which is, if this is what you did on Reddit, you asked the questions like this,

Starting point is 00:59:14 Reddit would become supremely valuable. Like, if I asked it, what movie should I watch tonight? Man, that could be really powerful. So shout out to the Reddit team. All right, what do you give it? You got to give it a letter grade.

Starting point is 00:59:24 I'm A. I give this an A. Like, I'm a really, really big fan of wirecutter. I think I probably end up using it, like, weekly, if not more often. And so having your own custom one was really awesome.

Starting point is 00:59:36 And like, I like the way they did it. So you know it went and looked at all those subredits because that's usually this crazy algorithm you do when you're looking for the best thing. And just saving all that time. Let's do one more demo here. So this one was totally, it's awesome. Basically, give slide decks, PDFs, whatever.

Starting point is 00:59:54 I did this beforehand just for speed here. I shared this glossary of American medical terms to it, which you can see here. This is like a, you know, several really, really long document here. I shared this. And what it does is it takes that and basically turns it into flashcards for studying. And so basically it creates this from that PDF, which is right up here, what is the medical term for this?

Starting point is 01:00:22 I mean, I don't know what it is right. And so it's going to be like, no, that's wrong. And it just basically goes through and you say next. And basically, you know, so it takes any piece of content you give it and creates a kind of flash card learning system from it for you. Yeah. I think we had a startup in our accelerator. It's going to come to me in a second that does this.

Starting point is 01:00:42 They do a really nice version of this. And they are making some decent money. They've been building in public. And they make study guides based on your notes from your classes and information like this. Because there's like a whole long tale of things you could do in education around this concept. Like you could, if you were the teacher, you could make a test that's unique for each student. if you're a student, you could do the test, and then whatever you didn't get right,

Starting point is 01:01:09 it could do adaptive learning and figure out why you're not getting it right, et cetera, et cetera, et cetera. And so this is the AI disruption of that. Yes. And so this is absolutely. Well, no, the one we did is doing it in AI, but they did the long tail of all the things. And so this is definitely a real space.

Starting point is 01:01:27 I think it's fantastic. I give this like a B, because they're doing one thing. But there are other people. Well, I like how they do it across videos and other. things, right, which is, I think, pretty cool as well. So I kind of give it an A. I thought it was very, very flexible

Starting point is 01:01:42 in its input, which I thought was super powerful. Basically, take data, make flashcards learn faster. Yes. Does it run the flashcards with you in an interface? Yeah. So that's also right. I was showing it right here. Like these were it, right? And so

Starting point is 01:01:58 yeah, fantastic. Just kind of go through it. Hey, which does this? I don't know. I'm just guessing these things. Right. Is it right? I mean, I think that this is fantastic and it's going to help people learn faster. And then just being able to say to people, teach me how to do, teach me to be better at my job, teach me about language, teach me about the states. You know, we used to have these kind of things. Like if you wanted to learn all 50 states, this is a great way to learn it.

Starting point is 01:02:21 There's always these people I remember are super impressed by it who could repeat all the states. Like Sarah Silverman can alphabetically just do every single state, boom, super fast. Awesome. Okay, one last one, JCal. I want to give this company, both of them actually are using grog, but these guys just integrated grog. This is a tool that we talked about before. You know, we talked about Globe Engineer,

Starting point is 01:02:43 and it's like, hey, I want to do a trip to Newark, and you can see it starting to build up. Oh, right. This is this, we saw this one last time. I think I gave it a B. It's a really cool one. Exactly. So what they've done, you know, you see it's doing its work,

Starting point is 01:02:54 but I want to show you here, they've gone and integrated. This is Globe. Not Engineer. Exactly. So what I want to show you, this one's still running back here, trying to build out this itinerary for,

Starting point is 01:03:02 for, for New York. But in parallel, I selected GROC over here, and I say, I want to go to London. And what you're going to see is see how fast it just basically goes and does the same thing. And so this is kind of just building on like sort of what we've been talking about is building really quick user experiences and having it come back quickly makes it more iterative for you. So I wanted to give these guys a shout out because they integrated GROC into it. And I've been using it all the time.

Starting point is 01:03:27 This is actually a product I've been using myself because it does that really wide search. now what I'll say is let's try your headphone thing in here. So what was your, you'd say, uh, use head. High fidelity headphones. Okay. Hi fidelity headphones. All right. And which are for short or high five, yeah, but let's see what it does here.

Starting point is 01:03:48 Over the ear. Okay. Yeah. So this is just another way to get the same data in a really great way. And yeah, it's, it's getting some of the great brands. Um, but it didn't get the best brand. So that's interesting. And what was the best brand?

Starting point is 01:04:02 It's this focal. So just say, most expensive next. Or most expensive. Oh, I should have done hypodilite. Let me do this again. Because you can refine it. That's what I just did, because I'm doing, you know, taking kids to Japan in the summer. Yeah.

Starting point is 01:04:19 And it really is like helpful. Most expensive. So here we go. Okay. Boom. So Sennheiser, Barodynamics, Noble. I mean, it's getting a lot of them. So it just didn't get the ones that I think are,

Starting point is 01:04:30 the best. But, you know, it's using the open web to do this. Type in focal next. See if it finds focal. Yeah, I found them here. Yeah, yeah, yeah. Okay, great. Amazing. Yeah. And then say sort by price. Yeah, under 500, between 1,000 and 2,000, over 5,000. Wow. Yeah, look at that. So, look at that. The over 5,000, I have the Bathies, which are the 800 that do Bluetooth and let you plug them into your USBC and get you, um, this, you know, digital audio conversion built into the headset. Then these other ones have these thick cables that plug into a DAC, like the really like each ear plugs into the DAC and has this like super high fidelity

Starting point is 01:05:09 experience. And those cost $5,000. And they're just extraordinary. The top of the line focals. Yeah, they're pretty amazing. Yeah. All right. This is great.

Starting point is 01:05:17 See how this like changes the research experience? It's really, really awesome. I really think it's like. I think it's just great like brainstorming. This was called mind mapping back in the day and you would do it manually. There was a company called thebrain.com. And there was a guy, Jerry Colonna. So if you type in Jerry's brain, he was a very famous analyst.

Starting point is 01:05:37 And he worked for Esther Dyson, Jerry McCowski, and fabulous editor of technology. You know, you might not know him in this generation. But over time, he kept, so like launched the brain here. So he's been keeping notes on technology for over 20 years. And so this is like a living representation of all of his notes. And like, you can see, you just click on stuff. of, you know, any topic. And good starting points might be a good place to start.

Starting point is 01:06:06 So click on good starting points. And what you see is this software, you hit good starting points. And laws, there's the electricity. Yeah, big questions. I love the idea. Big questions. You hit big questions. Boom.

Starting point is 01:06:18 And now here's all the big questions. Click on to Beaver's Know How to Build Dems? Yeah. How do Beavers know to Build Dems? What does the universe store this knowledge? Like, really interesting that he did this. But what this is doing is AI is doing it automatically. This isn't following one human.

Starting point is 01:06:33 Excuse my voice, everybody. I was at the NICS Philadelphia 76ers game in Philly yesterday. Yeah. Let's try this. Go here and. Oh, look at that. That's awesome. Service we invested in Monick, A-I, M-O-N-I-C dot AI.

Starting point is 01:06:50 M-O-N-I-C, okay. It's the one we did. Oh, here we go, yeah. Study the guy. That looks awesome. Yeah. And there's 100,000 students using it right now. So they're just, it's a lot more sophisticated.

Starting point is 01:06:59 the one we showed today, but the one we showed today is great. So I don't know, I give Globe engineer still like a B plus. I think the interface is a good start. The data looks relatively good. I think it's still pretty raw. So I'm going to just give it a solid B action. I'm going to give it a B plus here. I'm going to give them a little bit of a bonus points for integrating grok.

Starting point is 01:07:16 So they're going to get a, they're going to get a minus. Oh, okay. Great. I love it. Fantastic. Yeah. Go ahead and mention us. Sandeep will be speaking at the liquidity summit.

Starting point is 01:07:34 That's the summit I do every year. This is the six year. We used to call it Angel Summit. Now we have a lot of LPs from giant institutions. And Sendeep's going to be speaking. Liquidity Summit is June 2nd to 5th. Go to liquidity pod.com slash summit. Tickets are about to sell out.

Starting point is 01:07:50 You can apply 125 LPs, GPs, the world's greatest speakers, of course. It's just nuts who's going to be speaking this year. David Freeberg and Chamath and Sandeep are here. Ibrahim from Ubatala, Gavin Baker from a treatise, Phil Deutsch, Arbesty, Monique Woodard, Michael Jones from Science. They did Dollar Shave Club. They did Liquid Death.

Starting point is 01:08:13 Sky Dayton, another bestie. It's going to be talking about airline safety, Phil Deutsch on Energy, Sundip on AI, Peshman on startups, just all kinds of great people. And we'll look forward to seeing you at liquidity summit. It's not for founders.

Starting point is 01:08:27 I know a lot of founders want to come. try to raise money, which is a great idea. But this is just for the LPs and the GPs and the capital allocators to get together and break bread. I'm sorry, but that's what they want. It's great you're doing this. I think it's a very, really good environment. And then, you know, out of it will come opportunities for all these startups and entrepreneurs.

Starting point is 01:08:49 And you've been at it before. You've come a couple times. We have poker at night on the second, third and fourths of Sunday night, Monday night, Tuesday night, we have poker. Everybody hangs out late. It's really about bonding. And on Tuesday afternoon, people go do an event. They can go wine tasting, clay pigeon shooting, all kinds of great stuff.

Starting point is 01:09:02 Oh, I have an idea for you. I have an idea for you. Absolutely. And maybe this is like a little bit of a balance. Maybe you guys can put together like a little binder of interesting startups. That obviously includes stuff there, but other things as well. That's a great idea. What we're doing is we're giving everybody a profile page.

Starting point is 01:09:19 So each LP and GP coming gets a profile page. Did they ask you about your profile page yet to fill it out? They did. Yeah. Yeah. Okay, great. And we're giving everybody a giant PDF with everybody attending ahead of time. And then, you know, not this one, but the next one will probably have a way.

Starting point is 01:09:33 We might do like a speed dating or we might give the venture firms like a little networking area where they have a table and then let the LPs book meetings with them for 20 minutes and then 10 minutes to move from table to table. And we just do a little speed dating. Everybody gets to meet each other in 15 minutes. Say what they're working on and exchange notes. And I think it would be a lot of fun to do like a speed dating thing. I mean to the speed dating concept right now because I want to meet more interesting people. but I don't want to do one hour meeting. So I'm trying to figure out how to do like a lightning round speed dating.

Starting point is 01:10:03 Hey, here's what we do, share some materials. So these demos, these got you into these, like each demo is like 10 minutes long. You got me addicted. Well, you know, it's a TLDR. I'm giving a new framework to my team. I call it WTF. WTF, is it? Okay.

Starting point is 01:10:18 WTF, why should I care? WTF, what now? So you show me Grock and console, right? Okay, what is it? Okay. Rock does inference at a fraction of the cost and at a massive speed difference. Okay, why should I care? Oh, well, you're doing a startup called Inside and you're trying to do this inference really fast and give people fast data.

Starting point is 01:10:39 This is why you should care. Okay, what's next? Oh, you should get one of your developers to go to console.grock.com. Rock with a cue. Giving the final plug, go to grok. Console.com. Console.com. Yes.

Starting point is 01:10:55 Any other plugs? Sunday, Barry got out. No, that's it. We're good. That's it for the plugs. We'll see you next time. Bye, bye. Okay, everybody, I want to tell you about founder Fridays.

Starting point is 01:11:05 What are founder Fridays? This is an opportunity for you, if you're a founder, to get together with a half dozen, a dozen other founders on a Friday, the first Friday of every month. Why is this important? Well, if you meet with other founders, you can talk about the things that are working at your startup and the things you're struggling with, everything else in between. and then you can trade notes and you can make friends. It's really hard to be a founder, isn't it?

Starting point is 01:11:31 You're alone all the time. You've got to solve all these problems. And other founders are having the same experience you're having. It's isolating. It can be scary. It can be thrilling. And people don't understand what you're going through. If you go to a dinner party and there's one founder and seven other people,

Starting point is 01:11:47 you feel like a mutant. You feel like somebody who doesn't belong there. Nobody understands why you're doing your startup, why you're taking this risk, the problems you're facing, right? They're NPCs. This is a non-NPC event. Every first Friday of the month, we do Founder Fridays. We're doing them now in 71 different cities.

Starting point is 01:12:05 And if this sounds appealing to you, well, you're a founder. And you want to hang out with maybe seven other founders around a roundtable. You can have breakfast. You can have lunch. You can have dinner. You could have coffee. You can co-work. Dinner, however you want to do it.

Starting point is 01:12:17 And it's free. We've been doing this for a couple of months now. We've had 71 meetups around the world. And 929 founders have joined. I want you to join, and I want you to come to the next one. It's Friday, May 3rd. Now, if your city's not on the list, what are you going to do? You're going to apply to run your city with two or three other founders. Again, four founders by founders. From the number one podcast for founders, this week in startups, comes founderf Fridays.com. Go to founder Fridays. Tech to sign up, and you're going to meet all these great founders, and we give like a little prompt. And this founder Friday taking place on May 3rd, We want you to bring two things with you, your most significant challenge and one thing you wish you'd learned earlier. We're going to go around the table and each person is going to do that.

Starting point is 01:13:04 And then you'll get feedback from your peers. It's incredible. It's magical. And we don't want this to get big. We want to keep it small. So if you plan on going and you're interested and sounds appealing to you, just go to Founder Fridays. Tech, T-E-C-H, right?

Starting point is 01:13:20 Great domain name. Founderfriidays. And please take pictures and then share those pictures on Twitter at mention us at TWA startups at Jason. And it doesn't matter. You can be in San Francisco, New York, Chicago, L.A., Paris, Tokyo, Dubai. These things are happening all over the world. Again, 71 cities. Let's get it to 100 cities. We've got over 900 members. Let's get it past 1,000. Go ahead and sign up and you'll be in touch with my team and we'll see you there. Once again, founder Fridays.com. You know.

This Week in Startups - AI Demos: Llama 3 Innovations, vibecheck, Globe Engineer, Wisdolia and Tesla’s FSD | E1941

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.