Did Apple Miss AI? How To Monetize Your Tesla with a $30,000/Year Payback? Live Vibe Coding Demo!

Going.

There you go.

Hello and welcome to the AI Revolution Call. I am Cody Willard, that's Bryce Smith. We work together at a hedge fund. We've been investors in AI related stuff including Nvidia for like seven or eight years now. Google, it's our favorite AI play at this particular moment I would guess down below here is Johnny boy John Dio. You can find him on Twitter. He is an AI, some old soul AI guy. He's been doing AI for decades, long before it became trendy and cool. And I am, I'm honored to call him a 30, a friend for more than 30 years at this point. So I don't know if he calls me his friend, but I called him my friend.

So 31 years, 4 months, 8 days and no, just kidding.

I was just tweeting about the call and I was gonna say we are live and I, I typed we are friends and I caught it before I clicked tweet.

But no, that's, that's too, that's true.

We are live and we are friends.

There you go. There we go. I get to work with a friend every day. Bryce and I get to work with John about two or three times a month on these calls and anyway, look, let's talk some AI revolution stuff. I think uh, probably first and foremost we touched on it a little bit last week, but let's just hit on it again. The robo taxi rollout from Tesla continuing a pace they've had a lot of like success. No major disasters and no news is good news when it comes to robo taxiing.

Yeah.

So I, I, I have a friend who doesn't, I wouldn't say knows anything in particular about technology. You know, she hears what she hears but she was like, oh, have you seen this Waymo stuff? It's really cool and everything. I said, well you know about Robotaxi, right? She says no. I said it's you know, the same kind of thing of doing the automated driving. And she says yeah, but I think Waymo has far better technology.

Who said this?

Someone to be unnamed? No, just somebody I know was just, you know, we're out to lunch and.

Had they ever been in a Waymo or a full self driving Tesla or they were just pontificate.

Think of this as your social feed. Typical non technical middle class person who.

Has probably neither been in either one.

Right, right.

Well so John, that this is not an a marketing call, this is an AI call. But Tesla is the world's leading physical AI revolution company. Probably the leading, totally.

I'm just like, this is why, even though it's not a marketing call, I.

Just like, let me finish where I'm going with this. It's also like the. Probably the leading humanoid robotics revolutionary company. But do you know how much Tesla's market cap is?

Oh, as they said, marketing budget, the.

Market capped over a trillion. The marketing budget is. It's a million bucks, maybe five million.

It's crazy.

They do, you know, they pay some social media people like social media editor, they don't pay influencers, they don't do commercials. There are no maybe on X A. You know, on X Twitter, you might occasionally see a Tesla ad, but usually not just their algorithm. If you're a Tesla guy like me, it puts it into your thing. Should Tesla finally start advertising. I love that. They don't. I will just say it. I am old enough to remember that Google didn't advertise for about the first five or six years after it had come public because they just let the product stand on its own merits. That's what Elon has always believed in. With the marketing budget, he figures you make the best product that revolutionizes the world. You don't need to advertise. But at this point, I got people look at my cyber truck and they think it's a $200,000 vehicle. When I tell them it's 75,000 and you can get an even cheaper one for 55 grand, their eyes bug out. They have no idea. And I just wonder, should, you know, should Tesla be like, hey, look, we do robo taxis now? Because not everybody's watching CNBC and studying this all day like Bryce and I.

Well, they're really becoming a, like, conglomerate with all of these different areas that they have. And I think they should be flexible enough to look at those areas independently and be able to say, you know, with cars, it works. It worked for them to be word of mouth and it could continue to work, you know, with. And then with each one of them look at that and go see. I could see robotaxi being a very word of mouth thing because, you know, you, you go in an area and then you, it's. People start using it, it expands out, everyone talks to each other about whatever. So I could see robotaxi sticking to that model. But I do think it's kind of probably a, a per product type question.

Yeah, maybe the cybertruck needs advertising and the robo taxi doesn't. Bryce, your thoughts?

Yeah, a couple points. First is anecdotal, but I just want to say thank you. As a model S purchaser Cody, because you funded a lot of brand credit that I get as a Model 3 owner and I didn't pay a premium price to get that. And so like when I'm around town and I get out of my car, people like dude, that car's badass, blah blah blah. That's so cool, you must be rich. I'm like it's a thirty thousand dollar car. You know, it's like the same as your Kia. But I appreciate it so thank you Cody.

You're welcome.

Yeah.

30,000.

No, no bad timing on my purchase.

For the record on that one.

But yeah, then the second point is I was just looking back this weekend at like my model for full self driving and thinking about the pricing and everything because right now you can still buy used Teslas for $30,000 that have the full self driving on them. You could buy a new Tesla for you know, 45 to 50,000 and you're.

Paying 8,000 including full self driving, including.

The 8,000 for the FSD. And if you are a believer, I mean in this, if you think that this might actually play out in the next year or two, I think your, your ROI like your payback period if you were to, if you were to make your car at Robo taxi in a major city, I think you could get like a nine month payback on it.

So there's got to be people right now, there has to be people that are launching this business that are buying 10, 15, 50200 Model Y's with full self driving with the belief that they're going to be able to turn on a robo taxi and let it join the Tesla fleet a year from now. Maybe not 200 but there's got to be people buying 5, 10, 20.

Why buy ahead just because you're concerned it's going to the minute that it.

Truly solves full self driving. Tesla is a capitalistic entity. They are going to maximize their profits. And if the car can can generate 30 or $40,000 in earnings for you after you purchase it, that will be incorporated in the full self driving price.

Well, I want to go back to.

Will go from 8,000 to $30,000 next year.

I want to go back to the Bryce you were talking about. You could buy a used model three for and you said yeah, 30 to 35,000. Okay, so why is my wife telling me that they're offering her 10,000 dol? Tesla is offering her $10,000 for as a on a trade in for her Model 3 with which has full self.

Driving option once A new car sell. They would like her to sell that to them and get.

Yeah, but that's that, that, that, that paying 10,000 for it is like instead if she went and sold. You're saying if she sold it used, she can get well more than the 10,000.

Okay, I bet we're not talking doll. Like they're not offering to buy her car outright for 10,000 DOL. Saying they'll give her a $10,000 discount on the other car. There's probably 20,000, $25,000 of inherent value in what they're offering her for the Model 3.

I'm sorry, I got confused on what you were saying.

There's no way that they're saying, hey, her car is worth $10,000.

That's where I'm saying like I'm, I'm working a little bit game of telephone because you know she's saying and I haven't gotten involved or something. But I did say that sounds oddly low.

Yeah, I just.

$30,000 for a used Model 3 with full self driving. And we're talking a four year old Model 3, not a newer one.

I did Tesla had a promotion earlier this year that if you bought a new Tesla, you could actually transfer your FSD subscription if you've paid for it to a new one, which they, you know, they haven't allowed that in the past. And so I looked at it and I paid 30 for my car used with basically no miles and Tesla was offering me 25 and it's got 10,000 miles on it now. So.

Yeah, all right, well it's. So I just wanted to check because it sounded. Something was. Sounded off.

Yeah, something's off with what you're saying there. They're not offering her a value of $10,000 for a car. Probably it's washing out that it's a twenty five thousand dollar trade in value and that she needs to pony up ten grand to get the new one or something like that.

Oh yeah, that'd be awesome.

Point being that look, the Model 3 Model Y with full self driving right now is an opportunity for someone who believes that full self driving robo taxis from Tesla will be working in major cities around the country maybe a year from now. Bryce is optimistic about that. I would say it's two years from now, but whatever. A year or two. If you can buy a car for $35,000 and it generates $30,000 in earnings for you two years from now. Like, you know, as a Wall street guy you dream about a two year payback period.

Are they Already printing out Model twos, the actual robotaxi cars.

Probably not. They're probably revamping a line right now and within three to six months they're the Model 2. Let's just call it the Cyber Cab is what they're calling that. The Cyber Cab will probably be rolling out in. I don't even know mass production would be the right term. But in full production at least, you know, a few thousand next year.

And then what about the fact that, you know, my Model 3 has the older version of hardware and they said that oh, they're gonna have to upgrade.

You because let's just put it in context. If you're going to have to. If Tesla's going to and they do you have an A, an older version of the camera system in your car and Tesla will. If you pay for full self driving. Elon has said they're going to have to. Tesla's going to have to eat it and they're going to have to pay for a new camera system to be. They call it AI5. Right.

Bryce AI4 is what they're using now.

So AI4 and, or AI5 if, depending on when you go do it. But you're going to go get a new version of the hardware put in the car and it will cost Tesla probably somewhere between 500 to 1500 dollars for equipment and labor to do that for you. And again if the car is actually going to be full self driving robo taxi, $30,000 a year of potential revenue generation like The Waymo costs $150,000 but lidars radars and 56 cameras on it.

Yeah.

So it's like, you know, not that really big a deal to spend $2,000 on.

Right.

Tesla will spend it. Tesla will.

But they're not going to do it until they reach the equivalent of their, their AGI, their, their checkpoint of you know, it works.

Yeah, I, I don't know what that there's got to be some sort of threshold that is met that then they say yes. All of those people that purchase that deserve this now not sure what that threshold is.

It has to be them declaring them.

Yeah. It will be Tesla saying yes. Bring your start bringing make appointments go on the Tesla app. Make the appointment, bring it to a service center even they'll send a service agent out while you're working and change it out while you're, while it's in the parking lot.

But I'm saying the milestone will be them saying we have achieved full self driving.

Like I don't know, probably that would make sense, John. But it could be slightly before that or a year after that, I don't know.

And then if they're already in Austin and they're running, you expect them to be in beta for a year?

I'm gonna. My answer will be different than Bryce's. I, I think it'll be six months to a year of rolling it out around Austin and maybe San Francisco and maybe Atlanta. And so you might have three cities operating more than a hundred in total Robo taxis a year from now. That would be where I think would be a reasonably optimistic timeline. Even Bryce, you're much more optimistic than that. I don't think you're reasonable about your optimism.

Yeah, I'm biased. I own the car. I bought, I bought the car for this reason. I bought it last September. So, you know, I've been waiting nine months now to monetize my vehicle. But in the meantime, I have a great product that I get to use every day. My favorite part of the trade, in addition to the upside and maybe like a one to two year payback is I do think like your downside is like maybe a couple years of depreciation, but if you just bought the cars, put them in storage, you know, don't insure them, don't drive them, just put them in a garage, you're actually like limiting your downside, you know, to maybe the cop, you know, the value of, of the fsd, you're going to lose that or something, you know, so maybe you have a 100% upside in two years versus a 20 downside if it doesn't work out or something.

So actually, yeah, I mean, again, as a hedge fund manager, you look at the risk reward scenario, you're like, hey, let's assume the worst case scenario that Robo taxis are still just in Austin three years from now and that you don't get to make money with your Model Y and full self driving. Well, then you're probably going to go sell the fleet. You bought 50 of them, you go sell them, you eat 20% because the cars have depreciated a little and they've got the newer models are a little better and blah, blah, blah. Okay, so you lost 20% on a trade. I mean, that happens once a month over at the hedge fund man. Meanwhile, if it works out at all, you're going to have 50 cars that have fully paid themselves in full in your pocket and are now cash flow generating pure profit machines forever for years after, five or seven years after. So you have really, Bryce, not 100 upside potential you could lose 20 to 30% if doesn't work and you could make 3 to 500% over the next five years if it does work. That's a damn good trade.

Funny part of what you said was you have in your pocket and I'm like this is not animal crossing. I can't stick all those cars in my pocket. But no, seriously, what if I can buy just the self driving option ahead of time at the current price? Then it would be like a call option.

There's a freaking hedge fund business right there. Right. Let's start selling call options on fsd. Actually though we wouldn't want to sell them. We need a broker to sell them to us. I want to buy call options on FSD working a year from now. Yes, John, we could create an ETF that Trump's administration SEC would.

Right.

Absolutely. Not prosecute for being outside of the bounds of legality at all. Even though it and might be a little off.

I. I'm gonna push back on the car in the pocket. It. I do have it in my pocket. Oh yeah.

See if that'll take you to work.

The point is that I like earlier today I didn't drive the Tesla today because I had to run around horses. But the like I got a software update and it's like would you like to install a software play? Yes. Install right now.

Yes.

Right from my pocket. Yeah.

So John, I guess. But no, that's an interesting idea. Really. The call option on full self driving jumping from eight to $30,000 to purchase is a great like there's a business model right there that actually is.

Wow, cool. I said something.

Yeah.

Yeah.

All right everybody, thanks. We don't have to talk about anything else today.

No.

All right, but let's just finish real quickly. Any other comments on full self driving AI Tesla? The car.

Did they did deliver a car to the customer on Friday and there was absolutely no one in the car. Not a safety monitor, not a passenger. And it drove like 30 minutes across Austin on interstates, through suburbs and all the way to the customer's house. And I mean I think that's just downright amazing. You can go watch the whole video, you know and on a three times.

Feedback and apparently they didn't even have a robo operator on standby for it. And there's no reason to have done it that way except for the fact that Elon wants to take the risks that no one else will take. I will tell you every time I'm in my Tesla Model S or the cyber truck and I'm like, I don't put on my seat belt and it might yell at me for a minute and then it doesn't, it doesn't really like it. Like, it is amazing. I think often I'm like, you know that Elon at some point was like, guys, turn that notification off. We don't need them to be worried about that stuff. They're driving a car. They know how to. They know there's a. You have to pay attention to the road. I mean, how many times are you in an old fashioned Toyota or something in the. Every time you turn it on they're like, I agree that I will pay attention to the road. Like that stuff. Like Elon's not worried about getting sued for something random. He's actually, you know, like trying to solve full self driving. It's only going to take, you're going to, only you're only going to be able to get there. If you take a lot of risks to get there. I mean that is Elon, you don't.

Need to take unnecessary risks. Cody, I want you to be wearing your seatbelt.

Elon says that the cybertruck is actually just as safe unseat belted.

Okay.

They, they have so much protection and so many airbags inside that car. You don't. He literally, he's like, yeah, you don't really need a seat belt. But you're not going to argue with the regulators about it?

No. Good. I didn't know that. That's interesting.

Bryce. Yes. You know what the most amazing thing about the AI revolution at this particular moment in time is probably the fact that someone without any coding experience can build a web app that is useful and functional and exciting and actually creates value for the world you've been creating. What do you think? A dozen web apps so far?

I can tell you, I think I made like six that I published to the cloud and use so half a.

Dozen web apps so far including like our model and thesis on Google and why we think it's a good AI investment right now and you can interact with it and it understands Cody's approach to investing. And so it gives you feedback on what I would probably say. And it's incredible. And I think what would be fun on today's call, John, unless you have a problem with this, I would like Bryce to show us how he makes a web app in five minutes. Five minutes or less. The man can make can he can talk to Gemini from Google and make a web app. Would you like to see this happen in real time, John?

Absolutely. No pressure, Bryce. If it takes six minutes, go ahead.

No, it's pressure, Bryce. You got.

Yeah. Well, my original app idea, which I was talking with, speaking with Cody about earlier, I did not like that. Well, I wanted to build something that could listen to a song, build a chord chart for a guitar player. But I tried it and it made it, but it was not accurate. The chords were not accurate.

I thought you were going to tell it to just search the web for the chords of the song and make sure it picked the it thought was the most likely correct version. I didn't think you were going to ask it to actually listen to a song and make. That's freaking next level, dude.

I'll show you what it made.

I'm fascinated.

Okay.

Just, I saw in a newsletter that the. The first Vibe coded product that. Where a company was basing its product on Vibe coding. The company sold for 80 million.

Are you serious? No.

Yeah.

So, yeah, this is the Google AI Studio. You can chat with it if you want to. Just chat with it and play around with the parameters and things to change. Then you click here and you click on build. Here's some example apps that you could build to help you get started. Some things that people have made. Then you go here to your apps and then these are the apps that I've made. Some of them are useful, some are not.

Oh, your very first one there, Normandy Landing.

Yeah. I tried to make a video game. It's not very good.

Okay.

I don't even remember what it looks like, honestly. It was like, let's see, right, this.

Is front end and back end.

Yes.

Yeah. Well, it gives you a preview and then if you want to deploy it, you click this little button right here. It says deploy to cloud run and it uses a Google Cloud run. And then you link it up to one of your Google Cloud projects and then it will give you a URL.

Go ahead and leave it, Bryce. It's not loading. Anyway, go show us the other stuff that you wanted to show us.

Okay, well, I'll show you the. The chord, the sheet music maker. And this is what it's fed up the only. I want Google. I wish they would save your chat history here. They don't do that. So, like the prompts I gave it.

That'S an update they're gonna give in the next month.

You know, five minutes ago. So anyways, this is Cord AI. It named it. It says, you know, allowed access to microphone. Sure. Right. And so then like, it's like start recording. So I'll just play something and see if it picks it up. That's what I was doing earlier.

Okay.

Your mic's not great, Bryce. I don't know if it's working.

Well, it might be. My speaker's too far away, but anyways. Then it spits out the corporation, but that's it.

That chord, literally, you did this in five minutes. You were like, yeah, I want to do this.

The chords just aren't right. The song's in G. I know the chords of the song, but you can. And you can transpose it. The chord, it's close. Actually. It's giving E minor cd. So it's not really that close.

I mean, it's not telling you B flat, Right.

But yeah, and it made the ability to transpose it, which I, you know, I told it to do that, but I don't know.

On the left side. Can you say I make it so I can paste a YouTube music video link for you for the AI to process, the core of which the chords will be processed.

The cool part with vibe coding is that you just iterate with the AI just like this right now.

That was a vibe code. We just did a vibe code.

Right. If you want to watch it, think, you click through this and you can see what it's thinking about. I'm focused on integrating the user's request enable court analysis from YouTube videos. My latest. My thought is we should allow an input field for video URLs. Right?

Yeah, man. How long will it process that?

And then so it's, you know, creating this and then it'll start coding. Right. And it. John, you probably know what all these stuff says stuff is.

Well, that's all front end, right?

This is front end stuff. Okay.

Yeah. The typescript.

And then, yeah, you know, you can click through this and you can actually see the code. And you can edit the code manually if you know how to do that. I don't know.

Go back shows where it's about to make the code.

Yeah, it's still. It's thinking about what to. How to do it.

Do you have to click something else on that menu in the middle that.

You go once it's. See here. It's going now.

Of course I can.

Holy.

Right, well, let's see if it's always.

Gonna say, search YouTube.

No, it's not ready. First it's saying, Corsa can do this. And it's working on doing it. And also it always pats him on the back. It's like, this is a great idea.

Yeah, they will. The AIs will placate you, for sure.

So it's still writing.

You remember in the irobot, the book itself, where the robot, like, tells the woman that that man, that the doctor that she's in love with is in love with her, too, and he told her, and he's like. The robot lies to everybody, tells them literally whatever it is they wanted to hear.

So should I try that same song?

Yeah, exactly.

Let me grab it.

Where's the URL, Bryce?

I recognized it, but I don't.

It's Wagon Whale by Old Crow Medicine Show.

Where's the URL box? I don't see it. Oh, oh, it's on the. Right there. It says, from YouTube. URL. So you got.

Holy crap, look at that. It added it.

So share copy first.

Wait, it's not. I'm not seeing a change. It's. Oh, there you go.

Yeah, I was on another tab getting the YouTube.

Okay, so now.

Let'S see if this works. I'm gonna be shocked. No.

You don't know what key it's in.

Yeah, I do.

I do.

Switch it over, Bryce. See, switch it to key to the key to. To the original. It should be in G. Yeah. G, D, E minor.

Oh, that's actually right. What?

I don't.

I don't play the B. You don't have to play that B. But. Yeah, G. No, wait.

Cody, get your guitar. Cody.

That's right. No, that's. That's actually right. It's G, D, E minor, C. Okay, grab.

Grab a song I know. Rain by the Beatles. Rain by the Beatles. Rain, Rain.

I don't know by the Beatles.

Okay.

When the rain comes. You better. It's a. It's what? From the White Album they played. Or Abbey Road, maybe. They played it on top of the. That concert they did on the rooftop.

Wow. There you go. What do you think that is it?

It's ish. But, you know, I mean, again, this is a first iteration.

Yeah, I didn't know. I don't know that. Like, if you just use Gemini, the voice mode, or. Or chat. Gbt. Voice mode. I've tried that before and played music and see if it could transpose it, and it can't. So I don't know.

How about one more try? I don't think Hank done it this way. By Waylon. It's two quarts. Pretty sure.

Okay.

B flat. F. B flat. F, B flat.

Yeah, but it could always be, like, transposed, however it gets heard and then.

Yeah.

Shifted around. It should. It should say that.

All right, well, this is an interesting.

Are you sure, Hank. Done it this way.

Because I. I'll tell you this, guys, even just watching this right here, like, this is the whole point. This is what I wanted to show that it is fascinating that we can sit here and just come up with random ideas because, like, Bryce and I like to cover songs. Bryce is a very talented guitarist. Singer. I'm a good guitarist and a terrible singer, but I don't care. I. I sing, I play, I love to cover songs all the time. And even in my, like, Facebook feed, it has finally, I guess, figured that out. And now every. Almost every time I ever bother going on Facebook, at least one of the things in my algorithm is like, hey, here's the chords to Girls Just Want to have Fun or something. It's random stuff, but I'm like, oh, that's fascinating. I love looking at the chord sequences of songs. And so I find this, this would be a valuable end product. But again, like, this is where I call it the dial up Internet phase of the AI revolution. Because this didn't work perfectly right now, but you realize in six months it will, like, this stuff will happen. This is coming. It's getting. It's figuring out how to do all of this in real time.

Right. And there's real value here. Like, I have an app that I use to get chord charts, and I think I pay a hundred dollars a year for the app. And so, like, if you could build this, if it was possible to build this, that's a real value, right? To, you know, millions of people use that app every year. Pay hundreds.

Yeah. On that note, can you ask it.

To.

Go back to the other setting where it's not the YouTube videos, and ask it if it will just grab the chords that it. From the inner. The. The tab. The chords tab off the Internet based on what it thinks would probably be. Looks like it's the most common version.

So how would you. How would you say that?

Search the web for chord tablature.

For.

Any song requested and use what would appear to be the most popular chord sequence. Chord tabs. Chord tab literature. Pretty impressive. You could type all that.

I don't know how you spell table tableture. Oh, I just, I asked it to add lyrics to. To. Yeah, I like that too.

That's brilliant, dude. Again, it's vibe coding. You're like, it'd be nice to have the lyrics right in the app. Okay, here you go.

Let's see. While it's doing that, we could probably try it.

And look, by the way, when we finish this little part of the Exercise. We're going to wrap this part up and we'll go back to talking some news or AI or something. But, Bryce, let's make a note. This time next week, let's show. Let's. Let's have spent some time on this. Let's show them the product this time next week. And I bet we can get John to participate in helping us develop it, too. Bryce, you don't know the story. 20 years ago, YouTube going mainstream. Not even going mainstream. YouTube is like, everybody on TV, in the media and in the mainstream is like, oh, my God, plagiarism and copyright, and nobody's gonna listen to it. I'll come back to my story in a moment. I see that you just finished or it all has.

This is a hallucination. This is not the right song.

That is amazing. Hallucinations happen still.

Yeah.

The price. I'm like, you know what? Everybody's worried about copyright issues on YouTube. I will just guarantee that shit's gonna get figured out. Like, the demand for content on demand from any source is infinite. To go ahead and turn your code thing off so we can just get back to the call. And. And so I started a company that was going and literally did. What we did was we cataloged everything that was Main street, like Hollywood produced content on YouTube because everybody else was worried that all that stuff would get pulled off. And I was like, nah, man. It maybe sometimes, occasionally, but eventually most everything in Hollywood will be available, at least in clip form on YouTube. So I was working. I had. I had. I was teaching at Seton hall and I had about 20 revolution, Economic Revolutionomics students. And I was like, hey, we're gonna spend today's class, you guys, search Hollywood content on YouTube and paste the link into this program, this app that my developers in India have built for me. And it will put it in, it'll catalog it. And I get an email, I tell John about it randomly on a Tuesday night. And the next morning I get an email from John and he's like, hey, I wrote you some code here. Copy and paste this and put it in the JSON, AP something something place. Tell your coder guys to do that. And now you. All you have to do is type in, like an artist, type in Neil Young, and it will go find every car, every video with Neil Young on it and add it into your catalog. So just for the record, John has for decades been helping me do what he's just about to help you do right now with this. You're going to be like, hey, John, we actually Would like it to do this and this. And he's going to be like, ah, I'll have my AI coder program create the python for you and I'll send it to you. Bryce, you can paste it right in that one little file.

You're. John. You're muted, John.

It's a true story. I'm unmuting you. I didn't know you were muted, John. You got to unmute yourself somehow. You chose to mute yourself. It says I did.

It's a good thing.

Right?

But no, that's true story, but the part about doing it now is a little bit like over overselling. Before it's. Before we're doing it, which is like, take it easy there. I was thinking about this one here. You know, you could. Another thing you could do is take things like Suno or Yu, the ones where you can like make a song out of from nothing and go, right, all right, I want a country song about this, whatever, that does this, whatever, and then it does it for you, and then feed it into your app to say, give me back the chords. Because they don't do that on that now. Nobody knows what they are. And. And you go and get the chords for it.

Yeah.

And then.

And that's where the. This is where that stuff feeds upon itself. Right. There's a company that's create. There are companies that have created apps that allow you to vicode a song. And now you need an app so that the people who vibe code a song can actually play that song. And they go play it Rosie's Pizza on Friday night. And so it just created. What did it do, Bryce? It created more demand for coding, not less.

A flywheel. It's a demand flywheel for five coding. And for all of the. Yeah.

A piece of news on the copyright front is that there was a kind of a first decision from a district court you guys probably saw were involving a suit with Anthropic. And I'm not sure who the plaintiffs were. New York Times.

No, it was book writers.

Book writers. And anyways, yeah, the court said that the, you know, training the model with the books was a transformative work and therefore AI Anthropic did not violate their copyright.

But they have to buy the book.

Specifically, they said that because you did not make the content centrally repositoried that people could access the actual original works. It all like, it was both the transformative and the fact that it wasn't centrally located to be able to pull. So it was a double whammy. Go ahead, Johnny.

No, but it's true that they're allowed to use it, but also they have to pay for it. They're not allowed to pirate the book and then use it. But the question is, does that mean they can just get on Amazon.com and download, pay for and download a PDF and suddenly make it part of this thing?

That's what.

Or do they need to be making a license deal with HarperCollins and every other kind of whatever. And if so, you're, you know, it sort of was half of a victory and half not sure.

John, do you remember 20 years ago Google did a project where they created a central repository of every published book and they. And it lasted, I don't know, maybe a year or two before the copyright cops really cracked down and made them close it. Do you remember that at all? I do.

I don't right now. I remember there's, there's one that still exists today and it's got practically every book and I wonder whether, you know, the OpenAI etc are going in using it because they have access to every book there is and they didn't go and buy every book there is, so they had to get it somewhere. So it exists now. Yeah.

How much advantage to those does Google and Apple have about with the ability to create useful agents because of the personalized AI they have about each one of the people in their ecosystem versus, you know, chat, GPT or random Gemini interface interactions where you're just like, hey, teach me about how electrons work. And then you, you know, the way that most people are using AI right now is sort of helping them learn or correcting them or helping them build or create a product or a spreadsheet or some written content or something. But eventually the promise is that we're going to just, you know, hey, book me a trip to the Middle east, check my emails for the dates and where I need to go, make sure I'm business class everywhere I can be. I don't want to spend more than 15,000 bucks.

So I. Cody, I don't. It's kind of like, I don't think I haven't seen Google's head start from like knowing me for 20 years show up in what they're doing with personalizing AI. However, as the pieces like you described start to happen with the agent, you know, you wonder why you hear things like, oh, OpenAI thinks they're going to, you know, get into email. It's because they know that they need to be in the personal information of the that person, find out about them. But it gets them into the workflow.

That's what I'm saying, like that and that's where the edge that Google for sure and Apple's less so, but both of them have this edge about your calendar, your emails, your, your, the apps you use, when you use them, why you're using them, where you use them, what you search, what you like, what you don't like, where you eat. I mean that's the thing that Google has, you know, decades of search history from me knowing what I like to do and what I like. Sushi. So when I'm in a city it's going to book me sushi restaurant and I'm not even gonna have to ask it. I'll be like, hey yeah, you know, on Tuesday night I'm gonna want sushi at 7 o'. Clock.

I don't know. There's got to be a, a kind of a curve on this where it only takes the most recent three months to get to 95%. And the other thing is say you're OpenAI and you don't have this. It takes almost nothing to authenticate and give access to and be like oh Yeah, I use OpenAI and I use Gmail and OpenAI says click here to give me access to your Gmail.

Bryce the trade desk, it's got a unique like it has it, it solves a lot of that problem for a chat GPT it knows it's got unique identifier for everybody who's ever used Facebook and or web browsers or whatever. They, they use a lot of different things. Some things are definitely okay and some things might be a little bit privacy invasive. But that's an interesting thought. As John just said that right there. It almost like the trade desk would sort of be an end around for a chat GPT or an anthropic to end up with a lot of personalized data about you and you know, each person's personal habits and interests. Couldn't do some work on the trade desk. Remember we looked at that thing at the bottom and it was like at 25pe and we didn't buy it.

It's still down a lot from the highs. It's the high was 140, it's at 72.

I think you can tell that I will be working on the trade desk as a potential new idea after talking to John on today's AI call.

So there's another kind of thing going on which is like anthropic mcp, which is model context protocol. And so they're getting them into the game of getting themselves in the game of, of like being the API, the API for AI in a way. And like it gets in between everything. And what it does is it's like it can get in front of an API and allow an agent to, a software agent to talk to whatever that API was providing. It's at the front of that going, okay, here's what I got for you. You know, if you want it to be this or that or just ask me this. And so it, it lets you access these resources. So think about the Gmail or you know, other things or you know, draw Google Drive all those. It's like with an MCP server in front of each of them, your software, your AI software agent can talk to any one of those and make happen what needs to happen. Find out. Lake, do you think they like sushi?

I mean you not think that Google could strategically decide turn that off. We don't want anthropic having access to people.

People's Google, you gotta open source and you just need to inter anybody's using it. If they have an API, they put it on the front of it. And not really.

Yeah, it's an interesting dilemma then. I mean that, that would eliminate. Part of our excitement about Google as an AI play is partly because they've got infinite amount of personalized data on you.

But.

Well, I thought what Google had done with MCP is like, it's sort of like read only versus read write. I thought was kind of what they had done to be like, okay, you know, we can't just keep everybody in the walled garden. If they refuse, they'll just leave. And so they're like, okay, well we'll let chat, you know, maybe have read privileges to your Gmail, but not read. Right. And you know, the true agentic stuff is going to be able to take action on your behalf. And therefore that will keep Google's advantage in place.

Typically depends on how you authenticate. And it is possible to read right from into Google accounts with, with mcp. For me, I just looked that up as you were saying it.

Okay.

And. But yeah, it's a lot of this is an interesting area and one to you know, look at. It's like wondering like Cody got into this again, you know, 10, 15 years ago with when he was looking at apps very early on was like, how does all this stuff relate to each other? You know, it's like this, this whole new world and, and once again and it's the new world all over again.

And this one's bigger, this one's a new universe. That one was a new world. I mean, I think that's the thing about the app, the app revolution was that absolutely, you know, trillions of devices, let's call it tens of billions of devices doing trillions of interactions every day. And that will, that's still growing at this point. It might at some point start declining as we have different interfaces with talking and using AI. But the, the as big as it got, it's still, you know, like, it's still cumbersome. It still is so lacking and so dumb. And that's the difference between that and the AI revolution is the order of magnitudes, the orders of magnitude that the AI revolution can do economically. And then enabling the robotics revolution on top of that. I mean it's, you know, the world GDP can go up fivefold in the next 10 years and instead of $100 trillion global GDP, we're half a quadrillion GDP.

Just to throw a little cold water on this, just to kind of look at it from a different angle for a second. Unless you don't want me to throw any cold water. Nope, is there's throwing cold water.

John?

Yeah, there's some talk going on about.

The.

You know, how intelligent are agents really and is reinforcement learning actually resulting in more intelligence? And when you see the chain of thought and it says it's doing this and that is it actually. And what's that, that kind of error rate level. Have you seen some of this, Bryce?

Yeah, Apple put out a big research report which really spurred all of this discussion, I think.

About it.

John.

Who's that?

Gary Marcus or whatever guy. I mean he's probably like the, the lead font of all of this about all that AI versus AGI and we're never going to get to AGI because AI is stupid to this day and da, da, da, da. But Bryce had this epiphany about it that I've fully adopted and made him write about a lot. And I curious your thoughts on this if Bryce I'm gonna do my version of it and then correct any of it that I up Basically intelligence is on a spectrum. You got mollusks and then, you know, dumb mammals and then smarter mammals and then you get human being babies and then you end up with middle aged really brilliant people who have been training and are brilliant to begin with. And then, you know, you peak out somewhere on that spectrum and what difference does it make if AI ever gets beyond the right now conceivable spectrum or if it goes on beyond at all? Like it's Already incredibly smart. And it already is doing stuff that a 12 year old can't do. It's already doing stuff that a computer science scientists can't do. And whether it's perfect every time or whether it can actually be self aware completely and not just problematic, like, like, not just like predictive thinking stuff like it, it's already incredible. We're living in a moment where it's changing the world.

Yeah, well, yeah, no, you got all that right. And just one thing I'd add. I was watching an interview of Mark Benioff and they asked him about basically this question and he says, look, our AI as it stands right now is about 93 accurate. And the interviewer is like, wow, that's like not that great, right? He's like, it's pretty great. Because for our customers he's like, Disney, for example, right? He's like, if 93 of the time we can do for Disney what they need us to do, there's a lot of freaking value that we can deliver to Disney even If we're wrong 7% of the time. And the other point I make in that article, and I, I made it again last week with Tesla, is that like we're com. We've set the bar at like perfection when the real bar is not perfection. Because humans are very fallible themselves. You know, whether they're driving a car or you're a sales agent at Disney or a software developer at Disney, you might have bugs in your code, you might, you might be hungover and write bad code. Like, like humans screw up stuff a lot. And so it's like we think that the alternative is perfection. That. And that's just, that's just not the case. And so even if AI is not perfect, it can still deliver a lot of value as compared to the human.

John, hit us.

Okay, I say good. And it depends on, on what you want to be doing with it or how you want to take advantage of it. And the fact that what's there already is something that you can go take advantage of means you're right. That's all that matters. It's those people that are banking on it doing more than it currently does who need to pay attention to the main issue here, which I'm not sure how much Apple covered it, but I saw it in more recent stuff, which is the fact that the typical language model is very good at the knowledge side of things. Like you say, do you understand this and that about how blah, blah, blah. And it's like, yes, I do say, okay, then do it and it goes and does it and it's wrong most of the time on a lot of the things that. That it could go say that it, you know, it knew about. You know, you understand how to write this and make it rhyme at the end like that and so forth. It's like, well, then do it. And it goes to do it and it doesn't get it right. It doesn't even know it didn't get it right necessarily until it goes and tests itself or something. So, yes, there's a ton you can already do with AI, and of that, probably 99% of it is untapped to this point. So much that it can be tapped. Yeah.

And I just, you know, it's like everybody wants to be the smartest guy about all this. And they're like, see, I told you it wasn't going to get to AGI or ASI or. And in the grand scheme of things, we're seeing the world revolutionize in front of our eyes. We're seeing robots about to hit the road. We're seeing robots about to work in factories. We're seeing coders do ten times more stuff. We're seeing Bryce figure out that how to make chords and lyrics on an app so that we can use it when we want to play music. It is happening, and it is happening at an accelerated.