The AI Revolution Call: Space, the next frontier!

We, we tasked Craig with making a better intro video and I said every week, each week we would like to have a new intro, but that might have maxed out, man. Beautiful.

I'm glad you like it. And to go through a few different tools to get what I wanted. But even still, there's some things that I could tweak about it. But I think it's okay for now. I'll still work on it.

Pretty great. I gotta give you a song that Moses— my—

yeah, please. That would have been perfect for this.

Yeah.

Yeah, I've got a song where he says something like, uh, Cody Willard, New Mexico Lobo, and it would be funny. We gotta— we'll work that in. That's good stuff.

Um, what tools did you make? Twitter is down. Oh, like, across— for across the board, not just us.

Mine says X is on, YouTube is on, and Facebook is on.

Like the site, the x.com is literally down. Like, try to go to x.com. Oh, okay. Not, not that we're not connected to it.

I see.

Well, they don't have enough servers at x/xai, even though they have a gigawatt compute system, one of the large, probably if not the largest single compute cluster in the world.

Or they got hacked by the Iranians. I don't know. Okay, all right, Craig, what— it'll be replayed on Twitter anyway. We won't peak out at hundreds of viewers today, I suppose, on the live feed, but that's okay. Craig, what did you use to make that intro video?

So I used a Grok Imagine API. It renders the most normal looking videos like that. You know how whenever you come down, it looks like you're landing, bracing yourself on your knees and and Jordan's falling. It doesn't look like it's clipping everywhere. I did— I tried Veo, I just couldn't get it to work right. And I tried, um, uh, the animation thing on Replit, and it wasn't— it was very static. It wasn't like big cool video. So I used that, and then I used the music on Google. The Google Music Creator is actually pretty good, like, you know, for just making stock music stuff.

It's—

you can do all kinds of stuff. That was like my fifth or sixth. Did you—

did you prompt it to have me dunking on a robot Michael Jordan? Or did it make that up? And did you prompt it to be on the moon or whatever it was?

So yeah, so as soon as I just like, all right, Cody wants to dunk on Michael Jordan, how can I make this even cooler?

I was gonna tell you that though. You just know I like that picture I made.

Yeah, I guess you're right. And I was like, I'm gonna make that cooler. So I said, let's put him in zero gravity because we always talk about zero gravity basketball on the moon. And then if Jordan was a robot, you know, because that's, that's pretty sweet. And then you have to have the suit on. And when you watch the outro, you'll see you have a pocket square.

I saw that.

That was nice.

Yeah.

Yeah. What were you gonna say, John?

The moon doesn't have zero gravity.

I was gonna say, it's, it's the space station where they'll be playing zero gravity basketball. But I would also like to watch 1/6th gravity basketball on the moon.

That's good, 1/6th gravity.

There you go, I'll take the 1/6th gravity, yeah.

I think the moon basketball will be better entertainment.

You know, it's a good— it's an interesting thing. You probably want a little bit of gravity, just not no gravity.

Gravity is sort of fundamental to the concept of basketball.

I will have to ask Gemini to help solve this, unless you can, John, or one of the other guys. If I had a 40-inch vertical jump in college, which I did, did— what would that be at 1/6 gravity? Would it be like a 24-foot vertical jump?

Um, yeah, well, what's, what's— let me see, is it just times, uh, divided by 1/6? Well, divided by 1/6, like to make it go—

take the I don't know.

I know somebody that can answer this question.

I think so.

Anyway, we're—

that's a good segue to the fact that Elon over the weekend talked about the fact they're going to build SpaceX, Tesla, um, I don't know if you called it a joint venture, but partnership is going to build a terafab, which is fab, before that they're the only people on the planet who have the guts to take on TSMC and Intel and And, you know, other companies have been building fabs like Micron for decades, but they're not the most cutting edge. Um, and no one else has, you know, Google, Meta, Amazon, the rest of them all design their own chips now, but they send those designs to Taiwan Semiconductor and maybe someday Intel to build. And we've talked before, I think I told the guys, I wish I'd mentioned it on air because for a year ago I thought I said that, uh, you know, with Elon's obsession with vertical integration, maybe one day they'll build their own fab. And so they did announce they're going to do that. And as part of the introduction though, before Elon walked on stage, they showed an AI-generated video of a lunar launch, like, you know, great, not with no rockets. What do you call it? It's not a screenshot, but it's basically the premise. It launches it into orbit, deep space orbit from the moon, just using a launcher, not a rocket, because there's 1/6 gravity and you don't— and there's no atmosphere. And so you don't need, number one, as much thrust, and number two, you don't have to go through the atmosphere to get out.

To go—

this is when you're going from the moon to Mars?

Or the moon to the launch. Really, it's about launching, um, satellites, I think, not humans and ships.

Yeah, I have an answer for you on your vertical question.

All right, what is it?

It's actually more than 6x according to Gemini because you get the 6x from gravity, but you're— it's saying that, um, your, like, force applied does more work on the moon because, like, your body— like, it takes your legs' work to push your body to upright position. And so even that part's easier. So it's saying you'll actually be able to apply more force to the surface.

So I could have a 24-foot vertical jump.

20— it estimates 28.3 feet.

You got to worry about these days.

I'm afraid my vertical jump's probably more like 20. And so it's still like 12 feet jumping.

So, so where you'd have to put the basket at like 25 feet in the air.

I was wondering about that. Would you want to play 1/6 gravity basketball on a normal height hoop? Because that would actually be sort of exciting. But you guys are all saying no, you'd want it to be like a 20 or 30 foot.

Yeah, yeah, gotta be up there.

Have to be.

It doesn't matter. Colliding. Imagine nobody would ever walk. Oh man, the NCAA, it's as bad as the NBA this season. I hadn't watched it, but every video, every game, like the whole time I'm like, that's a walk, that's a walk. I feel like a grumpy old man. But there would be no walking on the moon. You could like, like one dribble across the court. So yeah, maybe it'd have to be like a 400-yard a quarter-mile court and 30-foot rim height.

Now we're playing soccer.

No, because it's still a little hoop and you're gonna—

no, you got to change the hoop too.

I just thought of something cool. What about whenever wearables become, you know, ubiquitous, like NBA players putting on like the pendant or whatever they're gonna have like while they're playing, and you can see them like from their POV, like when they shake somebody on the court and hit a 3 or something? I feel like in the short term, that's cool.

Craig, I don't even know if you'll need them wearing a wearable. The AI itself could re-render that from their perspective probably more easily than putting wearables on.

You get any kind of perspective. So I just saw this video about how much data is collected for all sports, Formula One, baseball, whatever, from every which angle, and how little is seen as we just go watch it on TV, and how all this innovation that's trying to happen to go give people new ways to, to kind of get at what's going on. Like even seeing, like seeing F1 in a tabletop in front of you with all the little AFX cars.

Yeah, that's super cars in the world.

Um, lots of cool stuff, which is the whole point.

I mean, you know, another thing that I think we should just hit on is this endless AI is going to kill the economy, it's going to kill all the jobs, it's All of this nonsense, and I say that without knowing it yet, I guess. I mean, it is possible that all that's, you know, like I tell everyone, like we can all come up with doomsday scenarios about war, but we don't walk around like thinking that nuclear war is the baseline. Like, yes, we can come up with scenarios where AI kills the economy, but I'm not sure that's likely, much less even possible, really. There's so much to be done with. I told Craig today to work on, uh, finding a list of all the industries that have years-long backlogs, because whether you're talking plumbing, flooring, construction, shipbuilding, plane building, CPU chips, GPU chips, everything is backlogged.

That doesn't even account for places where there isn't a backlog, only because people won't even try.

Yeah, they're like, why me? Like, yes, like that, that's the point, John. Thank you for saying that. That is the point, that we all want more. We are Jim the golden retriever dog who wants more pets no matter what, all the time. We all want more. We want more entertainment. We want more fun. We want more relaxation. We want more cool. We want more clothes. We want more travel. More of something.

Tokens. Manis created that website. I could share it with you real quick right now.

The one with backlogs? You already did it? Mm-hmm. I fucking love this thing. Like, sorry Craig, gotta lay you off because Manis did the thing I asked you to do.

Let's see.

Did you do it just now?

Making that joke? It's a rude joke. I feel bad whenever we say it. I hope everybody is fully on board.

He told me like 13 minutes before we had to come on, and I just crafted a prompt and it did it in 5 minutes. So the total backlog for these industries, the top 7 industries, just the top 7 industries.

$3 trillion.

Yeah, uh, 3.6 years. And there's 28 companies in this that's being tracked. And this is what it looks like. The majority is aerospace, then AI infrastructure, defense. Um, and we go down here. So this is what it looks like. There's— and they're— and I mean, we're not talking just 2 years. Like, this would be like 2 years here.

This is—

again, average is over 3 years.

What's so important about John's point is that it underscores the fact that this is what the backlog is for companies that are willing to get in a backlog of 2 or 3 years. Like, how many data centers are not even trying to buy GE Vernova transformers right now because they're sold out through 2028? So you're like, well, I'm not even going to try to buy that thing. And like, if there were more GE Vernovas built right now, the backlog might grow, not shrink.

Yeah, yeah, they're building another factory in North Carolina, um, a twin factory to the one that they already have there that sells Walkerswalt transformers. And they're— that's going to cut the backlog down. But I mean, you're still—

no, it's not.

That's my point, Craig. It will not cut the backlog down. It will allow them to get more units out the door, but the backlog is going to grow because people are thinking there be some more supply coming.

Okay. Yeah, yeah, that's, that's what I meant to say. Yeah, they're, they're going to get more orders out, but, you know, there's going to be more to be done.

Cody, you hear them talking all over about, uh, Jevons' paradox. That kind of comes up here.

Tell them what it is.

Well, I don't know. No, it's like, so the more, uh, the more supply we get, it actually results in more demand, which it seems to not fit the whole supply-demand curve thing when it does that. But that's because— what's that?

It's because the cost goes down. Yeah.

So costs go down and you get more demand showing up. Well, that is on the curve then, you know.

Yeah, but, but it, but it is like it's a paradox because you would think if there were so much supply, then the total of revenue wouldn't go up. All of this is the same point, really, right? I mean, this is what I'm saying. About like the GE Vernova stuff. It's like the more that they could make, the more that people will buy, and the backlog actually might exponentially grow if they exponentially grew the supply. Bryce, you were going to say something funny. I can see the smirk from here.

No, I just, I, I was thinking about the moon basketball stuff. I thought it was funny. I was, I was making jokes in my head.

Probably inappropriate jokes.

Is that— no, no, no, no, no.

Come on, well then tell us one.

Don't dump it.

I was just thinking like, like, uh, could like little kids or something like, like trying to play basketball on the moon, or like, might have a 2-year-old niece, like a toddler. I just think that would be funny, like picturing them like flying around like trying to accomplish something like in, in zero or 1/6 gravity. I think that would be really funny, you know.

Like, you know, that's funny.

You ever watch like City League basketball? It's pretty hilarious as is, right?

Yeah, but then Little League basketball— you mean City League is also funny, but Little League's horrible.

Yeah, but can you imagine with 1/6 gravity? Like, that's a hilarious image, you know, Bryce.

But on a serious note, the same thing, you know, guys, I have a medically fragile daughter, um, whom I think each of you have met at one point. Um, and she's 10 years old now, doesn't walk, doesn't talk, and, um, we carry her everywhere. And it would be sort of amazing. I've pictured this, like, I truly— like, how cool would it be to take her to the moon and let her— like, she could walk on the moon and it wouldn't hurt if she fell.

Yeah, good stuff.

And it'll be a normal thing in 100 years for someone who You know, even middle class will be like, yeah, I'm taking my daughter to the moon for some therapy for 2 weeks. And there'll be a Michael Jordan robot there to do therapy.

Make it a longer trip. It takes a week to get there.

One, going back to the Jevons paradox thing, John, I think like, um, one of the things we did this last week is for the first time we were able to actually get a large language model on a our own piece of hardware. So I used Manus to download— it's called Ollama. I don't know if you're familiar with Ollama. And we downloaded a couple different models to a Mac Mini we had laying around the office. And actually, we ran inference on it. And I'm like, essentially, that inference is already free. Like, the Mac Mini is a sunk cost. It was sitting there collecting dust. And it's the cost of electricity and I guess our time screwing with it, you know what I mean? Like, so, but then now, like, the models, like we put Gwyn 2.5 or Gwyn 3 on there or something like that, an Alibaba model. And it's better than ChatGPT was 2 years ago, right? And now we have, you know, what was formerly a frontier model running locally on the Mac for free. And what that opens up for us in terms of use cases is pretty extraordinary, right? Like, because we're using Manus credits all the time, and, um, like, you're like, there's a lot of things where Manus could probably, if it was smart enough to like dynamically allocate compute, which I think AI will pretty soon, it could be allocate— like, it's like, oh, this is a fairly simple task, I'm going to allocate it to our you know, our slave GPU in the Mac or whatever term you want to use for it and run, run the inference there instead of burning Manus credits, you know. So yeah, I think there's a ton—

there's so many more use cases there now that weren't formally available, which underscores yet another incredibly important revolutionary technology economic thought that when things are expensive and Manus credits are expensive— I burned through my first $40, $20 allotment, then $40 allotment, then $60 allotment. Now I'm on a $100 allotment heading towards whatever the max I suppose is, several hundred dollars a month, I guess.

And the highest tier on Manus is $20,000 a month.

All right, well, I'm not going there, but they're, they're going like, this stuff is going like The— there's a relentless march towards making it cheaper and finding alternatives. The customers like us of MANUS are desperate to find alternatives. MANUS is desperate to bring those costs down so that we'll use it more and the credits go farther, what the dollar goes farther. Maybe you buy more credits, um, or the credits do more, whatever their solution is. It's, you know, like it has to happen, and that's You know, Elon talked about this weekend that they would like to create a trillion— uh, 1 terawatt of compute per year for AI using the Moon launcher eventually. And the entire U.S. energy capacity, generation capacity, is half a terawatt right now. So that would be building And this is what I talk about all the time, the acceleration of this stuff, that it took 200 years to build a half a terawatt energy grid. And it'll take 10 years to build, you know, dozens of terawatts worth of compute.

Yeah.

We're not going to kill the grid. Backlog notwithstanding.

I think it's so interesting because this idea sounds crazy, like space-based data centers and moon launchers and things like that. But one of my favorite things Elon said in the last 3 months is somebody was talking to him there. They're like, I can't remember which podcast he was on, and they're like, isn't it like, this is sort of ridiculous? Like, why do you need to put, you know, we— you could just build solar panels. Solar arrays on Earth and all stuff. And he's like, go try to build a gigawatt capacity data center on Earth. Like, permits, easements, power connections, supply chain— like, all these things are like really, really hard. And he's like, if the cost of launch is basically de minimis, which it will be with Starship, um, there's none of that in space. Like, you design the materials and they have to be hardened for radiation and things like that. Like, The material design is the hard part, but actually executing and like is actually less onerous somehow. It's easier to build the data center in space than it is to build it in New Mexico.

And it's all about sunk costs, right? I mean, in some— to be able to get the launch costs down, when you say de minimis, de minimis, let's put it in perspective. If a data center gigawatt data center costs, what, $50 billion, right?

That's the rule of thumb.

So $50 billion for a major data center, and, you know, probably, what, 30 or 40% of that is construction, right? And 60% of it, maybe a little more than that, is chips and technology. And so you're talking 10 $15 billion of construction costs. If you can get one megawatt of a data center in orbit for, you know, some order of magnitude or several orders of magnitude less than $15 billion, it adds up pretty quickly. So 1,000 megawatt would get you to the 1 gigawatt. So you'd have to have 1,000 of those data centers in space. Do we have any idea what the average AI data center from XAI SpaceX is probably going to have as a matter of compute?

It's in the— measured in wattage, is it now?

Yeah, I think that he said that the first, um, satellites are going to be like 100 kilowatts each or something like that. I think maybe you would need 1,000 of those to get—

no, 100 kilowatts you said? You would need 10 of those to get to a gig, and then you need 1,000 of those to get to a terawatt.

And that helps.

A thousand.

Let's put it in perspective. Do anybody like to guess how many satellites Starlink already has in orbit that they've been continually putting up? And they'll just start— data center compute instead of just non- need a center compute.

8,000?

Yeah, somewhere between 8 and 10. So yeah, I mean, it's not inconceivable that they could put thousands. So there's a kilowatt, and again, that's their first one. So the first one's 100, version 3 will be 500. You get 1,000 of those up there, that's 5 gigawatts.

There's—

so, or more.

There are still a lot of, uh, firsts that have to happen. With this, things that have never been done before. The amount of data getting beamed back down, the radiating of the heat, which he says they can do, but it's still way more than what they have done so far. You know, this just getting it up and connected and having, you know, the bad chips drop out of the—

I don't know, John, I push back on all that. I mean, he's— what they've done with the Starlink stuff is proven out that they can get the compute and the chips to work just fine. They can get the data beaming back.

How many years?

Low latency and better economics than anyone else.

Yeah, but how many years did it take?

5?

Starlink started 5 years ago?

Probably not even. I remember when Elon tweeted, hey, I'm sending this through a satellite, and then he was like The next tweet was, holy crap, it worked, or something like that.

2019.

Yeah, so 6 years.

Yeah, I just had AI do the math. So each one of those satellites, 100 kilowatts, will weigh about 1 ton, 1 metric ton. And, uh, Starlink can carry 100 metric tons of mass to— or a Starship— Starship can carry 100 metric tons of mass to orbit. So you could potentially have 100, um, AI satellites, which would be, uh, a gig— would that be a gigawatt of— or is that 1,000 kilowatts? Yeah, that's a— yeah, on every single Starship launch. That seems like a lot.

Well, whatever the numbers are, I guess, John, like, that's the point too that we got to remember is in 2019 they launched a handful, in 2020 it was a dozen. When I first signed up for Starship, it had a few hundred, maybe breaking 1,000, and now it's got 8,000. The, the cadence, the acceleration of the cadence, the acceleration of the, the capacity of each satellite means that it wasn't a straight line of growth. Hockey stick. There's 10 or 50x more bandwidth and satellites up today than there was several years ago, which was only 10x of what it was before that.

It's 10— for the record, it's 10 megawatts of compute in orbit per Starship launch.

Yeah, you skipped meg, went straight to gig.

Yeah, that makes sense.

Yeah, you still mean interesting stuff. I will say I'm— I might be the only Revolution Investor tech space guy who is at all cynical, skeptical, worried about the SpaceX IPO. I think it'll happen, but I sure feel like everybody's counting their chickens and nothing's hatched.

You're much more confident about the technology side of things than about the IPO.

That's an interesting point. I, I'll finish the thought by saying that— I'll finish my thought by saying that part of it is I hope that Elon brings that thing public at at least $1.5 trillion or even more, and that he gets every dollar he possibly can out of the IPO to put in SpaceX's coffers to fight the wars that it's— of capitalists in space that it's going to have to fight in years ahead. Um, John, not necessarily. I, I fully expect years from now SpaceX will be publicly traded and worth trillions, if not more than $10 trillion. But I am less confident about the timing of it all I know we hope it's this summer, but that does mean we need to have a pretty strong, good market. We need an economy that's not crashing from a war. We need, you know, and even the war in some ways is— can be stimulative for the economy, number 2. And number 1, it could be like good for space defense, national defense-related valuations. And SpaceX is, you know, SpaceX is going to be the biggest, like, national defense contractor on the planet over the next 5 to 10 years.

Do you think they're still on the path to that vision that I heard at some point about, like, lots of, um, spaceports around the world and, uh, the, the rockets sort of delivering anything— troops, equipment— like, just around place to place very quickly?

Yeah, I mean, I think it's— we're going to need to extend the timelines out a lot because again, the permitting of all of that stuff, licensing, getting the locations built and all of that stuff is going to take many years. And it's many years away before there's enough Starships built that they would be— there's an— that there'd be enough excess capacity that it would be used to ship humans or goods from New York City to Beijing. But I would expect in my lifetime— I'm 54, so over the next 20 to 40 years, I would expect that I will at some point board a Starship at a spaceport and end up in China in 30 minutes. So there's still, um, spend more time in security lines than on the Starship.

There's still talk of, uh, doing the data centers in space will require like 2 launches per day or something, like every day.

Yeah, we're already doing one every other day with the Falcon. I mean, I, again, I think it's, you know, it, it could take years. Bryce, what is it that you so smartly said to me the other day that I keep quoting? What we invest in is inevitable. Yeah, the question is always a matter of timing.

Yeah, the biggest risk to our strategy is timing.

So the biggest risk to Starship flights being taking humans from New York to Beijing in 30 minutes is just time. I don't think it's any question that eventually we will be doing that.

Yeah, one of the things Elon says a lot too is that If they can make Starship fully reusable and reliable to the extent, like something close to an airliner, like Airbus A320 or whatever they are, hitting the launch cadence required to make costs that low to where humans can afford to fly from here to Dubai or whatever, you don't need that many flights compared— I just did the math. So the airline industry flies like 30 to 40 million flights per year. And for him to get a terawatt of compute into space would take 33,000 Starship flights per year. And 33,000 Starship flights sounds like a lot until you put it in the context of the airline industry. And Elon's like, hey, look, if we can make this thing fully reusable, basically all we have to do is refuel it. It lands, we refuel it, inspect it, make sure it's still safe, send it up again, right? You get the turnaround time to like 30, 40 minutes. And you're— it's actually not that remarkable of a number.

So 33,000 is 100 a day. And we have like thousands of planes in the air, probably 10,000 planes in the air per day.

Right. And I mean, it's like he's comparing himself to the entire airline industry, which is huge.

It's not just volume, because there's just the cost of fuel. I mean, how much fuel does it take to move? You can only fit 150 people on a, on a, um, Starship.

Yeah, but you're also expending that fuel for a matter of minutes, not hours. You fly from New York to Beijing, you got a 14-hour flight burning fuel, 14 hours consistently. I don't know what the numbers are. I did actually have a debate, I had a discussion with Gemini about all of this several months ago and trying to get some perspective of it. And I mean, this is where just sort of like you got to sort of bet on brilliance and believe that those— that SpaceX is going to deliver at least a large part of the vision. And I do. I mean, Elon has consistently done what no one else has— can possibly do, and SpaceX and Tesla are his tools to do it, his primary tools to do it.

What's the name of the SpaceX XAI combo?

SpaceX X, X, X, A, I, X, X to the 3.

Yeah, okay, X to the X, X cubed. That would be a good one.

No, isn't that his son's name?

X cubed is a—

I was just making a joke.

Close though. Yeah.

What did you say after you said all those random letters and stuff? I, I said, isn't that his son's name?

Oh yes.

So John, on your question, the raw fuel cost would be about 10 times higher for the Starship versus, um, you know, Boeing 777 or something. But the Starship can haul a lot more mass, so that cost per ton is actually about a fifth of the cost per ton of a You're telling me—

get out of here—

a fifth. So you're telling me they're gonna pack sardines onto that sucker and move like—

well, just even just say cargo, for example.

You said 100 tons, right, that it could carry? The space shuttle could carry 30 tons. So, um, so 100 tons—

what is a Boeing 8— or what you said the Airbus A320 carry, Bryce?

I don't know.

A ton is 10 people. So 100 tons is 1,000 people if you packed nothing but people in there.

It's, it's 12 Europeans, it's 7 Americans, correct?

Yes, the, the high-calorie, uh, yeah.

Bryce, what do you got?

Um, this is saying somewhere from 50 to 100 tons, which doesn't make sense based on its earlier math. Anyway, the largest passenger version carries about 67 tons, so I don't know, maybe the earlier math was wrong.

AI, never gonna work. It's complete failure. You can tell right now. It'll never get any better. It's all full of lies. John, Mark Zuckerberg today, at least the news was leaked, rumors that he's building his own super personal AI. And the anchors on CNBC, the reporter who reported it on CNBC tossed it back to the anchor by saying, wouldn't it be really cool if each of us had our own AI? And I was like, and the anchor was like, you know what, maybe someday we will. Uh, yes, yes, the answer is yes. That's the whole point.

What the hell?

When this came out, everybody was like, maybe someday we'll all have a smartphone and not use these flip phones. And before that, someone was like, Maybe someday we'll all have a cell phone and not just have to like go home and call our answering machine. Go back to the '80s. Yes. Like, yes, you're all going to have an AI agent. You'll have a lot of AI agents. Most of it's going to be like sort of like talking to Alexa and Siri. It's going to be the same stuff done over and over for most people. That's what the major agents will do. Google and Apple have all your data, all your emails, all your apps, everything right there. They're going to deliver the lowest-hanging fruit, most simple to use, very incredibly helpful and productive AI app agent things to each of us. And it will be unheard of. It will be as unheard of for people not to be using AI agents personally on a daily basis, hourly basis, as it is to think that these people do not use this on a daily or hourly basis.

I gotta raise my hand. So, uh, you know the famous example that is the guy from, uh, the IBM executive or president of IBM or something way back who said, uh, why would everybody need their own personal computer?

Yeah, literally, it was even worse. He said, no, nobody will ever need a personal computer at home.

Yes, that's right.

How many personal computers are at your parents' house, Bryce?

Mm, 3 or 4.

Yeah, I've got probably 3 or 4 at my house.

All right, your phone counts as one, so—

well, but no, I was actually going to say, and you probably have at least that many smartphones because—

and your iPad and your phone and your— these, these are computers. So, um, I don't understand what exactly he's going to be building with his, you know, personal AI because that's like kind of what everyone's already working on doing the—

I mean, well, I would think it's about tailoring it to the things that he needs done. Like we were talking about earlier, he'll— the, the root— the discussion is that it's like right now a lot of the information and data that— and charts and flow that he looks at, he has to go through layers of meetings and layers of bureaucracy to get all of that to him. And to have an AI that's constantly filtering that and presenting it to him in a very helpful manner would do that.

But he doesn't need to be Mark Zuckerberg and spend billions of dollars and kind of get his own pre-trained.

Nobody said he's spending billions. They just said he's making his own personal agent.

So he is, so is everybody else. I mean, it's like I'm saying it doesn't happen in the pre-trained model, It happens in the harness and it's what everybody's going and working on now, you know?

Yeah, actually, John, I'm glad you— you're right. Like, I mean, literally, like, everything we've talked about on these calls since Claudebot moment has been, hey, we're, we're going to make our own agents. We've been working on our own agents. We do— I made an agent do this for me. Bryce has got his Cauchy NCAA betting agent. Throwing money out the window. And we've got all kinds of other agents. Hey, Bryce, how's Texas Tech doing on your brackets beyond the Calci thing? Right now, Craig, how about KU? Rock Chalk KU!

They are gone.

They went home.

Yeah, Houston still in it.

I think you're in last place on the bracket though.

No, no, I'm now in third place, but you are temporarily still in first place, Craig. But your total points is double what my current point total is.

I don't know how I did that.

It's not going to work out well for you guys. That's my point. But the point actually is that, like, yes, John, you're so right. Like, all of us are working on these agents things. And it was really sort of silly for CNBC, number one, to even present this as news. You don't think Juan's got his own personal agent being created and worked on and everything else. And then second of all, my point is that the reporters could replay that clip in like 10 years and it'll look like that clip from the Today Show when Bryant Gumbel turns to Katie Couric and he's like, what is this internet webby thing? You apparently you type in w dot something and yeah, it's, it'll be as common as this. And by the way, I call it years instead of 10 to 15 years like it took this to go to a billion people. What, Johnny?

I call Michigan.

Michigan? Yeah, that's a good one.

Claude chose it for me. I had him do the whole bracket kind of, and I— and it concluded things like, uh, where did injuries happen, what are the most likely upsets, what are the whatever.

Yeah, my wife did the same and she's also been demolished.

So we'll see.

My, uh, my calcibot just bets for me. I don't even look at it.

And Bryce, what percentage-wise, how, how much is it down?

10%.

How much is your weather betting app up? And I know that one's fake money, but still, percentage-wise, how's it done?

Yeah, um, I haven't looked at it in a while.

He doesn't care about the fake money.

Yeah, yeah, I care about the real cash one because it has real money. So, so yeah, I made, I made another trading bot that trades the weather prediction markets and another one that trades Bitcoin 15-minute futures. And the Bitcoin bot was really bad in the early days, but it got better. And then the weather trading one was like a 75% hit rate. With like, it was hitting on like 4 and 5 to 1 odds. It had amazing ROI, but not nearly as frequent as the Bitcoin.

You're killing the climate.

Go ahead, John. Sorry.

I say you're killing the climate with all these funny things to use it for.

Actually, you should always bet the climate gets hotter because we're all going to use Calsci to bet on the climate. Bryce, did the Bitcoin Uh, bought ever short Bitcoin or only bought it when it thought it was going to go up? Because the two times I saw it, it did.

Yeah, it would— like, it's not shorting on KALSHI, but you're saying it's not going to hit this high level. Oh, right. Okay, let's go on the no.

Yeah, let's, uh, let's get the bet here on, on who gets the closest with just who's their final four. That way it's not just total fluke there. And then, you know, I'll put it for, for the record, uh, my Final Four: Michigan, Florida, Arizona, and Michigan State. So just whoever you got in your— it's now recorded, it's down for history.

I think I'm just gonna pull up my bracket.

My bracket's fully intact. Houston, I think, is my Final Four.

So mine's already— one of them's already gone, but I had Kansas against Houston and Arizona against Michigan.

I've got St. John's versus Houston and Arizona versus Michigan.

Wow.

So St. John's won. Nobody— I don't know anybody else that picked St. John's for the Final Four.

We've all got Arizona, Michigan.

You know, baby, come on.

No, I think I have Houston in there. Who does Houston— this season, who does— I don't think I have Michigan in my Final Four.

Anyways, well, this has been, uh, really good, uh, AI talk. And I saw Cody Willard.

Let's talk, um, about John von Neumann.

Yeah, sorry, just figured we should have a basketball on site as we're discussing hoops.

Catch, John. I just wanted to ask you about him, John, since you're an expert on the guy.

I'm not an expert, but I'm a big fan, right?

I think you read his biography, so you're more of an expert than anybody else.

You know, everybody talks about Einstein as being, oh, he's the most brilliant. I think maybe von Neumann was the most brilliant in the 20th century. Yeah, he did game theory. He invented the von Neumann architecture that we still use that separates memory from instructions in the CPU, or puts them in, uh, coming in through the CPU, goes and gets them and pulls them in there. Uh, it's, um, the way that whole— that whole thing is set up. He was a mathematician like ridiculous. And the story I remember from his, uh, biography— I've got his biography, I never read all the way through, but I've read from it. Uh, so he, it was— he worked on the bomb. Uh, he, um, uh, somebody— he was in amongst all these total geniuses, and they go and ask him, uh, you know, about this one, uh, mathematical equation. And they say, we just want to know if it's theoretically possible to have a result for this such-and-such, uh, um, you know, math problem here. And he goes off into a corner, talks to himself for a while, mumbles around, and he goes, yes, it's, it's possible, and the answer is—

and he gives The AI described him as a human CPU, like the, the computing power of a real computer in this guy's brain.

Yeah, kind of.

John, why have— so I don't know that I would have heard of him except I've heard you talk about him for years.

Well, the reason I talk about him at the time was he was, um, I think Hungarian, and, uh, there were 3 or 3 other people like in his high school who got the Nobel Prize. So now if anyone thought that, um, you know, the brilliance and everything was like genetic, here was the proof that it had to do with your environment and your schooling that, that could lead you to that. I was always blown away by that.

And that partly because your teacher, your parents were educators, teachers, administrators of education, and you probably had an interest in that.

Yeah, mine were. Yeah. And I did have a huge interest, still have an interest. And it's just that, and now, like, look with what's happening with AI, if only we took proper, you know, advantage of it, we ought to be able to have incredible advances in education now. And so then you have different problems like just all the mental problems we seem to have developed as a country. But, you know, if people could get back on the horse of just motivation for learning, you could learn like anything now. Karpathy, Andrej Karpathy from OpenAI talks about it a lot, you know, about just the potential for, for learning now.

I was, you know, I, I have, uh, I, I don't really watch college basketball anymore, but I do watch March Madness, or at least it's on mute in the appropriate season. So I've got the women's college regional second round over there going on, and I was just thinking about like The Vanderbilts and Brigham Youngs and states and towns and counties have spent centuries building these secondary education products, and it— I think we all know, like, the world's a different spot and I do wonder that traditional college education is peaked, secular decline in the way that we educate ourselves through universities from here.

Right.

Everybody's nodding.

Yeah. And, and, uh, you know, higher education debt is bigger than credit card debt. I just heard that one in, in, you know, Countrywide.

So yeah, I got on this today. It's related to what you are talking about, Cody, with education. I think the same thing in the legal industry. Um, we have these like gatekeepers, right? The universities— by law, you have to go to a law school to become a lawyer, to even be eligible to sit for the bar. That didn't used to be the case. Abraham Lincoln never went to law school. He just studied under an attorney and passed the bar. Um, I don't even know if they had a bar exam then. You could just be admitted to the bar if Literally, the bar was a group of people that would practice law and said, yeah, this guy is competent, therefore he can be a lawyer. And like the American Bar Association, law schools have spent the last 100 years lobbying states to keep their requirements and make laws like, like all of these barriers to entry they've created. And sort of the same in the education deal like this. You have to go to be credentialed and it's a status thing. And it's like, because they made the price of admission so high, they've opened the door for AI to just wreck their asses, you know, and that's what's happening. Like, the legal industry is being wrecked by AI. They don't know how to compete because they've spent the last 100 years being anti-competitive.

And, and Bryce, it's, it's first principles of our investing approach. It's the— it is the most economically and energy-efficient way of educating yourself will win out over time. And for a century or more, That has been university and classroom learning, and that is no longer the most economically and energy efficient. And because they have— because there's such a disjointing with the cost and the value that every student is doing the exact same thing that we're doing with compute, every Buddy, who's a potential student of university, is trying to figure out if there are alternatives. Maybe there's not yet, but over the next 5, 10, 20, 30 years, society is going to change and it will be very different how you educate yourself as part of that change.

Yeah. And why, why does a university, as many of them do, try to tell students you can't use AI in this class or working on things. It's like, it's because they're teaching the wrong things.

There you go. Yeah. It's said from someone who I've had dinners at John's house with his parents talking about, like they go, they went down a lot of like rabbit holes of theory about education. And so yeah, this burns near and dear to John's heart. I can tell you.

I forgot that you witnessed some of that. Yeah. You did, right?

Well, I think about the legal industry and it's like the bar exam doesn't measure your competence as an attorney. It measures, it sets some minimum floor that you can do this thing. And if you can do this thing, we probably think you're going to be a competent attorney, right? But it's, it's just a barometer. It's not an actual guarantee. And it's like the reason you can't use AI on the bar is because it defeats that little system they created, right? But like, your good attorneys going forward are going to be the ones that use AI the best, right?

Yeah.

You actually wanted to measure the competence of attorneys, like how well do you use AI to solve this problem?

Craig, write that down. That's a very important phrase. The best attorneys, the best hedge fund managers, the best coders, the best white-collar workers will be those— it's just such a simple thing to say— will be those that use AI the best. You use AI well, you're going to really succeed. If you don't, yes, sorry, there's going to be some disruption in your life. The amazing thing about AI though is how do you learn to use it well? You communicate clearly to it. You give it a good prompt in English. And if you can do that, you can do— you can use AI well. And it really is, if you could tell someone else to help you work on your job, well, that is the promise— not that it will be the delivery of AI.

Yeah. You gotta delegate well.

It's almost like you could You need to come up with like a new way of measuring competence because there is, there is a need to measure competence. Society values that, but you need like the systems we've created don't— are incompatible.

I'm glad you reminded me, Craig, write this down too. When I'm— when we finish this call, I'm headed to Las Cruces. I'll be working in our offices there for the next 2 days. So I'll be with Craig and Bryce in person. And one of the things I want us to do is play a— I found an AI prompt game and it scores you on how well you prompt the AI. Actually, I've seen 2 or 3 of them after I found it, but it sounds like fun. I think it would actually— and then I realized as I didn't play it yet, but I was like, the 3 of us should play that game and compete because it actually is a skill set that we should be refining. We should be practicing this skill set.

So I was— doing a thing that required, you know, doing a prompt. And, uh, because I'm thinking about how, like, the, the best prompts I write are the ones where I say, hey Claude, I'm going to tell you something that I want you to turn into a good prompt. So it's like I prompted to prompt, uh, but I went and I was working on this thing that involved, uh, multiple steps of turning a, uh, business plan into a software product case. So this— here's a business that's going to use software, and now the next step is to define the software itself in, in like a big document. And I go and I told it like all these steps I wanted to go through— look this up in here, check that about this, check the other. But what I did additionally was I said, while we're doing this, start to build it as a skill. And a skill is like another prompt, but it's one that gets saved and used over time to do that thing. Now going forward, when I need to turn— I'm making software for a company, when I need to turn a business plan into the software for the company, I've got skills that I'm making that it uses. The way it got those skills was by recording along the way what happened as we were talking.

It's amazing. We're living in the future.

Yeah, right.

Last words. Um, I'm headed to Cruces.

I got a 2-hour drive. On the same topic, uh, it's that last thing I was just talking about, just to say that, um, I'm— yeah, I'm personally experiencing the incredible, um, productivity boost that sometimes gets talked about, but then you actually see it happen. You guys see it when you're doing the Manus stuff. I've been seeing it when I use Claude and build these skills and so forth. And I've got like— I'm making for a small company an application that I never could have done on my own. It's quite detailed, all of what it includes, but I'm doing it with the help of AI and it's I just, I amazed myself. You know, it's like, like, did I just do that? You know, I did. Of course, the AI helped, but it was together with the AI. But it's like, the productivity is amazing.

Yeah.

I'll go next. So to Bryce's point about education, and just how it hasn't caught up, I was thinking, you know, when you do some of these interviews, and um, tests to apply to different, different, um, companies. They will give you a test, but they lock down your screen. You can't do anything. I think that's super counterintuitive. What they should do is allow you to have your AI chatbot open and give you problems to solve. Like coders, whenever you go and apply to a job as a coder, you have to do a coding test in front of people. So if I have a business degree, I need to show you that I can prompt and delegate and get a project done within a decent amount of time by the time I get there for my interview and by the time it's done. So you can ask me some questions, send me a lot, give me some time for a test, build out a prompt and everything like that, have the AI do it. And if I do it efficiently, then it shows you that I can handle the workload. It's kind of counterintuitive, but we do it now.

Get the hiring process is yet another thing that is ripe for disruption and completely revamp in a world of AI. Bryce.

I just had a thought today. I was doing something where I needed to extract some pages from a PDF. And we don't pay for Adobe Acrobat because it's stupid and expensive. And so I asked Manus, I was like, what would it take for you to build me in Adobe Acrobat clone, basically. And it's like, yeah, you know, it would take quite a bit of work. And it's actually more complicated than you realize, blah, blah, blah. And then I was like, well, can you just extract this page for me? And just do what I was going to do with Adobe? And it's like, yeah, sure. You know, 15 seconds, here's the extracted page from the PDF. And I just realized, and Craig and I have talked about this, part of this is from Craig, but it's like, there's never been this thing where like the marginal cost of of code is almost zero. And the, the Manus can just make this kind of ephemeral code that it uses once and can accomplish something. Because with Adobe, nobody buys Adobe because you want the Adobe thing, you buy it because of what it can do for you, right? And so if the AI can do all of the tasks that Adobe Acrobat did, you literally don't need it. And I just, I think, I don't know how to define that put the—

it transcends it, Bryce, right? It like the, the AI itself transcends the app, the—

right.

Yeah, I think it's, it's like the— it's, it's like a locksmith that has the key to every key in the world and it just poof, you know, I need to get through this door, I can just make this key and get through.

There still will be demand for like static software. I don't know, static isn't the right word, but like software that needs to go out and be used and be the same every time you log in. Permanent software, permanent-ish, but software also changes. Like, you know, if that's— I don't know, um, that there'll still be that demand. Like, Wall Street Journal is a piece of software that I— when I go to the site, I want to look at this and look at articles and things like that. Um, but for a lot of use cases that were formerly software, they're just tools that the AI can now do natively, and you don't need that sort of permanent software anymore.

My thought is this, you know, I talk about how the AI revolution is in its dial-up internet phase. And I try to picture like, trying to build an agent is sort of like, I don't know, like trying to figure out how to use email when you were being shown email the first time. And you're— back then there was like these steps you had to do to do that at home. You have to go to your computer, open this one app, and make it dial— make sure it's plugged into your phone and make it dial in order to get the connection turned on that's really slow, that maybe you can read an email. Then you open an email client and you read an email or you're on AOL and you're reading You've Got Mail. And but it was so like the steps to read an email, you had to be trained how to get to the point where you could even hit return. Like reply, here's my reply, send. And now, like your phone or your computer or like always connected. That step is gone. Even like opening an email client is sort of gone because you get a notification and it's pushed to you. I mean, like BlackBerry was revolutionary because it notified you when you got an email on a mobile device. And it's like, that's, that's the stage we're at. Like the things that we're doing with AI right now is still really cumbersome and hard. And people are intimidated by it. Like, we got an email from one of our subscribers and it says, Bryce, Craig, you have it in front of you. I sent it to you guys. There it is. He says, will you share your insights with me? What is the best way to just learn basic things we should be doing with AI? I do not expect to master it over overnight, but any wisdom you can share would be appreciated. Appreciated. There's so much probably garbage on the internet and Instagram and Twitter. Can you point me in a correct direction? I still work daily, we decorate corporate apparel, and I just turned 71 2 weeks ago. I so appreciate your wisdom and insight. Blah. But the— like, I forwarded that to Craig and Bryce because I'm like, it— that's like— that was someone emailed— it was a 71-year-old asking me in 1996, how do I access that email thing? It seems like I should be getting on that internet. Can you help me figure out what I should be doing on the internet?

But somebody asked about AI now.

You said AI is going to be talking to it and having it do shit for you, and everything else is just sort of like this stage. It's the dial-up internet phase.

Yes, John, are you saying somebody asked you that about AI recently?

That was an email I got this morning.

Okay, my answer would be go— there's this thing called ChatGPT, go talk to it.

Like, I mean, literally, like, that's the pro— that's a great point though, John. This email is the prompt.

Copy—

that should be my reply. I'll be nice about it, but literally, but copy this email text and put it in any Gemini, Google, even just go to google.com and type in that, and it will pop up an AI overview and answer this question for you better than I possibly can. The short answer is go use AI. That's why—

yes, or do that for him and, uh, like take a screenshot of having pasted it in and what the answer was and say, see how this did this? Now, now you do it. But one last thing is give, give them a link to the YouTube video Medieval Tech Support, because that's the one where the monk is teaching the person how to use a book instead of a scroll. It's like, oh, stuff disappeared! No, no, turn back, it's still there. It's brilliant.

That sounds great. Uh, Craig, will you find that video and send it to the three of us, uh, text on our chain, and we'll all get to watch it and write to each other, hello, lol and stuff. The other thing that I just remembered was— Bryce, you want to say something on that?

No, no.

The other thing I just remembered, like, I'm old enough to read— I've been on Twitter long enough, X long enough, to remember when I had to use a third-party app to upload an image that might show up in my feed. That's how bad even the first social networks that became $100 billion businesses were. You couldn't even upload a picture on Twitter, and you had 140 characters, obviously, and it was just characters. And my old Twitters, I bet, are hilarious because you would literally just try to get a point in about 140 characters, and usually it wouldn't really say much. And then you'd upload a picture and be low-res and crappy.

It was like haiku. You had to like—

that's not the one I used.

I don't know. No, I'm saying it was like writing haiku when you had to go write a tweet and you're like, how am I going to make this pithy but still get something across?

But, you know, yeah, mine were more like a limerick, but all right, everybody, that's the call. Sell everything. AI is dead. Space won't work. Robots are bullshit. I didn't mean to end the call on a cuss word. Thanks everybody, goodbye. Thanks, goodbye, gracias.

Closing video.

Yeah, it'll be there.

Okay, go ahead.

Oh, put it in chat, put it at the bottom before we log out. Post a comment with the link to the video. Craig, do you have it?

Oh, the medieval one?

Oh, you didn't do it.

That's all right, you said to put it in your, uh—

I know, but then I just thought if you had it you could put it in chat, but if you didn't like to get that Okay, all right, thanks everybody. Here's our outro video, which I'm gonna chuckle as it plays.

Mhm.