Will Manidis argues AI will create a 'sovereign frontier' of classified models and dollarize the world through agent transactions

Speaker 2: Coming on...

Speaker 1: Nowhere to be found

Speaker 2: He's right locked in. Is

Speaker 1: probably podcasting on one of his multiple podcasts. But we have...

Speaker 2: Will Manidis. Will Manidis. Let's bring

Speaker 1: him in.

Speaker 2: The author of the commanding heights, the hottest Here

Speaker 1: he is.

Speaker 2: To hit

Speaker 1: Sert it up.

Speaker 2: The internet in years. Congratulations. Omaha.

Speaker 1: It's great to have you.

Speaker 2: It's great to have you. I think we're looking at future future one of us. Future one of the

Speaker 1: Allocator.

Speaker 2: Future allocator potentially.

Speaker 4: In the Ultra Dome. Yeah.

Speaker 2: You are the only the only person that has fully co hosted this show from start to finish as a guest as a guest host who's not actually on the team. So thank you for your service.

Speaker 3: It is

Speaker 4: a tour de force. It's a terrifying thing to do.

Speaker 2: It is. But you pulled it off.

Speaker 1: You were you were sweating by the end. It's a lot. You kinda...

Speaker 3: No one warned me. No one warned me.

Speaker 1: No. No. It is as much as it doesn't look like work. It is. It's a real it's a real job. It's a real profession.

Speaker 2: Yeah. Only when the water goes out do you see who's swimming naked.

Speaker 1: That's right.

Speaker 2: That day Right. You did have some swim trunks on, so you you made it through.

Speaker 1: You survived.

Speaker 2: I will

Speaker 4: be applying for union benefits very shortly.

Speaker 2: Open AI. Yes. Yes. Yes. That's for sure.

Speaker 1: How's week been? Yeah. How's week been? Know this has been in the works for...

Speaker 2: Are you signing autographs yet? Are you getting recognized on Print the

Speaker 4: copies are en route. Mhmm. Commanding-heights.com request a physical copy. We have a thousand of them going out today.

Speaker 2: Cool.

Speaker 4: Leather bound? Leather bound, hardback, hand sewn, not America. You can't hire American folks anymore.

Speaker 1: Me too. Wow. There you go. Interesting. What... Talk about the process.

Speaker 2: Yeah.

Speaker 1: We've been hearing about it for for months now. Lighter's been...

Speaker 2: Well, it's been quiet.... Couple of things. Because for a while you were on a generational run of these essays that would go mega viral and it felt like chapters in a book and then you went quiet for a while. That was the book writing process?

Speaker 4: That was the book writing process. Joe and

Speaker 1: I have

Speaker 4: been convinced that AI is real. Right? Scaling laws are going to continually uphold us correct about almost everything. Mhmm. And we've been kind of shocked at how little discussion there's been about what this means for currencies and what those means for nations and what this means for populations. So about a year ago we said, what is kind of... What does the East Coast have to say about the the coming scaling laws that are going to happen? What is the finance take on everything that's going to happen? And we set out to write this manuscript. Joe Roberts and I, Farrell Gregory sat down, spent the last year drafting this thing and published this week. Very exciting.

Speaker 2: And what is the what is the TLDR on the knock on effects? Because everyone knows the bottleneck constraints like the the the situational awareness thesis has been digested. You're saying AI is real, but what are some of the knock on effects?

Speaker 3: Yeah. I mean, I think we need to

Speaker 4: move past bottlenecks and start to think about who is uniquely advantaged as a result of this technology. Permanently. It's staggering to me. Permanently. Yeah. Right? Like, it's staggering to me that you can use instinct every day at work and not process of like millions of people's jobs are within the capabilities cone of the models we have today. And if take the epoch AI scaling curve seriously, I think we should, these models are going to get so expensive that it gets to a point where there's only kind of one or two states that can continue to maintain a sovereign tier, meaning models that are kind of world class and continue to be trained. Yeah. So if we have a world where the best models are held inside of American skiffs, where the national security community is given Fable 5.5, right? Maybe some unreleased TBPN episodes, really kind of high security stuff. Yep. And the Chinese are able to maintain their own sovereign frontier in the East. Then what does this mean for the populations that are within the diffusion net? And we introduced a couple of new ideas as part of this PDF. The sovereign frontier is the big one. What I think is a really obvious idea that you will not soon be able to buy the best models, that labs will become something much closer to the defense industrial complex, that they will have models that are available to nation states for national security purposes. You will not be using them. Know, harkens back to the GPT-two release when Anthropic said, you know, it's too dangerous to release, but much more seriously. And then two, more importantly, as we think this will manufacture huge amounts of demand to transact on American payment rails. We introduced the concept of the neo dollar. Yeah. Which is this idea that if models are transacting neo dollar was going on. Yeah. If these models are transacting in the dollar, if all of agents are transacting in the dollar, if the thing that is backing US sovereignty is dollar denominated models, then the rest the world is going to dollarize quite quickly. And we introduced kind of two competing financial systems in

Speaker 3: the RMB and the dollar Yep.

Speaker 4: As two kind of different abilities to produce huge amounts of dollars to train the sovereign frontier.

Speaker 2: So how do you take that thesis into some of the more low risk mundane AI enabled technologies? I'm thinking of just like website builders. Like that's something like are you gonna have to go... Does Lovable have to go to the government for their inference or can they just use current models?

Speaker 1: These use cases are like pretty saturated.

Speaker 2: They are saturated. Yeah. But do they continue? Okay.

Speaker 4: Think like if you think like the the you know, the the the vast majority of workloads in The US Mhmm. Very soon will be kind of Chinese open source models. Open router makes this data very obvious.

Speaker 1: Mhmm.

Speaker 4: A bigger and bigger share of Frontier Lab revenues are just going to be from USG. It's the only possible way you can kind of finance these training runs with this So cost and the if you imagine kind of a a Chinese Open Frontier powering most US commercial operations, at the same time you imagine incredibly proprietary, incredibly close frontier intelligence powering kind of the the tip of the spear for US hegemony. I think you end up with a pretty stable market equilibrium. Like, I think it's all going to work. Right? Like, most discussions of this as a bubble, think, really miss the fact that this is the central economy of The US at this point. The entire empire is hinged on this. We got to make this work.

Speaker 2: So what are some of the other knock on effects? I mean, by dollars that be beyond that?

Speaker 4: I mean, dollar dollar devaluation seems really obvious, right? We will have to print huge amounts of dollars in order to fund those build out. Mhmm. One of our very strong convictions is that we will dollarize the entire world at the same time the dollar is inflated to basically zero.

Speaker 2: Okay.

Speaker 4: You probably want to own equities, right? A long equity is short to the dollar. Mhmm. And we've become particularly obsessed with the class of businesses we call commanding heights, is Yep. You know, where the title of the piece comes from, which are businesses that get better as intelligent scale.

Speaker 2: So you can

Speaker 4: think of it as kind of a mixed sleep shared economics or share of economics shared to the customer. But in the context of continuing AI improvement, what kind of businesses get better and better with size?

Speaker 2: Mhmm. And what are some examples of those types of businesses that should get better in the AI era? There's a article in the journal today saying that despite Google being down, Alphabet can still win in an AI pushback. How are you thinking about the different markets, the different tiers?

Speaker 1: Yeah. Thought I thought it was interesting that that this week we had commanding heights and we had Cliff's episode Oh, post episode on invest like the vest which was like somewhat complementary Yeah. In some ways, talking about the comp... What are the compliments of intelligence Mhmm. Which results...

Speaker 4: Yeah. I think Cliff and I are both like giant meta bowls at this point. Right? Like it's really hard to imagine a business that is better fit to diffuse intelligence to the global economy than Meta. All right. You have payment rails, you have WhatsApp which reaches the entire global South, you have frontier capabilities through Meta Super Intelligence Labs. Right. It is just an unbelievably well positioned business to both oversee the dollarization of the world and oversee the intelligence frontier across the world. Mhmm. We really like meta. In the piece we also support hyper...

Speaker 1: Yes. In all in all the conversations that you're having, you know, especially with people that you were giving early copies to, like how many... Like what is your estimate on on the number of people in the entire world that are truly like AGI pilled? Right? Like, it feels like it feels like especially today in the, you know, the markets this week and and, you know, the way that people question certain business models like exposes their real beliefs around the technology and how it's gonna diffuse. Right? And it it still feels like there's maybe like 10,000 people in the whole world that like are are truly bought in to let's say like, you know, the the situational awareness. Right?

Speaker 4: Yeah. I mean, I I jokingly say that New York is maybe more AGI pill today than San Francisco. And I think part of that is a willingness to accept on a big piece of the street right now that every equity is mispriced. If you think about kind of Leopold's progression through the markets, there was some sense in which like every six months we would determine something was mispriced, right? We were memory bros and we were storage bros and we were VRAM bros and you got the sense of this giant beast underlying the market could only eat one piece of it at a time and we would update our pricing on whatever and then we would sell it off rapidly to move on to the next thing. I think post SI blowup part one, there is some sense that we need to reprice all of this in a much more systemic way and we're seeing a lot more intelligence on the New York side of things taking the ASI pill and realizing that these things are going to have to move in big chunks. Right? And you're starting to see this happen with big sell offs and names like Reddit Mhmm. That are not just purely bottleneck trades.

Speaker 3: Yeah. Certainly, a lot of more

Speaker 4: New Yorkers pill today than it was six months ago.

Speaker 1: Talk about the push back so far in the piece. Think with some of the push back that I saw was this just feels dark so therefore it must be wrong.

Speaker 4: Yeah. I mean one of my favorite pieces of push back we got five minutes after we published the essay. It's a 140 pages. We got an incredible six page push back generated by AI, which means criticism itself is, you know, contained within the cast net of the models. Yeah. Look, it's not a rosy picture, but it's a picture that I think is inevitable. Technology progress has never been a thing that has happened easily. Technology progress has never been a thing that has diffused into economies without some degree of upset and turmoil. I think any kind of positive AI futurism is just going to have to process that these things have drag on effects. We've certainly seen pushback on how dark pieces of the piece are. We do think this technology will be massively destabilizing for Southeast Asia in particular in a way that is not rosy. And we've also seen pushback on particularly where command economics will happen. I think our exuberance for things like hyper liquid are not universally shared. I do think hyper liquid is an important primitive for where we're going, but certainly pushed back across the board and really fun to engage in it.

Speaker 2: Yeah. How are you taking this thesis to some of those more adjacent, less like bottleneck trades, but still like real world economy businesses? I'm thinking about like trucking, home building, lumber, things that aren't in the path of the data center construction, but aren't really affected by AI in an obvious way in the short term. How do you think about those?

Speaker 4: Yeah. I mean, like, a really great example trade was Thrive buying ADT. Sure. Sure. Right? ADT probably has the greatest sensor network on Earth. Right? Cameras, more infrastructure

Speaker 2: Yep.

Speaker 4: On the edge being able to capture novel data than almost anyone Mhmm. Else has. And I think the reaction to that has been very telling Also between who realizes. Yeah.

Speaker 1: Also, think that the upsell opportunity for ADT, I I believe really strongly that like robots are are gonna have like home security applications before they have like laundry applications. Mhmm. It just feels like much easier to... For a robot to like make it more difficult to break and enter a home than it is for a robot to be like extremely effective at like cleaning...

Speaker 2: Yeah. Not breaking your wine glass. Right.

Speaker 4: Like one... Mutual friend in the group chat this morning. I'm sure you saw it. He was in San Francisco asking people why they bought unitary robotics and this kid who had spent like $6,000,000 of his seat around buying unitary robots. Friend asked him what he has done with it and he said, don't really know. The robots can just kind of karate chop me. They can kick me, but they can't really do anything else. And, you know, Jeremy's riff on this. It's like, you know, singularity is here. Right? The robot can beat you up. Who knows what else it can do? So certainly, like, a karate chopping robot kicking people is probably the first use case. But like you have to be a d p... ADT pill even on the sensor modalities. Like, even if they're not able to get any lift out of the core business, that has to be the most valuable data set in the world.

Speaker 2: What's your view on interest rates in this neo dollar inflationary regime?

Speaker 4: Well, you can certainly track the great stagnation ending in treasury rates. Right? Like, you can literally look at the treasury yield curve and see kind of the twenty year stagnation that we lived through.

Speaker 2: Was it twenty or was it fifty? I feel like people go back to 1979 often or, like,

Speaker 3: forty years. I mean, I feel like the

Speaker 4: last time rates were at this level... I mean, yeah. Certainly, if if...

Speaker 2: Yeah. We did have high rates in the nineties. You're you're correct. High rates in the... But but at the same time, like, the stagnation thesis starts with, like, the moon landing in many tellings of it. Right?

Speaker 4: It's probably true. Probably not, you know, not adequately. We didn't have the Bloomberg terminal then,

Speaker 2: so it probably

Speaker 4: took twenty years to price in or something. Right now, Twitter.

Speaker 2: Sure.

Speaker 4: But certainly, like, if if the rate of technology progress continues at the rate that we think it will, then rates are gonna have to go much much higher.

Speaker 2: Yeah. And is that a crowding out effect, like just piling into AI investment that's driving up rates or something that's going on with the neo dollar specifically?

Speaker 4: I think partially neo dollar, partially I think a thing people have not totally processed is that when the risk free rate increases, the rate that risk demands does not increase linearly with the increase in the risk free rate. These things are actually exponentially related. So in a world where kind of yield increases two x, you may see a four to five x increase in risk rate on top of that. But certainly, just there's so much productive use for dollars at this point, where even the data center build out alone could saturate the monetary supply. Like, it's it's just... It's really important to be clear eyed about where all these dollars are going to go.

Speaker 2: How do you think the shape of the median or average firm changes in America or globally?

Speaker 4: We think there will be a lot less public companies in The US. A piece we developed at the end of it is what we call the forced settlement, which is if you can think about all of history as negotiated compromises between management, capital, and labor as kind of the three countervailing forces that make up the firm. Capital and management are going to become much more similar over time. You're going to see smaller management teams that are comped as if they are capital. You already see this with things like Meta where you have Alex Wang, basically a $20,000,000,000 employee. Certainly not a down, you know, 20,000,000,000 employees. Yeah. And you're going to see labor leave the firm, right? In the same way that The US offshored in the nineties and two thousands. I think we're going to see an offshoring two models, right? Maybe an onshoring into data centers Mhmm. Which will mean that firms will have probably much stronger but much leaner balance sheets as they push labor out of their doors and onto the models. And that means that capital management can begin to merge.

Speaker 2: And do you buy that returns

Speaker 4: to scale are really high.

Speaker 2: Yeah. Do you you buy that there will be more firms in the future?

Speaker 4: Probably more single proprietor firms. Right? You can imagine the the the kind of bottom end of that curve, like 1,000 entrepreneurs in a data center kind of thing. Mhmm. But certainly, we think in regulated industries, healthcare, banking, etcetera, there will be massive consolidation as a result of cyber vulnerabilities. But it just seems obvious that there's going to be a huge amount of merging happening in community banks, community hospitals, etcetera.

Speaker 1: Yeah.

Speaker 4: Just because the models are are so so good at security.

Speaker 2: At security. Yeah. And so you you you don't wanna... You wanna amortize the cybersecurity combating cost across the largest organization possible. Is that the theory?

Speaker 4: Well, yeah. Largest possible defense against the large... Biggest possible threat and right now it's too fractured to have a meaningful chance. Yeah.

Speaker 1: How do you think the investment community is doing underwriting cyber security risk? Because like right now, people are like, oh, I should own cyber security stocks.

Speaker 5: Mhmm.

Speaker 1: But, it feels like there's some risk of like broad internet shutdown, whether that be, you know, a day.

Speaker 3: I mean, I still

Speaker 4: think it's underrated that like internet banking is probably a thing that was like we're going to live to see the end of. Right? Like I could really imagine a world where you have to like go to the bank to get dollars out again just because it's it's too easy to clone your voice, it's too easy to break to a fight. Like, it's just... It's all too easy. You know, Corgi sells cyber insurance now, so maybe we should all feel good that Corgi is underwriting this risk at civilizational scale.

Speaker 2: Yes. Yes. It's a Fellow.

Speaker 1: Of the guys.

Speaker 2: Cornerstone of the market. Yeah. Where does it go next? Are you selling the rights? Is there gonna be a movie? Command and Plate's play?

Speaker 4: You know, the book's coming out. We will try to distribute the thing as much as possible. We're gonna try to produce it on stage as a fourteen hour production in New York City. Currently, Aaron Sorkin's bidding on the movie rights. Okay. Yeah. But the the book the book, you know, John Furrino has been even making bootleg copies of it. Sure. It's really incredible. You can buy kind of bootleg Commanding Heights copies.

Speaker 2: Yes. Yes. That's great selling them in Canal Street. I I think I saw a couple copies on laid out on the pavement there.

Speaker 4: I've been leaving copies around Central Park. You can go and get them.

Speaker 2: That's great. Well, congratulations on on the the launch and Yeah. Bring forth the most discussed... Yeah. A

Speaker 1: lot of text flying around yesterday. Cranding heights? Question mark question mark?

Speaker 2: It's great.

Speaker 1: Good stuff.

Speaker 2: Thank you so much for coming on the show. Happy Halloween.

Speaker 1: Congrats on the launch

Speaker 2: of Halloween.

Speaker 1: Great stuff.

Speaker 4: Have a good

Speaker 2: one. Have a good one.