前 Palantir 分析师谈社会主义世界中的情报与市场运作内幕

Alex Good

The first use of Palantir was like, “How do I know that I’m killing the right terrorist?” Because if you try to go to Afghanistan and kill all the people with beards driving a jeep, then you’re going to kill everyone, and you don’t know who’s who.

Honestly, I have no idea how to introduce this next guy, but his name is Alex Good, and he is one of the most interesting people that I’ve ever had the pleasure of interviewing. Our conversation spans a ton of different topics, ranging from what’s going to happen in the age of AI to the rise of socialism to how to benefit and put money in your pocket based on what’s to come.

1. Who The F*** Is Alex Good?

Alex Good

I’m in the Popcat Telegram group, and Morad is in there, too. But he’s the main event. It’s Morad in the Popcat group, and people are glomming on to him, and I’m like, “This is the community that you build.” That’s not the community that I want to build. I want to build something where my bar is, “Would I use this product every day?”

Would I show up every day and do what the AI told me to do publicly? That was the bar. The Popcat group just wasn’t the bar. So I’m like, “Okay, what would it take to actually make a system where a large group of people voluntarily opted into the system and had a real command-and-control architecture with an AI at the center of it?”

What was the issue with the Popcat group that you saw?

Alex Good

The issue with the Popcat group is 100% about the price of Popcat, and then it was about the promotion of Popcat. It wasn’t funny. The theme was, “Oh, we’re part of a community.” I didn’t see elements of a community forming there. There were not people helping each other out.

There were not people colluding usefully. It’s not like I’m in the Popcat group saying, “Hey, do you want to get into this venture deal with me?” It’s like, no, we’re all going to just post the Popcat chart and maybe engage with some tweets that Morad posted. That was the extent of the Popcat Telegram group.

The thesis that Morad had was more profound than that. It was like, “Look, in the age of AI, the only real identity is that we have these new digital cultures.” It’s very relatable because we’re on X, and to some extent, I would have never met you if it weren’t for X. So we natively understand that this is true.

The community on X is quite rich. There are DMs, meetups, in-person events, and deal flow, and it just wasn’t there for the memes. The meme coins were not useful social identity layers. But I was like, okay, I like the idea. I think it’s true, but you didn’t take it far enough. You didn’t have a real architecture.

That’s part of why I started a token. What would it take for me to join the group and not feel silly joining it?

Right. Yeah.

Before we get into that, you’ve created something insane. You were telling me before that the only point of it is to basically increase its own value. We’ll get into it. We’ll name it, talk about it, and dive in. But first I want to ask you: Who the fuck are you?

Alex Good

Who the fuck am I?

Who the fuck are you, man? I’ve known you for 4 or 5 years now. Yeah, we’ve met a few times in person. I’m a huge consumer of your content and tweets. I think that they’re brilliant. But you’ve lived a lot of lives.

2. The Advertising Edge: How He Read Stocks Through Ad Data

You’re a Wharton grad, former Palantir, former Balyasny. You started a company called Perpetua that sold, that got acquired. Then you became an independent trader. Then you started tweeting really actively. You started being an essayist. You have all these thoughts on the world, and now you built this thing called Post Fiat, which is a layer-1 cryptocurrency.

You’ve done more in a life than most people. You’ve done more in 10 or 15 years than most people do in a life. What are you doing? Who are you?

Alex Good

I was always really interested in trading and capital markets. I was very normal for a long time. I worked at Citi.

Do you consider yourself not normal now?

Alex Good

I think I got weird eventually, and I can tell you exactly what happened. I was always looking for an edge in trading. I did FX, and then I did big-data stuff at Palantir, and we found a lot of interesting stuff with big data at Palantir. Then I worked on SWIFT data in Singapore.

When I showed up to Balyasny, I got hired because I had access to a data set that got cut off because of the SWIFT problems at Stan Chart and HSBC, and I had to invent a new trading strategy. That trading strategy was based on advertising.

So I traded Google, Facebook, and Twitter. Then I had to advertise things. Next thing you know, I’m advertising video games. I’m advertising OTA, Booking.com, and Expedia, and I’m trading those stocks.

When I started Perpetua, it was just a very simple observation that I knew, because of the advertising that I was running, that Amazon was incredibly undervalued. At Balyasny, risk was constantly calling me: “You have 15% of your book in Amazon, which is way too big.”

“Why do you own all this Amazon?” I’m like, “Because everyone is saying the retail business is worth zero. I know it’s worth a cajillion dollars because I’m running all these ads, and they’re breaking even on ads.”

It took a long time for this to adjust itself, but what ended up happening was that I just got really in the weeds on what it took to acquire a buyer of a thing—an advertisement. Eventually, we got so big at Perpetua that we started booking Crocs and Kimberly-Clark.

We were still selling data to funds, but when we started booking these big companies, it was no longer viable to trade. That’s how I got into crypto, because I was like, “What can I advertise that has an edge, that has a direct feedback loop?”

By this time, I’d already started seeing stuff like Tesla, way before meme stocks were in our common parlance. We’d be running ads for Tesla cars, and all of the clicks were coming from a search for “Elon Musk.”

Can we take a step back for a second and talk about the strategy and how you found it and generated that edge? I think one of the most important things that you do as a trader is figure out what your edge actually is. Half the time, it’s searching for an edge, and then it’s monetizing the edge. So what were you doing, and how did you find it?

Alex Good

Yeah, it was actually a pretty funny thing. I showed up to work. I had been hired because I had this data set—or was supposed to have this data set—that I didn’t have, and they were like, “You have to come up with this strategy.” So they moved me from Singapore to San Francisco because I started the strategy.

I talked to a guy who owned $200 million of Facebook stock, and I asked him point-blank, “Have you ever run a Facebook ad?” He’s like, “No, I’m just in it because Zuck is a Chad.”

That was it. He was right, actually, in hindsight. He was right. But I was like, this is a huge opportunity, because people are trading these advertising stocks, and you can get real-time data about how well their ads are performing.

Then they’re like, “What do I actually advertise?” You have to advertise video games, or stuff on Amazon, or travel booking sites, or Groupon. All of a sudden, you’re advertising airlines, and suddenly you have a view.

A very simplified version of the strategy is that, at the time, about 70% of your margin at Booking.com was spent on ads. If you talk to the CFO of Booking.com, they’d be like, “Yeah, we’re very actively managing our advertising budget.” You’d say, “Look, dude, you can sell a $50 video game for $4, right? Why aren’t you spending all of your money? Why are you spending anything on TV ads? Why are you doing any TV ads?”

He’s like, “Well, we know the digital ads work, but we like the margin story. We can progressively increase our digital ads.” And you’re like, holy crap. At the time, Electronic Arts was cheaper on an EV/EBITDA multiple than Booking.com.

You’re like, okay, company A can acquire customers for a $50 video game for $4, and company B is spending $5 out of its $6 of profits, and they’re trading at a way higher multiple. So you go long Electronic Arts, you short Booking.com, you’ve got a pair trade, you’ve got the valuation on your side, and you’ve got the data on your side.

If those dynamics change in your quarter—let’s say, “Oh, Call of Duty ads are way more expensive this quarter”—then you’re like, “Okay, EA is not cheap anymore because their product is hard to acquire.”

So at the time, it was very normal. And then, I guess, to your question, it only got nonlinear once you started getting into meme stocks. Because at some point you’re like, “Wait a second. The alpha I’m measuring with Tesla is not how cheap it is to sell a car. It’s how cheap it is to sell the stock, right?”

And that’s how I got into XRP. That’s how I got into Cardano. That’s how I got into Binance affiliate programs.

That's how I got into X because I was running affiliate ads, and they were like, “Yeah, we're getting rid of our affiliate programs if you don't have 10,000 Twitter followers.” I'm like, “How am I going to get 10,000 Twitter followers? I need to start saying things on the internet.” And so that's how I ended up starting. Everything that I became was a sort of result of economic necessity.

3. The Goldfish Theory & Where The Model Breaks

And I think it's crazy because you've actually witnessed the degradation of our markets and society in real time from the inside.

Alex Good

Just based on that framework, I used to use this data set to figure out if things were expensive or cheap, and I used to use it to express real views on the real world. Now I just use it to express views on attention.

Yeah. Do you think that's right?

Alex Good

Yeah. And I think you realize, over time, at first you think that, and then you're like, “Okay, well, how much has Tesla gone up in terms of its real cash-flow generation since that time?” Enormously. So, at the time, Tesla is a meme stock, but they gave Elon capital, and he did useful stuff with it, so it's hard to really assess it that simply.

Even GameStop now has enough money that they're going to acquire eBay, and now people are really trading Pokémon cards. When GameStop was a meme stock, it had no fundamentals—zero fundamentals. Now it actually has fundamentals, right? Michael Burry was actually a GameStop investor because he saw the trading-card turnaround. These things aren't that simple.

In crypto, we see it too. At some point, Ethereum was worth $7, and it was ridiculous, and then eventually now you have Tether and Circle trading on Ethereum. So I think memes aren't necessarily just degradation. Sometimes these memes are hallucinations of our society, or hyperstitions, if you will, where capital flows into ideas and people that society wants to advance. So maybe we can talk about this hyperstitional vortex as an example of what you're getting at.

I think at the core, to me as a trader, it just reminds me of that simple graphic from The Alchemy of Finance—the George Soros graphic of reflexivity—where perception actually does influence reality. Then you can get a flywheel effect: If people believe something is valuable, then it actually becomes valuable.

But there are 2 different outcomes here. It's like a meme—no matter how high it goes, it's not generating any value for anybody, right? It's generating value in the sense that you can sell it at a higher price point, but Tesla, as it trades higher, you sell the stock, you get the cash, and then you can actually invest and grow the company. So the perception of it as more valuable actually does make it more valuable. Or do you think that applies to memes as well?

Alex Good

Well, I think the frame that I learned to take originally was—I call it the goldfish theory. It's the idea that I noticed this pattern with Amazon sellers, where there was a guy selling dandelion tea, and every day, 8% of people who went to the page bought it. It was always 8%. It never went to 6%; it didn't go to 10%. It was always 8%, and it was like this for literally years, right?

And I'm like, “Why do 8% of people always buy this? They're different groups of people.” It turns out that certain things, like attention vortices—Elon Musk being Elon Musk—are actually very predictable, right? So, in a way, predicting people's affinity to buy assets is far more predictable than predicting the future.

That was sort of the frame that I took: You don't know in advance who is going to be the vortex of attention, but once you know that they are, it's very predictable that they're going to continue to be that. And so that allows you to create a more predictive mental model of the world than a lot of people in macro trading have. They're like, “I think this is going to happen. This is where the puck is moving.”

I would rather be like, “Yeah, I think 40% of people who see GameStop come up with a Pokémon card release are going to buy the stock, and that's going to drive the price up.” That's something I can wrap my head around. I don't like predicting the future when it comes to trading. So I think that's how I got into it, and I don't know if that answers your question.

It does. It does, but there's just so much to cover here. I also want to understand where that model breaks, right? Because at some point, that model falls apart. You can't say 40% of people who come across Cardano are going to buy it ad infinitum. At some point, they stop buying it, right? So how do you assess that risk? Actually, Michael Saylor is a great real-time example of where the model breaks.

Alex Good

Okay. Fragmenting liquidity. For example, the dandelion example: If you launched another dandelion that was very, very similar to that dandelion, it would break the model. If you launch STRC in addition to MSTR, you break the model, because you fragment liquidity and fragment attention into 2 different things, which suddenly become comparable and previously weren't.

Trust matters a lot, too. The other thing about the dandelion tea example is that it was entirely based on product reviews. So if their product review went from 4.5 stars to 3.8 stars—if you start to lose trust in Michael Saylor, it's like his star rating going down—then the conversion rate drops.

There are things in real life that affect these conversion rates, and the most direct way to drop your conversion rate is to drop a competing product. Elon dropping SpaceX is unambiguously bearish for Tesla stock, because it's an attention fragmentation. A likely Anduril IPO would be bad for Palantir stock, because it's the same e-commerce dynamic with competition, and that's the most predictable fragmentation.

4. Why He's Bearish On Palantir

Actually, this is why altcoins have such a hard time: They're playing for the same pie. Bitcoin doesn't have any competitors, right? Bitcoin doesn't—there's not another store-of-value asset that has successfully argued that we just have a fixed supply and that's what we do. Ethereum and Solana are fighting, and then all these other L1s are fighting for their pie, and it's very competitive. So I think that's how the story breaks down.

That makes sense. So you actually worked at Palantir?

Alex Good

I did.

Was that out of college?

Alex Good

I worked at Citi in FX and equity derivatives, and then got picked up at Palantir.

Right. And so when you were at Palantir, what were you doing for them? Was this data analysis?

Alex Good

Yeah. Peter Thiel and Bridgewater built a product called Palantir Finance, and I was one of the first users of the product. I was very fortunate to have a group of people train me at a young age to do all these quant trading strategies I had no idea how to do. Eventually, I realized that, because I was such an early user, I had a lot of leverage, actually. So I was like, “You have to hire me because I'm the only person who knows how to use this product.”

I worked on capital-markets stuff. We worked on 2 big projects. One was the application of credit-card data to predicting large-scale purchase decisions at the CIO-office level of a bank—basically, trying to predict macro slowdowns. I also worked on SWIFT data and processing: both compliance for SWIFT data and converting SWIFT data into, once again, macro signals.

It was always kind of like the 2 business lines. There was a revenue side where you're like, “How can I take this data to make more money for the CIO office?” That's the fun stuff. Then the bread and butter is, “This guy is selling barrels of oil at $130, and the price is $80, and it's probably transfer pricing, and you should do something about that.”

So that's always the 2 sides of a Palantir deployment: Help the business lower its future fines on one side, and then the other side is, once we've lowered your fines, here's the cherry on top—here's the revenue boost.

I didn't actually realize that Palantir had a finance side to it that you were working on. Where is that today? I mean, you look at what's happening in the world of AI and as AI applies to finance—is that sort of what Palantir was doing back then? Was it just aggregating data, or were they building models to actively trade?

Alex Good

They were really early on machine learning. They would work with really big oil companies early on to make these lead-lag signals, where it's like, “Okay, you have this many shipping assets. Load this into the Palantir ontology and predict if there's going to be a blowout and a spread of oil.” That would go to their capital-markets desk, but then it would also go to their shipping groups.

In the early days of Palantir, people hated the stock. People hated the company because it was just a bunch of guys coming up with these insights, pseudo-productizing them, and giving them to the CIOs. Eventually, what Palantir evolved into was a much more elaborate product, right? They eventually productized all of this, and it stopped being consulting and started being delivered via Foundry and other products.

There are different products. One of them is Gotham, which is a security product, and then Foundry and other ontology products are more revenue-generating. I think their business is split roughly 50/50 between the commercial business and the government business.

But you're kind of bearish on Palantir now, is what I'm hearing from you. You're tweeting a little bit about the fact that they're in a fight with Anthropic and OpenAI. It almost seems like everybody is in a fight with Anthropic and OpenAI. Every week, you see a new release from these frontier labs that could take out companies in this world. What's your view on what's happening in the world of AI? Are we seeing a brawl out in public between all these companies?

Alex Good

I think you're seeing a really specific brawl because Alex Karp called out Dario Amodei—basically called him out publicly in an interview. Dario was saying, “Oh, we're going to have 10% unemployment.” Karp was like, “If you think you're going to have 10% unemployment, well, guess what? News flash: you might have a high IQ, but you're actually retarded.” They're starting bigger and bigger fights.

And, yeah, I am bearish on Palantir now. The reason is that the assertion of Palantir is that LLMs don't work out of the box. They say you need Palantir for LLMs to work, and I'm just like—

Is that because of the underlying data?

Alex Good

Oh, yeah. They're saying that in order for your business to properly use LLMs and avoid hallucinations with your large data sets, you need to have an ontology baked in so the agents know what to interact with. I'm like, I just know that's not true, right? I use LLMs all the time without any Palantir.

Can you clarify what you mean when you say you need to have an ontology baked in? What do you mean when you use the word “ontology”?

In the Palantir world, there is a model. The origin of Palantir is actually interesting: How do I know that I'm killing the right terrorist? If you try to go to Afghanistan and kill all the people with beards driving a Jeep, then you're going to kill everyone and you don't know who's who, right?

How do you even know that Osama bin Laden is Osama bin Laden? For the Toyota Hilux, you need his license plate number, you need his known associates, you need his interactions with other people, and you need his bank accounts. This idea of these things that comprise Osama bin Laden is a, quote-unquote, ontology.

In order to target the right things, whether they be financial assets or military personnel, you need to have a higher-level ontology to map onto it so that you know you're working with the right stuff. In the world of AI, they're using the same argument. They're saying, “In order for your AI agent to know that it's interacting with the right data sets, you need to label the data sets effectively. You need to have a model for when they've been acted on by a certain employee at a certain time in order for this all to work.” That's the argument.

But if you work with AI in real life, you know that's not true. You can point an AI at your codebase and just be like, “Yo, figure this out,” and it will.

As AI models have become more capable, the need for ontologies has decreased linearly. There are all these studies of prompting and how effective it is, and all the major labs are basically saying prompting is less important than it's ever been. You don't need to paste a 2-page prompt into your model for it to do the right thing anymore. Now you can just be like, “Build me a great website,” and it'll do it.

That's real-time evidence that the use of ontologies is a hack to make AI models more performant. The other thing is that there's a lot of fragmented attention. OpenAI has forward-deployed engineers. Everyone uses the phrase “forward-deployed engineer,” right? Anthropic has forward-deployed engineers with Goldman Sachs.

Palantir used to be special because it was the only one on the block with an intelligence product, and now there are 3. It goes back to what we were talking about with fragmenting attention: When there are 3 things in the store instead of 1, guess what? If you're trading at 100 times sales, you're going to face investor headwinds.

Would you agree with the statement that structured data is less important, but raw data is still as important as it was before?

Alex Good

The ability to turn raw data into structured data has never been cheaper.

Right. That's what I'm getting.

5. AI Won't Make You Useless — But It Will Take Your Job

You could argue that the value of raw data is definitely going up a lot. That's something you're seeing in the market, right? The new drop-shipping is data acquisition. Everyone is starting a data firm to sell data to labs because it's much, much easier to do it.

What are some opportunities in that? What are you seeing? Are you working on anything in that world?

Alex Good

My sort of view is that we've seen a crypto protocol do this, right? Grass. They sell data to labs. I'm not interested in that. My premise is that most people in crypto are here to speculate, and the information that they generate is very rich and very valuable. If you actually think a data source is valuable, you shouldn't sell it; you should monetize it.

That's what I'm working on: How do I generate the richest possible data set with a group of pseudonymous actors and monetize that information instead of selling it to an AI research lab? Just monetize it internally.

Right. But that's, I think, pretty difficult for most people. One thing that I always come back to is that AI, for high-agency people who really want to get things done, is this massive tool of leverage. Your average person isn't taking advantage of it in any meaningful way. I think most people are using it as just another chatbot or an assistant, and people don't really go past that or beyond that.

What does that mean if you can start to have individual people build billion-dollar companies by themselves? Where does that leave us? Do you agree with the statement that AI is making the vast majority of people useless?

Alex Good

I don't think so. I think it changes the use case. We're not yet at that point—we're still at full employment, actually. A lot of these AI-unemployment things have yet to kick in.

What has yet to kick in? Do you think it will?

Alex Good

I do. I think a good example is Figma's stock. It dropped 85%, and their headcount went up. Or Accenture dropped 60%, and their headcount is up. They have 800,000 employees, right?

We know that Accenture hasn't fired anyone, and the market has crushed their stock 60%. It's somewhat safe to say that the employee reduction is yet to come because it just hasn't happened, and the employment statistics don't reflect it. There's this huge disconnect between what Dario Amodei and Sam Altman are saying about this existential crisis and the need for MMT, and then the employment report comes out and you're like, “We're at full employment. You can't hire people.”

Do I think that people are going to lose their jobs? Yes, I do. Have they lost their jobs yet? No, definitely not. I think the most interesting primitive is not that we're going to have these solo shippers. It's more that there are a lot of people in the same boat. There are a lot of Accenture employees who are about to get laid off.

And then the question is: How can you use collective action to extract economic advantage? I think that's the interesting question, because there are different ways to approach it.

Some people are MMT maxis. They're like, “The way that we use collective bargaining is that we all vote to give ourselves money.” You're like, “Okay, well, what would that look like?” MMT is actually quite hard to distribute. The taxation that would be required is difficult because global interest-rate markets are not in a happy spot in terms of government spending. The idea of doing MMT on top of the current JGBs blowing up is hard.

You say, “Okay, well, you need taxation. What does that look like?” That probably is central-bank digital currencies. I think the thing that's going to happen is that more and more people are going to be focused on saying, “Okay, we tried to escape the permanent underclass. We didn't. We're all in the permanent underclass together. What are we going to do about it?”

That's a more interesting question. That hasn't kicked in yet. It's like a meme right now.

6. Unemployment In Three Months

Right. But people really do take it seriously. I think that's actually a premise of why a lot of people listen to this podcast in particular, and to other financial media and news: People are trying to figure out, one, how much time do I have left? You're kind of saying it might actually be more elongated, I think, than what people are claiming right now—or what Dario and Sam are saying.

Then, how do I—where do I put my money to escape the permanent underclass? Is it even possible to escape? I want to start with the first question, and then you can walk me through what you think is going to happen. How long is it going to take before we start to see these unemployment numbers kick in and we start to see this permanent underclass actually form?

I think, literally, unemployment should start kicking in within 3 months, and it's just because of Accenture.

Within 3 months.

Yeah. Because the Accenture situation was so extreme, the stock dropped 26% in a day. It's one of those moments where the management—

If they weren't awake before, they woke up real fast.

The software indexes are not bouncing back. The really egregious overhiring firms, like Salesforce, are getting their stocks crushed by the market. The market is being very, very clear with companies: You need to do RIFs. People didn't want to do it, and they're going to be forced to now because their stocks are down so much.

Do you think there's a trade in that—that once they start laying people off, maybe these stocks bottom for a little bit?

Alex Good

I think so. I think some companies with credible AI turnaround plans, where they could genuinely deliver their software for much less, could have crazy terminal EBITDA margins. The joke is that commodity companies right now are oftentimes twice as expensive as software companies. The big diss used to be that, oh, you're trading like a commodity. It's like, well, dude, now commodities are twice as expensive as software.

So software is kind of bottom-of-the-barrel multiples. The bar for them to start firing people and delivering a product is really appealing. Stuff like Nintendo is not just software; it's also people competing with enterprise for things like memory. Nintendo just couldn't launch consoles because it couldn't afford memory, and so its stock is down 60% year-on-year. You're like, “Okay, well, what if memory breaks?” Then it can rebound. So, yeah, I think there are going to be a lot of opportunities.

One of the interesting AI developments is that you saw all this really, really hyped stuff with Etched. All the VCs were saying, “Oh, there's a new fundamental chip breakthrough.” It's essentially like what happened in crypto with ASICs, where you're like, okay, rather than serving a model that's massively memory-intensive with inference, what if you just literally put a model directly into an ASIC? Then you could run a physical model. Rather than having a generic GPU, you have a specialized GPU with a specific model running on it, which would drastically cut memory usage.

The interesting thing is that if you remove the overhang of a lot of these so-called bottleneck trades, a lot of companies that have been destroyed by the bottlenecks are going to bounce. So, yeah, there's going to be a ton of trading opportunity.

Maybe you can talk about some specific companies and how they're going to do, I guess.

Alex Good

Yeah, Nintendo.

Yeah, Nintendo's big. Okay.

7. The Nintendo Trade & The Coming IP Goldmine

Alex Good

The other thing is that we saw the LibGen settlement with Anthropic, right? Essentially, LibGen scraped all the books in the world, and then Anthropic trained on LibGen. There was a big copyright settlement. People have proven, for example, that you can get 97% of Harry Potter out of ChatGPT. You can just prompt ChatGPT to get Harry Potter, and that's a problem from a copyright perspective, right?

There are all these companies with great IP franchises. Nintendo has a great IP franchise. If you're bullish on non-enterprise or more consumer-type stuff, there's a world where all future AI Marios pay Nintendo. Games Workshop would be another one, like Warhammer 40,000. That stock has done a lot better than Nintendo.

That would actually be a huge value unlock.

Alex Good

Yeah, for Nintendo, if they actually allowed that to happen.

Yeah. Right. If they allowed people to just create games with their own IP using AI, that would be massive.

Yes.

And I assume that you're thinking that this could apply to other areas as well.

Alex Good

Anyone with core canonical IP, Lindy IP.

Right?

Disney, Nintendo, Games Workshop, and many others, like Hasbro and its D&D IP. A lot of these things are very, very popular—Star Wars—and they all have really radically different valuations, right? Some of these charts are up and to the right; some of them are not, right? So there are different opportunities.

I want to go back to the point about whether you think there could be mass unemployment coming in 3 months—or the beginnings of unemployment. Take that a step further and talk about, let's say, 6 months, 9 months, 12 months, or 24 months down the line. I mean, you've written about the Blackprint. Maybe get into that and talk about how you think the world is going to change because of these new tools that we have.

Alex Good

Well, yeah, I think there is a point, and I think we've already crossed the point where this has stopped being a technological phenomenon and has started becoming a political phenomenon. It's already showing; it's going to show up in the midterm elections, right? The next big checkpoint that I think would validate my worldview—and I guess, to succinctly summarize it—is that people won't vote for a right-wing accelerationist system which doesn't benefit them economically.

The math is actually crazy, right? There are 50,000 NVIDIA employees, and NVIDIA is worth, single-handedly, more than the entire Russell 2000. The Russell 2000 employs millions of people. So there is this extreme imbalance between the number of voters at NVIDIA and the number of voters in the Russell 2000.

Right now, the market kind of assumes that this will go on forever, that there's no Democratic outlet, that it's kind of Trumpistan all the way, and that the right wing will keep winning for some reason. I'm like, no, I don't think so. I think what's going to happen is that the typical Accenture employee who loses their job is going to be quite effective at coordinating grassroots Bernie Sanders votes, right? Those are competent people.

Yeah. And you're sort of seeing that, I think, with the DSA. The DSA is extremely effective at what it's done right now, and it's installing socialist candidates across the board. The competency levels are going up among these socialist politicians. But it's true that on the right, you also see the rise of populism.

It's a little bit of the horseshoe theory here, where it's just different constituencies that are getting the benefits. On the right and on the left, I think both are appealing to people—whether it's giving subsidies to farmers or giving subsidies to people in the cities. It almost seems like that's where our politics is headed.

Alex Good

It's also Trump. Trump went in and regulated Anthropic. He already kind of started the process of the political system blocking major model releases. That's a big deal, because now that the gloves are off, it becomes acceptable to regulate and sequester AI in the name of the so-called public good.

It's normalized now on the right and on the left. Previously, that would have been a Biden policy. That would be a Biden move of, “Oh, we're going to stop this AI model from launching.” Now it's a Trump move. So I think you're going to see that the core thesis of the Blackprint is that the political system in the United States is not designed for hyper-acceleration.

8. The Sovereign Margin Call: Why AI Gets Regulated Like Nukes

The fantasy that people have that we're going to allow peptides is a good example, right? The FDA is like, “It's not very clear that peptides are even going to be legal,” right? Everyone's shooting these things up, but it's like, okay, what about human testing? What about human trials? What about the process of how we approve drugs in the United States?

If you look at the history of U.S. medicine, it's never been like, “Yeah, let's YOLO this vaccine.” Except for the one time we YOLOed a vaccine, and then it did not go that well, right? The one time we YOLOed a vaccine, there's myocarditis. It's a problem, right?

The backdrop is that, actually, right now, the consensus trade is biotech. Everyone is long biotech because they're like, “Okay, what the AI companies have to do in order to prove themselves is launch a cure.” Who doesn't want to cure cancer? What if Anthropic cures cancer?

Do you think that trade is truly baked in already?

Well, I think it's gone the wrong way. I'm bearish on the application of biotech and AI.

The reason is just national security. The tail risk is so high after we saw Wuhan. It's basically like all it takes is one guy running an unsanctioned experiment on a bat. There's a guy doing stuff in his garage with an AI model, and I just don't think that type of stuff is going to be tolerated in the medium term. I think you'll probably have—the second you have one event, just one event—it'll be banned.

Right. All right. But don't you think that these pharmaceutical companies are, at this point, quite highly regulated? Wouldn't you assume that they would be able to figure out how to integrate AI for drug discovery in a way that's in the lane of safety? That would just accelerate their drug discovery process, but it would be overseen by the same regulators that we have now?

Alex Good

Well, it depends on what they’re doing and who has access to the AI models. Then it’s also, okay, what’s the uncapped risk of this behavior? If the uncapped risk is something like COVID developing, or some sort of virus or unintended consequences, it’ll be really hard. I think people are overly optimistic about deregulated AI acceleration.

Actually, the reason is just because of Fable. If people are already worried about Fable debugging your codebase, which is why they pulled Fable, imagine if Fable is creating a new peptide. Wouldn’t that argument also apply to the development or deployment of AI in consulting firms and law firms, sort of across the board? If the government saw that it was actually going to cause tremendous job loss, wouldn’t they try to step in and stop that before we at least have social programs in place to deal with that fallout? Because our government’s not stupid. I mean, they’re incompetent, but they can see that this is going to happen. So, wouldn’t you expect them to act?

Alex Good

Yeah, you would. And you’d think they already would have, but they haven’t. That’s why I think the job loss is actually going to kick in, because Figma is already down 80%, and it’s for a good reason: the current models, like Fable and GLM 5.2, can actually deliver a good experience. So, the AI that we already have is kind of going to, I think, go after the laptop jobs, right? I don’t know if AI is going to be able to, for example—like Andrew Kang says—we’re going to have humanoid robotics. I’m totally on the other side of that.

You don’t think humanoid robotics? I mean, this is good. Let’s dive into that. Everyone is talking about humanoid robotics. I had Andrew Kang on this podcast, and he gave me the whole spiel that specialized robotics are not nearly as useful as generalized robots. Therefore, we’re going to see robots in the next 5 years doing construction. We’re going to see them on assembly lines. We’re going to see them in people’s homes as cleaners. We’re going to see them as personal assistants. And you disagree with that?

Alex Good

Yeah. I mean, Trump banned port automation, right? So, even port automation was a no-no. And on the right—now imagine it on the left, right? The idea of the government on either side of the current political spectrum rugging blue-collar workers by allowing humanoid robotics, given the history of self-driving, is laughable.

But wouldn’t that put us massively behind in our fight with China? If we’re not willing to adopt automation, we’re kind of messing up.

Alex Good

No, but I don’t think that’s how you win elections. I think you win elections by telling people you’re going to keep your job. I don’t think the vision of humanoid robotics is probably exciting to a lot of people. I think it’s deeply unpopular, actually. If you look at societal perception of AI right now, it’s already very unpopular, and that’s before people have started losing their jobs.

Once people—we’re at, like, 4% unemployment. The second you get a kick in unemployment and it’s due to AI, and people already hate it, it’s going to get way worse. Then there’s going to be regulation on just raw AI, let alone humanoid robotics. This is the area that’s going to be regulated.

We already know from self-driving cars what that looks like. It means that they’re literally not allowed on the road. And when they are allowed on the road, periodically they get pulled. It’s like, okay, someone attacked a Waymo, and we don’t know why they did it, so now we’re pulling Waymos off the road for a month, right? And that’s with cars. It’s going to be worse with humanoid robotics.

So, this kind of implies—and maybe I’m putting words in your mouth—that the AI bubble is close to an end. What we’re seeing with a boom in data centers and the massive rise in the stock prices of all these hyperscalers almost implies that we’re close to the end of that, because you also think that we’re close to the beginning of regulation.

This is kind of the opposite of the doomer thesis in a way.

Alex Good

Well, I don’t think the world’s going to end. I think the world is going to get more distracted and more speculative. I do think the economy is going to get crushed, because I think people assume that AI is going to get us out of the debt bubble, right?

The assumption people have is that if we have a massive productivity increase, then we’re not going to have to worry about our World War II level of spending. And it’s like, okay, if that doesn’t happen, then it’s basically a sovereign margin call. That’s sort of why I’m in crypto.

I don’t think I actually disagree with that, because back in crypto world, right when Truth Terminal came out, everyone thought we’d have all this AI entertainment. We’re starting to see it with Seedance, right? There are these cool videos, but I don’t know about you—I don’t actually watch any of this AI stuff.

No, I mean, I’m not really interested in fruit videos.

Alex Good

It’s fun. It’s funny, but it’s not that good.

I’ve definitely been got by an AI video before. I watch it and then I watch it again. I’m like, “Wait a second.”

Alex Good

Yeah. Some of them I like—the inspirational ones, where they get the athletes or the animated Mike Tysons. I get a lot of historical AI on my slop feeds.

Well, that’s pretty good.

Alex Good

Where people are recreating Napoleon’s battles using AI. I’m like, okay, that’s actually pretty compelling. But, I mean, I think we’re pretty early on AI. I don’t think we’ve fundamentally seen entertainment or generative worlds really kick in.

I think one thing you know for sure is that personalization delivers massively. If you do the math on the cost of an AI video right now, a personalized AI video is something like $55 for a reasonable video, and it’s not that good. Now, if that goes down to $2, which is believable, in 2 years you’re basically going to have mass-generated personalized videos for everything, whether that’s entertainment or video games.

Also, the tokens per second are going up exponentially, which means that you’re going to have compelling computer-game NPCs and improved social-media algorithms. My core bet is that AI just makes society look more like it already is. We’ve already seen what happens: the government regulated nuclear technology in 1945, and they basically said, “You’re not allowed to have a nuclear weapon, and if you do generate a nuclear weapon, we’re going to bomb you. You’re not allowed to do certain types of physics research, and if you do, we’ll probably kill you.”

Then what happens? It’s like, okay, computers are allowed, digital economies are allowed, and entertainment is allowed. The same exact thing is about to happen with AI, where you’re like, what will be allowed? How do you pay for the data centers? It’s like, you cook everyone’s brains, right?

The actual thing that will happen is that we have a society which is more online, more digital, where things get more expensive and productivity doesn’t necessarily increase. Right now, annualized productivity, as of the recent quarter, is 6% annualized. I’m like, dude, Microsoft’s CEO said that we’d see 10% annual productivity increases. That’s a lot different from 6% and 2.8% year on year, right? It’s not 10%.

That’s why things like Robinhood and Interactive Brokers are kingmakers, or Hyperliquid is a kingmaker, right? People aren’t actually generating more work. They’re generating more capital, and that capital is being recycled into an attention economy. I think that’s just going to continue.

I’m completely on the other side of Andrew Kang, or a lot of these people—and even Elon, right? With SpaceX, they’re like, okay, it’s trading at some ungodly multiple on the premise that we start colonizing Mars or launching space data centers in 2029. I’m not sure that’s going to happen. I think it’s easier to underwrite the government cracking down on things and having more of the same things that we’ve had since 1945.

So, I mean, I think that’s as good a place as I need to wrap up. Sovereign margin call—that’s, I think, this is—I’ve been pretty bearish on Bitcoin. This has maybe actually made me a little bit more bullish on Bitcoin, just this entire talk. I thank you for sitting down with the 1000x podcast. I mean, this was awesome. If anything, I think this was an advertisement for better financial planning out there if we’re heading into a collapse. And thank you, Alex, for joining us.

Alex Good

Thank you, Avi. Cheers.