Is AI Replacing Workers Faster Than We Think? | We Break Down the Viral AI Doom Loop Article

Summary

---

Podcast Summary: Excess Returns – “Is AI Replacing Workers Faster Than We Think? | We Break Down the Viral AI Doom Loop Article”

Hosts: Jack, Kai (plus mentions of Matt and Justin, not present in episode)
Date: March 1, 2026

---

Episode Overview

This episode dives deeply into the viral “AI Doom Loop” article by Citrini, which recently stirred vigorous debate across finance, tech, and investing media. Hosts Jack and Kai analyze the scenario posed in the article—where AI adoption sets off a rapid cycle of worker displacement and economic upheaval. The discussion blends macroeconomic theory, technology adoption, investing strategy, and real-world trends, with a thoughtful back-and-forth about risks, skepticism, and practical takeaways for investors and workers.

---

Key Discussion Points & Insights

1. AI and Economic Growth: Substitution vs. Complement

[00:28, 07:11]

• Jack: Economic growth balances on the formula of “people times productivity.” The worry is whether AI is replacing people or making them vastly more productive.
• Kai: Core question—Is AI a substitute for human labor (replacing jobs) or a complement (amplifying productivity and boosting output per worker)?

2. Dissecting the Citrini Post’s Influence

[01:10, 01:44]

• The Citrini article sparked intense commentary, including Bloomberg headlines, counter-responses (e.g., Citadel, Jeremy Siegel), and even a Fed governor’s statement.
• Jack: “You don’t see that typically from like a Substack blog.” [01:10]

3. The Euphoria Phase & Lack of Immediate Data Signal

[04:52, 05:37, 06:02]

• “Everyone’s pumped up about AI, but we haven’t yet seen dramatic labor or productivity impacts in the data.”
• Productivity measurement is particularly hard in the tech age, and economic data is backward-looking.

4. Framework: AI’s Effect on “People x Productivity”

[07:11]

• Possible dual effect: AI could reduce workers employed but boost their productivity drastically.
• Understanding the net effect for markets and GDP is central.

5. Software Disruption—Winners, Losers, and Moats

[09:08 – 14:10]

• Will firms actually insource software? Unlikely for large incumbents with high switching costs and strong moats.
• Kai: “The reason the DoorDash app is worth billions of dollars is because of the network effects... not just the software.” [21:12]
• Companies with “only a software moat” are vulnerable; companies with brand, relationships, stickiness are better positioned—even if their code is easier/cheaper to produce.
• Margin compression is likely: cheaper software means possibly lower prices, but could also massively expand TAM if barriers to app creation fall.

6. The “Doom Loop” and Job Losses

[16:26, 17:18]

• Article envisions a cycle: firms lose pricing power → lay off workers → need more AI to compete → others follow → repeating negative loop.
• Real-world examples: recent Block layoffs (attributed partly to AI but also overhiring and cost-cutting more generally); tech companies have run leaner without notable business fallout.

7. Limits to Immediate Impact: Who’s Really at Risk?

[19:17, 49:31]

• Jack notes: Only about 10-15% of jobs are in tech directly. Most of the economy still consists of non-tech, small business roles.
• Many service jobs (plumbers, electricians, small retail, etc.) are insulated—at least for now.

8. Friction, Price Discovery, and AI Adoption Lag

[19:41 – 28:43]

• Will AI-driven price-shopping “agents” truly erase firm moats? Not so simple: most moats are not mere UI/software, but network effects, regulation, brand, etc.
• Kai on DoorDash: “It's not the software itself, it's the network effects...that make it as powerful as it is.” [21:46]
• Pace of adoption matters: tech moves fast, mass adoption moves much, much slower. Older generations and large enterprises notoriously slow to switch.

9. Constraints: Compute, Power, Physical Adoption

[28:57 – 29:52]

• Constraints like computing power and electricity are real. AI’s cost curve may limit runaway adoption or enforce a measured pace, “giving people time to adjust.”

10. The Self-Correcting Mechanism Debate

[31:43]

• Does rapid AI deployment break the economy’s self-balancing feedback? The article argues that if firms are forced to lay off en masse, the usual correction (laid-off workers filling new roles) fails if AI can do everything better and cheaper, triggering a negative spiral.

11. Creative Destruction—Historical Parallels and Differences

[34:48 – 35:18]

• “60% of jobs today didn’t exist in 1940” (citing David Autor), but it’s hard to see future jobs now, which biases us toward pessimism.
• This time may be different: AI competes on intelligence, not physical strength, possibly endangering broader white-collar swaths.

12. Wealth Inequality and Potential Remedies

[39:12 – 41:17]

• AI likely to accelerate capital-over-labor share trends, widening inequality.
• Potential solutions: government intervention, UBI, “sovreign wealth funds”—but these come with other risks and tradeoffs.

13. Adoption S-Curves and Market Response

[54:35, 55:01]

• Historically, new tech takes longer to diffuse through society than its inventors expect.
• The current labor market data does not show meaningful AI-induced job loss yet.

14. Practical Investing Implications

[59:49 – 62:11]

• Tech disruption produces winners and losers. Aggregate wealth/trend is likely positive.
• Investment focus: Intra-sector selection—find which firms benefit most or are “AI positive.”
• Human capital: Individuals should lean into AI, becoming users/experts to stay relevant. “AI won’t replace you, but someone who uses it will.” (Rob Arnott, cited at [61:29])

---

Notable Quotes & Memorable Moments

• Jack, on the blog’s influence:
  “You don’t see that typically from like a Substack blog.” [01:10]

• Kai, on DoorDash’s moat:
  “The reason the DoorDash app is worth billions... is the network effects... Even before AI, it wouldn't take that much time to put together a prettier app.” [21:12]

• Jack, on network effects:
  “If the barrier is software related, I understand why we have a problem. I don’t know if that is the case here.” [23:40]

• Kai, on creative destruction:
  “60% of jobs that exist today didn’t exist in 1940.” [34:48]

• Jack, on technology optimism:
  “I choose to believe that, in the long run, this is going to be like the other technologies and this is something that’s going to make our lives a lot better.” [59:18]

• Kai, on agnostic investment strategy:
  “My thought would be trying to figure out within each sector, which are the companies that are well-positioned and trying to kind of concentrate in those stocks and kind of steer away from the ones that might be value traps.” [60:40]

• Rob Arnott (quoted):
  “AI isn't going to disrupt you, but somebody who uses it is.” [61:29]

---

Structured Timestamps of Important Segments

• AI’s Economic Impact Framework – [00:28, 07:11]
• Viral Reaction to Citrini Post – [01:44]
• Euphoria Phase and Data Reality – [04:52, 05:37, 06:02]
• Framework for Productivity with AI – [07:11]
• Enterprise Software, Moats, and Margin Compression – [09:54, 13:34, 14:10]
• Doom Loop Dynamics & Labour Displacement – [16:26, 17:18]
• Tech Layoffs and Broader Economic Effects – [18:04, 19:17]
• Limits to AI Adoption – [19:41, 28:43, 29:12]
• Constraints (Compute, Power, etc.) – [28:57, 29:12]
• Self-Correction and Creative Destruction – [31:43, 34:48]
• Socioeconomic Consequences & Policy Talk – [39:12, 41:17]
• Debate on Adoption Speed – [54:35, 55:01]
• Practical Investment & Personal Advice – [59:49, 61:29, 62:11]

---

Arguments For and Against the "AI Doom Loop"

Arguments For

• AI is qualitatively different from earlier tech – threatens intelligence-based jobs across all sectors. [44:44]
• Incumbents (big software companies) could be more vulnerable than expected as moats (sometimes assumed durable) may not be. [46:05]
• Accelerates job losses among top earners who drive most economic consumption. [53:11]
• Wealth and power could further concentrate among AI capital holders. [39:12]

Arguments Against

• Adoption will be much slower than technological progress; enormous institutional and practical inertia. [26:46, 55:01]
• Major constraints: computing power, energy, capital, workforce adaptability. [29:12]
• Most jobs have bundles of human tasks; few will be 100% automatable soon. [48:34, 49:31]
• Labor market data doesn't support “doom” yet; similar past fears were overblown. [57:20]
• Technological disruption has always created entirely new work and needs we cannot currently envision. [35:18]

---

Tone & Takeaways

• Tone: Thoughtful, skeptical, nuanced, with a strong “let’s think it through, not panic” approach.
• Investor & Worker Advice:
  • Don’t fear-monger, but don’t ignore AI either; those who use it will leapfrog others.
  • Instead of trying to time the market, focus on understanding which companies and sectors are truly AI-resilient.
  • As a worker, become an “AI whisperer” in your domain—learn and adopt rapidly.
• Ultimately: The doom loop is a useful thought exercise, not a reliable prediction. It helps stimulate the real, granular questions investors and workers must answer about the next technological wave.

---

Concluding Quote

“I choose to believe that, in the long run, this is going to be like other technologies... and will make our lives a lot better. But... it’s not going to go smoothly.” – Jack [59:18]

---

Transcript

[Sponsor/Ad Voice]

Starting a business can seem like a daunting task unless you have a partner like Shopify. They have the tools you need to start and grow your business. From designing a website to marketing, to selling and beyond. Shopify can help with everything you need. There's a reason millions of companies like Mattel, Heinz and Allbirds continue to trust and use them. With Shopify on your side, turn your big business idea into Sign up for your $1 per month [e-mail] specialoffer when

[Jack]

I think about the economic growth, I always think about this idea of people times productivity. So it seems like economic growth is to some extent a function of the number of people that are employed. And on the other hand, how productive are those people?

[Kai]

The big question with AI is is it a substitute for human labor? Will we just automate away all our jobs or is it a compliment? Is it a force multiplier on giving, say, an individual software engineer 10x more productivity through cloud code? The whole point here is that AI is so good that it crashes the economy. If AI is so good and the cost of goods fall 90%, then you know, if you were a worker who makes say 50k a year and you still make 50k a year, but now you can buy 500k worth of goods, that's pretty good, right?

[Jack]

The speed at which technology moves is different than the speed at which adoption moves. And it's one thing to say, all right, the technology will be able to have this, you know, AI that'll make all my buying decisions. It's a different thing to say. Like you'll probably be one of the first people I know who will probably do that if it's available. Like my dad will probably not do that for a very, very long time. So Kai, I think it's fair to say this Satrini post has kind of taken the investing world by storm. Like I saw there was like a Bloomberg headline which was basically like market crashes or whatever due to Satrini post. Like you don't see that typically from like a substack blog.

[Kai]

Yeah, no, I've been getting actual push notifications from my Bloomberg app to inform me about the Citrini article and I've probably had five to ten people email me saying, hey, have you seen this article yet? So yeah, it's a pretty big deal.

[Jack]

Yeah, what we wanted to do is we wanted to. Whenever you see an article like this, you always get people polarized. You get people with this huge concern about we're about to have a disaster with AI, then you get people on the other side who pick it apart and say, there's absolutely no validity whatsoever in this. And what we want to do today is we just wanted to work through this the way we work through it. If you and I were on the phone, like, work through this and say, what are the pros and cons? What validity is in here? And we're also going to sort of tell the story of the article as we go. For people who haven't read it, I mean, I think a lot of people probably have read it, but talk about what he's saying might happen. And even he admitted, I think, when he did this, that this was sort of a thought exercise. I don't think this is like Citrini's base case in terms of what's going to happen in the world. I think he's looking at a series of probabilities, and I think he's looking at this as a thought exercise about one possible thing that might happen.

[Kai]

Yeah, look, I mean, it was a well written article, and we'll get into the points. And at the very least, it did spark a lot of conversation that I think as AI takes off and continues to be transformative, it's good for us to have that conversation. And so I'm excited to get into it. Any opportunity to talk about AI, I'll take it.

[Jack]

Yeah. Yeah, me too. You know, a lot more about AI than me, but I still like to talk about it. Yeah. And I was kind of thinking about when this came out. I was thinking about that Mike Reed piece, I don't know if you remember that, the $140,000 poverty line piece. And it was. It's kind of the same. Like, I look at this kind of sort of the same way I looked at that. Like, that piece, you know, people picked it apart and there. There might have been issues with it, but there also were sort of directional things within it that I think were important for all of us to think about. Like, the idea that it is harder to raise a family, like, on a certain wage than it was in the past. Like, the American dream is harder to get to. So, like, even though people picked apart the poverty line stuff, there was some truth to it. And I think it's the same thing with Citrini. Like, I don't think anybody thinks the world's going to melt down tomorrow, but I think there this. There is this battle between, like, AI is a different technology. And so some people might view AI as something that's going to enhance workers. And if you look at Citrini's ideas, maybe it's something that would replace workers. And if it does that, there's all kinds of ripple effects to the economy. So I think it's just important to think about it that way and to say, even if it's not this extreme, maybe we get a reduced version of this and we all have to think about it in terms of the way we're investing.

[Kai]

Yeah, I think that's right. I mean, again, I wouldn't say it's my base case for what happens here. I think most folks, and even Satridi himself, I assume, would say this is a kind of tail scenario that it's worth considering and is potentially more plausible than others are willing to concede. And I think directionally, by thinking about some of the points that are being raised in this article, while as you point out, they are obviously very extreme scenarios, you know, some of the kind of underlying mechanics are not wrong. Right. They likely exist probably in smaller size and maybe kind of counterbalanced by other forces in the economy, but are certainly worth considering, you know, as investors and economists.

[Jack]

So as we talk about it, we'll start at the beginning, which is he's kind of starting with where we are now. And there's this thing he calls the euphoria phase, which we've sort of seen in the markets. Like, everybody's all pumped up. I mean, maybe not today, but in general, everybody's been very, very excited about AI and what it might do. And what's interesting about it is and Citadel wrote a counterpoint to this and they pointed this out in their article like, we haven't seen a ton yet in terms of actual impact in data. Like, Citadel may have committed somewhat of a chart crime, I guess, with their I'll put this chart up and then maybe I'll have you share your screen after so you can show me the other chart. But Citadel had this chart about software job postings, and they were basically saying they've actually been a little bit up. But I think when you looked at that, maybe that's not the case. But either way, like software job postings have not changed dramatically, at least yet because of AI, right?

[Kai]

Yeah. And I think more generally, if you step back and ask how is AI impacting the data? Right. Whether it's the labor market, data, productivity, any of these things, you're not seeing it yet. Now, that doesn't mean you won't see it over the next two years or longer, but at least I think it's fair to say that we're not seeing massive dislocations in either way from AI so far.

[Jack]

Yeah. And we see it in productivity, although you and I, we talked before we started, like, productivity is almost impossible to measure in a technology age. So like, you can't necessarily trust the productivity data. But we're not seeing anything. We're not seeing mass layoffs, we're not seeing like massive economic growth boosts. We're not seeing anything yet. But we also have to know that obviously economic data is backward looking. So by the time we see that stuff, probably a lot of other stuff will have happened in the stock market. Yeah. So I want to start with maybe a framework of how I think about this. And you can tell me whether this framework is solid or not. But when I think about the economic growth, I always think about this idea of people times productivity. So it seems like economic growth is to some extent a function of the number of people that are employed. And on the other hand, how productive are those people. And I think it's interesting to think of AI from that perspective because you could argue AI has sort of different effects there on the number of people employed. It could actually ultimately have a negative effect on that part of it. But on productivity, people expect it to boost productivity, to have a positive effect. And so I just think as we consider all the different this article and all the different impacts, it's interesting to maybe keep that framework in mind. Like, how does this affect this people times productivity thing?

[Kai]

Yeah, and this goes back to what you were saying initially, which is the big question with AI is, is it a substitute for human labor? Will we just automate away all our jobs or is it a compliment? Is it a force multiplier on giving, say, an individual software engineer 10x more productivity through glide code? And I think those forces both exist, and they exist in varying amounts throughout the economy. But I think as we think through this piece, and just in general, the disruption caused by AI, trying to kind of frame it in both those ways and trying to understand in what parts of the economy is one versus the other dominant. I mean, I think that's a helpful framework for us to be analyzing this situation.

[Jack]

And what Satrini was talking about in this euphoria phase is AI boosts productivity, margins improve at businesses, those profits end up getting invested back in more AI. And then markets sort of extrapolate this into a boom that's going to go on for a long period of time. And that third one is important, that profits are funneled into more AI, because that'll come back to it later as we think about Some of the negative parts of this.

[Kai]

Yeah, I think it's worth mentioning that this piece is written as a kind of fictional retrospective. So it's like basically in 2028, we're sitting here and the economy has done X and we're trying to explain kind of X post what happened and why it did. And so that 2028 number, I think it's actually June 2028 to be precise. That date's pretty important because as we'll go through this, I think that the timeline is a really important component to how all this plays out. Right. People focus on the end game. Okay. When technology is at this new equilibrium, where will we be? Whereas I think the path dependency matters a lot. Whether we end up Getting there tomorrow versus in 30 years gives us a lot more time to adjust and for things to kind of equilibrate relative to this shock. So let's just keep that kind of two, two and a half year timeframe in mind as we go through this piece.

[Jack]

Yeah, that's really important. As we talk about the pros and cons of this, we'll talk about that more because that timeframe, like you said, is so important. But I want to start with. He sees the disruption first in software and that's, you know, a lot of people think about that because we've seen the disruption in software in the market right now. And I think part of it is this question, I think you probably know this better than me is like, how much do we have to worry about this idea that like we have all these pieces of software that tons and tons of businesses and people are using in the world? Like, how realistic is it for people to just say, all right, I don't want Salesforce anymore. I'm just going to build an in house solution to Salesforce and I'm going to claude code it. Like, how realistic is that? Because part of this, you know, this argument being valid relies on this idea that a lot of this stuff, you know, will just be able to be built with this technology. These huge software companies that they make will be able to be built easily with this technology.

[Kai]

Yeah. So I think the idea of insourcing enterprise software. So imagine you're like a tractor company or something, you make tractors and now you're going to, you know, fire Salesforce and build your own in house CRM. I think that's unlikely. I mean, there are business model reasons why firms outsource software to vendors who can specialize in the maintenance and kind of operating of those systems over time. But I think if you step back, I mean, they are making a fair point, which is that with AI coding tools like cloud code or Codex, software engineers are significantly more productive and therefore to produce a fixed amount of software you need a lot fewer workers and that could potentially put pressure on margins for these businesses. So I think it's fair to say that there is a disruption happening in SaaS stocks and software stocks and you know, that's why their prices are down 30% or so over the past few months. How that will affect the incumbents, I think the Trinity article doesn't, I don't think it captures it quite correctly. And this is why. So let's imagine, let's frame it this way, which is we can both agree that AI coding tools are pretty powerful and they allow software engineers to be a lot more productive and they push down the cost of producing software to, let's say, let's take the endgame argument and allow that to play out towards zero. Okay, so now imagine two types of companies. Company A is they're basically a software company where their only moat is their software. In other words, they produce really nice software and that's why people buy it, but that's the only thing they have going for them. Well then yes, of course it will be the case that if someone else can vibe code an app that looks exactly the same as what they have built, that their motion is basically gone. Whether that happens to be an in source, whether it's a competitors, sorry, their customers competing against them, or new startups that are kind of, or existing incumbents encroaching on their territory using their moat is basically gone. So I think we can agree that those companies are imperiled. But now imagine the second type of company which, you know, large enterprise, let's say a CRM company, right? And they have these big Fortune 500 companies as their customers. Well, it's not that they have the best user interface. In fact, most of these, you know, kind of larger incumbent firms have pretty poor user interfaces, especially compared to the startups that tried to compete with them. Right. These larger firms, their advantages are not their code. It's actually the other stuff. I focus a lot on intangible assets. I think a lot of intangible moats. It's the brand equity of these firms. You don't get fired for hiring Salesforce. It's the customer relationships, the distribution, the lock in, the switching costs, the network effects in some cases that make these firms so powerful. And if that's the case, it's almost the opposite where it's like, well, these guys will have their customers no matter what because of these other reasons. And hey, guess what? They used to have to spend all this money on software engineers in order to support and maintain and upgrade the software. Well, you can now cut your workforce in half, and that's helpful. Right. That means that their cost base is much lower and they manage to maintain their revenues in this hypothetical scenario. So I think you have to think a little bit more about the exact competitive dynamics of each industry, because I think both these scenarios are plausible. But. But again, it's going to be very case by case, depending on what the kind of, I guess, the competitive dynamics of each of these businesses.

[Jack]

Yeah, to your point, I mean, I don't think Ford is going to like, Vibe code their CRM tomorrow. And if they did, who's responsible for when this thing crashes or when it falls apart? Or like, there's a value to having a sales force sitting there behind the whole thing. Right.

[Kai]

I mean, I think that's the liability point, which is, you know, in all these professional services, whether it's, you know, hiring McKinsey to CYA or hiring a doctor or who can use their medical license to stand behind the work that their AI tools may produce, I think across the economy, you need someone to be able to sue. You need somebody to blame when things go wrong. And that's part of, I guess, the business model reason why this exists and why humans ultimately at least need to be behind the software or any of these tools being created.

[Jack]

But one of the things I thought was interesting in your point here about the idea that these things are way or much less expensive to build than they used to be is part of what he got into the article, which is the idea of software firms lose pricing power. So I am wondering, will there be big compressions in margins? Some of these software programs are really, really expensive. And if they're cheaper to make, if they're cheaper for other people to make, if they're cheaper internally to make, do we start to see margin compression inside of software?

[Kai]

Yeah. So I think we're asking two different questions. The first question is, will there be new winners and losers? Is there a distributional shift where potentially the incumbents lose against the new entrants or their own customers competing against them? My general view is that I think the incumbents are incumbents for a reason, and that reason is not necessarily their software skills. But I think you're raising a second point, which is the overall pie, which is the overall software sector, as opposed to to the Rest of the economy, obviously that's been a growing sector. Now that this disruption is happening, will that put downward pressure on margins? I think this is where the Jevons paradox idea comes into play. So this is the idea that when the price of a good falls, intuitively the first order consequence is that that should put negative pressure on margins. But there's the potential counter argument that maybe it's the case that there are so many apps that people want to create that they cannot because it's too expensive to create it. But if the price of Software goes down 90%, suddenly we're all building our own apps, or apps are being created to solve problems that otherwise wouldn't be software problems. So the total addressable market of software is actually far larger than that's currently being realized. Again, this is not an argument that I'm saying is definitely true, but there is some intuition behind it. For example, when Excel was created or whatever, imagine you were a banker and you said, I'm going to spend the entire week doing this one model. And then Excel comes out and you're like, wow, I can do this model in one day. Well, in theory you could cut your work hours down 80%. But what actually happened, as we know, is bankers still work 100 hour weeks. It's just that they make five models now. Right. So the fact that these tools create more productivity and efficiency may just lead to an increase in quantity even though price falls. And that kind of is an offsetting or even perhaps dominating effect.

[Jack]

Yeah, and the point he made in the article, and again, we're sort of counteracting some of this, but this idea that once they lose pricing power, these firms are going to have to lay off people, they're going to replace the people with AI and you've sort of got this doom loop that starts forming. And he talks about it with software firms, but then with the overall economy as well, which is everybody to compete as we start, you know, as this problem starts to come to compete, you have to use more AI and you have to lay off workers and then your competitors are doing it and it's just, just creates this circle that doesn't end.

[Kai]

Yeah, I mean, look, it's an interesting dynamic and I think he does a really good job laying it out. I think it's something that is worth considering. I think that there are mechanisms that could serve as brakes or governors on this loop actually forming in reality. But I think it's reasonable to say that we should take this, you know, seriously.

[Jack]

Yeah, and I was thinking about the Block thing, you know, we just talked about that just happened yesterday. You know, Block, I think, was it 40% of their workforce they're laying off?

[Kai]

Yeah, yeah, yeah.

[Jack]

And I do wonder like is that start, you know, because one of the things Jack Dorsey said when he did it is he expects everybody to have to do this or a lot of firms in his space to have to do it. And he wanted to do it all at once and he wanted to do it quickly. And you do wonder does, do we get some sort of situation where. Because one of the things I think we learned even before AI, like when Meta had their huge job cuts, there was not a huge negative impact on Meta's business or Elon Musk ran Twitter with whatever 20% of the workforce or whatever that was there. So there probably is a lot of bloat in these technology jobs. And so the question is, with AI and combining all that together, do we start to see more and more of these tech companies say I could do this with way less people?

[Kai]

Yeah, I think in his case there might be something specific going on in possibly over hired during the COVID period and they are just too slow to downsize. And I'm not saying that this is going to be him specifically, but we do know it is a fact that a lot of CEOs will use AI as a cover in order to do layoffs. They want to lay off employees anyways because either their business is struggling or they realize that they just overhire to begin with and they just blame AI. It's an easy escape Goat. But I do agree with you. I think that AI productivity could potentially lead to software job losses, assuming it is the case that there is not a corresponding increase. The Jevons paradox in demand for software. That could certainly be the case. I think the fact that the Dorsey layoffs happened the same week as the Citrine article is kind of interesting and it may be intentional on his part to use this again cover for something he knew he had to do at some point and just get it over with. But it's not clear to me that what happened at Block is systemic and that it's representative for the rest of the software and broader economy.

[Jack]

But we also have to keep in mind tech jobs are certainly important to the economy, but the vast majority of people are not working at Meta. I think tech jobs, it depends on how you define it, but I think it's something like 10 to 15% of jobs across the economy are in technology directly. And obviously when you get to knowledge workers, it's Much broader than that. But this is not like they make a lot of news. But the vast majority of people work for small businesses and do other things and they're not like working for Google.

[Kai]

Right, That's a good point.

[Jack]

So this next section was when friction went to zero. And one of the interesting premises of this, and I wanted to talk it through with you, was this idea that eventually we're getting to this point now where we are trusting AI effectively as like our buying decision maker, where instead of me deciding, you know, I'm going to buy this airline ticket from Delta or I'm going to buy this from this person or you know, I'm going to use Uber or Lyft. Like I've got this agent that's just making these decisions for me and this agent is basically just price. All they care about is price. They're just buying whatever's cheapest. So like what did you think about that idea?

[Kai]

Well, first of all, I think it's pretty cool. I mean I spent, yeah, if it ever happened, right. Several hours over the weekend booking flights, which was kind of a waste of my time. And so yeah, having a personal assistant that could just go around and do my shopping for me would be really nice now. You know, but obviously the question is, I guess more, you know, do we think that it will compete on price? Do you think it is the case that it will, you know, break down all these moats that other, that these companies such as like the bookings or doordashes or real estate agents is the other example. Right. You know, they, they, they have, I mean I thought it was actually quite interesting that they, that he picked DoorDash as like the canonical example as to how with these price shopping agents we'll be able to just intermediate DoorDash. But I think that almost perfectly, I think it's a really interesting choice of an example because in my mind DoorDash is almost the best counterexample to the idea that we can disintermediate this with AI. So what is DoorDash? It's not a fancy app, the DoorDash app. Even before AI, it wouldn't take that much time to put together a competitor that looked just as nice, if not nicer than the DoorDash app. The reason the DoorDash app is worth billions of dollars is because of the network effects that they built. They've managed to put together a three sided marketplace between drivers and restaurants and consumers, which is a really, really hard thing to do. I mean even just creating a two sided marketplace is really hard, let alone three. I mean, imagine trying to compete with them. I mean, Uber Eats was like really the only company that was able to do that. And they already had two sides put together. All they had to do was build that third one. So this is almost a perfect example of a case where it's not the software itself, it's the network effects and the other kind of intangible moats that make it as powerful as it is. So I thought that was kind of an interesting choice. That would be his example of what he felt would be disrupted.

[Jack]

Yeah, and he seemed to almost be making the case, Right. That the AI would start its own doordash competitor or something and like pay drivers more. It seemed pretty extreme that it would get to that point.

[Kai]

Yeah, that seemed a little bit implausible. I mean, I think look at the broader point that AI reducing friction could potentially change and disrupt firms who serve as kind of these intermediaries or middlemen in terms of matching buyers and sellers. I think that point is reasonable, but you have to be careful in terms of how exactly it plays out. I think even real estate agents are another interesting example where it's like people complain about the fact that real estate agents charge 2.5% to 5% on a transaction, but when the Internet came around, we have Zillow now and Redfin, that model still exists. So in a way it almost again serves as a counterpoint to the idea that technology and the Internet or AI, you know, are able to instantly displace a model that's been around for a long time. And maybe it is the case that people do value personal relationships and hand holding in terms of the actual purchase of, you know, a home, which will be the largest asset that they, you know, most people own.

[Jack]

Yeah, he ended up picking like some of the places that have been the hardest for technology to disrupt. Like he talked about MasterCard as well. Like that technology basically done nothing to that, like real estate transactions. The same thing. Like I don't know if it'll work or not, but obviously there's a bunch of barriers that have prevented that from happening if it hasn't happened already.

[Kai]

Right. And those barriers. And the barriers are not software related. Right. If the barrier is software related, I understand why we have a problem. I don't know if that is the case here.

[Jack]

I'm wondering, do you have any thoughts on the MasterCard thing? I'm just wondering because this whole idea of MasterCard being replaced with stablecoins, do you think long term that is something that might happen?

[Kai]

Look, I'm a fan of stablecoins. I think that it's an interesting technology and certainly we'll put pressure on these companies. Now many of these companies are themselves adopting the technology and that's another point that should be made too, which is, okay, so you're a salesforce, the CRM and you're getting pressure from AI native competitors. Why don't you just, you could also put AI into your own product and embed it there and use your distribution across thousands of large companies to cement your advantage. We talked about this. Jack Gemini and Google, they kind of came late to the game, but they had distribution through the entire G suite product. That gives them a pretty big advantage. So it's not always the case that incumbents will lose in the case of a disruptive new technology to the extent that they are themselves willing to adopt it. I think in the article they talk about how Blockbuster is a case that when the Internet came out they didn't adapt, whereas their expectation is that these incumbents will actually lean into AI because they have no choice but to do so. Now they think that that will be futile and kind of too little too late. But I'm not so sure because I do think that the incumbents do have advantages that smaller businesses or, you know, startups don't have.

[Ad Voice 2]

Predator Badlands now streaming on Hulu and Hulu on Disney plus here.

[Kai]

You're not the predator, you're the prey. Prey, prey, prey, prey.

[Ad Voice 2]

Critics are saying it's epic, stunning and breathtaking.

[Kai]

Many have come here, none have survived.

[Ad Voice 2]

Predator Badlands now streaming on Hulu and Hulu on Disney plus rated PG13.

[Jack]

There are also reasons, even if I'm using an AI to make my purchases, there's things I'm going to tell that AI to use besides price. So like for instance, if I'm on like Delta's Sky Miles program and I can go in that like fancy lounge before I fly every time, like my AI is going to be aware, don't just pick the cheapest flight and have me like on, you know, Aello or something like I'm going to go on Delta. So like that, that part of it is going to come in, it's not just going to be a price based decision. Even if AI is making it right.

[Kai]

And also the underlying data, where are they getting the data from? If it's like the inventory that a travel aggregator has in terms of hotels or any of these things, the AIs need to plug into an underlying data source and those things already pre built, it's a lot easier for the AI to just price compare Uber versus Lyft as opposed to say, oh, I'm just going to build my own network now and just cut them out and go straight for the random driver, you know, in Brooklyn or whatever.

[Jack]

And I think it's important to make this point now because it's one we'll talk about later too. But this idea that the speed at which technology moves is different than the speed at which adoption moves. And it's one thing to say, all right, the technology will be able to have this AI that'll make all my buying decisions. It's a different thing to say. You'll probably be one of the first people I know who will probably do that if it's available. My dad will probably not do that for a very, very long time. So part of the premise of this was in mass, people were using these AIs to make pricing decisions and that's probably a long way in the future.

[Kai]

Yeah, I agree with that. I think if you look at the 2028 timeframe for this, I mean that is very aggressive. Even if the technology is good enough by then to do that, which it's not yet, but let's imagine it were. It's always the case that adoption happens with a lag. There's the early adopters and then kind of people, the S curve forms and then your grandpa or whatever will come at the end. And on the business side in terms of enterprise adoption is even slower. It takes a very long time for even startups to change their business models, let alone enterprise companies. Anyone who's worked at a large company knows that it's a multi year process to make, even to switch in CRM or any of these things, just because there's so much institutional inertia against any of these things. So yes, you're absolutely right that the speed of this diffusion is not always the same as the speed of technology. A lot of folks out in the, in Silicon Valley are, you know, too, too in the bubble, kind of focused on the speed of the pace of technology, which is, you know, pretty impressive but you know, are kind of ignoring the friction around the actual adoption through the real economy. And that, you know, is, is a multi year process. I think it would be.

[Jack]

There also are some things, and I know you were talking about certain things you don't like to buy, but there are some things that people actually enjoy being part of the buying process. Like I was just taking my daughter yesterday and she was picking out shoes and she was like, she Loves, like going and looking at all the different shoes that are available. And like, there's a lot of things people aren't going to. They have personal preference, they like to have their hands on it. There's a lot of things people aren't just going to hand over to AI in terms of their purchasing decisions.

[Kai]

Yeah. And that's right too.

[Jack]

And the only other point I want to make in terms of the speed of the technology, and I'll put up this Gavin Baker tweet that you and I both saw. The other idea is, even if the technology is capable of moving fast, we are constrained here.

[Kai]

Right.

[Jack]

We're constrained by compute, we're constrained by power. And Gavin Baker's point here was like, even if this is a feasible outcome, we're a lot further out than 2028 or even the2030s in terms of when there's going to be enough compute for this to actually happen.

[Kai]

Right. So that's right. So even if technology were good enough and there was the will amongst corporate America and households to adopt this overnight, it wouldn't even be feasible because we'd require one, maybe two orders of magnitude more compute, which we just can't build. It's just not possible.

[Jack]

Yeah. In a separate thing. Yeah. He had another tweet he wrote where he was saying, my best guess is that we would need roughly 1000x more compute for the unlikely hypothetical scenario described by Citrini to be remotely possible. And the time it takes to get us there will give humans time to adjust and maximize the many potential benefits of AI. So that's what we were talking about before, which is what you mentioned at the beginning, which is the more time this takes to play out, the more time we're going to have for these adjustment mechanisms to kick in.

[Kai]

That's right, Yeah. I think the best case scenario is that, and it's kind of weird to say this, but the best case scenario is that we want some natural limiters and governors on the speed of the build out. If it happens too fast, it'll almost certainly end with a bubble like a boom and a bust, which can be destructive. The Internet still ended up being game changing. It just was very painful along the way. To the extent we can, various forces, whether it's compute constraints or institutional constraints, maybe unions and lobbyists, which generally people think about negatively, could actually be helpful in this case because it could help kind of put some brakes on too aggressive an acceleration of this process, which could actually lead to a lot of problems. We'll get into the Political issues later too. But yeah, a lot of unrest and displacement as you say, Jack, the more time we have to put into place and to adjust to the shock that is AI, the more likely we are to be successful in doing so.

[Jack]

So we won't get too much of the doom stuff here because obviously this spiral is out of control and it goes to the whole economy and then software's private equity's lent in a bunch of money to software and that gets worse. But I do want to put up this feedback loop thing because this idea that there's sort of a self correcting mechanism in the economy, the AI feedback loop, a non cyclical disruption, is the chart that he has in the paper. And this idea that there's this self correcting mechanism that exists and that sort of the brake on that self correcting mechanism is going to go away. And I think part of that is because the more this happens, at least he's arguing, the more there's an incentive for companies to use more AI, to lay off more workers and use more AI. And so some of the correction mechanisms that exist don't exist here because of that incentive to keep using more and more and more and more AI. Do you have any thoughts on that?

[Kai]

Well, okay, so he's saying the companies want to use more AI and they're laying people off. And then the problem is that when people are laid off, therefore they have no income and no consumption power. And that's what forces companies to lay off more people. So I think that link in the chain is interesting. I think that's something I think we should zoom in on because two things are happening here. So one is he's implicitly assuming, or I think he talks about it more later on, that these people who are laid off from a job as a software engineer at Google, that they'll either not work or they'll go drive Uber. Right? The second best option for someone who's making 200k as a software engineer is to go drive Uber. And I think that kind of ignores the kind of creative part of the creative disruption process, which is that all throughout history it's been the case that jobs have been destroyed, but others have been created. I think I was listening to that thing with David Autor and the other economists where they cite this thing from Autor, this is the MIT economist who says that 60% of jobs that exist today didn't exist in 1940. Right. So most of the things that you, I and other people are doing, like podcasting right now, that didn't exist as a job 80 or whatever years ago. And, and I think it's possible to say this time is different. It's possible. I can understand the argument that AI is different from past innovations in that it dominates what humans have historically been our best feature, which is our intelligence. So I can see that. But it is a strong assumption. I think it's worth just highlighting that is a critical assumption that you may or may not disagree with, but is kind of required for this to be, for this doom loop to be the case. I think the other key assumption that they're making here is one around the benefits of the deflationary benefits of massive productivity gains are not passing through to the consumer. So in other words, if it is the case that AI makes us X times more productive, that should in theory drive down the cost of all these goods. So the cost of consulting services down 90%, tax services down 90%, health, healthcare, pharma, go through the entire list, kind of the whole point here is that AI is so good that it crashes the economy. If AI is so good and the cost of goods fall 90%, then if you are a worker who makes say 50k a year and you still make 50k a year, but now you can buy 500k worth of goods, that's pretty good. So that doesn't necessarily mean that you'll be in a position where you feel like you need to cut back spending, thereby driving down the margins of of companies, thereby forcing more layoffs. So I think there's a couple links in this flywheel that are making very strong assumptions that again may or may not be the case. But I think it's worth thinking through and just interrogating each of these pieces so that again, it's a very clean narrative. The piece is very well written and very persuasive. But I think you have to just think through a little bit of what assumptions are being made underlying the narrative.

[Jack]

And this is one of the challenges too, because AI probably will, or definitely will create new jobs. The question is, it's hard to imagine when you thought of it, computer was a job at one point for somebody who actually did that. It's hard to imagine what those jobs are when you're sitting where we are today. Because a lot of the jobs were created by other technological revolutions. You didn't really see what they were going to be. We can see the hard job loss or the potential for the hard job loss and we can't see what gets created. So I think maybe that makes us too pessimistic. Some of the time. I don't know if you agree with that.

[Kai]

Yeah, I completely agree. I mean, again, as you say, it's easy to see the jobs that we know exist no longer existing. It's very hard to imagine the future jobs that will be created. I mean, it's hard to imagine the future. It's easy to think about the destruction that will occur in the world that we currently occupy. And so I think there is a bit of a bias there, whatever it's called, that does tend us to be a little more overly pessimistic when it comes to this kind of creative destruction of jobs issue.

[Jack]

How about the thing you mentioned about intelligence though and about this? This is different because this is effectively able to almost do everything a human being can do. And not physically yet, but maybe physically in the future. Does that make it different? Does that mean that it's going to be. Maybe the short term pain is going to be harder or it's harder to figure out what jobs can even be created because this can just do all the jobs. What do you think about that?

[Kai]

Well, look, I think the ultimate solution if that is the case, is government intervention, right? Is UBI or I think they mentioned in the article, we all get a share of sovereign wealth fund, so we're all capital owners. Now I know Trump's doing that already in a small scale. That's kind of the trump card, so to speak. That's kind of the backstop. We can pull it out if we need to. At the very end, if there really is no need for human work, I think that that's like again, an extreme assumption and assumption that will certainly take more than two years, will likely take more than a few decades, but who knows? I think it's important to recognize that that is always a solution in the back end if need be. Because in that world you just outlined, it's a world of abundance. There is so much wealth and so much productivity out there that, that we could all live without, we can all live very well without having to work. Again, there's significant issues around self worth and self actualization. Do people actually want to just sit there and be paid? But I think there are at least economic solutions to these problems if they do present themselves. Not saying they necessarily will.

[Jack]

And that's kind of the world we talk to, the Elon Musk's and the Marc Andreessens. That's the world they're seeing, right? And people like us are thinking like that's not even possible, but they're seeing a lot of possibility. They're better probably at seeing the possibility this can unlock than someone like me might be.

[Kai]

Yeah. And they're also focused on the end game. They're focused on what does the utopian dystopian future look like after this process is played out. I think for us as investors, we need to get there first. And that again will be likely more than a two year process. That is a multi year process. That is probably an entire investing career maybe to kind of work your way through this entire process of adjustment before you kind of need to sit think about these larger, I guess, long term topics.

[Jack]

So we've debunked a lot of the. Maybe we've got a near term massive catastrophe on our hands. But do you think because of how disruptive this technology. I would think to some extent the more disruptive the technology is, probably the better long term benefits, but maybe also the more short term pain in terms of we might have a period where we're adjusting here where we do lose a bunch of jobs and maybe there is a negative hit. Do you think that's a fair way to look at it?

[Kai]

Yeah, I mean the reason why so GPT also stands for General Purpose technology. And the reason why AI is such a powerful and transformational technology is because it's general purpose. It's not like a tool that only helps in one industry. It's a tool that helps across the economy. And then with robotics, literally even the physical world too. So extremely powerful, extremely pervasive, which is why there's so much upside. But you're right, it's also pretty much in that solution. In that scenario, pretty much everyone is exposed, the entire economy is exposed to disruption. And so yes, I think that is fair characterization.

[Jack]

Do you think this is sort of an aside, but do you think this is worse for wealth inequality at least in the short term? I mean you talked about the idea that the government may have to give people benefits. Like does do we end up, we're obviously not going to end up in a world where like Elon Musk is our overlord and like we all just answer to him because all wealth is accrued to him. But like you could argue, wealth is going to accrue to the people like behind this technology, maybe more so than your average person or workers because of this. And we'll probably have to make adjustments because of that. Do you think that's right?

[Kai]

Yeah, man. I think that's been a trend unfortunately for the past several decades. Like if you look at the data on like labor versus capital share value add it's been a pretty much a one way trend where whether it's globalization or just automation in general, and just traditional software technology workers or human labor, their share of the pie has been shrinking relative to capital for a long time now. And I think it's quite likely that AI accelerates that trend further. The owners of the AI models, or even on the human capital side, the folks who know how to operate the models as opposed to those who don't, there's going to be dispersion there too. So there's dispersion at all levels. If you're a tech worker and I'm an AI tech worker, there's going to be a massive dispersion amongst our incomes and then all the way down through the economy. So again, there's going to be winners and losers, unfortunately, as there are. And so I think that there is all else equal, more need for something to be done. I'm not saying it's like redistribution necessarily, but I think, and part of the reason, I mean to the credit of Citrini, part of the reason for this piece was to create this doom scenario with the idea that it would hopefully stem some sort of response amongst policymakers, economists and investors. How do we think about solving this problem? Let's make these issues very acute, even if we're kind of overselling the point, and then try to think through what are the solutions that could potentially come into place to help address and ensure that these scenarios don't actually transpire.

[Sponsor/Ad Voice]

Starting a business can seem like a daunting task unless you have a partner like Shopify. They have the tools you need to start and grow your business. From designing a website, to marketing, to selling and beyond, Shopify can help with everything you need. There's a reason millions of companies like Mattel Heinz and Allbirds continue to trust and use them. With Shopify on your side, turn your big business idea into sign up for your $1 per month [e-mail] SpecialOffer this

[Sponsor Voice 3]

episode is brought to you by State Farm. Listening to this podcast Smart move. Being financially savvy. Smart move. Another smart move. Having State Farm help you create a competitive price when you choose to bundle home and auto bundling. Just another way to save. Save with the personal price plan. Like a good neighbor, State Farm is there. Prices are based on rating plans that vary by state. Coverage options are selected by the customer. Availability, amount of discounts and savings and eligibility vary by state.

[Jack]

Yeah, to his credit he did say at the end, like we still have time to adjust. Like he was Saying like, this is. This is not some like high probability scenario we can't get away from. This is something. And I think these pieces, like, I know people bash these doom pieces all the time, but I, I do think this is useful for me. I know a lot more now about how this. I've thought a lot more about how this AI might play out. And sometimes throwing a really extreme scenario out there makes you think and sort of back into a middle of the road scenario. So I think for me it's been helpful in terms of as long as you don't take it as that and you don't go out and start panicking. If you take it as something that

[Kai]

the problem is that people actually did do that. Right.

[Jack]

No, that is the problem. The stock market, the Bloomberg headlines and.

[Kai]

Yeah, yeah.

[Jack]

Like, I'm just. As an aside, like, I've been surprised that like the. I said this before we started recording, like the number of people who have written responses to this. Like Citadel, we mentioned before wrote a response like Jeremy Siegel's out with a response like, this was something that hit a nerve with people because like some very big name people felt the need to sort of. Noah Smith was as well. Like, people have felt the need to combat this. So obviously this is, this is kind of a hot button issue for people.

[Kai]

Yeah. And in a way, that's the most valuable takeaway from this piece, which is, first of all, no one knows anything, that we're all trying to figure this out and the opinions can vary wildly on this topic. And second, the market's on a knife's edge. I think people are legitimately, because of this uncertainty around how AI will play out when they see something, they panic, which is not good, of course, But I think it does highlight this larger dynamic of where we are in the cycle.

[Jack]

I mean, I wish I had wrote my substack would have way more followers. I can only imagine the number of substack followers that have come from having a piece that. This is widely read. I mean, I think like a Fed governor. Right. Responded to this. Right. I think it's pretty crazy. I think Waller had a response to this. So, yeah, it's insane how far it got. But as we wrap up here, I just wanted to, for the last third of this, I just wanted to maybe go through the arguments for and against this because I think we've touched on a lot of them already. But I think it's helpful for all of us when we think about this, to think about what are the arguments in favor of something bearish like this, even though we all agree it's probably not going to happen this quickly. And then what are the arguments against it? So I think the biggest one of the arguments in favor of this is this is a different technology, and this is really the first technology that's replaced a human being. And if it's the first technology that can replace a human being, then we can have all kinds of problems related to unemployment and other things like that that we haven't seen with other technologies. So what are your thoughts on that?

[Kai]

I agree with that. I think that there is an argument that it's a little bit different from past technologies. And so just looking at the historical data isn't going to give you an accurate read on how things will play out. I think related to that is the fact that we have lived through a very nice cycle, especially in software stocks. Multiples are pretty high, profit margins are pretty high, and disruption can, when things are really good, disruption can only make things worse. Or to put it differently, there's a lot more downside than upside potentially to the extent we don't get things right. So I think that there is that piece as well.

[Jack]

It's funny how you sort of see, when you look at history, you see these big dominant companies, and at the time, people thought there was no way they would ever be disrupted. And you see that in the past and you know what happens. But then you can never see it, like in the current time, you look at the dominant companies and you're like, they can't be disrupted. But I think that's part of the lesson of this is like, so, for instance, Google search, like, I was like, you know, if you ask me, like 10 years ago, it's like, there's no way. Or five years ago, whatever, there's no way anybody's ever disrupting Google search. And then AI comes along and it's like, wow, Google search is like a threat. And the same thing with the software companies. You're like, salesforce is like the bulletproof company. Nobody's going to ever touch it. But we never anticipate what these, you know, these new breakthroughs could be that actually disrupt things that we think are undisruptible.

[Kai]

That's right. And I think going back to the whole point of the article was, is it possible to be an AI bull and think that AI is a technology will be so dominant yet to think that that could actually lead to a worse outcome for investors. I think that overall premise is quite interesting. And I've argued this in the case of the hyperscalers, where I said they're kind of inventing the technology, but they're kind of positioning themselves as utilities. Right. Which is not where the profits accrue historically. In a way, what Citrini is doing is making that same argument, but even more broad across the economy. And again saying that you can believe in AI as a technology, but not necessarily think that it'll be good for financial markets and for the economy in general. And again, I think that the argument's being made very strongly, but that general view isn't necessarily wrong just as a framing and is potentially an interesting way of thinking about things. Yeah.

[Jack]

And that gets kind of into the second argument I had here for his piece, which is the idea that AI will replace jobs rather than enhance workers. And I think that's maybe the big question here is how much of this AI is going to enhance workers and make us all better off and how much of this is going to take over for workers? I think that's a huge question we all have to try to answer and the answer is probably somewhere in between. Right?

[Kai]

Yeah. And I think again, this is something we've discussed, which is that a job is a bundle of tasks. Right. So on a day to day basis, I do a few different things. I send out emails, I like do some coding, I'll do some research, I'll talk to some clients, I'll hop on and talk to you. And if you think about it that way, then certain tasks that I spend time doing I no longer have to do or I can do it in a fraction of the time. Coding is a lot more efficient now. When I used to spend my time scrubbing data or scraping websites, that's pretty much a solved problem too. Other things I have to spend more time on, it's not saving me time talking to clients. In fact, all else SQL, I should spend more time talking to clients because that's one of the things that I as a human have a comparative advantage in. So I think again a combination of both substitution and augmentation, both at the economy level, at the job level, but also within the job level, at the task level where our jobs, whatever you want to call them, will look very different. Our day to days will look very different in 10 years than today because a lot of the things we spend our time doing now will go away and then new things will come into play or we'll just do more of certain things which isn't necessarily bad. That's just the nature of the world. And in many ways it's Good. A lot of things I spend my time doing that I can now automate away with AI are things I hated doing anyways. So it's actually kind of maybe a net benefit for me.

[Jack]

Yeah. And for me, just thinking about the podcast, there's so much stuff that I was doing manually, if you go back even a few years, that I'm not doing anymore. So I have no idea what that means for the economy, but I do know that it allows me, at least for now, it allows me to do tasks that I'm better at as a human and it replaces that stuff. So I think from that perspective, it's only a positive.

[Kai]

Yeah, Yeah. I mean, again, like, there are going to be certain jobs where 100% of the things that that person does is automated. I think that's where things get a little bit tricky, you know, so. And that those are. Those are the jobs that, you know, we need to be more careful about that. That's the. Yeah, I don't think. I think Citrini in his article basically made it sound like most jobs are in that category. I think that's a minority of jobs. But they still. But they are still jobs and there are still people who are in those positions. And so I think we need to be thoughtful if the adjustment occurs quickly, you know how we want to deal with that.

[Jack]

Yeah. I was just driving around the other day and I was thinking about, like, all the jobs you see as you're driving around. Like, you see a store, you see a plumber's truck, you see an electrician. Like, those types of jobs are not going to be disrupted by AI for a very, very long time, if ever. Like, so we lose track. Like, I think when we sit in the stock market and we talk about Google and Amazon all day, like, that's not the real world. I mean, that's part of the real world. But there's so many jobs out in the real world that are not going to be affected by this too much. Like, they will probably eventually, but not too much in the short term.

[Kai]

That's right. Yeah, I agree.

[Jack]

So two more of the arguments. For one, we talk about a lot, which is this idea that it's such a fast moving technology, but we have to keep in mind that adoption moves slower than that. And also we have to keep in mind, like Gavin Baker's argument, which is there's going to be constraints outside of the technology itself in terms of how fast it can grow. So I think we've covered that one enough, but I think that's really Important to keep in mind here because it might not disrupt as fast as people think.

[Kai]

Yeah. I think the other point to add on to the compute constraint point is that even if we don't max out compute as a hard limit, there's still going to be the issue of prices adjusting. Right. The price per token could increase just because supply and demand. If demand increases 100x and supply only increases like 10x, the price per token could get so high that even that just naturally equilibrates the market between human and non human labor. Maybe you end up with a point where you're like, all right, well for $0.01 a token I'd be willing to use an AI instead of using a human. But for $0.10 maybe that doesn't make much sense to do that because the cost of these tools can be quite expensive. I mean, you're talking about for some enterprises, they're spending tens of millions of dollars on anthropic, that's a lot of money. And maybe they're saving more money than that in terms of productivity, but it's a non trivial amount of money. And if we do end up running into constraints on compute and that jacks up the price of tokens, then that could just be a natural governor on the kind of flywheel that the treaty is laying out.

[Jack]

That's really interesting because I've been watching this week in startups with Jason Calacanis and he inside of his own business, he's gone crazy aggressive on this. He's got openclaw and they're doing all kinds of stuff. And one of the points he made is that this idea that he has to now balance the cost of tokens against the cost of employees. So it's not like this is costless. It's not like when you're using AI, there's not a cost associated with this. And he's running a massive cost on tokens. So it becomes like this balance against labor cost that you have to consider as you're considering what to allocate to which one.

[Kai]

Right?

[Jack]

Yeah.

[Kai]

I mean in the same way people route the different models, they say, hey, this is a pretty difficult problem. I want to use the highest end model for this one. And this is like a lower end problem going to outsource this to a cheaper model. You can add humans, I guess to that.

[Jack]

They're like the highest end model or something. At least the most costly model right now.

[Kai]

Right. Depending on the human, I guess.

[Jack]

Yeah, I guess that's true.

[Kai]

I'm only going to buy an alpha, not if it's me, Einstein's time or whatever, if I really need it, maybe

[Jack]

we'll all price ourselves in tokens Kai. Down the road the smartest people will price themselves the most and I'll be down towards the bottom pricing myself somewhere down there.

[Kai]

And it'll charge more for the best hours of your day. You're like ah, you know if I'm just like sitting around like on the, on my phone like you know in, on the train anyways, yeah, I'll do some like low end tasks but if you're taking up my 9 to 9 to 12, you know, Monday morning time period, that's going to be expensive.

[Jack]

So the last one I want to get into before we get into the arguments against this is this idea that the jobs that are going to be replaced are more high end and that those people do most of the spending in the economy. So like is that a risk here in terms of like a lot of other things that have caused maybe unemployment haven't necessarily caused it at the absolute upper end but you could argue most of the, many of the people that get paid the most are going to get the people like lawyers, accountants, you know, people in technology, people that make the most money. So is it going to be more economically disruptive because it's disrupting that high end?

[Kai]

Yeah, it's a really interesting point. I mean this goes back to the David Artur, another thing he wrote about was talking about how past, like the past 30, 40 years a lot of the disruption has been more like factory workers or clerical jobs. Kind of your middle class America has been hollowed out. Right. What's unique about AI is that it's going after the knowledge workers, it's the lawyers, accountants, doctors, financial analysts that are kind of in the crosshairs of disruption now. Yes, it is the case. I think the data that suggests that a significant percentage of consumption is driven by the kind of high end of the market that just makes sense, that's just mathematically the case. I think one countervailing argument could be the marginal propensity to consume. So in other words the marginal propensity to consume is higher amongst less rich people. Like if you're a lower income person and you lose your job then your spending goes down proportionately with that. Whereas if for a very high end high earner you lose your job, yeah you might take down your consumption a little bit but you got plenty of savings to buffer that. So I think that might be an argument to partially counteract maybe not fully that effect. But yes, I Mean, I do agree that this will be a unique, this is a unique situation because relative to the past several decades, given that the jobs that are most at risk are your kind of white collar jobs.

[Jack]

Now arguing the other side. Now, in terms of why this might be wrong, one of the things that we talked about this already is the speed of adoption. But I want to put up this chart from citadel. They've had PCs and the Internet and other things and they talked about this idea of an S curve. So what do you think in terms of the adoption of this and the speed of adoption of this relative to other technologies? I mean, should we expect a similar S curve type thing or should we expect this to be markedly different?

[Kai]

I think we should expect faster. I think if you look at the history of these S curves, there's some data on this. They've been compressing over time. So in other words, 100 years ago a new technology would come around and it would take, you know, 15, 20 years for it to diffuse through the economy. And then 50 years ago it would take, you know, 10, 15 years. Right. And these S curves have become shorter and shorter for a variety of reasons. But that seems to be the trend now. Will it be two years? I think that's. I still think that's aggressive, as I was saying before. But yeah, I mean, I think that AI is a powerful technology and there's competitive reasons why in a capitalist system it should be adopted. If I adopt it and my competitors don't, then they're going to lose market share. And so there's a forcing function to make them also try to adopt it too, otherwise they kind of fall behind. And I think that pressure does exist in the economy.

[Jack]

What do we know? And this is another chart from the Citadel about the percentage of people using Gen AI, the percentage of people using gen AI at work. They're saying that's not inflecting. So, and that's part of the whole thing is like we, those of us that use technology, those of us that are in these industries, we think differently about these things because we use them every day. Like your average person out there. Like my parents are saying, you know, whatever, the plumbing's not working. They might ask a question to AI about it, but it's not going to overtake their entire life. They're not going to install OpenClaw or whatever, like on the server they've got in their house, which they don't even have. So like, how do you think about this in terms of adoption, in terms of like people using this in their everyday life.

[Kai]

Your average person. Yeah, look, I think everyone is different. And again, we do live in a bubble because we are kind of quantitatively minded. We're kind of more tech savvy, I guess. But yeah, I mean, I would expect it to take a long time and it may even just require kind of a generational turnover and handover of companies and such for these things to be pushed through. Right. It is the case that even Gen Z, I'm assuming, is more tech savvy than our generation, you know, and that's just the way these things play out. And then again that, that the human like preference for just kind of doing things the way it's been done, you know, with, with your father or whatever, like that's just kind of another, another natural limiter on, on the speed of, of the fusion.

[Jack]

So the, the last argument here we'll talk about is this idea that the labor market's not breaking. And we, we talked about that before, but we are not seeing this. Citadel made that point. We're not seeing anything in actual labor market data yet to indicate anything close to this. Now that doesn't mean labor market data is backward looking. It's notoriously not great. We know all those things. But it is a fair point to say even if you're looking at job postings and stuff, you're not seeing anything yet to indicate that this massive disruption is coming.

[Kai]

Yeah, if you look at job postings, you're definitely seeing a increase in the number of AI linked job postings. Of course, companies are looking to bring in AI human capital to kind of meet and stave off the disruption or take advantage of the opportunity with AI. Now, is that leading to a corresponding decline in jobs elsewhere? Is it crowding out software jobs or just traditional jobs? I think so far the answer is no. I mean, at least we're not seeing it in a statistically meaningful way in the data.

[Sponsor/Ad Voice]

This episode is brought to you by Indeed. Stop waiting around for the perfect candidate. Instead, use Indeed Indeed Sponsored jobs to find the right people with the right skills fast. It's a simple way to make sure your listing is the first candidate. C. According to Indeed data, sponsored jobs have four times more applicants than non sponsored jobs. So go build your dream team today with Indeed. Get a $75 sponsored job credit@ Indeed.com podcast. Terms and conditions apply.

[Ad Voice 2]

It's not just something you made. It's the privilege that you get to work with your hands. It's building something that serves a purpose, proof that you have the grit to keep Going at Timberland, we understand you take your craft seriously and we do too. Which is why our products are built to the highest quality. We put in the work so you can perfect yours with purpose, in every detail, and crafted with intention. Timberland built on craft. Visit timberland.com to shop.

[Jack]

So as we wrap up, like, I mean, what do you think our takeaway is like? I choose to be pretty optimistic about this, but I also understand, and I think what this piece was good for, for me is understanding that the rapid adoption of AI is going to have negative consequences. It's not just going to be, you know, the world's going to be beautiful and everyone's going to be happy in the world we live in. I mean, we already have issues with income inequality, we already have political issues like this is not going to go smoothly probably. But I choose to believe that in the long run this is going to be like the other technologies and this is going to be something that's going to make our lives a lot better. Like, how do you think about that?

[Kai]

Yeah, I think any new technology has winners and losers. It creates disruption and creates dispersion inequality amongst companies, amongst stock returns, amongst individuals, amongst workers. I think two things, I guess, which is at the aggregate level, look, I think it's likely to be a net benefit. I think for all the destruction that occurs, technology is generally net beneficial for society. So I'm optimistic as well that AI will help create more abundance and create more wealth. Keeping in mind that we have a lot of challenges too, like demographic challenges, for example, like our workforce is shrinking. We kind of need this in a way to offset some of the headwinds we're facing as a society. And so yes, I think I'm excited about the economic investment applications or implications of AI. And what I'm focused on more is more just kind of trying to figure out on a cross sectional sense. So within the economy and within the stock market, which sectors which companies are best positioned to survive and thrive in this disruptive period as opposed to may go away, they may be kind of you're falling knives. So I think as an investor, just to make this more investment related rather than focus on is this going to crash the market 50% or not, which I think is a really hard thing to do. Market timing is really hard. Macro forecasting is really hard. My thought would be trying to say figure out within each sector, within the across sectors, which are the companies that are well positioned and trying to kind of concentrate in those stocks and kind of steer away from the ones that might be value traps.

[Jack]

Yeah, I think that's great advice. And in my personal life, what I'm doing. And Rob Arnot had this thing when he came on the podcast, he said that, he said to all their employees, which is basically, AI is not going to disrupt you, but somebody who uses it is. And that's the way I'm trying to do it. I think all of us have to look at this and say, use this in every possible way we can to make ourselves better. Because I do think if we look forward years in the future, people who just ignore this and don't do anything with it are going to be at a huge disadvantage against people who are at the front end of the curve and are really trying to use it in the right way.

[Kai]

Yeah, that's right. So what we're saying is that your investment portfolio should tilt towards AI positive companies, and your human capital portfolio should also. You also want to steer your human capital towards things that will likely be beneficial in this AI world, whether it is protected things. So like human empathy, let's say, or it's trying to embrace the technology, get really, really good at it, so that you can be the guy at your company who's like the AI whisperer and your boss who's like 80 years old or whatever is like, yeah, I do things this way. But you're the guy who can help translate all this amazing insight from AI into my team. So I think trying to lean into understanding and being at the frontier of technology would be pretty cool.

[Jack]

So I guess the idea is, whether I want to or not, I'm making a massive bet on AI with my future. So hopefully it pays out for me. Thank you, Kai. I appreciate this. This is a lot of fun. I enjoy doing it.

[Kai]

Yeah, this was cool.

[Jack]

Thank you for tuning in to this episode. If you found this discussion interesting and valuable, please subscribe on your favorite audio platform or on YouTube. You can also follow all the podcasts in the Excess Returns [e-mail] if you have any feedback or questions, you can contact [e-mail] no information on this podcast

[Kai]

should be construed as investment advice securities discussed in the podcast may holdings of the firms of the host. Close your eyes, exhale, feel your body relax and let go of whatever you're carrying today. Well, I'm letting go of the worry that I wouldn't get my new contacts

[Sponsor/Ad Voice]

in time for this class. I got them delivered free from 1-800-contacts. Oh, my gosh, they're so fast and breathe oh, sorry. I almost couldn't breathe when I saw the discount they gave me on my first order.

[Kai]

Oh, sorry, not namaste. Visit 1-800-contacts.com today to save on your first order, 1-800-contacts or their clients.