AI Managers, US vs China Analysis, and AI Art w/ Byrne Hobart

Hey, Upstream listeners. Today, we're releasing a conversation I had on Turpentine's show, The Riff, with my friend, Byrne Hobart. We analyze the dynamics of the U.S.-China trade relationship, the potential for AI agents to replace managers in companies, the discourse around AI art, and more. Please enjoy.

You wrote about how there's a bit of a realignment here in how China is wanting a bit more free trade and the U.S. is the one that's a bit more protectionist. What do you make of this realignment?

Oh, I think, I think it's really positive. Not, not so much like free trade in general has been, has been very beneficial on net. There are certainly some costs, certainly some trade-offs, certainly some things that the U.S. in particular could have done differently, but But, you know, overall, overall, just from a utilitarian kind of standpoint, the specialization is helpful and it seems to be the most effective way to get countries out of poverty, et cetera, et cetera. But the main reason to be more near-term optimistic about this is that for China in particular, participating in trade as a way of legitimizing the state is a relatively new part of their political mix. And I don't know that it's one that will last forever. So I've, I've always been kind of paying attention to whether it seems that the CCP views political repression or making people in China richer as better avenues to, to stay in power. And to the extent that they do view trade as the right thing to do, that's, that's a whole lot better than the alternative, both for people in China and for people elsewhere. So it, but it's also just, it's just kind of funny. It is kind of funny that the chairman, you know, the, the ruler of of the longest lasting, I believe, communist ruled state is this global champion of capitalism who's meeting with CEOs and telling them, please continue doing more business, especially business with other countries. They've, they've come a long way since Mao. So I, I like that. It's just, it's one of those funny things. It's like, I, the fact that there is an institution called the Ho Chi Minh Stock Exchange, I think it's just hysterical. I think it is, it is this total victory for, for the idea of capitalism that we have these communist insurgent war heroes and we memorialize them by naming a stock exchange after them. You know, eventually, and I guess, you know, it is a testament to just the fact that trade, trade works does not necessarily have to be completely free trade, but just a system where countries can import some of the things that they need. They manufacture things the rest of the world wants. Everyone is able to specialize a bit. And ideally a system where countries do choose to do choose to retain some control over their futures. They do choose to keep some critical industries onshore, even if there's a temporary economic case for moving them offshore. Like, I think, I think that's a better world than a world where what your country produces is what your country gets. And where trade is, is rare and is mostly high value goods consumed by elites. So like that, that system does exist in some parts of the world. There are some poorer countries, like for. For a lot of a lot of developing countries, when they're growing and they have export-led growth, one of the complaints they get from their trading partners is that these countries are intervening to keep their currency cheap because that keeps their goods competitively priced. And sometimes they do that very directly. Sometimes historically they've had fixed exchange rates. Sometimes you have places like Taiwan where they they nominally don't, but there's there's been some interesting forensic accounting on the financial statements of Taiwanese the Taiwanese Central Bank and the financial statements of their big life insurers that indicates that. Implicitly they are buying dollar-denominated assets in order to keep the Taiwan dollar cheap so that their exports remain competitive. So like, that's, that's the usual thing that you see, but there's this other pattern in countries that don't develop where they actually keep their currency artificially expensive. They keep it artificially highly valued. So it means they can't export very much, but it does mean that when they're purchasing luxury goods as imports, they actually have a lot of purchasing power. So it's, it's the case for a lot of countries that are natural resource exporters where that's their main business and where the state is kind of either the state is controlled by natural resource interests or the, or the state is kind of exists in part to extract some of the benefits of that. Like in a case like that, you know, it's very hard to have an exchange rate that makes oil a bad product to sell if you have a, if you're a country with a lot of oil. So your, your economic policy at that point can be, okay, conditional on us having a lot of oil revenue, conditional on some of that oil revenue making its way towards towards people who are connected to the government. How, how much, you know, expensive property in London can they buy? And how many, how many nice cars can they get? Things like that. So yeah, you, you do see, you do see both of those, those kinds of patterns show up in, in terms of how countries set up their exchange rate system.

Do you think this posture that we've been describing the U.S. and China having is a temporary posture, as in, you know, this administration, or do we think this is likely more of a medium-term I, I think it's best to bet on it being temporary.

I think that tariffs as they're being rolled out now are just not going to be all that effective for achieving America's economic goals, but the costs are probably going to show up sooner than, sooner than the benefits. And I, I don't know if that just means that we, we will just have tariffs for a while and prices will be a bit higher. The U.S., I think, would be less affected than a lot of other places, but especially, especially because In the event that you have global economic slowdown, that's partly caused by slowing trade. The U.S. U.S. consumption actually becomes a more important stabilizing force for the rest of the world, and so other countries may just have to produce things and sell them to Americans, and we may just end up buying them. Like it's it's a kind of messy messy solution there, but I like that is that is a case where it is a worse scenario than what you get without just blanket. Fairly high tariffs, but maybe a better scenario than some of the counterfactuals that you could dream up. But yeah, I do think that probably there's still some element of bluffing or some element of retro— what will retroactively be seen as bluffing in the tariffs. But it is, as we've discussed before, it's really hard to talk about bluffing and the level of bluffing as this stable quantity, because literally every time you speculate that the tariffs are a bluff, you are making it Slightly less effective a strategy to actually do the bluff, you know, actually do threaten tariffs as a pure bluffing mechanism. And by the same token, if people do think that okay Donald Trump absolutely does not want to just put a 20% tariff on every single product coming into the United States, that would be crazy, and therefore we should assume that this goes away. Well, if you do that, and Trump knows he's going to be in power for a while, and he also has this. has this vision and maybe the vision is not 20% tariffs specifically, but it is a bit more of a protectionist vision and a bit more of a vision where other countries are doing things. They're, they're forming more of their economic policy in light of things that the U.S. tells them to do. Maybe in that scenario, he says, okay, we are doing tariffs because people have to not think that this is a bluff. They have to not think that they can make some tiny concession and the tariffs won't be a problem.

Howard Lutnick on the All In podcast was talking about how If you have tariffs combined with tax cuts, that should stimulate a bunch of growth. Does, does that make sense or seem plausible to you? Or how do you think about that?

There are probably configurations where that's true. I think the case where that has to be true is a case where the U.S. has a lot of investment in manufacturing capacity that we could make. We don't make it because there's too much competition from foreign producers and therefore we just import the goods instead. So you could combine those, if you combine those 2 and suddenly A U.S.-based manufacturer does actually have a viable business selling to Americans and maybe exporting globally, although that's gonna be harder if there are retaliatory tariffs. So let's let's stick with okay U.S. manufacturer does decide that they're going to build a factory because now they they don't have to worry about competition from China or Germany or wherever else. So they do it, and then we that soaks up some surplus workers who would otherwise not be working, and they they do end up producing things and so on. Yeah, there's. There are scenarios where that does work well. I think in a scenario like that, you're probably also betting on the spillover effects of if we do have more of the supply chain in the U.S., then U.S. companies can iterate a little bit faster. And, and maybe it does make sense to assemble some, like there are some products where it makes sense to assemble them because more of the components are being built in the U.S. and you can source them more quickly and you're operating under a consistent legal structure, et cetera. But I think in, in a case where that's the pitch, it's probably an even better pitch to actually choose which industries we want because there are some industries like it just doesn't make sense to pay Americans American wages to make t-shirts. There's just not, not really, not really a plausible world where we do that, except a world where there's some very capital intensive way to automate almost all of it. And we just, for whatever reason, want the t-shirt machinery all to be in the U.S. and, you know, the, The one technician who's just looking at the machine and seeing that all of the, all the display lights are green. And so the machine is working fine. Like, yeah, that is, that is a job that we would get, but that doesn't really move the needle on bringing manufacturing jobs back. It does move the needle on bringing some manufacturing capital back. But I think if you, if you look at the space of industries where the U.S. either A, could be competing more and isn't, so things like chips and a lot of the equipment that goes into chip fabs, I think that is a case where, yeah, the U.S. actually is plausibly globally competitive in that space. We have been within living memory, a major producer of those products. But the, if you have non-targeted tariffs, then that doesn't really work very well. Chips, chips are a weird case though, because they, they do have very high incremental margins. And so tariffs are just less of a consideration. It's the cases where the tariff actually exceeds the end profit margin that the supply chain expects to get. So, which could be the case for some low margin products, like in a case like that, tariffs do have a really big impact. Everyone has to respond to them, and they they'll often respond to them just by raising prices because like a lot of this stuff, a lot of the the imports especially from China, and this is less gets less the case every year. But China does still have this huge manufacturing base and very low value added products that are just incredibly cheap, and those are products where they're still just absurdly cheap even if there's a higher tariff. Like there's that famous chart of the century that AEI likes to put out, which is just tracking. Cumulative inflation since 1990 in various categories. And the way they usually cast it, which is true, is that you see very high inflation in very regulated products. So anything that touches real estate, education, healthcare, things like that, prices have been rising for a long time. Actually, a lot of those prices have leveled off since then. So it's not as extreme as it used to be. And then if you look at things like toys and furnishings and electronics, then you see this rapid decline in prices. And that, you know, these are less regulated sectors, but these are also sectors where they're tradables. And that's the other thing the chart is showing is that if you have a very rich country, you should expect prices to go up disproportionately for non-tradable goods because you, if you're buying a haircut in America, like you can't outsource that haircut to a cheaper labor market. You can outsource the labor to someone whose labor is worth less than yours. And most of us do that, but you, you can't do the full outsourcing thing, whereas If it's, I don't know, a new action figure, like that's just going to get built in wherever the cheapest labor is. There, there isn't a lot of complexity to building things like that. And so they just tend to go wherever they're cheapest. So I think in, in, in categories like that, like there's just, there isn't really a lot for anyone to do except raise their prices and accept that maybe on the margin, there is a little bit less consumption in these categories. I did think about the chart of the century a while ago when I was considering what kind of allowance to give my kids and the allowance situation did is just in constant flux. It is, it's a lot like Trump's trade policy in a way. And has a lot of the same political dynamics. We, we're, we're always unsure about the extent to which we're making a really hasty decision or doing something as a negotiating ploy or making a deliberately bad choice entirely to preserve our ability to bluff in the future. But regardless, one of the things I thought about was the child consumption basket is very heavily weighted towards toys and electronics. And these are the things that have actually had the most deflation. So If anything, parents should actually be giving their kids, like modern parents, parents my age, their kids actually have more purchasing power for a dollar than I did when I was a kid. So parental, like the allowance should actually be much lower in nominal terms for modern kids than it was when I was a kid. But my, my kids really don't appreciate the nuances of this kind of macroeconomic argument for, you know, especially, you know, for that or for Here's why the tooth fairy is not as generous as you may have heard from other sources. You know, the sources could say what they want, but the quarter under your pillow doesn't lie.

We'll continue our interview in a moment after a word from our sponsors. There's a growing expense eating into your company's profits. It's your cloud computing bill. You may have gotten a deal to start, but now the spend is sky high and increasing every year. What if you could cut your cloud bill in half and improve performance at the same time? Well, if you act by May 31st, Oracle Cloud Infrastructure can help you do just that. OCI is the next-generation cloud designed for every workload, where you can run any application, including any AI projects, faster and more securely for less. In fact, Oracle has a special promotion where you can cut your cloud bill in half when you switch to OCI. The savings are real. On average, OCI costs 50% less for compute, 70% less for storage, and 80% less for networking. Join Modal, Skydance Animation, and today's innovative AI tech companies who upgraded to OCI and saved. Offer only for new US customers with a minimum financial commitment. See if you qualify for half off at oracle.com/turpentine. That's oracle.com/turpentine. Every time I hop on the 101 freeway in SF, I look up at the various tech billboards and think, damn, Turpentine should definitely be up there. And I bet I'm not alone. In a world of endless scrolling and ad blockers, sometimes the most powerful way to reach people is in the real world. That's where AdQuick comes in. The easiest way to book out-of-home ads like billboards, vehicle wraps, and airport displays. The same way you would order an Uber. Started by Instacart alums, AdQuick was born from their own personal struggles to book and measure ad success. AdQuick's audience targeting, campaign management, and analytics bring the precision and efficiency of digital to the real world. Ready to get your brand the attention it deserves? Visit adquick.com today to start reaching your customers in the real world. Again, that's adquick.com.

Parenting classes should include, you know, some, some basic macro to like get better, better deploy accordingly.

Well, like part of what you want when you're arguing with kids is you just want the ability to deploy overwhelming argumentative force. I first read about this in the context of, it was a book about Enron that was talking about Enron's COO, Jeff Skilling, who was kind of the architect of a lot of what Enron did. Including a lot of the good stuff, but also a lot of the bad stuff. And he, he came from McKinsey and someone was describing him as this prototypical McKinsey consultant where if he doesn't know the answer, he's just going to recite in perfect, complete paragraphs, endless statistics about every little data point that implicitly supports his argument. And he's hoping that if he talks fast enough, you won't be able to trace this back and say, wait, those numbers, those are accurate. And it's really impressive that you knew them to that many significant Significant figures, but they don't actually have any connection to what you're telling me we should do. Anyway, I, I thought of that as like actually a decent strategy for arguing with the kids. It's just, if it sounds like you just know a whole lot more about this process and that you will just bore them to death if they want to argue about it, then sometimes, sometimes they give up.

That might work with adults too.

I feel like I've seen it happen.

The, let's say we're back to China while we're on the topic. There's a lot of narratives around China and sort of their sort of medium term. People talk about the demographics, people talk about sort of their technology advancements, people talk about sort of their precarious sort of food and energy situation. People talk about their sort of credit crisis or real estate bubble. What narratives around China do you think are overrated and which narratives do you think are either underrated or correctly rated in terms of the doomsday or big challenges? that they're facing?

So I would say demographics, probably very slightly underrated. Just, it is very easy to underrate how much the demographic situation in a country affects what the optimal balance sheet is for the country as a whole, as channeled through households and the government, to a lesser extent companies, but companies, companies don't seem to change their balance sheet policy all that much in response to demographic factors. But your basic, the basic way to think about it is that young people do lots of consumption. When they save, they should be putting a lot of money into equities, they do end up putting a lot of money into real estate because that one step of household formation and having kids and things is often owning a house, but required. And the cultural norm for that might actually weaken over time, but it is, it is kind of a default, but housing, housing is this really interesting asset class where houses tend to last for a very long time and you, you need them when people want to have houses. And so you can't, It's not really economically optimal for people to save up enough money to just buy as much house as they want. What you'd have is a lot of people living, living with their parents until they're like 45, and then they finally buy a house and it's perhaps a smaller house than they really want to live in. It is a better system to have people save up enough to borrow a lot of money to buy a house. And taking the other side of that borrowing, you have older savers who should not be putting most of their money into equities. They should be putting a lot of money into fixed income and Because they're, they're going to be alive for a long but uncertain period of time, they tend to, it tends to be best for them to have very long duration income sources. So you have, you have this, this older generation that wants to buy duration. And then you have this younger generation that because they're buying an asset that lasts for a very long time, they want to sell duration by borrowing for say 30 years. And so you can, you can intermediate between those 2. And it works really, really well if those 2 sides are really balanced. And then if you have a surplus of young people, as in the baby boom, it often pushes interest rates up when they get old enough to actually start buying houses. That was, you know, the big, big headline component of 1970s inflation was the oil crisis and then all the ripple effects from the U.S. having to rearrange a lot of industrial supply chains not to rely on oil for power, but then, and also to be more, more careful about using oil for transportation. But then you also had a lot of baby boomer household formation. And then you also have this rise in divorce, which has this really interesting effect in wealthy countries where if you have 2 households, they just have a large consumer durables purchasing footprint. So every time a couple gets divorced and they each want to maintain a kind of similar standard of living, now you have 2 households that each need dishwasher and a dryer and an air conditioner and all these kitchen gadgets. And of course they need a car. So it does end up driving a lot of one-off spending on consumer durables and Those can be very inflation sensitive because the, the manufacturers of those durables, they can recognize that sometimes demand is really high. Sometimes demand is really low. And if you're running a manufacturing business that is selling some asset that has fairly cyclical demand, you pretty much, you should be thinking to yourself that if you are never at the point where you would be selling more, but you just can't find the labor and you just can't run your factory for that long. Like if you never run into that point, you are overproducing. On average, and what you want is sometimes your factory is underproducing, sometimes you're even losing money, but then during boom times, you have some pricing power and you can produce flat out. And there's just no end to how many cars and appliances people will buy. So that all of that is to set the stage for saying that when you, when you have the opposite phenomenon, when you have an aging society, what you have is a lot of people who are looking for long duration assets and simply can't find them. So that tends to push interest rates down. And when interest rates are down, that means that assets that have some stream of cash flows over a very long period tend to go up in value. And if you think of a house as a financial asset where it basically pays this continuous dividend in the form of your, sometimes we call it owner's implicit rent, or sometimes you could just think of it as the value, the equivalent rental value of that house. Like that's, that's a very long duration stream of cash flows. So its value also goes up. So you can end up in this pretty uncomfortable situation where those assets are being bid up. They're out of reach of a lot of people, and it's mostly or entirely driven by demand from older savers. And just anytime, anytime you have a case like that, you, you have an economy that's actually kind of hard to manage. We, we have one example of a country that was able to get through an aging crisis in, in all arguably pretty decent shape, which is Japan, where they started aging a lot earlier than many Western countries. And the U.S. is still, still, still a fairly young country and is not aging as fast as many economic peers. But Japan did have a quasi, like a sort of drawn out banking crisis that could have been a real crisis. And that probably would have been better for them. But instead, they had this very long period of economic stagnation in terms of total output, but If you break down the numbers and just look at real GDP per working hour, Japan and the U.S. were actually pretty much even. There was just, there were just a lot fewer working hours, both from older people retiring and then from younger people taking a while to start fully participating in the workforce. And in an environment like that, you can, you can see this weird thing where rates are low, but now one of the things that drives housing prices is just there aren't really opportunities that push people to live in particular expensive places. And then Japan is, it's also. Uniquely weird, and it is not a good guidebook for any other country because Japan has this very unique property market where houses tend to depreciate a lot faster than they do in the rest of the world, and they tend to just get torn down at the end of their useful life and then rebuilt. So Japan actually turns over its housing stock really, really rapidly compared to the rest of the world. So Japan is this one place where housing is not this long duration; it's not nearly as much of a long duration asset that has corresponding long duration liabilities associated. With it when people buy it. So there is not really, we just don't have in modern history that I know of, we don't really have an example of a country that went through shrinking population under modern, under modern technological constraints and with a fairly modern economic system. It's just very unclear exactly how we would manage that. My guess is if I were a Chinese policymaker and I were trying to figure this out, I would be offering very large down payment subsidies to new families buying first homes. And I would be trying to run a larger budget deficit to basically soak up some of this demand for excess savings. I'd probably be shutting down some banks, recapitalizing them, and just getting, trying to get things to the point where if you're looking at a given bank, you assume that the bank is solvent. But this is a case where you can, I think it's easy to overrate the extent to which China is actually at a risk of a real financial crisis because going back to that regression point, They don't really need to have bank runs. Like they, they, they do not need to allow people to actually line up in front of a bank and ask for their money back. And they, they don't need to allow a lot of other normal market behaviors that happen when you have financially dubious financial intermediaries where people stop giving them short-term wholesale funding, et cetera. Like China can just tell banks to keep one another alive when those banks would otherwise naturally fail. And they can have a more orderly wind down of different parts of the financial sector. So they also, they also have some room to just restructure some of the local government obligations because a lot of those are not on the balance sheet, but they are pretty real obligations. And that seems like a case where China, China as a country issues a lot less long-term government debt than many other places of comparable size, or, you know, they, they issue a lot less long-term debt than you'd expect given the size of their economy and given how interconnected it is with the rest of the world in trade terms. And one pretty good excuse to do that would be recapitalizing some of these local government financing vehicles that are just insolvent, won't be paying back their money, and that are thus making some of the lenders to them also potentially effectively insolvent. So I do, I do suspect that they will, they will get through okay, but I think it's, it's totally reasonable to expect something akin to the Japanese lost decade, just with Chinese characteristics.

Fascinating. The going back to the original sort of China, that's a great. Is there anything that you think is particularly overrated as a as a concern that we haven't yet got into?

Aside from financial crisis, not really. I think it is it is interesting to watch Chinese companies move up the value added, you know, move up to more value added things, and it's certainly really interesting to see that happening in electric vehicles where they've done really well on. On the angle of mass producing quite cheap cars that seem quite good. And also building, building good software for these cars. And now they have fast chargers, apparently. I, that is, that is one where I would like more information before I decide that that's as transformative as it sounds. But, you know, my, my, if I had to bet, I would bet that that's pretty real. And so pretty big deal.

That makes sense.

I want to segue to a couple other things you wrote about.

One is we, you looked at sort of the, the bottom of organizations or sort of more junior employees and asked sort of, you know, wrote about how to think about this idea of like, is AI going to replace a lot of junior work? Why don't you talk about that?

Yeah.

So the main way I wanted to turn that around was to say, what if AI actually replaces a lot of senior people? Because if you think about the, the junior senior setup, the idea is that There's some grunt work where as long as you know exactly what you're doing, you can just crank out whatever this is. And it's fairly straightforward for someone to look at the output and say, yes, you did a good job, or no, here's what you missed. And then the job of the senior person is basically to look at large projects and break them down into discrete tasks that are scoped appropriately for more junior people who, A, have less context on the overall project, and B, have less context about the industry. But LLM just has a lot of context generally. It knows a little bit about a whole lot of different things. And so you can ask it to do things like, give me a checklist. Like, I'm a junior developer. Here is the large project I'm building. Please break it down into some specific steps that I could work on. And then, you know, you can ask it to review the code and not just say, does this seem like it works or help me fix this bug, but also ask it things like, Okay, if this is, you know, I'm building an e-commerce platform for some reason, I'm not using Shopify.

I want to roll my own.

Here is my customer database. Are there anything I should change about this before I start building the checkout feature? And maybe it will tell you like, this decision you made sounds really reasonable, but here's why you should do it differently. And here's how things will break as you're growing your, your product. So all of those things are the things that the senior person is supposed to do. And I think that if, if someone is working with an LLM and has the right kind of humility and some level of judgment about, you know, sometimes the LLM is going to be wrong about these things, like it seems like an interesting substitute. It seems like, um, in some ways it is a very bad substitute for mentorship. Like probably the LLM is not going to be able to persuasively encourage you that yes, you're frustrated. Yes, this bug is hard. I'm sorry, I can't figure it out either. We should both keep going. Pour yourself another cup of coffee and let's get this solved. LLM is probably not going to do that all that well, but it can do a lot of the context providing and of basically providing some kind of synthetic wisdom. And I, I did want to make sure that when I'm talking about this, I'm not saying, you know, every company should fire every layer of management below the CEO and should just have junior employees who are asking LLMs what to do and then have the CEO. Come up with a strategy and then just tell GPT-4.5, okay, break this strategy down into different instructions for different groups within the company, et cetera. Like that's, that's probably not going to work. Like it is probably strictly worse than having, having an actual competent human being in the loop. But if it's say 70% as good and we don't all have access to mentorship in all the domains where we would like it, then I think it's actually pretty powerful. Like in my case, I am, I I do some programming, but I'm more of a finance person by, by nature and by career background. So I haven't really worked in an environment where I've been writing code that a more experienced software engineer is looking at and giving me good feedback on. Just isn't something I've really done. And I haven't really been in an environment where there is a whole process for going from you've written a script that does what it's supposed to do to We actually have production code that customers can interact with. I did, I think I've written a lot of scripts that were just for my own use and used them to draw conclusions about things, but it hadn't really shipped something that, that end customers would use. And it, I think it would be a very poor use of my time to start applying for jobs at tech companies and telling them, I want to be a junior software engineer. And I want you to teach me, you know, how, how professionals think about GitHub or whatever. But a lot of that information, how professionals think about GitHub, how, how someone with 5 years of experience designing such things would have architected this particular system, like that is, that is in a lot of these models. So you can get to some level, you can tap into some level of collective expertise on these things without actually needing to be a part of large institutions. And I think the places where that shows up most are going to be places that are Fields that are largely permissionless and don't really don't require lots of expensive equipment to get started. So basically, we're back to finance and tech. So in particular, very low barrier to entry for opening a brokerage account. There are plenty of brokers with APIs. So if you have some idea for a systematic strategy, you and your LLM-based mentor can can walk through how to actually implement that, and then. In actually building tech products, again, very low barriers to entry for putting an app on an app store or for registering a domain name and having it point to some service. So I, I wouldn't be surprised if we end up with a lot of fragmentation in those spaces. I think both of them have other things that are pushing more towards scale. So you do have higher fixed costs in a lot of stock picking adjacent businesses, and you also have higher. At least not necessarily higher fixed costs, but much higher cost to scaling, but a lot more visibility into how that scaling works in a lot of consumer-facing and, and B2B software businesses. So in some ways it will mean more prototypes, but not necessarily more companies, but I think it's an interesting phenomenon. And I also think it's just, it's often useful when you're looking, when you're looking at how technology will affect some kind of relationship between 2 different kinds of people. It is often very natural to say, here's how it affects the side that I'm actually most similar to. And I am, I am not a junior at this point, or at least I'm, I'm not a junior in the things that I spend most of my day on. So it is very natural to me to think about, okay, how, how do I as the boss leverage this tool to get more out of junior people or hire fewer of them in order to get the same output from them? But I think it's, it's actually a very good idea to think about how this would affect things from the other angle. Especially because if you, if you are trying to manage human capital, you're always thinking about adverse selection. You're always thinking about how do you end up with the opportunities that you have? Why are you getting the opportunity to work with this, this person instead of somebody else getting that opportunity? And I think it's often the case, like you, you want to think about those things because there is often a good answer. And in particular, if you, if you have some kind of media-like enterprise, some kind of public presence, there is a really good answer, which is you have fans, your fans Some subset of those fans would be delighted to work for you. And if you are hiring, you will almost certainly hear from them. So like, it's, it's good to have a good answer on the adverse selection question, but that does mean you do want to ask it. And so I do think there, there will be people who seem to just come out of nowhere, starting, starting little companies that scale surprisingly quickly. And we'll kind of wonder, how did this person do this without that, without the relevant experience and without the relevant experience in scaling? And how did they do it without having, you know, without raising venture capital and having a venture capitalist kind of tell them what the lay of the land was? And, and what, what obstacles to watch out for. And I think in some cases the answer will be they actually asked an LLM those kinds of questions and then they didn't need the advice. Or they, they asked the LLM those questions, they got advice that was not as good as they could have gotten otherwise, but they also did not have to give up like a quarter of their equity in order to have someone give them a little bit of advice derived from experience.

Yeah, it could be. Yeah. Yet, yet another example of a situation where we thought the low end was getting replaced, but actually it's the white collar stuff or the, or the, or the higher end. So yeah, interesting to see.

We'll continue our interview in a moment after a word from our sponsors. Hey everyone, Erik here. In this environment, founders need to become profitable faster and do more with smaller teams, especially when it comes to engineering. That's why Sean Lanahan started Squad, a specialized global talent firm for top engineers that will seamlessly integrate with your org. Squad offers rigorously vetted, top 1% talent that will actually work hard for you every day. Their engineers work in your time zone, follow your processes, and use your tools. Squad has frontend engineers excelling in TypeScript, React, and Next.js, ready to onboard to your team today. For backend, Squad engineers are experts at Node.js, Python, Java, and a range of other languages and frameworks. While it may cost more than the freelancer on Upwork billing you for 40 hours but working only 2, Squad offers premium quality at a fraction of the typical cost without the headache of assessing for skills and culture fit. Squad takes care of sourcing, legal compliance, and local HR for global talent. Increase your velocity without amping up burn. Visit choosquad.com and mention Turpentine to skip the waitlist.

I want to segue to another thing you wrote about, which is StubHub as a broader example of a business that's a stack of monopolies monetizing a layer of perfect competition.

Why don't you talk about that?

Yeah.

Yeah.

So the stack of monopolies is that Taylor Swift has a monopoly on being Taylor Swift. It's a lucrative monopoly. She is a billionaire because she controls this monopoly and ruthlessly monetizes it. And as far as all the Taylor Swift fans are concerned, that's great. And then venues also have these kind of localized monopolies where if there is only one place where you could have 10,000 people watch a show at once in this given city, That place does have a monopoly on doing the big show. And then you have other kind of monopolistic things. Like if there is a particular venue that is really well regarded in the blues community, it is probably the place where all the best blues musicians want to play when they're in town. And so it does get first pick at them. So yeah, the live music business and live entertainment in general, it is this series of small monopolies. They aren't really socially destructive. In fact, they're socially constructive in the sense that There's just a lot of money that would not go into things like Broadway, broad, like putting on new Broadway musical productions or getting a band to the point where they can record a demo in a good studio. Like that money only flows in because occasionally it means having a stake in the equity of someone who is just a runaway success story in, in that, in that field. So those monopolies, the existence of those monopolies is what subsidizes the next generation of those monopolies. So it is, it is probably a good thing, but it does mean that there's this very tricky dynamic in pricing where for some weird reason, a lot of times, especially in music, the tickets are just underpriced. Like you, someone like Taylor Swift could have sold those tickets for a whole lot more money. And it is this outstanding economic puzzle that they don't, that they don't just do an auction. Like why? Like, I think the economist question is if people can buy Taylor Swift tickets and flip them for 5 or 10 times as much. Why doesn't Taylor Swift say, okay, we're going to sell tickets tomorrow and you put in a number. And if your number, we will take the top, you know, top 10,000 bids and everyone will pay the bid number 10,000 and then you will all buy your tickets. You'll all pay the same price. And it is a price that is at worst equal to, but probably much better than the price you expect to pay. And part of the answer is that their fans would just be really annoyed with them. Like, if you, if there's a musician you like and you have your parasocial relationship with them and you really want to see them live and you find out they're playing in your town and you go to the website to buy tickets and you're basically prompted with this page that says, how much do you like this person? Answer in dollars. It's just kind of unpleasant. And then there's also within that, there, there is this fun dynamic where There are people who have a lot of money and very little cultural footprint. So I think in, in something like live, live music, I basically fit into that category where I don't think anyone, I don't think I could really influence whether or not people went to see some musician if I tweeted that I was going to go see them. So I would be the person who is mostly paying money to, to see this person, but some of their fans are very online, very fanatical, and they will talk nonstop about the show before it happens and they'll tweet about it during. And post to Instagram, and then they will be talking about it for a long time afterwards. And it's basically marketing for that musician. Granted, some of that marketing feeds into just selling more tickets. So it doesn't, doesn't work forever. At some point they do reach some kind of critical mass, but it's valuable to them to have, have people in the audience who are not actually paying the market clearing price because they can't pay the market clearing price, but they are absolutely going to make up for it by generating lots and lots of online attention that will lead to more ticket sales. So yeah, you have this, this messy situation where the price at which tickets could be bought on the open market is a lot higher than the price at which the artists are generally willing to sell them. And then stepping into the middle, you have StubHub. They make a market increasingly. They, they, they're actually, they were originally like a pure broker model. Now they have a little bit of a broker dealer thing going on where they do have inventory that they own, but only a little bit of it. And they resell that. But a lot of their business is just if you want to sell tickets, if you own tickets, you want to sell them, they will be the intermediary. And so if you're buying on StubHub, you are paying pretty close to just what the fair market value is for those tickets. And that's, that's what the seller is getting. And a lot of people just find that really annoying. They find it really unjust that one, they're paying a lot of money and two, they're paying a lot of money to someone whose sole function as an intermediary was just buy the ticket that you would have bought at the, at the original offering price, like buy it instead of you and then sell it to you for a large amount of money. But if, if you have a situation where a product is underpriced, it's going to get allocated somehow. And it usually gets allocated based on latency or some kind of access, but often just latency, like who clicks buy faster. And so you will inevitably have people who do that. They click buy faster or this, you know, this can be kind of an industrialized ticket scalping operation where you're writing a bunch of scrapers and you're renting out different residential IP addresses. So you're able to continuously ping this site without getting flagged as a bot or something. But some of it is also just, if you know that tickets tend to be underpriced, the game theoretically optimal move is book them now. Just buy them if you think you might go. And then if you sell them, you may actually end up making money on the sale. If not, you still at least got some of your money back. So, so like there, there are a lot of people who have reasons to be annoyed by this entire system. And you can view the cut that StubHub gets as basically their compensation for being the brand name on the part of the transaction that is most likely to annoy people. So StubHub allows Taylor Swift to not be blamed by Taylor Swift fans for the fact that those fans spend as much money as they did on tickets. And And StubHub gets a cut in exchange for that. And it's also, I found it just an interesting market because they actually have to price a bunch of discrete assets. Like there's shows happen on particular nights and Thursday night is not the same thing as Friday night. And every seat is different. And there, you have to have some view of what the value of all these things actually is relative to other seats. Like anytime you're getting any of that wrong, you're leaving some money on the table. And They are trying to be a little bit more ruthlessly Econ 101 about this stuff because that is, that is the environment that their customers have opted into by buying a ticket on StubHub. So they can kind of get away with that. And yeah, it gives them just a lot of, a lot of room to take, take a process that is not perfectly optimal and then make it for some value of optimality. Yes, pretty much optimal.

That's a good articulation. Maybe in closing, let's end with the biggest sort of story was last week, at least in terms of on all of our X feeds, which was sort of the new image model or sort of the Ghibli-ification of everything. It's sort of ironic you're seeing this video of Miyazaki saying that, you know, he thinks AI art is a travesty or sort of, you know, he would hate to be, to have his work transformed in this way. You wrote a little bit about AI art and how some artists are actually getting paid more because of it.

One you sort of share some broad reflections on this.

Yeah.

Yeah. So I thought it was a really interesting phenomenon. I, you know, one thing that it pointed out to me was that there is in fact a lot of demand for AI services. You had Sam Altman tweeting things like our GPUs are melting. And I think there, there are a number of people who at least should have said, wow, that's really surprising. And it changes my mind about the demand picture for AI because a couple of months ago, the big AI story was that the cost of incremental, like the cost of reasoning models is plummeting and therefore there won't be any demand for GPUs and the price of all of this stuff is just asymptotic to zero. So sell everything. But now, now there's, there's clearly room for demand. And it was interesting because the demand was not, hey, here's a totally new use case. Like people have been doing generative AI images for, for many years. And it was an exciting thing for many years. You know, there was that, there was like the excitement when DALL-E first had its demo blog post and was generating things like, here's an armchair shaped like an avocado. And we all thought that was amazing. And now you can go back and look at those announcements and they, they look a lot less impressive than they did before. Because of course the model quality has gotten better. And it did hit some kind of threshold where pretty much any photo that you ran through that you told it to generate in Studio Ghibli style just looked great. And you can, it was interesting actually to look at some of these, some of the side-by-side images really closely and look what made it great, because there are all these little details. Like if you, if you Ghibli-fy a photo and there's some clutter, some of the clutter just magically vanishes. And then the rest of the clutter actually looks a little bit more like it's supposed to be there. And if you Ghibli-fy a photo and someone's in profile, it will tilt their face towards the camera. I think some of that is just that if the model is trained on lots of of screencaps and probably not trained on just every single frame of every single animated movie. That seems a little bit aggressive, but it seems actually quite hard to say we're going to train a model on all these images on the open web and we will just make absolutely sure that there's never an image from that, never a screencap someone took from Spirited Away or something like that's probably not going to happen. So a lot of these images that it's trained on are actually the better looking images. Like they are the one frame where someone is looking most directly at the camera. So it is doing a lot of things other than just copying the art style. It's actually copying some of the composition as well, which I thought was kind of cool. In terms of the Miyazaki video, I had seen that quote floating around for a long time and I forget who, someone on Twitter pointed out, and then I tried to highlight this, that he's not actually talking about AI art as we understand it today. And you can go and look at the documentary on YouTube. He is talking about a very specific thing, which is a video that, a demo video that someone at, at his studio showed him of these horrifying zombies that are crawling on the ground and flopping around. And the guy who's describing this video is saying, we built, you know, this model, it looks very realistic, but the, the, the zombies don't treat, they, they treat their head as just another limb. And it just looked horrific and unpleasant. And Miyazaki is talking about how He has our friend who's disabled, and this seems like a just deeply cruel thing to to bring into existence in light of that. So I think yeah, if you if you look at the video that he's responding to, it is actually a weird disturbing video. He he probably is somewhere between just kind of confused by this trend and actually pretty upset. I think it'd be reasonable either reaction would be pretty reasonable. But he he he was talking about something different, and so I do I do think there there is a kind of fun irony where. People are taking something Miyazaki did, they're reappropriating it, putting it in a context that has very little to do with what he was originally trying to create, in this case, a monologue, I doesn't like this video, and then using it to just score points on social media. So we're all doing the same thing. We're close to morally, you know, morally similar in this regard, but there is a difference, which is that there are, I've seen a ton of people criticize posting this stuff and say that we should, that either A, they make art and it's better, or B, you should just commission art from human artists and it'll be better. But I, I don't think I've seen anyone actually do that and show any of their own art. So I think the people who are saying that vociferously online are actually probably not as good as GPT-4o in terms of making cozy little animated whatevers. So that, that was, that was another thing that, that came up. And it was also interesting to look at the reaction to that because I, I didn't actually view this as making art. I did not think of myself as an artist who's creating some artistic work. I thought of it as a fun thing you can do online. I thought of it as much closer to playing a casual game or much closer to writing a tweet. And I think, you know, if, if you, if you wrote a tweet and someone replied to you and they said, This is the shortest novel I've ever read. You don't even have any characters. There's no plot. Also, you wrote it on a website. You didn't even print it. Like, you're clearly never going to make it as a novelist and you should give up. Like, you wouldn't be like, wow, yeah, that really cuts deep. You'd be like, no, I'm not trying to do that. Like, the entire, the entire conflict here is in the head of the person who's criticizing this. So like, I didn't, I didn't think of it as art, but then I was thinking about just the question of what is art? And I realized people have debates about the nature of art all the time. And there, there are a lot of historical periods where we, people are asking, is this actually art? Whether it's like, is hip hop actually music or is it something else? You know, is this book art or is it porn? Is, is Dada art? Is it like a commentary on art or is it just someone realized that you can take everyday objects or You know, yeah, you could basically take everyday objects and call them art and then sell them to gullible rich people. But what I realized was that it's hard to think of any case where we had that debate and didn't eventually settle on, yes, it's art. Like, I think it's, it's much more contrarian now to say, no, hip hop is not art than it would have been back when that was really a live debate. So since the debate usually does settle on, yes, it's art, even if the person who created it didn't try very hard, even if they use mostly use tools that somebody else had built. Even if it's all just borrowing from somebody else's idea, you know, even if it is just a painting of Campbell's soup cans or whatever. Since, since those debates usually end up with, yes, all of us saying it's pretty much art, like it's, it's not art I like, but it is art. I think that the, the critics of gibblification are actually turning it into art. They, they, by, by discussing whether or not it counts as art, they are making it art. So that's cool. They, I didn't think of myself as an artist, But now I am.

Awesome. Well, that's a, that's a good note to, to, to end on. As always, Byrne, this has been a great and wide-ranging conversation, and until next time.

All right. Talk to you next time.

Upstream with Erik Thornberg is a show from Turpentine, the podcast network behind Moment of Zen and Cognitive Revolution. If you liked the episode, please leave a review in the Apple Store. Hey everyone, Erik here. At Turpentine, we're building the first media outlet for tech people by tech people. We're the network behind the show you're listening to right now. We have a slate of hit shows across a range of topics and industries, from our AI and investing cluster of podcasts to shows that drive the conversation in tech with the most interesting thinkers founders, investors, and influencers like Econ 102 with Noah Smith. We're launching new shows every week and we're looking for industry-leading sponsors. If you think that might be you and your company, email me at erik@turpentine.co. That's E-R-I-K at turpentine.co, and let's partner together.