This Era’s Inflection Points with Byrne Hobart | DOGE, Elon, OpenAI

Hey, Upstream listeners! Today, we're releasing a conversation I had with investor and writer, Byrne Hobart, on Turpentine's show, The Riff. We discuss the future of the Department of Government Efficiency, leaked emails revealing some of OpenAI's early history, Trump's strategies, and more. Please enjoy the episode.

Burn. Vivek just tweeted that he wants to make the USA like, like Malay. We talked about this many months ago, if what Malay is trying to do in Argentina could possibly be feasible in the US. I think, you know, one thing that Zach Weinberg and others have been tweeting about is how payroll is just a, you know, a small fraction of our budget on under less than 10%, I believe. And so there's a question as to how much could the Department of Government Efficiency even achieve in the first place? And you wrote a little bit about what could the effects be on GDP. Why don't you pontificate on a few years from now what we think the impact of this admin might be on the budget and more broadly as it relates to the financial situation? Yeah.

So the payroll thing, like the direct payroll costs to the US government are not a huge fraction of the deficit there, but then There are these indirect costs associated with a lot of those payrolls where you have a lot of people where their, their cost to the government that like the accounting cost is, you know, very low 6 figures. And then the actual cost of the country is orders of magnitude higher than that, because if their job is to either their job is to speed something up and they're slow, or their job is to slow something down and they're effective, then they, they do tend to, to slow, slow growth a bit. And so like when I was, when I was thinking about this Department of Government Efficiency, like one of the things I was thinking was just all else being equal. Well, if you do cut expenses, that is deflationary. And, you know, over the last few years, inflation has been a bigger concern than deflation, but deflation is in some ways harder to deal with. I think if you have an independent central bank, there is a lot they can do to get a handle on inflation. They could just keep raising rates until the inflation goes away. And at some point that does work. But if there's deflation, there are just limits on what they can do because they are, they, they, the channels that they have to provide capital don't provide capital in equal amounts to everybody. And it's just hard for them to engineer a monetary policy that leads banks and other financial intermediaries to put more cash in the, in the pockets of people who will immediately go out and spend it, which if you have a deflation problem, that is ultimately what has to happen. There's some combination of that and just productive capacity gets, gets cut until there's, until there's something closer to an equilibrium. And also like a lot of the rhetoric around Department of Government Efficiency is talking about these really, really tiny projects that do not matter for the budget deficit at all. And don't seem like an especially big problem. There's like, there's a general problem in political discourse of decimal point blindness that pretty much any number that ends in -illion seems to just mean a lot to, to many people. And so, you know, is, is $1 million studying the, the sexual habits of some particular species of frog? Is that A lot of money for that? Is that a little? Well, it ends in illion, so it probably feels like it's too much. And I don't know, assessing all of those individual studies, it actually seems like that's a case where your return on time is extremely low because the lazy thing to do is just find the silliest looking studies and talk about them. And, you know, you do that, you come up with your list of like the 2 dozen lamest sounding things and you kill them all and you have saved the the government, you know, tens of millions of dollars. And that is just like, we don't actually measure the deficit to that many significant figures. So you don't even make a measurable difference if you do things like that. The big chunky expenses are entitlement expenses. I mean, there's a lot of defense stuff where I do think that there is a lot of room to make that procurement simpler and faster. And it does, there are companies that are pretty optimized around getting very high-priced contracts and That, I think that raises a point about the payroll question of like the direct payrolls are not really high. There are a lot of indirect payrolls where there are, you know, if you work at the compliance department at a bank, like you're, you don't show up in the headcount of the federal government, but your job exists because of the federal government and because of the rules that they have. And so those are, that's a case where I think if you measure it, if you actually measure the economic impact, then maybe something that makes compliance departments, like slims them down or reduces the need for other companies to have them, et cetera. Like maybe that is actually a big improvement in economic efficiency, but it doesn't actually show up in the deficit unless you express that deficit as a percentage of GDP and look over a long period. And even if you do, like you won't actually see the direct impact. You, you will just have to believe that growth was happy. Growth was higher in this scenario than in the counterfactual scenario. It's like, I'm, I am all for a more efficient government. And I do think that it spends a lot of money in, in very dumb ways, but like the main dumb ways it spends money are just continuously transferring money from young people who are building up savings and potentially having families to older people who aren't doing any of those things, but do vote. And that is like, it's a political reality that as long as voter turnout is, as long as voter turnout goes up with age, the government is going to cater more to the interests of older people rather than younger people. But that's not, that's not so much an efficiency thing. That is some combination of like a long-term prudence thing of maybe you want some sort of inverse Social Security where you tax high in like high net worth demographics, such as over 65s and transfer that money to young people who are saving up a down payment for a house. That would be better for long-term growth, but just politically, you know, you get a few more votes from the handful of politically engaged young people who have not made up their minds and then You lose the over 65 vote forever, and one of those counts a lot more than the other. So yeah, I'm I am generally not not really impressed with what I've seen from Doge so far. I do think it's a really hard problem. On the other hand, really would not want to bet against Elon Musk. He's he's tackled other really hard problems before. In fact, his resume is basically one hard problem after another, and you know one near death experience after another. And so it almost feels like what you actually want to wait for is like the moment where Doge is. Totally discredited. And, you know, that's, that's actually the point at which it will start to work. But I don't know quite what that looks like. I do think that people, people have this view that a Trump presidency is, is pretty inflationary. That does seem more likely than not. But then a lot of the specific stuff that they are talking about is like on the, on the Doge front, at least, is a little bit more deflationary than inflationary.

Yeah.

Yeah. I heard Saagar sort of pontificating or sort of claiming that he thinks this administration will actually run a deficit or continue the trend of running a deficit. Someone said that Elon hopes to save $2 trillion or something like that. Larry Summers said we'd be shocked if we even saved $200 billion. Yeah, it'll be interesting to hear. And then many people are speculating that the Elon-Trump sort of partnership won't last that long to, or long enough in order to implement the changes, even if they were, if they were able to, but, but we'll see. Yeah.

Yeah. Trump has cycled through partners. Elon has had fallings out with various people. So like, yeah, they're, they're both very, very high agency people who tend to focus on a handful of projects at a time and then not focus on other stuff. And so I sort of think. Like, Elon seems to be having just a really good time being the probably most powerful advisor to the next president. But yeah, it seems better for his long-term career growth if he finds some way to do this not at Mar-a-Lago and not in a way that detracts too much attention from Donald Trump.

Yeah. Speaking of falling outs, let's talk about Elon and Sam Altman because some people are Speculating that this is a really tough time for OpenAI. If, if Elon in fact holds the, the keys to the, to the kingdom, so to speak. And the internal emails came out and you wrote a little bit about it between Sam and Elon. Why don't you unpack some of your thoughts there?

Yeah. So the main thing that I came away from looking at those emails was just like, things get negotiated fast by people who are busy with other stuff. And like that, I think the deal, like the deal to Get some money from Microsoft in order to scale up OpenAI's training. It was one of the more significant tech deals of the last couple of years, but it's, it was done by 2 people who I think both had pretty demanding day jobs at the time. If I have the timeline right, like Altman was still at YC, still running YC, and Musk was of course running his businesses. So you have 2 people who this is. Even though this is really interesting, even though it's like philosophically close to the top of their priority list, um, as a day-to-day task, maybe it isn't that close to the top of their priority list. So they both sort of tried to come up with some kind of hacky solution that would sort of work. And it looks like Musk just got tired of it and washed his hands of it. And then, and then Sam Altman was, you know, did his own thing. And, um, we'll see how that goes. You know, so far, so far so good, at least in terms of market value created.

Yeah.

But there was, there was this interesting detail about the, the possibility that Tesla and OpenAI could have merged in some way because Tesla's doing a lot in AI. They have the budget, they have the GPUs, they already have a lot of staff. They have a lot of unique data. So I think there, there was an interesting case for that. And that just, it feels so that, and then Sam, like the emails allude to, they don't, there, I didn't see an email that actually lays out this plan in detail, but the emails allude to Sam Altman considering an OpenAI crypto offering and Musk being apparently pretty disgusted with that idea back then. And so like, it just, history is just really, it's full of weird contingencies. You know, there's a very near possible world where OpenAI is a Tesla subsidiary. Maybe it does worse, maybe it does better. We don't really know. There's a near possible world where there is actually an OpenAI cryptocurrency and people are trading it. Um, it is kind of funny that, um, scams imitate art, so. Periodically someone will hack the account of a prominent, uh, hack the Twitter account of a prominent OpenAI person and they will post a tweet announcing that OpenAI is selling a cryptocurrency. And it's very funny to me that this thing that scammers periodically do when they break into Twitter accounts is something that actual Sam Altman also seriously considered during back in the day.

Well, I mean, Worldcoin was also a crazy idea at one point and now is actually a real idea.

Yeah.

Yeah. And maybe Worldcoin, and I guess Sam has talked about this, that, that Worldcoin sets up infrastructure for providing universal basic income in a hypothetical world where we are much richer. Like the mean GDP per capita is way higher because of AI and the median GDP per capita is way lower also because of AI. That we have like a handful of people who have those performance-linked equity-like units in OpenAI and They're all billionaires and there's nothing economically valuable that the rest of us can do. So yeah, it's, it's good to see that he is thinking ahead, that it's not all just like late night negotiations and just like seat of the pants, you know, arrangements with various people. And there's, there are all these little fun little details in, in that, in that email dump. Like there's, there's this point where OpenAI is getting started. They've made offers to a bunch of people and everyone is taking a pay cut, but they believe in the mission, et cetera. And then it turns out that Google is trying to poach every single one of them. Offering them massively more, so OpenAI just unilaterally gives everybody a pretty substantial raise. Pretty much right at the beginning. So, like it, I guess that does indicate that someone out there was taking OpenAI seriously. But it does feel like of all the people involved, like Google was actually taking OpenAI more seriously than many of the people who were making these key decisions for OpenAI. But that's kind of, I guess that is sort of how it has to work. That there's Whenever you, if there's something that you do with your life, like some career track that you get on, there's a really good chance that it was a vague interest for a while, and then you started getting into it, it started getting more interesting, and at some point it just swallows all of your free time. And that's true for a while until you've kind of exhausted the topic. So maybe that is, that's sort of what we're seeing a snapshot of in those emails is Elon and Sam both realizing as these negotiations get more contentious and as more issues come up that this is actually a really big deal and that OpenAI is either one of the handful of companies that has a shot at building real AI or it's not, and it's something else. And they came to very different conclusions there. And I think, you know, so far, I, it's very early to say that Sam Altman has been vindicated by history and also early to say that Musk got this one totally wrong. But like, so far, OpenAI has been able to raise a lot of money, has been able to commercialize some parts of their product and then has so far avoided any really big embarrassing catastrophes on the, on the AI side. They've had governance catastrophes, but that's different. And that, so I, I think part of the lesson there is just taking, taking those things pretty seriously, pretty early on and just spending some time thinking about different scenarios and trying to flex your assumptions a little bit. But it, it also struck me reading this that they, If they had known how expensive it would be to train some of the models that are being trained today, they would probably have had a different structure. Like it probably would have had to be for profit from the beginning. But also if they'd known that, maybe they wouldn't have actually done it. Like maybe if you could go, like you have a time machine, you go back in time to 2018, you show Sam Altman the OpenAI profit and loss statement for the last 12 months. And this is the reason he doesn't do OpenAI because it's just this giant money pit. And yeah, you built a really, really cool advanced autocomplete product and you can make an image of like a hamster on a skateboard fighting a ninja, but also you burn billions and billions of dollars doing it. So was it, was it worth it? Yeah. So very, you know, we, we always live in, we live in the result of, you know, some, some inevitabilities, some coin tosses and some things where it wasn't either of those. It was actually that a small, one person or a small group of very willful people made a decision about what the future would look like and then made that happen.

And it is in terms of how they made that happen. There's this broader question that tends to happen around people like Elon and Sam and Thiel and others who are sort of constantly making waves is a question of like, how much of it is premeditated versus how much of this could have been foreseen in advance versus how much of it is really just like gut-driven. You know, speaking of contingent, there's a world where Elon didn't take over Twitter, right? Like, didn't he try to get out of it at one point? And if he got out of it, perhaps You know, maybe Trump doesn't win the election. I don't, you know, it's, it's, yeah, it's pretty fascinating. It's kind of like my instinct is they're, they're actually just acting from the gut kind of. There's not a ton of premeditation because you can't see all the things, you know, you can't see how all the things will play out.

So there's this funny thing about, about these gut decisions, which is that if you read business biographies where it was a friendly writer and a cooperative subject, Those biographies will often be able to trace this very, very linear path where, you know, someone is born at a particular place in time. And at some point they hear about, you know, this computer stuff sounds interesting or, hey, what's a stock or whatever. And they get really into it. And then it's just this pretty straightforward path that goes from the kid who could have been that person to the adult who you already have some conception of in your head. And then when you read business biographies that are a bit, a little bit less cooperative and, you know, they couldn't get their subject to sit down for an interview. So they talked to everyone that subject has ever done business with and got a lot of, you know, off the record quotes from people who say damning things and on the record quotes from people who sound a little bit too, too interested in offering really high praise, et cetera. You tend to get a more complicated story where you could see things moving in a very different direction and where the things that they worked on didn't really look like they, you know, they, they thought they were on a different path than they were actually on. And I think the synthesis for that is that sometimes your gut is, it is some combination of just like you are, you are fated to accomplish something and you've put a lot of the pieces together, even though you don't really, you didn't really know what you were building when you assembled those pieces. And I think some of it is just when you, when you have a lot of the relevant context and you know that you have more context than other people, sometimes you should just trust your instincts and try to move very quickly. Like if you have an information edge and you're confident in that edge and you have an opportunity to exploit it, then that is probably what you should do. Like you probably shouldn't try to expand the degree by which you are better informed than the next best informed person. If you are the best informed person, like a lot of this is a relative game. So I think, I think that kind of, kind of ties those 2 things together where maybe it felt like a gut decision at the time. And then you look back and say, actually, this person was pretty much preparing to do this for most of their life up to this point. Like Facebook, for example, Mark Zuckerberg was programming really early on. He was doing little contract programming gigs for people. He had that Winamp thing. You know, he had a music player that was like, you know, it at least said it was AI based. I don't know what they actually used, but like, you, so you have someone who even in high school is asking, how do we use artificial intelligence to show people the content that they would be interested in engaging with? And, you know, you probably would have like puked if you use that to describe what he was doing with his nights and weekends. But it is an accurate description. It's the description you use today for, for that kind of activity. And that's very much what the current meta is like. It's constantly trying to collect all the information it can about people in order to show them things that appeal to them, whether those things are status updates from their close friends and family or just, you know, funny, funny reels or ads. So there is this deep connection that's really only apparent In retrospect, but some of those gut decisions I think are partly composed of that. They're composed of just having a lot of the information that you'd want to use to make the decision and just kind of instinctively knowing which way you ought to go.

We'll continue our interview in a moment after a word from our sponsors. You know, like many of you, I spend a good chunk of my day managing emails and phone calls. And despite being careful about where I share my contact info, somehow the robocalls keep multiplying and my inbox is flooded with spam. It turns out there's a reason for this: data brokers. These companies don't just collect basic contact info. They're gathering everything from your Social Security number and financial records to your shopping habits and employment history. Some health insurance companies are now working with data brokers to track your online activity, which could actually affect your insurance rates. Another huge privacy risk are people search sites. These websites publish detailed profiles about you, including your home address, family members' information, and even sensitive details like religious beliefs or political affiliations. This got me thinking about how exposed we all are. Whether you're sending a wire transfer, want to protect against identity theft, or just trying to keep your family's personal details private, that's why I started using Incogni. Here's what makes it different: they automatically handle 5 different types of data brokers— marketing, recruitment, financial information, risk mitigation brokers, and those people search sites. Instead of you having to track down Incogni contacts them on your behalf and demands they remove your information. And here's the key: they keep monitoring and removing your data automatically, because these brokers often try to collect it again. Incogni protects your data across all your devices, whether you're working from home, a coffee shop, or anywhere with public Wi-Fi. Plus, you can add up to 4 additional family members under their Family and Friends plan. And they're so confident you'll see the difference in your day-to-day, they offer a 30-day money-back guarantee. That's a pretty impressive offer. Take your personal data back with Incogni. Use code Upstream at the link below and get 60% off an annual plan. That's incogni.com/upstream. 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.

And relatedly, you know, Thiel made this bet on Trump in 2016, but it's possible he was too early or didn't stick with it enough to get like almost like a VC who invests in the first round, but doesn't do the pro rata or kind of gives up on the company. And then the company gets really big and, you know, doesn't have as much loyalty to the first investor if they didn't, you know, back them the whole time. Because Thiel could be sitting where Musk is right now, where now maybe he's in a good place if Vance has loyalty to him and maybe it's, you know, That, that's the long-term bet. But, but in the same way, then in tech, you know, being early is as, as useless as being, as being late, too late. Maybe there's a similar lesson here.

Well, The Laws of Trading has a very good point on this, which is that you will be dissatisfied with every trade. Like every, every order you put in, you regret because let's say you, you like the stock, you put in an order, a limit order to buy 1,000 shares at You know, like a little bit above the, a little bit above the current bid. So one possibility is your order doesn't get filled. And the reason that happens is the stock started moving up. So you were right about the direction, you got greedy and set too low a price, so you didn't get any of it and off it went. I had this happen where I tried to short a stock right after it de-SPAC'd. And I think it went from like $10 to $35 and change. And I put a limit order to sell at just under $36 and it never reached That it never reached 36. It just went pretty much straight down from when I put in that order. So in order to save like a very small, you know, in order to save like a couple percent on that trade, I missed taking advantage of the stock pretty promptly dropping by half. So that stings. On the other hand, if you do the same, you're in the same scenario, you put in your order and it immediately gets filled. What you know is that there is someone who is highly confident that the stock is going down. They are looking for liquidity providers and You are it. So yeah, you're, you're always, you know, for all the good trades, you could have done more for all the bad trades. You could have not done it. You don't really know until you actually do the trade. And by then it's, it's too late now. Now you just have to do another trade where you have exactly the same regret dynamic. So yeah, I think backing, and I think the other thing is like backing Trump in 2016 was just, or 2015 and again, '16, it was still like quantifiably more contrarian. We, our sample of election outcomes did not include any elections where Donald Trump was running a serious campaign. And so we just had no idea how big the polling error could be and what, you know, how big, how correlated the swings in different states' results could be. And then in, in 2024, Trump's odds just, they never got down to where they were ahead of the 2016 election. You know, you, I think, wasn't Silver saying like 8% chance that Trump wins or something like that the day before the election? And he was like the most, the most accurate of anyone who had a serious model. There were plenty of people who were more accurate, but just didn't, didn't publish their approach. And so, you know, the, like when they were right, it was that either they got the vibe right, or they did have some view that the errors in these polls are more correlated than people think. And your swing states are all close enough to flipping that If, if one of them flips, all of them do. And so the number, the odds should have been closer to 50-50 all along. So, so yeah, I think Trump, like, it was just statistically less contrarian, but I think maybe, maybe in, in intra-elite terms, in some ways more contrarian because we did just have more information on what Trump would be like as a president. And there was the whole January 6th thing. So like, there were, there were a lot more reasons to say like, He's not a clown. He's actively dangerous, or he's not a clown. You know, he's he's doing he's going to actually do some of these things. On the other hand, you could also say like he he does say these crazy things. He mostly doesn't do them, and that on the pandemic, he he actually got the important part right. Like he did push for a vaccine, and he did get that funded, and that was a huge boon to America and the world. He doesn't want to take any credit for it because his base no longer likes vaccines, and now that he has RFK in the cabinet, he can take. Give him less credit for it, but that is, he, he does have this kind of Cassandra-like thing going on where he did actually get the important thing right in the pandemic. There was a ton he got wrong. On the other hand, a lot of the pandemic critiques, like part of, part of the big critique of Trump in 2020 over the pandemic was like, you basically have people saying, so I have 2 problems with Donald Trump. One is he didn't look at this crisis that started in a different country and use it to Seize significant power for himself and radically curtail civil liberties. And also, he's a fascist. Like he he had a really good shot at being a fascist. He had a he had you know the best opportunity that any American president has had since the 1930s to bring fascism to America. If any if anyone could have pulled it off, and if there was any time to pull it off, the man met the moment in 2020, and he did not want to do that. He he seemed to think like this is just really really annoying and not what he wanted to do with this job, and. So delegated a lot of it to the states. Different states did different things. We had some quasi-controlled experiments in how those different things worked out. And then, you know, he kind of, he bet a lot on this magic bullet thing. Like I did not think that a vaccine would happen nearly that fast. I didn't think it was going to be nearly that effective. And I thought the, the mRNA stuff, like people were talking about Moderna very early in the pandemic as having a really good shot at this. And it always felt kind of like a gimmick of like, every time something crazy happens, there will be some company that becomes this fixation for investors, whether it's like I remember after 9/11, there was a company that was, they were, they didn't really have a good version of this, but they were working on facial recognition cameras. And the stock went from like $3 on 9/10 to like $15 by the end of the week or at the end of the next week. And, um, it never really worked out. So, and that happens when there's a big hurricane and Generac just rips, or when there's a pandemic somewhere, like when Zika virus was happening, there was some Some vaccine stock and some PPE stock that was doing really well at that time. So like, I just had the, the general sense of people, people always bet on magic bullets and they usually don't work out. And then in this case, yeah, it turned out there was a magic bullet. Like it's a really, really effective vaccine, does save a lot of lives. And it, it, there, there were a lot of changes in how the government approached vaccine development that made it actually happen on a timely basis. And that's very good.

Yeah. You mentioned earlier that a Trump admin will be, might, you know, be inflationary. Is this because of tariffs or can you talk about, you know, why would I be in the—

Well, so the tariff thing, like I, tariffs are just not, not a great way to raise revenue for either a, um, a very trade-centric country because you're messing up your key industry or for a comparatively autarkic country. And the U.S. compared to basically every other rich country is incredibly autarkic. A huge fraction of the stuff that we consume is produced domestically. We are much less dependent on trade than a lot of other countries in terms of total GDP dollars. Like there's, there's a lot of specific stuff that we really like having that does come to the U.S. from, from China on ships. And we would be paying more for that stuff. But a lot of what we pay for is either non-tradable or hard to trade or can actually be made in the U.S. And so there just isn't enough stuff for us to fund all of Social Security, Medicare, and defense entirely with tariffs unless we managed to significantly globalize the economy and massively increase both imports and exports, which is something that you can do, but it's hard to do that if tariffs are going up rather than down. So tariffs are not, not great as a revenue source. They, they will provide some level of revenue, but it's, it's not super efficient. Where they do actually help is if you have some industry that you think is strategic and has some returns to scale or has some positive geographically constrained spillover effects, it can be worthwhile to use tariffs to protect that industry for a while until it grows, scales up, and can actually stand on its own. And in that case, you do want to be very deliberate about how you are imposing a deadweight loss. Like consumers in year 1 will be worse off if we have a high enough tariff that the screws that you were importing from Shenzhen, you're actually making domestically. And the bet has to be that In 10 years, being able to manufacture these screws domestically and not have to worry about what happens if China decides they don't want to trade with us anymore, or you know embargoes some particular component that turns out to be really important to say building new Boeing planes or building new MacBooks or whatever. You you do win from that, but it is an investment, and it's an investment with an upfront cost. The place where I think inflation is more likely is that Trump likes to cut taxes. The rest of the Republicans like to do that too. There is a Republican majority, and the The pre-COVID Trump administration sort of reached this arrangement where Trump gets to do a couple actually Trumpy policies. He gets to say Trumpy things nonstop. And he does stuff that Mitt Romney would have done, you know, cutting, cutting taxes for, for higher earners, cutting corporate taxes. And I think those policies, they are, it's actually the, the, the tax cuts for really high earners and tax cuts for corporations don't have a hugely inflationary effect because high earners don't have a very high marginal propensity to spend. Companies do, they have a variable marginal propensity to spend. If you have a combination of even lower corporate tax rates and there's some tariff protection for key industries, then a lot of corporate America will reinvest its tax advantaged profits in those industries. And that's temporarily inflationary and then long run deflationary because it is adding more productive capacity to the economy so we can make more stuff. And so that part, that It's a little bit less less inflationary, but I think to to pass a tax cut like that, and so sorry for the for the really rich people when you cut their taxes, they just invest more. They don't actually consume more, so it doesn't have an effect on inflation. It can affect asset prices, but even that is kind of it's kind of you know a mixed bag. Like it's it is a lot closer economically to just having having the central bank fund deficits by buying debt directly from the government, where you are making these accounting like these bookkeeping entries, but They don't actually correspond to real purchases of goods and services. So they just don't have a huge effect on price levels. But to pass a big tax cut for high earners, you typically need to do something for other earners. And people in the middle and lower ends of the income distribution do tend to spend a lot of the money that they get from tax cuts. So that part is inflationary. And that is realistically the main thing that I think you'd want to bet on if you were thinking about the inflationary impact of a Trump presidency is that Lower taxes for people who actually spend more money when they have more money does tend to lead to higher consumer spending, and that leads to inflation.

Yeah, interesting. Peter Thiel went on Barry Weiss's podcast and talked a bit about the tariffs among other topics. And one thing he said is, yeah, it might have some economic downsides for, for consumers, but if we can hurt China more than we can help ourselves. It's a win, basically, in sort of the, you know, the realpolitik of our situation with them. And we could, you know, instead work with other countries more, et cetera. Do you expect this to play out that way?

There is very little that China makes that other countries are incapable of making. And there are many things that China makes that other countries are incapable of making at the same cost. So, yeah, China, you know, there's always this debate about which side has more to lose from the US-China relationship. And I think that it's just, Unequivocally China, like, but I think if you switch it to politicians and you ask who has more to lose, Donald Trump or Xi Jinping, Xi has a lot more control over what the media environment is in China, what the narrative is, and even what the criteria by which you judge a leader is. Like it was, it was a policy decision on the part of the CCP to say that their, their work should be judged based on real GDP per capita over time. And before that was their policy. They were judged on the fact that they had won a civil war and they conquered China. And, you know, there wasn't much you could do about that. So if they've changed how they want to be judged in the past, they can probably do it again. And they are just gradually moving in that direction. Like they've, they've emphasized GDP growth a lot less as that GDP growth has slowed down. It doesn't look like China really graduates to a high income country anytime soon. Or at least if they do, it takes much, much longer than previously anticipated. There are a lot of industries where they've moved very high up the value chain, but there just aren't, aren't really cases where there's a product that you couldn't imagine making outside of China. So it would be, it would be really expensive to do that. It'd be very expensive to shift all of those supply chains over. And I think it probably takes longer than one election cycle to, to make that happen. So you, you could definitely imagine a scenario where Trump And you have to have a serious imagination for this. But if Trump decides he's going to, he's going to do the right thing for America, even though he will be hated for it, and that he's willing to leave office as the most unpopular president since Nixon, or, you know, take a, you know, set a new record after that. But he is going to ensure that China does not dominate global manufacturing anymore. And he's going to push really tough tariffs, convince other countries to do it, basically tell everyone they have to choose between America and and China, and that if they don't choose America right away, they are de facto you know a Chinese vassal as far as the U.S. is concerned. He could do that, and it would be very inflationary. Like the U.S. would probably face a significant recession. A lot of people would be very upset, but it would also mean that China would be less of a threat for a long period thereafter. Although it could also you know there's there's definitely the scenario where. The Chinese government realizes that this is their peak of influence and material power for their foreseeable future. And so if there's any time to take back Taiwan, now's the time. So yeah, lots of, lots of interesting uncertainty over the next 4 years. But my, my, my suspicion is that Trump is not actually going to do that, that he is not going to make that kind of deeply unpopular, difficult decision that has a payoff that won't really be known until after he's dead. That he will, he'll probably be a little bit more chill, relaxed, focused on the legacy, et cetera.

Well, and it's also interesting to think about how, because he's effectively a lame duck, you know, how what happens this election might set up, you know, J.D. Vance 2028. Is it as easy as, you know, if the economy is bad, the Democrats will be like, how should we think about this?

I mean, it's not, not quite that easy. The economy was still doing fine in November of 2000. When Gore lost, barely, but he lost. So yeah, it's, it's, it's not just that there is, and you know, economy was also improving in 2016 when Trump won. So things, things don't just follow that path. There is some dependency on who is running, et cetera. So I think it is, it's very early to speculate, but I think the base case right now is that it is like J.D. Vance is the most plausible 2028 President of anyone you can name, but he's not. He doesn't have 50% odds at this point. There's just a lot that can happen between now and then. And there's also the question of will Trump serve a full term? He's not in not in exceptional health, has slowed down a bit. And I think if like Trump ran in 2016 and won, and there was there was a lot of uncertainty and controversy there. It was a very very close election, and then. I think like 2024, I do think that he was really, I think, more excited about Election Day than about Inauguration Day. And that he's going to be like thinking about whether or not this was actually the right decision because he got the vindication of like, yes, a majority of the voters actually voted for him this time. And so, you know, by, by that measure, he is the most popular person in the U.S. So by a very specific definition, and it's pretty clear looking at Trump's long-term track record that being a very popular person is really important to him. So maybe he's actually gotten everything that he wanted out of running already. And we'll, we'll let, let Elon and Vic and everyone else just do their own thing and actually run things. He does have the, he does have the 2026, you know, 250th anniversary of the country to, to look forward to. And he does seem like he would have been a good mascot for that, even if he weren't president. I tweeted about this a while ago, that there's this near-possible world where he retires from The Apprentice, doesn't go into politics, and then he is brought back as the emcee of the 2026 America's 250th anniversary celebrations and just dies as this universally beloved entertainer, the personification of American bombast and weirdness and just in-your-faceness and unusual fashion choices and whatever. And I think, yeah, he's like a good good character for that, but that's also a pretty good time for him. Like, it's a pretty plausible time for him to take stock and to say like, he's done what he wanted to do and wants to pass the torch. And so yeah, it's President Vance from then on.

Yeah, no, that's real.

We'll see if that happens. I don't know.

Yeah. Yeah. It'll be interesting to watch. Let's segue to a couple other things you wrote about this week. Maybe first we could start with your great Capital Gains newsletter. You covered selection effects. What do you think is so important to understand here? What can be sometimes a bit non-obvious about how to think about them?

Yeah, so I noticed that I just use the term selection effects a lot in The Diff and in Capital Gains, and that I've never really written about why I use that term and what it means and how to use it. But I think like the really high-level view on selection effects is that Everything you perceive about the world is selected in some way. Like you are never interacting directly with reality. You are interacting with the things that you personally encounter and those things undergo constant selection. So your view of what's normal is very dependent on who you are, what social group you select into, what city you live in, et cetera, what media you consume. And someone else has a totally different idea of normal. And it's just really easy to fall into this trap of believing that the things that you see day to day are a representation of Average reality as it's experienced by the average person, and that's just never true. So in that piece, I ran through a bunch of specific examples of where you should think about selection effects, and but in particular, like it's it's very hard to know what average behavior is for a lot of traits because you will like if you let's say you are in an industry where people are generally pretty smart, your experience is almost certainly that within that industry there are people who are really really smart but kind of flaky or really obnoxious. Noxious to work with. And, you know, they're only, they only have that job because they're brilliant. And then there are other people who are not just not as brilliant, but they work a lot harder than everybody else. They're really, really friendly, easy to get along with. They do the, you know, the, the annoying stuff that has to get done. They're more likely to do it, et cetera. And so you develop this sense that these things are inversely correlated, but actually a lot of that stuff is pretty positively correlated. Like smart people do tend to be more friendly and conscientious, et cetera. Like a lot of these traits are positively correlated, but In any given organization, there are people who get selected, like they barely make it in because they have the good trait that offsets the bad trait. And then you have people who, they have a lot of the good trait and then they have, they're really good on the other trait and they just get selected into a better organization. So if you actually see that in your life, all of these things are positively correlated, like the, the really smart people are also more friendly and they're also harder working and they're also more attractive than You know, you've actually made it to the tail of all of these distributions, but that's almost never the case for anybody. And also, as a corollary to that, the more useful any given trait is, the harder it is to get some kind of arbitrary sample of people who have that trait, because if it's a useful trait, that trait probably keeps them really busy. Like, if you're really good, a really good athlete, then you are probably, um, part of a professional sports organization. And so you're spending some time training, some time playing the game, and you You know, you're, you're a very expensive person. You're very scheduled, et cetera. So normal people are just not that likely to encounter you. And similarly for intelligence, like you, there are a bunch of different jobs in different sectors of the economy that select really strongly for intelligence. And so you can kind of get a sense of what smart people are like by having one of those jobs, but you also get a sense of what that particular kind of smart person is like. So like, someone who is an actuary and someone who is an AI researcher, like, They're both in jobs that are very, very strongly selected for just raw intelligence, but they have a totally different set of stereotypes about what do really, really smart people do. Um, in fact, like on, on social things, it's going to be pretty much the opposite stereotype. So like the actuaries, I think actuaries have like one of the lowest divorce rates of any job. So their view of smart people is smart people are just very, very, like a very, very middle class values and very predictable, have pretty normal personal lives. And then OpenAI, your view of smart people is like, They're usually in molecules, but there's a lot of variance in how big the molecule is and how weird the molecule is and how formalized it is, et cetera. So that, I think that that sort of forces you to have some level of epistemic humility that like the more, the more interesting your observations are, the less likely they are to apply more broadly than in the narrow niche in which you're actually making those observations. There's also, I guess, going back to the point about how you're You're disappointed in every single trade you ever make, or you should be if you know what you're doing. There's there's also a selection effect about which opportunities you see as as an investor or as a potential employee, where they are they're always selected as like someone someone else who wanted this didn't either they weren't shown it or they said no to it. So you're either selected into it because you have a really good network; you're selected out of it because or you're selected into it because there are more talented people who've all said no, and as someone was working their way down the list, they got to you. This also applies to things like just which companies are publicly traded, that the, the better quality a company is, the more likely it is that there's some investor who knows this and doesn't want to give up the upside. So the companies that do go public, it's some combination of the market is really, really high right now. So they can get a really good price and they should sell and they'll buy back some of that company later, or the company looks better than it is and investors are temporarily enthusiastic about it. And then the maybe one exception there is If it's in an industry where there is a longstanding tradition of equity compensation and the company's been around long enough that people are, their options are getting close to expiring, maybe the company's kind of forced to go public through those circumstances. So a corollary to that, which I didn't go into in the piece, is that you should expect the median quality of public companies in industries that have a tradition of equity comp. You should expect the median quality of those companies to be higher because they, they will go public for Not strictly economic reasons or for, for reasons that have more to do with how they pay their employees than whether or not the stock is worth selling at this point. So that is maybe that is part of why tech has done so well as an investable sector is that the tech companies were forced to go public at a time when you know say an oil company with similar unit economics somehow just wouldn't be tempted to go public. They'd rather just enjoy their profits, continue to reinvest them, grow fast, knowing that they aren't going to run into any big obstacles. schools there and just be happy with the dividends that they get without ever having to have a shareholder meeting.

Yeah. The, another thing you wrote about is why the news business is perhaps the hardest media business in the stack or not as perhaps monetizable as it used to. As we're both media businesses, we think a little bit about how much should be news versus analysis and other things, et cetera. Why don't you talk a little bit about the news portion?

Yeah.

Yeah.

So they're like the newspaper bundle was a bunch of different things. It was like national and international news. There was local news. You've got sports, you've got business, you've got entertainment, you've got classifieds and this bundle worked really well, but there were different parts of it that monetized differently. And classifieds did really well as an economic engine, although that is actually a fairly recent thing. So in the first half or so of the 20th century, most big cities had more than one newspaper. And so there was actually price competition on classifieds and you just, you couldn't jack up your prices arbitrarily because even if your newspaper has 60% share and the other newspaper has 40% share, like you can charge 50% more for your ads. That's fine. But you can't go much beyond that. Like no one needs total universal coverage for their classified ads. But once the newspaper is the only newspaper in town, then if someone is selling a used car and Craigslist doesn't exist, then this is the way to do it. It's like either the newspaper or flyers and you quickly reach the point where flyers have diminishing marginal returns. So for a while they had a really good economic engine where people would buy the newspaper because they wanted different parts of that bundle. And then they would also, like, there was a lot of ad revenue from local businesses that had few other options and then from classified advertising who had basically no other options. And as a lot of that has been unbundled, we've had to be more conscious of which parts of that bundle actually are valuable. And one of the obviously valuable ones is business and financial news. So if people are making financial decisions based on the, based on the things they're reading, then you, you, you have a lot of room to charge them a lot for that. If it's sports news, people are very loyal to their sports teams. They want to know what's going on. So they will pay for that entertainment. It's entertaining. People will actually pay for that. And then like hard news about national and international developments. Well, most people don't have a very close attachment to an arbitrarily selected country. So sometimes there's a country where a lot of people have opinions and the news is important because of that, or there is just something happening and everyone wants to talk about it. But advertisers don't always want their ads shown. They don't want their brand name right next to some really horrific, tragic stories. So maybe it doesn't monetize as well. And those important news stories They tend to be good for one side politically and bad for another side politically. And so if you are just going straight down the middle, you are always either irritating all of your Republican readers or all of your Democratic readers because you're telling them things that they really don't want to hear. And a better model just financially is you pick a side and you just tell that side everything it wants to hear. And you absolutely never mention any story that makes your side look bad. And that model has worked reasonably well. Like Fox seems like a pretty solid investment that, that has done well over time and, you know, has its, has its franchise, et cetera. There's plenty of that on the other end of the political spectrum too, but that is, that kind of shades into commentary rather than, rather than pure hard news. So yeah, news and, and then in a, in a social media environment, that same polarization thing means that people will just endlessly fight about this stuff. Like everyone. Automatically has an opinion about whatever this news story is. It's a pre-cached opinion and, you know, it doesn't, doesn't necessarily connect that well with the facts, but it does connect well with their hobbyhorses. And then they'll argue with each other and that's really good for engagement in the short term. And then they start to realize that Facebook is the site that makes them waste an hour and feel really bad about themselves afterwards. So they stop logging in. So a lot of these sites have basically decided that those, those are kind of toxic page views. Like you can get a lot of them and you will pay for it later on in, in losing some of your users forever. So yeah, a lot of them have moved away from it. And as that's happened, regulation has also tightened up a bit. So, you know, now you, you have, you, well, for a while you had people tracking which stories are most popular on Facebook, like which ones get the most comments and what's the partisan valence of that. You have some apparently restrictions on political advertising, digital political advertising in the EU that are sufficiently confusing that Google is just not going to run those ads for a while until they figure out whether or not it's legal. So yeah, if you, if you have this space where everyone has an opinion on it, these opinions are always pretty close to 50/50. It's not like everyone agrees and it attracts a lot of attention to your business and advertisers don't actually want to advertise on it. And if it is national and international news, any one local news provider does not have a huge advantage there. It is most, there are scale advantages to just having a correspondent in every city, you know, every, every part of the world where something interesting could happen, like There's an advantage there, but does the world need, how many, how many networks like that does the world need? It's probably more than one. You probably don't want one company to have a monopoly on always having someone on the ground everywhere something interesting would happen. But I don't know that you need like 20 of them. So yeah, over time, that industry just like that, that part of the industry, as you extract other things from the bundle, just turns out to be the worst one. And the one where you have like, The worst economics and it's the biggest pain in your business. And, you know, like when you look at the big court cases that determine like the limits of what the government can and can't do to punish media outlets that publish things that the government doesn't like or that they weren't, you know, that weren't necessarily supposed to be published, et cetera. There aren't a lot of those in the, that cover the sports pages or entertainment pages. Like it's typically the hard news stuff. Like it's It's Watergate and not a celebrity divorce that causes a lot of these precedents. So yeah, you, if you can get away with it and you're running a profit maximizing news business, do everything except actual news and you'll probably be better off.

Yeah, it is interesting. I'm going back to your, your point about, you know, people tell people what, what they want to hear or sort of confirm their biases. I think of the Free Press as an example, as a company that helps former liberals be comfortable becoming conservatives basically, or, or becoming, you know, sort of less liberal today. Like it's interesting what kind of sort of, you know, like psychographic profile, you know, people, people appeal to and what kind of sort of, you know, it says about somebody to read a certain thing and how they continue to, you know, make you feel good about where they're not just where they're at right now, but where they're headed perhaps.

Yeah, I think, I think there is some, like there is something to the general idea that Part of, like, part of media consumption is this form of therapy that you are trying to just feel a little bit better after you've consumed the media than before. And they can provide that in different, different ways. Like more purely entertaining stuff is closer to just injecting dopamine. And something like reading, reading a, a book from the, the literary canon is generally just less fun than taking out your phone and playing playing a mobile game, but it also, like, you feel better about yourself for having done it. You will, you will absorb ideas that stick with you for a long time. You are kind of inhabiting this world that this author has, you know, through, through their selection effects, they've, they've identified the salient details of their world. And so, you know, it has, it has a lot of these positive aspects. So yeah, that, like, a lot of what people are doing, it is like the, the Homo economicus thing. People make fun of it a lot, but it is to a large extent, like, a useful working assumption that People do things that make them happy over some timeframe. There's some variance in what the timeframe is and there's some variance in people's preferences, but that is like a general guiding indicator to why people do what they do is they like it.

Yeah. The gearing towards closing here, maybe the last question I ask you about is about prediction markets. I mean, you talked about Polymarket and what happened to the CEO this week, but also you talked about sort of internal prediction market experiments that were run by Google a while ago. And, you know, Robin Hanson has been pushing this idea of governance by prediction markets, what he calls futarchy, both I think on a governance level, but also on an internal company level for, I believe, decades now. But he says that people don't want to adopt it because it might make them look bad or they have other reasons beyond sort of optimizing the success of their, of their organization. Do you have any commentary on, on Robin's, uh, analysis or the broader idea?

Yeah, I think he's happy. I think, I think it raises a really interesting question because there are a lot of cases where there's an idea that sounds like a good way to run an organization, but it would make some people uncomfortable. And that could be anything from, we're going to have quantifiable goals for everyone and we're going to give people bonuses if they reach their goals and we're going to fire them if they don't, or like Like we will have a norm in our company that you can always raise concerns and that, you know, if the CEO is wrong and that the CEO sends a company-wide email and makes a mistake, like you can reply all and correct the CEO and that's fine and praiseworthy. Like this is something that Bridgewater has said is true. And there's a, there are questions on how true that actually is and how politicized that is. I have also read some anecdotes about early PayPal that this was also true, that people would just reply all to their manager's manager and say, this is stupid. Here's why, and get praised for it. So like that stuff seems to work. It seems like it should work. And, and yet a lot of large successful organizations, people are not quite that blunt. And there's not quite that much information sharing. And so maybe like one possibility is that we have a finite supply of really, really disagreeable people who are willing to earnestly argue about that stuff. And they don't have hurt. Their feelings are not especially hurt when someone argues with them. Because that's just what you do. And, you know, once, once that supply, like that supply of people, they all work at these organizations and then everyone else works somewhere else. But maybe there is actually some value in not widely sharing information and in making slightly suboptimal decisions that do keep people happy. I hope that's not true because I, I would definitely thrive more in a world where you can argue and don't just have to agree with everybody. But it also. It seems just empirically hard to defend the claim that more openness and more disagreeability is always going to lead to better decisions, because if that were true, the world would be ruled by much more obnoxious people. And then, yeah, that, so that the, the piece I linked was this backstory of some of Google's internal prediction markets. And this was the exact problem they ran into was that these prediction markets give a real-time live view of everyone's best guess as to Whether or not project X will be delivered on time, whether or not customer Y will actually sign this contract, et cetera. And maybe a lot of people just did not actually want a real-time metric of whether or not they were succeeding. It's, it's a lot more stressful. And there is this return smoothing thing where if you are running a sales team and let's say you have some quarterly goal and that's when you're going to get, that's when the check-in happens. If your first month doesn't do that well, maybe there are some levers you can pull, like maybe there, maybe you can push people work a little bit harder. Maybe you can cut a slightly better deal with some customers. Maybe you can tell someone, please sign by the end of this month and we'll make it up to you next time you renew, et cetera. Maybe, maybe people want that level of flexibility and they just don't have it if there's a prediction market. And so in the absence of a prediction market, the implicit graph of will this work or will it not work is always just a straight line to either 100% yes or 0% yes. And That straight line is just a lot more comfortable for people to deal with, especially if it goes straight to yes. It's also possible that prediction markets make it harder to sandbag. So if you have 2 managers and they both have their goals for the year and the prediction market opens up and for one manager, everything's 50/50, or it's like one thing's 70/30, one thing is 20/80, et cetera. It's a big mix. And then for the other manager, it's like everything is between 85 and 95% going to happen. Well, you know that for that manager, they're just sandbagging. Like they just picked very, very low, easy to clear goals. And so if you have a prediction market and you are predicting goals that are some specific person's responsibility, then for the company, it's probably optimal to set goals where they average out to 50/50 once you weight them by importance. But if you, if you actually have a set of goals where the reasonable expectation is your odds of hitting all of the, hitting any one of them is about 50/50. Then it's, it's very stressful. Like half the time you will have failed your goals at the end of the year. And a lot of people just dislike that feeling of, I spent a year working on this and I failed. And when you look at, say, how companies guide, when they, how companies issue guidance to investors, they will typically issue guidance that is below what they think they can do. And it's often by a very predictable margin. And there's, there's this implicit understanding that the company's going to say, We'll earn 46 cents a share next quarter, and analysts are going to pencil some things out, and they're going to guess that the average expectation is actually 48 cents, and the company is actually going to hit 50 cents, and 50 cents was the company's internal estimate all along. Netflix says they don't do that, but they've had just really, really good results over the last decade or so. So it's hard to say if they are actually secretly sandbagging or if they just keep just destroying expectations. And I think the success of Netflix is maybe evidence that a little bit more disagreeability, a little bit more of a data-driven approach does actually help even in businesses that are very connections-driven like media. But it is, it is also, it is also just striking that companies seem a lot nicer and people within companies seem a lot nicer than is truly economically optimal. And so maybe there is an ongoing cost to meanness and Maybe if you select the right group of employees such that they're all very high tolerant, highly tolerant of meanness, you have like a less pleasant day to day, but your company does achieve enormous success.

Yeah. And on that note, perhaps that's a good note to, to, to wrap on as we're, as we're, as we're at time, but this was a great, great discussion on what, what Elon and Trump could look like, OpenAI, media prediction markets, sort of tariffs, inflation, and, and, and more. Byrne, as always, until next time.

Yes, indeed.

All right.

See you next time.

Upstream with Eric Thornberg is a show from Turpentine, the podcast network behind Moment of Zen and Cognitive Revolution. If you like the episode, please leave a review in the Apple Store. Hey everyone, Eric 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 erik@turpentine.co, and let's partner together.