Google’s Monopoly Money + Is the A.I. Bubble Popping? + The Hot-Mess Express

“Maybe this actually is a chance to press the reset button on the internet.”

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure, and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner, Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 0:34Kevin Roose?

  I was very sad to learn that one of my favorite pieces of trivia about the Olympic horses was misinformation. So I got very excited because I saw a post going around on X that said that they have a plane that flies all of the horses to the Olympics.

• 0:55Kevin Roose?

  And that the plane is called Air Horse One.

• 0:58Casey Newton?

  Oh, come on.

• 1:00Kevin Roose?

  And so I repeated this fact to everyone I met in the span of like three or four days. And then I looked it up. And Casey, would you believe that that is made up?

• 1:10Casey Newton?

  Oh, well, you hate to hear that. I would love for that to be true. How do the horses get over there?

• 1:14Kevin Roose?

  So there are planes that fly the horses. That part is real.

• 1:18Casey Newton?

  Okay.

• 1:18Kevin Roose?

  And there is actually a plane called Air Horse One. But the company that runs Air Horse One did not transport any of the Olympic horses.

• 1:27Casey Newton?

  What does Air Horse One do?

• 1:29Kevin Roose?

  I mean, it transported horses, but it's not being used in this Olympics.

• 1:32Casey Newton?

  I mean, I think by the naming conventions of the United States, Air Horse One has only one job, and that is to fly the president's horse.

• 1:39Casey Newton?

  So if it's not doing that, I don't know what it's doing. But I'd love to find out.

• 1:49Kevin Roose?

  I'm Kevin Ruess, a tech columnist at The New York Times.

• 1:51Casey Newton?

  I'm Casey Newton from Platformer.

• 1:53Kevin Roose?

  And this is Hard Fork.

• 1:54Casey Newton?

  This week, Google loses the biggest tech antitrust case in a quarter century. The Times' David McCame joins us to discuss. Then, are we in an AI bubble? We'll hear the case for and against. Finally, it's time for a new segment, All Board the Hot Mess Express. Choo choo.

• 2:15Casey Newton?

  All right, Kevin, it's time to play a game of tech monopoly.

• 2:18Kevin Roose?

  Yeah.

• 2:18Casey Newton?

  Namely, which tech company is a monopoly? And did it maintain that monopoly illegally? Well, this week, a federal judge ruled that Google does have a monopoly in search, and it has been maintaining that monopoly illegally.

• 2:33Kevin Roose?

  Yeah, this is a big deal.

• 2:34Casey Newton?

  Very big deal. So Judge Amit Mehta, in a 277-page decision this week, wrote that, quote, "After having carefully considered and weighed the witness testimony and evidence, the court reaches the following conclusion: Google is a monopolist, and it has acted as one to maintain its monopoly." So Kevin, how did you react when you heard this news?

• 2:55Kevin Roose?

  Well, I immediately went to Google and performed a search through the Google Monopoly search product to learn more about this case. And it was truly, I mean, we've been talking about the various antitrust investigations and cases against Google. This was a big win for the Department of Justice, which has been suing Google for years. It also has another active suit against Google. But this was sort of the big one. This was the one that I thought had the best chance of succeeding. And yeah, the judge agreed.

• 3:26Casey Newton?

  Well, you know, for me, Kevin, I have been so worried about the state of the web, and I've been so worried that Google in particular has too much control over it, that when I saw this ruling, I thought, well, maybe this actually is a chance to press the reset button on the internet, at least to some degree, and maybe improve competition, at least in search. So we want to get into what this decision says, what it means for Google and the tech industry more broadly, and we wanted to bring in an expert. And so we've asked David McCabe, who has been covering this trial from the beginning for The New York Times, back onto the show to join our conversation. David, welcome back to Hard Fork.

• 4:03Speaker 03

  It is a pleasure to be back.

• 4:04Casey Newton?

  So heading into this decision, what was the conventional wisdom about how Judge Mehta might rule? And how did the decision that we got this week compare to that?

• 4:14Speaker 03

  So I think in some ways, it's hard not to be surprised or struck by the basic fact of like, a ruling like this hasn't happened against a tech company in 20-ish years. Right? I think the conventional wisdom was that he is a very rigorous and deliberative guy, and he would probably try and kind of split the baby, right? He'd give the government some stuff, he'd give Google some stuff. Ultimately... He gave the government the core of what it wanted. He did give Google some wins on some more minor questions. But when it came down to the center of the government's case, he ruled with them.

• 4:50Kevin Roose?

  So this is a big win for the DOJ. And I think we should just step back for just a second to remind people what this case was about. So this case was primarily about these sort of default agreements that Google has struck with Apple, with Samsung, with the manufacturers of other phones to pay them lots of money. In Apple's case, billions and billions of dollars a year for the right to be the default search engine on Apple devices, on Samsung devices, et cetera, et cetera. And the DOJ basically alleged that this was anti-competitive, that Google not only had a monopoly in search, but that it was preserving that monopoly through the use of these agreements. Google had argued essentially, and David, correct me if I'm wrong here, that these

• 5:35Kevin Roose?

  kind of placement deals are standard operating procedure, not just in tech, but in things like grocery stores where a company might pay to have their beverage placed on an end cap, that it's sort of not illegal to pay for promotion. And that was essentially what they were doing.

• 5:52Speaker 03

  That's right. And I just add, right on top of that, they said, like, we are being chosen as the default because we have the better product. The sort of quote from their lawyer that stuck in my mind was, "Google is winning because it is better." And that was at the core of their case.

• 6:06Casey Newton?

  And why, David, did the judge rule that it was illegal for Google to be making these payments to ensure that it is the default on iOS, on the Mozilla browser, and on so many other places?

• 6:20Speaker 03

  Well, he found that it did have anti-competitive effects. So he first determined that these agreements were effectively exclusivity deals. And so that kind of crosses the bar into exclusionary conduct. And then he said that there had been these, what they call in the law, anti-competitive effects, that they had hurt competitors in various ways, that had done things like limit innovation or allow them to raise ad prices far beyond where they should have been able to. And so those things kind of all added up to breaking the law.

• 6:46Casey Newton?

  According to the ruling, by 2020, Google had nearly 90% of all search queries and on these sort of general searches, and on mobile devices, it was almost 95%. So it seems clear that, you know, whether you think that there's illegal anti-competitive stuff happening or not, you cannot deny that no one has really been able to compete with Google.

• 7:11Speaker 03

  So the Justice Department had argued that these agreements helped drive like all of that search volume to Google, and that that gave them a lot of data, which allowed them to make their search engine better, which allowed them to make more money, which they then used to pay for these default agreements, right, and lock the process in. And this was like a central theme of the Justice Department case, this kind of like flywheel or feedback loop, whatever, you know, phrase you want to use from the business books.

• 7:39Casey Newton?

  Right.

• 7:40Speaker 03

  And the judge basically agreed with that theory, like he said that he is happening. You know, he clearly bought into the idea that the default payments were central to the decision by Microsoft not to invest in its search product being, obviously, they have invested some, but they haven't invested as much as maybe they could have. And he seemed to buy the idea that Apple, one reason they might not have invested in search, was the default payment. You know, he was a little bit less accepting of the GOJ's argument that Apple had like fully stayed out of search as part of some, you know, complicated arrangement with Google. But broadly, right, he found that that there was less investment in the search space and therefore less innovation.

• 8:23Kevin Roose?

  Right. If you are a young company that's maybe got aspirations of building a new search engine, you might not get very far because Google is just so dominant. And so why would VCs fund you? Why would you, you know, raise a bunch of money to go take on a company that has effectively locked itself in this position of dominance? That's the argument for sort of the pro-competition fallout from this ruling.

• 8:47Speaker 03

  Yeah. And he heard from a company, Niva, that had tried to do this, right? So he heard throughout the trial from people who had or either operating search engines and saying this is really hard, right, without the default position, or from startups that had said like, we tried to do it. We had a lot of money. Maybe we even built our own index, right? The index is the the set of data that makes up the results in a search engine, and we just couldn't do it. And he seemed persuaded by that.

• 9:11Casey Newton?

  I think there's a really interesting story here about the history of the Internet. So when you think about the dawn of the web, when Google is first formed, I think it's unquestionable that they were the most innovative. They did become the most popular because they were the best, and that turns them into this monolith. And then mobile devices come along, and other people want to try new things, and a lot of folks try. But Google just kind of has all of these prophets sitting around. And so they can go to all of these device makers and these new browser makers and say, hey, why don't we just pay you some money, become the default, and that'll be good for everyone? And it works. And there are these stats in the ruling, David, that just struck me so hard.

• 9:56Casey Newton?

  In 2017, Google estimated that its default placements, so these deals that it struck, drove more than half of its overall search revenue, which was a percentage that had grown since 2014. And in 2020, Google's internal modeling projected that it would lose between 60 and 80% of the queries it gets on iOS devices if it lost its default agreement on Apple devices. So why is that fascinating to me? To me, that is the DOJ's case in a nutshell. Google keeps saying, no, no, no, we're the best. The reason that people keep using us is because we're the best. And yet their own internal modeling has shown, if we do not make these agreements,

• 10:37Casey Newton?

  we will lose a staggering percentage of the queries that we're getting every year and of the revenue that we're getting every year.

• 10:44Speaker 03

  Well, a moment that really stuck out to me during the trial was DOJ brought in a behavioral economist at the very beginning of the trial. And he sat there and he told the judge all about the ways that defaults are really powerful. He just walked through the judge in very plain terms, like, you go to a coffee shop and the tip screen is there, and it gives you these options, and so of course you're going to choose one of those tip options.

• 11:10Casey Newton?

  And Kevin always chooses zero.

• 11:13Kevin Roose?

  But I do feel like they've gotten a little greedy with those options. I saw one for 25% the other day at a coffee shop. I'm like, I'm not tipping 25% on a coffee. Anyway, we're getting off track.

• 11:24Speaker 03

  I'm not touching tipping discourse.

• 11:28Speaker 03

  But like, he just like made these very plain spoken intuitive explanations of like, defaults are extremely powerful. And you could tell that the judge was into it. And, you know, the judge heard from a lot of witnesses. He read a lot of paper. He seemed extremely deliberative. I'm not saying he was like convinced in this moment, right? But you could tell that it was an engaging piece of testimony. And I think it spoke to something that people intuitively understand, which is that these defaults are extremely powerful in getting you to do something. Because why would you change?

• 11:56Kevin Roose?

  But how does this square with another thing that I understand that Google argued in this case, which is that, you know, in places where there is a choice given to users, such as the EU, which has tried to sort of make this, what they call a ballot screen, where basically you're setting up a new device. And it's like, okay, which, which search engine do you want to use as your default? And users actually have a choice. Google, from what I understand, argued that in those cases, people overwhelmingly choose Google because they prefer Google to Bing or DuckDuckGo or some other search engine. So how does this sort of square with the argument that you just laid out, Casey, which is that if these default deals went away, Google would lose some huge percentage of its search traffic and its search revenue?

• 12:39Speaker 03

  I mean, here's, Casey can respond to his argument in defense, but I like, there was an interesting kind of element to this trial, which was there was a little bit of an argument of a catch 22, which is like, the other search engines aren't as good as Google because they don't have the volume of queries every day that Google has, right? That search engines get better as they get more information about what people are looking for and how they're looking for it. And that without that, you can't be as good as Google. And that they can't have that volume without access to the default, right? And so this was basically what Microsoft was saying, was like, the default position is uniquely powerful in helping us build that kind of corpus of information that's at the center of a really good search engine. So.

• 13:23Speaker 03

  You know, they would argue, I think, that the quality isn't there on the other search engines, in part because of Google's behavior. And then, of course, Google comes in and says, well, of course, people choose Google because it's better. And they would say, well, it's better because you did this thing. Now the judge's role was against the law.

• 13:36Casey Newton?

  So Kevin, on this point of, hey, didn't the EU kind of try to fix this and it didn't work because everybody just loves Google? I do think there's something to that. I do think that if Google were not the default and you presented most users with a screen, they probably would just pick Google. But I was talking with somebody who works at DuckDuckGo, which is this, you know, rival search engine. And they were telling me, Google's been really good at just kind of getting around that restriction that they have to show the screen. For example, if you set a different default, but then you ever use Google, it'll be like, hey, do you want to switch back? And it will sort of always be prodding you to do that. And, you know, she sort of laid out some other ways that Google tries to kind of undermine that ballot screen. And that's one reason why they say that if the court decides that we should move in

• 14:20Casey Newton?

  this direction, you might actually need some kind of monitoring body that is keeping an eye on Google, saying like, hey, are you actually acting in the spirit of this agreement? Or are you trying to undermine it at every turn?

• 14:30Kevin Roose?

  Hmm. So, okay, now we have this big ruling. Google is a monopoly, according to this judge, and has acted illegally to maintain its monopoly power. What happens now? Like, is this going to be trapped in appeals for years and years and years? Like, what are the actual consequences of this?

• 14:50Speaker 03

  Trapped in appeals for years and years, for a thousand, Alex. No, I mean, Google has said they plan to appeal. A lot of people expect this or one of the other tech antitrust cases to go to the Supreme Court. I mean, that's a years-long proposition. Just to talk sequencing for a second, the judge has ruled that they broke the law. There will be a second phase to determine how to fix it. We'll know in about a month, like, how Google and the Justice Department want that to play out. In terms of this sort of remedies process, the judge has a lot of options, right? There's sort of two buckets here of traditional kind of fixes for antitrust problems. One is what are called structural, right?

• 15:33Speaker 03

  That's you break up a company, you have to spin off some unit. And then there are, quote-unquote, behavioral fixes, right? That's a judge says, you got to change X, Y, or Z about your practices.

• 15:44Casey Newton?

  Mm-hmm.

• 15:45Speaker 03

  You know, something that I think stuck out to me, and we all know this, I mean, covering it intuitively, but like, reading the ruling...

• 15:52Speaker 03

  Man, search is complicated.

• 15:53Casey Newton?

  Yeah.

• 15:54Speaker 03

  Like, there are so many interlocking pieces in these systems, both like internally at Google and externally between Google and other companies. And so like, he digested just a huge amount of information about a very complex thing.

• 16:08Casey Newton?

  Mm hmm.

• 16:08Speaker 03

  And I kind of expect he has to do that again with the remedy.

• 16:13Casey Newton?

  Yeah.

• 16:13Speaker 03

  Right? Like, you have this complicated set of contracts and they feed into these technical systems and you're thinking about all this while trying to devise a fix. And of course, the more complicated the fix is, the more monitoring you have to do, the more potential for unintended consequences. Because there are like all these deals and all these elements.

• 16:31Kevin Roose?

  Um, I want to talk about remedies here because I think that's a big question that I have is like, how do you actually fix the problems that the judge has identified? DuckDuckGo has released a list of remedy requests, including, uh, Google can't do these deals anymore. They also want, uh, Google to be forced to share, uh, API access for search and ads with competitors so that they have sort of the same data that Google itself has. They also asked for something called public education initiatives inside and outside of product. Basically, you, you can't just give people a choice about which search engine they use. You have to actually educate them about the costs and benefits of various search engines.

• 17:13Speaker 03

  So like a legally mandated like video that's like,

• 17:16Kevin Roose?

  Yes.

• 17:17Speaker 03

  Here's what?

• 17:19Casey Newton?

  No, we have to go, we have to go into elementary schools and teach children how to change their default browser. Why wouldn't we do that?

• 17:24Kevin Roose?

  They also asked for a choice screen that reappears regularly. So it's not just once when you set up a new phone or something that you are asked like, which search engine, like every month or so, maybe it'll, it'll pop up a new thing that says, are you sure you want to keep using Google?

• 17:39Casey Newton?

  And I will say, Apple is perfectly happy to show me a screen like every month that says, hey, do you still want to share all your photos with Google Photos? So don't tell me that that's a ridiculous thing that Apple would never do.

• 17:49Kevin Roose?

  So what do you make of these remedy requests, David?

• 17:52Speaker 03

  Well, so I obviously have this list in front of me. I think, putting aside for example, what the judge is going to do. It's an interesting list. I think the API access, is that, is that idea that it would be like access to the full search index?

• 18:06Casey Newton?

  Yeah, so you know DuckDuckGo actually uses the Bing Search API. So they want something similar for Google where they have access to the Google search index and they're able to kind of build around that.

• 18:17Speaker 03

  So this gets at, I think, a really interesting question in the world of antitrust law and this kind of space, which is can these, these laws and these remedies be used to sort of encourage innovation? The judge has kind of identified innovation, harms to innovation as a problem here. Right? But it's a different question of like, can, can these laws be used as a tool to make an economy that is not just like more competitive on price, but, but more competitive on innovation and the quality of products?

• 18:47Speaker 03

  And something that I've been thinking about a lot in this case is in the 1950s, the government required through a settlement with AT&T. AT&T has had, I mean, decades of antitrust litigation with the government. We all know about the breakup, but like it went on forever. And in the 50s, the government reached a settlement that said they had to make licenses and access to some of the Bell Labs technology, particularly the transistor, available to other companies on reasonable terms. Right? And there is an argument that that was really decisive in, in allowing the semiconductor industry to get built and, and thrive. And so I do think that if you look at how the Justice Department argued this case, they

• 19:31Speaker 03

  argued there was this cycle of dominance that Google was engaged in. And I would expect that they will try through this remedies process to do something they think will disrupt that cycle.

• 19:43Speaker 03

  And so something like other search engines could basically have Google's results quality.

• 19:51Speaker 03

  Tomorrow, when yesterday they only had Bing's kind of quality of results? Like, it is a really interesting question of like, would users be more likely to consider other search engines if they had access to Google's quality on a DuckDuckGo or another search engine?

• 20:07Casey Newton?

  I mean, personally, I would love to see this. One of the big tech conversations over the past couple of years has been this growing perception that Google Search is getting worse over time. And of course, one reason why a product will get worse over time is that it has no competition. You know, Ben Thompson made this great point in his newsletter this week where he talked about what happened when Apple decided to stop relying on Google Maps and stop making that the default on iOS and build its own Maps product. Now, at first, of course, the Maps product was very bad. I think there are still some people who are lost because they were using the Apple Maps product when it first came out. But over time, it got to be quite good. Many people believe it's now at parity or maybe even better than Google Maps.

• 20:50Casey Newton?

  And it is, because it is the default on Apple devices, more popular on those devices than Google Maps is. How has Google responded to that? Do they give up and walk away? No, they've invested massively in making Google Maps better and better. And I don't know one person who thinks that Google Maps is worse today than it was five years ago, right? So to me, that is such a powerful story about what you get when you have a competitive market. And we have not had it in search in a generation.

• 21:19Speaker 03

  Yeah. Something that was really interesting to me was the judge. It's not that he didn't agree that that could happen, but he did not find that that happened here. He was like, Google has continued to innovate, even in the absence of competition. And then he said something that I'm sure like struck some people who are supportive of Google, which was like, and that actually is behavior consistent with a monopoly, that they would continue to put money into R and D because they want to like, kind of maintain their advantage. And then he said, like, I do note that, like, they introduced Bard in response to ChatGPT, and that's an example of responding to competition, but notwithstanding, like, I largely find that they have continued to build and innovate, even in the absence of competition.

• 22:00Casey Newton?

  To me...

• 22:00Kevin Roose?

  One of the clearest pieces of evidence that Google is a monopoly is that they were totally caught flat footed by Bing, by ChatGPT. They had this like, they were late to incorporating all this generative AI stuff into their products. And yet Bing did not take market share from Google to any appreciable degree. Like, Google still maintained its dominance, even though it had more competition theoretically than it has had in years. Like, the numbers just didn't budge all that much.

• 22:29Speaker 03

  I mean, generative AI kind of hung over this case, but because it originated now four years ago, basically, and concerned conduct from like the last 15 years, it was not really a part of like the record in the case, right? They weren't fighting over ChatGPT. So occasionally someone would like reference it, but they also like couldn't get really deep into it. But it's the obvious thing that hangs over this remedies discussion, right? This is clearly a sort of pivotal moment for Google and for the search business. And does the remedy kind of change the game in a way that either allows, you know, new AI players to thrive or or does not?

• 23:08Casey Newton?

  I just want to say, every time one of the two of you says remedy, I think of the song Remedy by Jason Mraz, and it's driving me insane. Okay, let's talk about what this means for the tech industry more broadly, right? Because as this remedy, my brain goes again.

• 23:27Speaker 03

  We could say fix, too.

• 23:33Casey Newton?

  As this remedy gets decided, many companies are going to be affected. So, I mean, let's start with who could maybe be the biggest loser here, which is Apple, which is getting around $20 billion, I think, at last count from this deal. What do we think happens to them if the judge says, hey, you're not allowed to take that money anymore?

• 23:54Speaker 03

  I mean, I think, do they keep Google as the default? Is the real question, right? Like, does not getting the money, if they did, they stopped getting the money, change their behavior and their thinking, right? Or not?

• 24:07Speaker 03

  Obviously, it's not good for their, it's some percentage of their services revenue, right? I mean, you may have that number in front of you, but I, I think more broadly, the question is like, do they decide to invest in search in some way, or, you know, work with OpenAI or something like that?

• 24:20Casey Newton?

  What do you guys, okay, so let's talk about the Mozilla Corporation.

• 24:23Kevin Roose?

  Yes.

• 24:23Casey Newton?

  So Mozilla makes the Firefox browser, that's the thing that it is best known for. And in its heyday, Firefox was, you know, one of the most popular browsers on earth. It's really kind of fallen off, I think, over the years. But the Firefox community is relatively big, and Google pays Mozilla hundreds of millions of dollars a year to be the default. And that is by far the largest source of Mozilla's revenue. Like, 80% of Mozilla's revenue is coming from this search deal. What happens to Mozilla in a world where it loses 80% of its revenue? Is that just the end of Mozilla?

• 24:56Kevin Roose?

  I think it's the end.

• 24:56Casey Newton?

  Yeah.

• 24:57Kevin Roose?

  Yeah, I mean, David, do you think they can survive? I think it was like 80% of their revenue, you know? How do you survive that?

• 25:04Speaker 03

  I mean, it's an obvious question. I think they were obviously like,

• 25:09Speaker 03

  they were witnesses at the trial. My memory is they were pretty favorable to Google in their testimony.

• 25:15Kevin Roose?

  We would like the company to be able to continue paying us the vast majority of our revenue.

• 25:19Casey Newton?

  It's so interesting, because, you know, the Mozilla, like, it is a nonprofit. And, you know, I think there are a lot of really, you know, conscientious civic-minded people who work there, who believe in a sort of a healthy and open internet, and, you know, who I think I sort of share a lot of values with. And yet, it has become such a sprawling and weird company that, you know, every couple of months, they're emailing me with some new initiative, you know, they tried to build a phone at one point. Like, these folks completely lost the plot. And I almost see it as kind of a parallel to Google, where just as these sort of monopoly profits drove Google, you know, into laziness, I think, over time when it came to search, so too do I think Mozilla, just having hundreds of millions of dollars coming in the door every year for doing nothing,

• 26:03Casey Newton?

  and not really having to compete on the basis of the quality of their browser anymore, actually ruined the product.

• 26:08Kevin Roose?

  Yeah.

• 26:09Casey Newton?

  That's kind of a side rant, but I've been wanting to say it for a couple of years. Okay.

• 26:14Kevin Roose?

  I have sort of a big picture question, David, because there's been this feeling in the tech industry over the past couple of years that this sort of aggressive antitrust stance that the DOJ and the FTC and all these federal agencies have taken under the Biden administration with people like Lena Khan and Jonathan Cantor at the DOJ, that they've been not getting a lot of wins. Does this ruling sort of invalidate that? Does it make you think that actually this approach to antitrust is working?

• 26:47Speaker 03

  You know, it's a question I think about, not surprisingly, a lot, right? And particularly, how do you evaluate if any of this is quote unquote "working" or what it's doing? And I think at the very least, this is undeniably a massive win for the Justice Department, right? I mean, they went in, obviously, the case was brought in the Trump years, but it was tried by the team under Jonathan Cantor, who is sort of the equivalent to Lena Khan at the FTC. And there has not been a win like this for the government since the Microsoft case. So undeniably, on the question of their record and the narrative around their record, I think this is like, they've got to be thrilled, right? I think the question of, has it worked or what has it done?

• 27:31Speaker 03

  Really will depend a lot on the remedy the court decides to impose here and just how the industry reacts, right? I mean, something that, if you went to the FTC and the DOJ today and you were like, "Seems like you're losing a lot, like, can you show me why you're winning?" One thing that I suspect that they tell you, and certainly the people who are very close to them would tell you, is there's a deterrent effect, particularly on mergers, that there's less deal making happening, that more companies are sort of like trying to do it on their own versus getting bought by a big company. It's an interesting question of whether or not there'll be a similar effect here, right? Just the fact of the ruling causes people to act differently. Does it cause some investor to be like, "Oh, maybe there is an opportunity in search, maybe there is a window here"?

• 28:15Speaker 03

  Does it cause Google to say like, in the millions of product meetings that make up the soul of a big tech company, right? Like, does it cause them to be a hair more cautious and potentially create an opening for a competitor? Because that's part of what has been argued in the Microsoft case, was like, one of the reasons it opened the door for Google, was that they just had gotten more cautious and more careful, and they were so freaked out by the experience of tangling with the DOJ for years.

• 28:43Speaker 03

  You know, this hasn't, not to be too, I'm not trying to dodge the question, but this hasn't happened in more than two decades, right? And so like, we'll have to look and see how Google responds and how people at Google respond.

• 28:54Casey Newton?

  So if there's no further business, I need to go listen to Remedy by Jason Mraz.

• 29:00Kevin Roose?

  All right, David, thank you so much for joining us.

• 29:02Speaker 03

  Thanks, David. Thank you for having me.

• 29:07Speaker 03

  When we come back, bubble, bubble, toil and trouble.

• 29:10Casey Newton?

  People are worried we might be living through some inflated valuations in AI, Kevin.

• 29:30Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner. Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 30:05Speaker 04

  AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real time guardrails by team and model, helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.

• 30:36Speaker 05

  Ripple has been powering modern finance for a decade. It's not our first trip around the blockchain, and it won't be our last. We're leading the charge to bridge the gap between existing financial systems and the future of finance. You can have it all. Legacy, stability and digital innovation. Systems that work with your other systems. Because with the right partner, the question isn't crypto or traditional finance. The question is simply, why choose? A new standard of modern finance. It's happening with Ripple. For informational purposes only. Availability of products can vary and may be provided by different Ripple entities.

• 31:11Kevin Roose?

  Well, Casey, the other big news this week is that the markets are going a little nuts and people are starting to worry about it. And what makes this a story that I want to talk about with you is that a lot of the sort of panic and worry right now is about this question of whether we are in an AI bubble that is bursting.

• 31:32Casey Newton?

  Yeah, this discourse has gotten really hot this week, and of course, that is a question right in our wheelhouse. And so I'm excited to hear the news and what you think about it.

• 31:42Kevin Roose?

  So just for some background, at the start of this week, markets in the US and around the world were really volatile. The most valuable tech companies, what they call the magnificent seven.

• 31:52Casey Newton?

  Which, by the way, didn't we know that they were called that before this week? That was a new one for me.

• 31:57Kevin Roose?

  That was a new one for me, yeah. But it's also like a great name for like some sort of a figure skating troupe or something. I don't know.

• 32:03Casey Newton?

  Without looking, how many of the magnificent seven can you name?

• 32:06Kevin Roose?

  I think I could name all of them.

• 32:07Casey Newton?

  All right, let's do it.

• 32:09Kevin Roose?

  Microsoft, Apple, Google, Nvidia, Meta. Anyway, we're getting off track here. Okay. So at the start of this week, the magnificent seven lost roughly $1 trillion in market value. In the US, the S&P 500 fell 3%. In Japan, the Nikkei had its biggest one-day point drop in history. So it's just a lot of volatility. People are getting worried about their stocks and people are trying to figure out what the heck is happening. And some of it has nothing to do with tech, right? That like there's this sort of jobs report for July showing higher unemployment than a lot of economists expected. There's also some factors about the Fed and interest rates that people are saying have

• 32:53Kevin Roose?

  played into these market moves. But the thing that makes it a story for us is that a lot of the focus this week has been on AI and the role that it is playing in all of the market chaos.

• 33:04Casey Newton?

  Yes, now, I mean, the one thing that I would say is that while it is true that stocks were down earlier this week, it did seem like they more or less immediately bounced back, at least part of the way. In fact, as we're recording this, I think Meta is actually up for the week. So what you said is true, there is a lot of volatility and this has been, I think, like, one of the biggest subjects of the week is just what is going on with AI. I did just also want to note that.

• 33:29Kevin Roose?

  Yeah, so I want to sort of spend some time kind of deconstructing the argument about an AI bubble, because I think it's really interesting and important. And it is sort of the big question that is kind of hanging over the industry right now is like, is any of this sustainable? Is any of the hype that we're seeing around generative AI actually pushing the tech industry forward? Or is it just like, you know, the metaverse and crypto and all these things that sort of come and go? So let me just sort of lay out the steel man arguments for the case that we are in an AI bubble. And I want to hear your responses to them. So the first argument that you'll hear from people who think that we're in an AI bubble that is about to pop or that might pop soon is that the economics of all this are just unsustainable and they're actually getting worse.

• 34:14Kevin Roose?

  So far, in the first six months of the year, Microsoft, Alphabet, Amazon and Meta have spent a combined $106 billion. And that's probably too low an estimate because that only includes their capital expenditures, things like building data centers, acquiring GPUs, things like that. That doesn't include like all the engineer salaries and stuff like that.

• 34:34Casey Newton?

  And the ad that Google ran during the Olympics.

• 34:36Kevin Roose?

  Yeah, that was probably at least a couple billion right there. And that spending is not slowing down. In fact, it is expected to grow. We know that AI companies are going to spend a ton more money in the coming years, maybe as much as a trillion dollars, according to some estimates, to build out these AI systems. That is just a lot of money. And it's not clear that they have a plan to recoup those costs, right? You can't sell enough $20 a month ChatGPT Pro subscriptions to recoup the costs of spending hundreds of billions of dollars training these models. And unlike some technology where you spend a bunch of money upfront to build something and then you can kind of sell it over and over again with no marginal costs to your company,

• 35:20Kevin Roose?

  AI spending actually scales as the products get better. Like, it costs more to train the better models as they get better and more powerful. The costs actually go up. What do you make of that argument?

• 35:32Casey Newton?

  Well, I think that that argument has some strengths to it. I would say a few things in response. One is that I think some companies will win and some companies will lose. It's almost certainly the case that some of this vast amount of money that is being spent now is being wasted. I think it's also very likely that one or more of these companies is going to figure it out and be able to make some of that money back. And in fact, the way that this capital is being deployed is out of a sense that essentially one company is going to be the first to create this artificial general intelligence and that company might become the most powerful company that ever existed, right? So that is purely speculative and maybe we never get there. But if it turns out to be true and you were like Andy Jassy

• 36:17Casey Newton?

  or Mark Zuckerberg and you said, eh, we're gonna sit this one out, history is gonna look at you as like one of the worst CEOs of all time, right? Like, what was the point of making all these monopoly profits if you weren't investing them into trying to build the next thing?

• 36:29Kevin Roose?

  Right. And this is sort of what tech CEOs have been bringing up on earnings calls over the last few weeks. Google CEO Sundar Pichai said, quote, "The risk of underinvesting is dramatically higher than overinvesting." Basically, we'd rather spend too much to develop AI than not spend enough and get beaten by someone else. Mark Zuckerberg also made a version of this argument, too. So this is something that is very much out there, but the tech companies are saying, basically, the real danger here is missing the boat, not spending a few billion more than you need to.

• 36:57Casey Newton?

  Yeah. Now, where the economics do seem to be unsustainable, Kevin, is for the smaller companies, right? The startups, like the inflections and the adepts and the character AIs, companies that had raised tens or hundreds of millions of dollars and were once valued at a billion dollars or more, they're throwing in the towels, saying, hey, we're gonna figure out some weird way to sort of get reverse-acquired back into these tech giants because as they look at their own balance sheets, they figure, we're not actually gonna be able to hit AGI before we run out of money. So in that case, it does seem like there is a bit of a bubble bursting.

• 37:31Kevin Roose?

  Yeah. Okay, the second argument that we are in an AI bubble is just that AI progress is slowing down. You'll hear people saying this, like, we're just not seeing the kind of rapid capabilities gains that we saw a couple of years ago. And so maybe we've already kind of gotten most of the juice out of these systems. We're close to tapping out the potential of this technology. What do you make of that?

• 37:54Casey Newton?

  Well, so I mean, I am sympathetic to this argument in some ways because that has also been my frustration is that we have been promised these, you know, perfect digital assistants and so many other things and they are kind of stuck where they were in 2022, or at least it feels that way. I think if the, you know, AI companies were here, they would point to a million things that they've improved since then. And I think there's a lot to that. But to me, like, the real test of this question is, what happens when these next generation frontier models come out? Once we see what GPT-5 looks like or Llama 4, do we actually see that step change in capability that reassures us that like, oh, no, we actually are on a glide path to creating a super intelligence? Or no, no, no, that actually was just hype

• 38:39Casey Newton?

  and we've run out of ideas. That is just truly a question mark. I don't know the answer to it. But again, it is in the nature of business people and of venture capitalists to try things. And when you think about like, what could a transformative technology be? Hard to think of one that would be much more transformative than that. So it doesn't really surprise me that people are continuing to invest in it despite that uncertainty.

• 39:01Kevin Roose?

  Right. And this, I would say, would be like contingent, this argument on what happens in the next six or 12 months, right? If OpenAI comes out with GPT-5 and it's as much improved over GPT-4 as GPT-4 was over GPT-3, if we continue to see like large capabilities increases from these models, I think that will be a signal that we just needed to kind of wait and give it a minute. But yeah, I think if the next models that come out of these big AI companies are just sort of incrementally better than the ones that we already have, I think a lot of people are going to start to say, well, why are we spending hundreds of billions of dollars to train these things again? Okay, that's the second argument. AI progress has slowed. The last argument that I hear a lot from people who are sort of

• 39:43Kevin Roose?

  skeptical of all the hype around generative AI, who think that we are in a bubble, is just essentially that we haven't found the right uses for this product yet, that it doesn't have product market fit. Elliott Management, which is one of the biggest hedge funds in the world, sent a letter to investors recently that basically made this point. It says, there are few real uses. For generative AI, other than summarizing notes of meetings, generating reports and helping with computer coding. People will also say, you know, look at how many people are just using these tools to sort of cheat on their homework. There have been some people who have observed that the usage of ChatGPT tends to go down during school holidays.

• 40:24Kevin Roose?

  Basically just saying like, this thing, it's maybe useful if you're a student, but it's not useful to sort of real professionals doing real jobs because, you know, it makes too many errors, it's too unpredictable and the tech just isn't very useful yet. What do you make of this?

• 40:37Casey Newton?

  This one I just think is silly, you know, it's like when you're like, well, you know, all it can really help with is computer coding. That's actually really important. And as it improves over time, the goal of these companies is that you create an AI system that improves the AI itself, right? They call this recursive self-improvement or RSI. But once you're able to get into that zone where the software itself is improving the software, then you enter this sort of liftoff state where it really does seem like, you know, we might achieve AGI. So when Elliot comes along and says, well, I don't know, you know, it seems like all it's really good for is like creating virtual software engineers that do the job of a human being. It's like, well, that actually seems sort of important to me.

• 41:21Casey Newton?

  I would also say if you don't think it's useful, well, ask a high school student if they think it's useful. Ask a high school student if they would miss it if it went away, right? So I think we've already discovered a pretty significant utility space for this stuff. I am a lunatic and probably pay for six or seven different AI services right now. And like, there isn't one I would cancel, you know?

• 41:42Kevin Roose?

  Wow, you you pay for six or seven different AI services?

• 41:45Casey Newton?

  Gemini, ChatGPT, Claude, I pay for it in Notion. I pay for it in Readwise, which is like a read it later app. I pay for it in Lazy, which is another kind of reader app, and I pay for it in Raycast, which is a launcher app. And, you know, are those services redundant and competing with each other? Yes. And am I doing this in part because I think it's part of my job to understand, you know, the state of the art in software? Yes, but again, I like all those things and I think the AI makes them better and I think they make me better at my job.

• 42:14Kevin Roose?

  Yeah, that's interesting.

• 42:15Casey Newton?

  Yeah.

• 42:15Kevin Roose?

  I do wonder if there's sort of a mismatch between what people want from these tools and what the companies are marketing them for. Like, there was this study that I wrote about that came out of something called the Data Providence Initiative at MIT a few weeks ago. And they analyzed a bunch of different transcripts of conversations that people had had with chatbots that were part of this sort of data set. And they found that a significant number of conversations that people had were like sexual or romantic, right? It was like that was what people want out of these things. But that is not what the companies want them to be used for. They want these to be your office copilot, your productivity enhancer, your coding assistant. And so maybe there's just a sort of a mismatch there. But I do think we are starting to see

• 42:57Kevin Roose?

  some more questions about the utility of these things. And I would just point to the ad that Google ran during the Olympics for Gemini, where it's like not quite clear what they want this to be used for. Is it writing letters to your favorite athlete? Is it helping you with some problem at work? Like, the sort of use case for generative AI, I don't think the companies have been articulating that very well.

• 43:20Casey Newton?

  I think that is true. And yet, we cannot assess the quality of AI based on whether Google's marketing is good, Kevin.

• 43:28Kevin Roose?

  Yeah.

• 43:28Casey Newton?

  Because if we've learned one thing about Google over the years.

• 43:32Kevin Roose?

  That's true.

• 43:32Casey Newton?

  It is that their marketing is absolutely unhinged.

• 43:34Kevin Roose?

  Right.

• 43:35Casey Newton?

  I mean, you know, here's where I will agree with some folks who say that we're in a bubble. I think it is true that lots of people try ChatGPT once or twice and can't figure out what to do with it. I think it is true that it is unreliable in ways that make it quite bad for many different use cases. And so people who've tried it and either sort of couldn't make heads or tails of it, or just felt like it sucked and they didn't want to use it anymore. I'm actually like very sympathetic to it. But our job is to try to see in the future a little bit and imagine what things will be like assuming it does get five or 10 or 50% better, because it is our experience that it almost always does get five or 10 or 50% better. And so I would just not draw too strong of a conclusion based on the state of the art in August 2024.

• 44:19Kevin Roose?

  Yeah. And I'm not sure which side of the argument my next point is actually going to support here. But I would just point out that tech adoption is often very slow, right? Like, one of my favorite stories about slow tech adoption is the microwave oven. So the first microwave oven was invented in the 1940s. Do you know who invented the first commercially available microwave oven?

• 44:38Casey Newton?

  Was it Frank Microwave?

• 44:41Kevin Roose?

  No, it was Raytheon.

• 44:42Casey Newton?

  Okay.

• 44:44Kevin Roose?

  So they actually dropped them over Berlin to try to win World War II.

• 44:48Casey Newton?

  Exactly.

• 44:49Kevin Roose?

  So Raytheon invents this microwave oven, the first commercially available one goes on sale in 1947. In 1971, almost 25 years later, less than 1% of US households had a microwave oven in their house. They didn't achieve sort of mass adoption in the US until the 1990s. So the point is that technology, even if it exists, even if it makes a sort of measurable difference in people's lives, people's habits are just very ingrained. And often it just takes a long time for people to cycle through one technology to get to the next thing. And so ChatGPT was one of the fastest growing apps in history.

• 45:29Kevin Roose?

  It got to 100 million users within two months, which is like insane by sort of historical standards for these kinds of things. But I think it could still take a lot longer for people to sort of adopt this in any sort of mass way than the AI companies think. Because we know from history that it just, it sometimes takes a long time.

• 45:47Casey Newton?

  And that reminds me, Kevin, I've been meaning to tell you, you really need to cancel your AOL dial-up subscription. You're not using that modem anymore and it's time.

• 45:55Kevin Roose?

  But I like the CDs they sent me. So Casey, after laying out the arguments for and against the existence of an AI bubble, like where do you stand on this? Do you think we're in a bubble?

• 46:06Casey Newton?

  Um, I think, I'm like, I'm sympathetic to the idea that we may be in some sort of AI startup bubble. Okay, I think that it's seeming less and less likely that a venture-backed company that wants to sort of rival a Google or a Microsoft in creating AGI is actually going to get there. And I think that's like kind of sad. My hope was that there was going to be a lot more opportunity there and so far it doesn't seem like it's been manifesting. On the public side, I don't think we're in a bubble at all. Again, I assume that a lot of this money is going to be wasted, but I assume that corporations are wasting billions of dollars every single day. Like, that's not a surprise to me, right? But if you ask me, do I think that they are going to get somewhere that like makes them a lot of money?

• 46:51Casey Newton?

  Like, yes, I do. What do you think?

• 46:53Kevin Roose?

  So I think the synthesis here and where I'm sort of coming out on all this is that you can have a bubble, but also that bubble exists within a sea change.

• 47:05Casey Newton?

  It exists within a context.

• 47:09Kevin Roose?

  So I would compare it to like the dot com bubble of the late 90s where you did have a bunch of silly overvalued companies that were raising huge amounts of money and didn't really have a business that was making any of that money back. And a lot of those companies went broke or got sold for pieces or whatever. And that was a bubble. But at the same time, the internet was not a bubble, right? The internet was a transformative technology. And so we did not see the internet go away when the dot com crash happened. And I would compare our moment in AI to the dot com bubble in the sense that yes, there will be a washout. I totally agree with you. Like some of these, these sort of mid-level

• 47:50Kevin Roose?

  AI startups will never make back the money that they raised and will probably end up crashing out or being bailed out. But I also think that we can't ignore the sort of sea change happening underneath that. And so, yeah, it may take a while. Investors may lose some patience along the way. But I don't think you can sort of put this one back in the bottle.

• 48:10Casey Newton?

  And to be clear, you're not saying that we are on the verge of like a dot com style wipeout, right? Where like it hits the public markets and like really affects the economy?

• 48:18Kevin Roose?

  No, because most of the companies that are sort of washing out, these mid-level sort of AI startups are private companies. They're not, you know, being traded on the stock exchange. The companies that are public, the Metas and Amazons and Googles of the world, they have, you know, these massive piles of cash that they're sitting on and they have these other businesses that are kind of subsidizing their investments in AI. So will Nvidia be able to sustain its current stock price? I have no idea. But I think the idea that these companies are all going to come crashing down because they spent too much developing large language models is a fantasy.

• 48:49Casey Newton?

  The last thing that I'll say is that I had this formative experience, which is I started covering tech in 2010, which was right when the, the mobile boom hit, right? Like everyone started to get a really high-powered smartphone and we saw dozens and dozens of companies get born, right? And all throughout this time, people were saying, are we in a bubble? Are we in a bubble? And some startup would go under and people would say, aha, we're in a bubble. And as someone, and I can't remember who, but I recall someone making a joke during that time. And it was something to the effect of, you know, reporters have successfully called 25 of the last three bubbles. And the idea is people are always saying there's a giant bubble, right? And I think honestly, it just shows you how people like still are traumatized to this

• 49:33Casey Newton?

  day by the dot com bust. But I think it should also give us pause when we walk around saying we're in a bubble because I think recent history has shown that the tech industry is like here to stay and has generally, you know, been pretty good about figuring out where to spend its capital.

• 49:47Kevin Roose?

  Yeah. I do wonder though, if the investing community has enough patience to wait this out.

• 49:56Casey Newton?

  Sure. But also it's like, you know, what choice do they have other than to wait and see? You know, it's not like there's some other hot technology on that they're passing, you know, while they spend all their money on AI. It's like, what are they going to invest in? Podcasts?

• 50:07Kevin Roose?

  Let's spend a hundred billion dollars developing a podcast. I think that could be good.

• 50:13Casey Newton?

  Can you imagine the studio we'd have?

• 50:14Kevin Roose?

  Oh, my God.

• 50:19Kevin Roose?

  When we come back, we're going to take a ride on the Hot Mess Express.

• 50:35Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure, and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner. Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 51:09Speaker 04

  AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real-time guardrails by team and model, helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.

• 51:41Speaker 06

  Today's AI builders need infrastructure that scales with their ambitions. Irene is a vertically integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of LAN and grid-connected power in renewable-rich regions globally. Irene's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. Irene, AI cloud for the next big thing.

• 52:16Kevin Roose?

  Well, Casey, all aboard!

• 52:18Casey Newton?

  The train's pulling into the station, Kevin.

• 52:20Kevin Roose?

  Yes! So today, we are playing a brand-new game called "The Hot Mess Express."

• 52:29Kevin Roose?

  And we even have a theme song for it.

• 52:41Kevin Roose?

  So Casey, this segment came to us during a brainstorm recently because-

• 52:46Casey Newton?

  A trainstorm, even.

• 52:48Kevin Roose?

  Trainstorm. Because as we were looking at the tech news of the week, there were just so many stories that struck us as just being very messy.

• 52:55Casey Newton?

  Very messy week.

• 52:57Kevin Roose?

  And so we decided to call it "The Hot Mess Express."

• 53:00Casey Newton?

  Yeah, and so how this is gonna work is we have a number of stories that are gonna be coming down the tracks, and when we arrive at the station, we're gonna sort of tell you a little bit about them, and they weigh in on what kind of mess this is, and we have a scale, and we'll be using it to assess sort of what was the week's biggest mess. And that's "Hot Mess Express."

• 53:17Kevin Roose?

  And if you're watching us on YouTube and you see the toy train set in front of us, and you're wondering, "Where did that come from?", well, I'll tell you. I stole it from my child.

• 53:27Kevin Roose?

  I'm not proud of that, and if you're wondering, "Did he melt down as I was leaving the house this morning with the train set?", the answer to that is yes.

• 53:35Casey Newton?

  Well, I'm sure I'll have a lot to say about that in therapy in a few years, but in the meantime, we thank him for his sacrifice.

• 53:42Kevin Roose?

  The train conductor hat looks great on you.

• 53:45Casey Newton?

  It looks great on you, yeah. Again, if you're not watching the show on YouTube, you're really only getting half the experience of this for it. All right, well, shall we see what's coming down the tracks first here, Kevin?

• 53:55Kevin Roose?

  Let's see what's coming down the tracks!

• 54:00Kevin Roose?

  Casey, it's so good. I love this. Oh, man. Okay. So I'll reach into the box car here.

• 54:08Casey Newton?

  Oh, yeah, there is a box car that has slips of paper that contains the week's stories.

• 54:13Kevin Roose?

  First up on the Hot Mess Express, this one's called Open Desks at OpenAI. This one, of course, we should preface with our standard disclosure that the New York Times has sued OpenAI for copyright infringement. So this came out of OpenAI this week. There was some news that at least three high-level leaders were either leaving the company or taking a leave of absence. John Schulman is one of them. He is an OpenAI co-founder who had been pretty central to ChatGPT. He is leaving to go to Anthropic, the rival company where he's going to work on AI alignment. And he said in a statement that he was not leaving OpenAI due to lack of support for

• 54:55Kevin Roose?

  alignment research, alignment being the sort of area of AI research that is concerned with how to make these models behave. But he's still leaving to go over to Anthropic. The information also reported that another OpenAI co-founder, Greg Brockman, is taking an extended leave of absence through the end of the year, but plans to return. They also reported that Peter Deng, who was one of their product leads over there, has left. And obviously, this came on the heels of a bunch of other people who've left OpenAI in the last few months, including Ilya Sutskever, one of the co-founders. Putting aside Greg Brockman, who is just sort of on leave for now, only two of OpenAI's 11 co-founders are actually still working there day to day. So Casey, what is going on at OpenAI?

• 55:40Casey Newton?

  You know, I wish I knew. And in fact, a former OpenAI employee messaged me this week and said, what's going on at OpenAI? So that's, it's not just the reporters who are asking that question. You know, at this point, Kevin, more co-founders have quit OpenAI than most companies have co-founders.

• 55:59Kevin Roose?

  That's true.

• 56:00Casey Newton?

  And not only that, but this departure of Greg Brockman, I think, is super interesting. You know, Brockman is a really integral part of- This company working, you know, hand in hand with Sam Altman, the CEO and Mira Maroti, the CTO. And, you know, I generally find him to be a pretty voluble person, got a lot to say. And so when he just showed up on X and said, uh, by the way, I'm going on leave of absence, I'll be away for a while, but AI is still super important and we got to complete the mission, which is what he did. I thought, this is weird. Like, why take this leave of absence now? And I have some irresponsible speculation, but I wonder if you have any to share first.

• 56:37Kevin Roose?

  No, I want to hear your irresponsible speculation.

• 56:39Casey Newton?

  I'm just going to tell you what this sounds like to me. Okay. Because I think, look, if they wanted to get rid of him, they would have. And if he wanted to quit, he would have quit. What this sounds like to me is, he wanted to quit and somebody said, don't quit, take a break, and then we will revisit this later. And he sort of walked out the door and that sort of explains why this felt sort of abrupt and why he did not have a lot to share. You know, he didn't share any warm and fuzzy feelings about OpenAI when he said he was taking a leave of absence. So that's just kind of what it felt like to me, but I'm curious what it felt like to you.

• 57:08Kevin Roose?

  Yeah, I think there's a story here that will come out in the coming weeks. I've been asking around people at OpenAI. Greg was the surprising departure here. John Schulman, people sort of suspected that he might have been kind of, you know, disaffected or just sort of had some questions about leadership since their big board drama last year. And so people are obviously sad to see him go. He was well-liked there, but they were not shocked. But Greg, I think, was, it's safe to say, that many people did not see coming. You know, I think it's possible that he is just taking a leave to kind of recharge himself because he thinks that, you know, 2025 is going to be a really big year and he's going to be really busy. He said in his statement that this was his first time to relax

• 57:51Kevin Roose?

  since co-founding OpenAI nine years ago. So maybe there is an innocent explanation, but it does seem like they are losing a lot of key personnel recently. And I don't really know what to make of that, except that it seems to be a drama factory over there.

• 58:03Casey Newton?

  Well, I keep a file on my computer called OpenAI is a Weird Company and I added this story to the list because I thought, once again, I'm over here scratching my head.

• 58:11Kevin Roose?

  What else is on your OpenAI is a Weird Company list?

• 58:12Casey Newton?

  Oh my God, well, let me pull it up for you right now. Let's see.

• 58:17Casey Newton?

  Where's my subtitles?

• 58:20Casey Newton?

  Here are some things that I have. Um, OpenAI illegally stopped staff from sharing dangers. Whistleblowers say, OpenAI employees say it failed its first test to make its AI safe. Microsoft and Apple drop OpenAI board seats amid antitrust scrutiny. Employee says he was fired for raising security concerns to the board. The opaque investment empire making OpenAI's Sam Altman rich. It's just, again, a lot of this stuff, it's like fairly innocent. One, but it's like, oh, this happens, that happens. But you put it all in a file, Kevin, and you say, it's a weird company.

• 58:51Kevin Roose?

  It belongs on the hot mess express. This company, for many of its victories over the past few years, it is a hot mess. How would you rate its mess? Is it on our scale?

• 59:02Casey Newton?

  Okay, so look, I don't think this is quite as messy as some of the OpenAI stories we saw last year. So Kevin, I'm going to give this a rating of a warm mess. I think it's a warm mess.

• 59:10Kevin Roose?

  Okay, yeah, I agree with that. I think it's a warm mess that has the potential to become a hotter mess, depending on what we learn about these executive departures.

• 59:18Casey Newton?

  That seems very fair. Now, Kevin, do we have another mess coming down the tracks? Oh, there it is. Let me just run down to the station. Now, this was a very fun one, Kevin. And this was X introduces its innovative new forced advertising plan. Do you hear about this one?

• 59:34Kevin Roose?

  I did.

• 59:35Casey Newton?

  So X filed a lawsuit this week in a US district court in North Texas against GARM. That, of course, is the Global Alliance for Responsible Media. It is an advertising industry group that tries to make sure that ads are brand safe, right? That, you know, your ad isn't going to run next to hate speech or violence or gore, that sort of thing. And X filed a lawsuit against GARM as well as individual companies like Unilever, Mars, and CVS. And you're probably wondering, well, what did GARM do?

• 1:00:07Kevin Roose?

  I assume they caused.

• 1:00:10Kevin Roose?

  Irreparable Garm? Is that a thing?

• 1:00:12Casey Newton?

  There have been allegations of irreparable Garm. That's what's being alleged.

• 1:00:19Casey Newton?

  So, Garm had encouraged advertisers to avoid X after Musk bought it. And this happened in November of 2022, which was like the month after Elon bought the former Twitter. During this boycott, 18 members of Garm stopped advertising on the platform altogether, according to the lawsuit, and dozens of others reduced their spending by 70% or more, according to the filing, took that from the great story about this issue by Kate Conger in the New York Times. Now, what's weird, Kevin, is just last month, X reinstated their own membership in Garm and said, "X is committed to the safety of our global town square and proud to be part of the Garm community."

• 1:00:59Casey Newton?

  But the CEO of X, Linda Iaccarino, posted a video statement about this situation on X this week, Kevin, and well, she had some troubling things to say.

• 1:01:09Kevin Roose?

  Yeah, we should watch it.

• 1:01:10Casey Newton?

  Yeah, let's watch this video.

• 1:01:13Speaker 07

  Hey, everybody. I was shocked by the evidence uncovered by the House Judiciary Committee that a group of companies organized a systematic illegal boycott against X. It is just wrong, and that is why we are taking action. Today, we filed an antitrust lawsuit against the Global Alliance for Responsible Media, four of its key members, and the World Federation of Advertisers.

• 1:01:41Kevin Roose?

  Okay, let's cut it off there.

• 1:01:42Casey Newton?

  That's enough of that. The first thing you need to know about this video is that Linda Iaccarino is wearing two necklaces. One says mama, and the other says free speech.

• 1:01:51Kevin Roose?

  I saw.

• 1:01:53Kevin Roose?

  What do we make of the free speech bling here?

• 1:01:56Casey Newton?

  Well, you know, I found it ironic because when someone chooses not to advertise on a platform in protest of the awful things that appear there, to me, that would be a pretty pure expression of free speech.

• 1:02:07Kevin Roose?

  Yes.

• 1:02:07Casey Newton?

  You know? And so for X to turn around and say, "No, if you're organizing with others and saying, you know, don't advertise on this website because of its policies," that seems like a pretty anti-free speech thing to do, Kevin.

• 1:02:20Kevin Roose?

  Sure does.

• 1:02:20Casey Newton?

  What do you think?

• 1:02:21Kevin Roose?

  Yeah, I mean, that was the sort of widespread mockery of this lawsuit comes down to the fact that these advertisers, they are choosing not to spend their money on X because they don't like what goes on on X. They don't want their ads for, you know, beauty products appearing next to hate speech and divisive political content. And that is their right. It's a free country. Advertisers can and do make choices all the time about where to spend their money. And in this case, they are choosing not to spend money on X. And instead of saying, well, maybe we should make our product more compelling to those advertisers by removing some of the hate speech, instead, they are taking those advertisers to court to, I guess, sue them into changing their mind. Is that a strategy you think is likely to work, Casey?

• 1:03:04Casey Newton?

  You know, I talked to a First Amendment lawyer who said like, no, basically, this is exactly what the First Amendment is designed to protect. This is obviously just an expression of speech. And of course, Kevin, the whole thing is very funny, given Elon's appearance at the Deal Book Summit last year, when he was asked about the advertiser boycott. And some of the things he said in response were, one, don't advertise, and two, go fuck yourself.

• 1:03:31Kevin Roose?

  And they did. They went, they didn't, they stopped advertising. And all of a sudden, this is shocking news.

• 1:03:38Casey Newton?

  Yeah. And by the way, Elon's summary of those events that I just described on X this week, quote, "We tried being nice for two years and got nothing but empty words. Now it's war." So that was what being nice looked like. And I guess the lawsuit is not... But look, this obviously, I mean, I do think this whole thing is a joke, but I do think there's also a more serious dimension to it, which is Elon found allies in the House of Representatives, in particular, Representative Jim Jordan, who held a hearing about Garm and has sent letters to lots of big advertisers, asking not just, hey, why aren't you advertising on X, but why aren't you advertising on Breitbart and The Daily Wire, these sort of conservative-aligned

• 1:04:21Casey Newton?

  media. And at this point, X is absolutely conservative media. So I just think that is something that people should be aware of, because, you know... The folks like Linda Iaccarino who want to brandish the banner of free speech, I just find it truly upsetting, given that, you know, she is working to threaten advertisers who are just truly exercising their free speech.

• 1:04:41Kevin Roose?

  Yeah, I agree. I think it's also a sign that X's business model is just in real trouble, right? They they're not.

• 1:04:47Casey Newton?

  Wait, are you saying that healthy businesses don't normally sue the people who have chose them out to be their customers?

• 1:04:53Kevin Roose?

  I'm saying it's not what I would do if I ran a business and wanted to make it profitable. But but clearly, they have sort of realized that their future in advertising is not bright. And so they are making a pivot to subscriptions and they're trying to convince people to pay up for X premium, but that doesn't appear to be going all that well either. So this just reeks of desperation to me. Casey, where do you rate it on the hot mess scale?

• 1:05:17Casey Newton?

  Kevin, I have to say this has all the bearings of a classic hot mess. This company has no idea what it is anymore. It is just acting based on Elon's whims and they're nonsensical and desperate. What do you think?

• 1:05:32Kevin Roose?

  Well, I think they are creating some business for the free speech pendant necklace market, and I think we should celebrate that. That is business that didn't exist before.

• 1:05:40Casey Newton?

  All right, what's the next train coming down the tracks?

• 1:05:44Kevin Roose?

  All right, pulling into the station, we've got From Friends to Enemies. Casey, this is a story about a feud that has roiled the AI companionship market.

• 1:05:58Casey Newton?

  This is so sad.

• 1:05:59Kevin Roose?

  Yeah, so you know the AI company Friend?

• 1:06:02Casey Newton?

  You know, I didn't know about it before this story, although the founder had pitched me about writing about it. And I thought, I don't know. And now I kind of wish I had.

• 1:06:10Kevin Roose?

  Yeah, so there's this guy named Avi Schiffman. He has an AI startup called Friend. They just announced a wearable pendant. Basically, it's like, you know, an AI that sits on a necklace, speaking of necklaces, and sort of listens to your conversations and then will text you sort of, you know, affirmations or, you know, tips or whatever throughout your day.

• 1:06:32Casey Newton?

  You look great today.

• 1:06:33Kevin Roose?

  Yeah, exactly. So this got a lot of attention over the past couple of weeks, in part because of the $2.5 million that this company Friend had raised, it spent nearly 1.9 million of those dollars on the domain name friend.com.

• 1:06:50Casey Newton?

  Is that high?

• 1:06:51Kevin Roose?

  That's high.

• 1:06:52Casey Newton?

  Okay.

• 1:06:53Kevin Roose?

  That's a lot. And they produced this teaser for this friend pendant that got a lot of attention. People said it was dystopian or weird or whatever. Some people thought it was cool. Anyway, there's now another friend because

• 1:07:07Casey Newton?

  A second friend has entered the chat.

• 1:07:09Kevin Roose?

  A second friend has entered the chat. This one is run by a guy named Nick Shevchenko. It is also an AI pendant company. This is something called the Friend Necklace that supposedly comes out later this year. It's all open source. And after Avi Schiffman released friend number one, Nick of friend number two responded by saying that he is the real friend and accused Avi of, quote, riding my wave. And Casey, he made a diss track.

• 1:07:37Casey Newton?

  Oh, my goodness. Can we possibly listen to the diss track right now?

• 1:07:40Kevin Roose?

  Let's listen to the diss track.

• 1:07:42Speaker 08

  Yo, I'm Nick. And for those who don't know, I built Friend. Renamed your tab, jacked my style, built nothing real. It's been a while. Investor cash, you're very quick. I'm running hard, chasing my peak. Stop big game, what is the proof? All your guests with no roof. I'm the founder, harder than a hand. User shishero, can't understand. I view your jokes per meal and a name. Can't ship a thing, just Twitter fame. Friend and yours, riding my wave. Open source fam, digging your grave. Take a seat, I've learned the grind. You're wasting cash, I'm building in time. Respect the game, respect the source. I'm the real friend, the AI force.

• 1:08:14Casey Newton?

  You know, Linda Iaccarino could learn a lot from this guy.

• 1:08:18Kevin Roose?

  Yeah, bars.

• 1:08:19Casey Newton?

  Yeah. If you want to, you want to make the case that you've been wronged, maybe try to make it rhyme. Look, do I understand one word of what that man just said? No. Do I instinctively support him in this and all of his future endeavors? Yes. Nick, I'm sorry for whatever happened to you, and I hope you get justice.

• 1:08:35Kevin Roose?

  You know, I feel like we haven't had a good founder drama in a while.

• 1:08:39Casey Newton?

  True.

• 1:08:39Kevin Roose?

  You know, back in the last wave of tech, you had, you know, the Winklevoss twins accusing, you know, Mark Zuckerberg of stealing their ideas. There was like a Snapchat founder drama. People are always accusing each other of stealing each other's ideas. We haven't had one of those in a while. And so for me, this feels like nature returning.

• 1:08:55Casey Newton?

  Well, what about the nine founders who have quit OpenAI? Does that count?

• 1:08:59Kevin Roose?

  Well, they're not arguing about who actually started it.

• 1:09:01Casey Newton?

  That's true. Everyone agrees that it was Elon Musk's idea.

• 1:09:06Kevin Roose?

  It was Elon Musk's idea.

• 1:09:07Casey Newton?

  Partial, it was partial, his idea. So what kind of mess do we think this is?

• 1:09:11Kevin Roose?

  I think this is not a mess. I think this is nature healing. I think that founders arguing about who stole the other founder's idea is just a natural part of the Silicon Valley ecosystem. And frankly, I for one, I'm glad to see it returning.

• 1:09:24Casey Newton?

  I agree. I look at this and I just think this is beautiful to me. So congratulations to the friends. They brought us together on that point.

• 1:09:30Kevin Roose?

  All right. Casey, we got one more.

• 1:09:32Casey Newton?

  One more. And this one says, Mr. Beast Carnival of Doom.

• 1:09:39Kevin Roose?

  This was one of my favorite stories of the last couple of weeks.

• 1:09:42Casey Newton?

  And we will be sharing some details here from willing to die for Mr. Beast and $5 million. A story in the Times from Madison Malone Kircher. The contestants in the internet stars Beast Games expected outlandish challenges and signed contracts that acknowledge risks of serious injury and death. Still, few were prepared for the conditions on set. Kevin, do you want to tell us a little bit about what happened? Over at the Mr. Beast Carnival?

• 1:10:06Kevin Roose?

  So Mr. Beast, the sort of famous YouTuber who does these like viral stunts and you know, gets billions and billions of views and is just sort of like the Willy Wonka of YouTube.

• 1:10:15Casey Newton?

  Famous philanthropist.

• 1:10:16Kevin Roose?

  Yes. He struck a deal to develop a new reality show called Beast Games with Amazon MGM Studios in which 1,000 contestants would compete for a $5 million prize.

• 1:10:29Casey Newton?

  Sounds great.

• 1:10:30Kevin Roose?

  And this story is basically about how the Beast Games kind of became a nightmare. Among the things that happened to contestants in these games, when they showed up to this football stadium in Nevada where they were supposed to compete in these games, not only were there twice as many contestants as they expected, there were 2,000 contestants, not 1,000 contestants, but they had to turn over their phones and personal belongings, including medications. Things got very fire festival, like they were given these meals that were described by several contestants as including a tiny portion of cold oatmeal, one hard boiled egg and a few pieces of raw vegetables. Apparently several contestants said they were served food they were allergic to.

• 1:11:15Kevin Roose?

  And that no alternatives were provided. They also were having trouble getting medication. One contestant had trouble getting insulin who needed it.

• 1:11:25Casey Newton?

  Do you really need insulin?

• 1:11:26Kevin Roose?

  I think you do. I think that's one of those ones that you do. One contestant said that after repeatedly asking for food that was required to take some medication, she was eventually given half a banana.

• 1:11:39Casey Newton?

  What'd they do with the other half?

• 1:11:41Kevin Roose?

  I don't know.

• 1:11:41Casey Newton?

  Interesting.

• 1:11:42Kevin Roose?

  We'll see. This thing appears to have spiraled out of control. According to this story, contestants in the Beast Games were told to put five days worth of underwear in a bag and give it to the production staff who would sort of like dole it out to them as needed. But several contestants said they had waited hours or even days to receive their clean underwear. One woman said that when she and a group of other contestants who were menstruating during the event had asked this production staff about getting their underwear more quickly, she was told that it was, quote, "not a medical emergency." And she recalled that a nearby member of the production staff had laughed at the woman's pleading. There were also reports of various injuries among contestants.

• 1:12:23Casey Newton?

  Wait, pause. Let me say something. Release the damn undies.

• 1:12:26Kevin Roose?

  Release the undies, indeed. Other contestants reported being injured by these competitions and not being treated well during the competition. One contestant described watching other contestants vomiting and appearing to pass out. There were apparently several hospitalizations. This appears to have just been a real mess all around.

• 1:12:51Casey Newton?

  Yeah.

• 1:12:52Kevin Roose?

  And my favorite part of this is that when asked about all of these accusations, a spokesperson for Mr. Beast had said that the Beast Games shoot was, quote, "unfortunately complicated by the CrowdStrike incident, extreme weather, and other unexpected logistical and communications issues." The spokesperson also said that Mr. Beast had started a formal review and had taken steps to ensure that we learn from this experience.

• 1:13:17Casey Newton?

  Kevin, did you find that the CrowdStrike incident affected your supply of underwear?

• 1:13:24Kevin Roose?

  Only slightly, although I do keep many backups, so I was, I was able to weather this storm. What about you?

• 1:13:29Casey Newton?

  I mean, look, you know, on one hand, I understand that lots of people are rooting against Mr. Beast because he is, you know, the most famous YouTuber and he's like sort of corny and like, yes, he does all this philanthropy, but I think people are sort of always like suspicious of his motives and want to tear him down. And even given all that, though, this is just a very funny story.

• 1:13:52Kevin Roose?

  It's very funny.

• 1:13:52Casey Newton?

  And like, I assume that going on any reality show is actually kind of a nightmare, you know, like they really do take away your phone and every, I, you know, I'm sure these are not the-

• 1:14:00Kevin Roose?

  Wait, that's the part that sounds the worst to you?

• 1:14:02Casey Newton?

  I mean, to me, what is it, what sounds the worst to you?

• 1:14:03Kevin Roose?

  Not the vomiting and the, the hospitalizations?

• 1:14:06Casey Newton?

  That's fine. My, my friends are always doing that around me. But, you know, I think the, the sort of sheer accumulation of incidents here makes you wonder like, how low was the budget on, they're giving away $5 million, like presumably they had a lot of money to spend, but you know, maybe Mr. Beast kind of saved it all for the prize and hoped that everyone else would get by on half a banana.

• 1:14:27Kevin Roose?

  Yeah, I think that Amazon has a real opportunity here because they can not only release the Beast Games show as they intended, but they can also release the documentary about the disasters behind the scenes at the Beast Games, which would be an even bigger hit.

• 1:14:39Casey Newton?

  Yeah, I was gonna say, I'm gonna guess that that one gets way more views than other Beast Games itself.

• 1:14:42Kevin Roose?

  How would you rate this on the, on the hot mess thermometer?

• 1:14:46Casey Newton?

  This feels like a hot mess. Like, I don't think it's gonna truly boil over. And if anything, I think it's just built anticipation to watch the show. But it does seem very messy and I mean, I'm sure there'll be, you know, lawsuits and so on. What do you think?

• 1:15:00Kevin Roose?

  Yeah, I think so. But it's, you know, Mr. Beast, I think is too big to fail. I think he's gonna be fine. But I do think, I am very excited to watch this show now.

• 1:15:08Casey Newton?

  You know what I would say if, if I were running this program and someone was like complaining to me that they were hungry all the time?

• 1:15:14Kevin Roose?

  What would you say?

• 1:15:15Casey Newton?

  I'd say, hey, the name's Mr. Beast, not Mr. Feast.

• 1:15:24Casey Newton?

  And then maybe they would laugh and calm down.

• 1:15:26Kevin Roose?

  Yeah.

• 1:15:28Kevin Roose?

  Have a hard boiled egg. All right, that is

• 1:15:35Kevin Roose?

  the Hot Mess Express.

• 1:15:37Casey Newton?

  We hope you enjoyed that segment. And if you have other ideas for segments that would also work well on a children's show, please email [email protected].

• 1:15:45Kevin Roose?

  Maybe a puppet show next time? What do we think?

• 1:15:48Casey Newton?

  What about one of those bubble shows, you know, where it's like these sort of

• 1:16:11Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure, and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner. Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 1:16:45Speaker 04

  AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real time guardrails by team and model, helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.

• 1:17:17Speaker 06

  Today's AI builders need infrastructure that scales with their ambitions. Iron is a vertically integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. Iron's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. Iron, AI cloud for the next big thing.

• 1:17:52Kevin Roose?

  Hard Fork is produced by Rachel Cohn and Whitney Jones. This episode was edited by Shriya Sinha. We're fact-checked by Caitlin Love. Today's show was engineered by Daniel Ramirez. Original music by Marian Lozano and Dan Powell. Our audience editor is Nell Galugli. Video production by Ryan Manning and Chris Chot. You can watch this full episode on YouTube at youtube.com/hardfork. Special thanks to Paula Schuman, Kwing Tam, Dahlia Hadad, and Jeffrey Miranda. You can email us as always at [email protected]. Send us your disc tracks.

• 1:18:52Speaker 09

  It's time to innovate at the speed of AI without compromising control. That's Outsystems. Take a look at what happened with one global retailer during a period of massive expansion. They used Outsystems to build AI agents to automate invoicing and translation processes, shifting 95% of their daily store operations to a new digital app, saving thousands of hours of manual work. Outsystems gives you the flexibility to deploy how you want and the governance to do it safely. Own your agentic future. Learn more at outsystems.com/NYT. That's outsystems.com/NYT.