Welcome to the Vergecast, the flagship podcast of unfair questions. We ask them, we steadfastly refuse to answer them. This is what we do here. I'm your friend David Pierce. Nilay Patel here.
You know, I've been deposed like two or three times in the course of running The Verge.
Have you really? I've never even thought about that.
There's a very funny transcript of me being deposed.
The important one was there was a game streaming service that Sean had called defunct, and they sued us saying that they were not defunct and that...
Don't put it in the newspaper that we're defunct.
And it's a long story that ends with the judge in Delaware realizing that some obscure law of republication did not have enough precedent. So they allowed the case to go forward so that they could say that us linking to an old story in which we had called the company defunct did not constitute a republication that would then bring it into the statute of limitations.
It's true that I've been deposed, and it's, you know, because lawyers are like, just say you don't remember. So it's just an hour of being like, I don't recall.
Which I really... this is where lawyer Nilay and journalist Nilay run into each other in my favorite way. Like you love to gossip. You love to share information that you know. You like to be a person who knows things, and yet you just have to sit there and you're like, I know the correct lawyerly advice is to sit here and say I can't recall.
You just don't remember. Whatever happened in the past has escaped your memory. It's...
So long ago. Who's to say?
That's very good. So we do have a bunch of legal stuff to talk about today. The big news of this week obviously is Elon Musk versus OpenAI. We talked about kind of the basics of this trial and the run-up to it with Liz Lato on Tuesday. Go listen to that if you haven't yet. We got some unexpectedly spicy early testimony from Elon Musk himself.
Elon's... he's been on the stand. He's on the stand as we speak.
He doesn't like it. He doesn't like being the main character of the lawsuit that he filed and insistently drove to trial.
Yeah, you would think this would be the thing that he wanted the whole time. Like, let's just get into this. Actually, we got a lot of stuff to get to today, but this is clearly where we have to start because the dynamic of this is so strange. So, Elon Musk was called as the first witness. They spent a couple of days getting a jury together, which was very funny, and we should talk about the ways in which the jury came together. But the first witness in this trial is Elon Musk himself. And to your exact point, this should have been the moment he was waiting for. All of this, this obviously losing battle, this weird process of discovery that has been kind of damaging to everybody, this whole lark of a thing Elon Musk has been on, you just have to assume is so that he can get up and publicly say a bunch of rude things about other people in the AI world. And he just didn't pull... he just did a bad job.
Yeah. I mean, everyone thinks they know how a trial will go. I will remind the audience for the 10 millionth time that the law is not a computer. It's not deterministic. There's no predicting what will happen, especially in front of a jury.
But you can do yourself some favors. You can be charming and warm and receptive and act like a human being on the stand when you're giving the testimony in the case that you filed and insistently drove to trial.
And also, if you want to win the battle everyone assumes you're going to lose, doing all of the things you just described strikes me as even more important. Like, make the jury like you, Elon, becomes a very important part of this whole process.
It's just very obvious that Elon Musk, in an environment where his money doesn't make him important and away from the social media platform that he controls that algorithmically brings people around to tell them that they agree with him, is kind of an uncharming, unpersonable dick.
Yep. And especially when he's under cross-examination from a lawyer who's being paid a lot of money to make him look like a dick. He is not prepared for that circumstance in any way, shape, or form. And you can just see it happening on the stand in this trial in ways that are potentially bad for like the entire AI industry. Just now as we're recording, as we sat down to record, OpenAI's lawyers got Elon Musk to admit that xAI distilled OpenAI's models as part of its training.
Distillation is a process where a sort of less powerful model asks a more powerful model like what to do. I don't know how else to describe it. That's more or less what's happening. It trains itself by prompting the more powerful model. Yes. And you can feel a lot of ways about distillation. Like Google calls it an attack. Anthropic has anti-distillation protocols and they're all like, 'You can't steal our property.' And I think anyone who is rational or sane or kind is like, 'But you already stole everything.' Right. This is like distillation is just very funny for these companies to get worked up about.
Yes. But they all do get worked up about it.
They all get totally worked up about it. They're like, 'We scraped the entire internet. We fed them into our huge data centers and took all the energy away and we didn't pay anyone for any of that. But now you can't distill the model.' We know...
Like Anthropic's source code leaked and they're like, 'Oh no,' and it's like, well...
You might have had this coming.
So there's something very funny about distillation just generally from that perspective, but you're not supposed to do it. The companies don't want you to do it, and you're especially not supposed to say OpenAI is horrible and then try to catch up as fast as you can with xAI and then do it by distilling OpenAI. And so just on the stand minutes ago, they got him to admit that xAI has distilled OpenAI's models, and he's trying to pass it off. He's like, 'We all do it, everyone does it, this is how you train a great model.' And it's like, no, most of the companies train their less powerful models on their own more powerful models. This is a scandal, and I just don't think he was ready to walk into these kind of scandals.
Evidently, he wasn't. And it seems like... so Liz Lato on our team has been in the courtroom through all of this. She's chronicling it all over the website. This is like we invented quick posts as a format just for Liz to live blog things that Elon Musk says in a courtroom. You know what I mean? And so we should just start at the very beginning because this becomes delightful from like minute one of jury selection, which is sort of a weird thing to do, right? You ostensibly need a group of people who are going to be impartial. And the list of things that it is hard for people to be impartial about in this particular case strikes me as like astonishingly high. You have to get a bunch of people who are impartial about AI, which is not a thing most people are. You need people who are impartial about Sam Altman, which most people are not. You need people who are impartial about Elon Musk. This is just like this is a sort of unicorn type of person you're looking for in the jury. And basically very quickly, Liz seems to have discovered all of the possible jurors don't like Elon Musk.
People hate Elon Musk. This is what I'm saying. If you get him away from X, a platform he controls that is algorithmically designed to just make him feel cool.
Yes. In the real world, people do not like him and he does not have the charisma to overcome it. And we're just looking at the answers on the juror questionnaire. So, these are questionnaires the jurors had to fill out. And admittedly, a lot of people are probably just trying to get out of jury duty.
Sure. Yeah. These are not the people who were selected. These are questionnaires. And I'll just read you some of the quotes. 'Elon Musk is a greedy, racist, homophobic piece of garbage.' That doesn't seem like he's likely to win that shirt over. 'Elon Musk is a world-class jerk.' Another one: 'I very much dislike Tesla. As a woman of color, I am very aware of the damaging statements and actions Elon Musk has enacted and been a part of.' And this keeps going and going and going until the judge, Yvonne Gonzalez Rogers, who by the way is a recurring character on The Verge.
Yeah, she was the judge in the Apple-Epic trial, right?
Yep. These cases all get filed in a handful of courts. We have a handful of judges who keep showing up. So, YGR is back for Musk v. Altman. And she eventually had to say, 'The reality is that people don't like him. Many people don't like him, but that doesn't mean that Americans nevertheless can't have integrity for the judicial process.' So, she's basically saying, I trust the jurors that we have selected to set aside the almost universal disdain for Elon Musk that we all feel in this courtroom and do a good job.
Rough. This is a lot to ask. This is before we've even begun, before he's even had a chance to open his mouth.
The Doge master has shown up and everyone's like, 'Yeah, we hate that guy.'
YGR also, I think, made the main characters in this story agree to stay off of social media during the trial, which is a wild thing.
It's because Elon is posting 'they stole a charity.' This is going to come up again. He's got these lines he thinks are going to be the winning lines and you can't just say them to a jury. And OpenAI has filed motions to strike some of his testimony because he's doing tweets.
And so she basically is like, I have to be able to trust you're not going to leave here and corrupt this process.
So she's told everyone to stop tweeting.
Yeah. So before we get into all of the insane stuff that happened on the stand, I want to try and very briefly explain the two sides of the argument and I want you to tell me if I more or less have it right as we get into this because I think it ends up being important again for all the lines Elon Musk has. So the Elon Musk stance is: I helped fund and start a thing that I believed was a charity. This whole thing was my idea. I was the driving force behind all of it and then they stole it from me essentially. They forced me out. They took my money and they built this gigantic for-profit company on top of it. How dare they do so? Is that a reasonable...
Yeah, that's his version of the argument. I think that's his version of the argument.
There's some daylight between what he thinks the argument is and what's in his legal briefs.
Sure. And there's a lot of daylight between his argument and what may have actually happened, which is that he just hates these guys.
Right? So the other side of the argument is: as OpenAI is coming up, Elon Musk tries to basically take over the thing for himself. He wants all of the control. He wants all of the board seats. He wants everything. And when he doesn't get what he wants he decides to take his ball and go home. Leaves. Says screw this. Tries at some point to subsume OpenAI into Tesla because he thinks that's the only way they can take on Google and win at AI. This doesn't work and he gets angrier and angrier as OpenAI gets more and more successful and thus then decides to sue because he's mad that he didn't get to be in charge of this thing that is huge. Like in a real way the argument on the other side of Elon Musk is he's really mad that Sam Altman is the face of AI and not Elon Musk. There's that. And straightforwardly, you didn't read the contract, which also comes up in these cases when we talk about them on the broadcast over and over and over again. Did you do the reading? And the answer in this case, we now have evidence, is no. Elon insistently did not do the reading.
Yeah, it's good stuff. So we have a bunch of... there's just a lot of good lines from the testimony. Liz has been in there talking through it all. We pulled some favorites. Nilay, do you want to start with... let's just run through the testimony and do some greatest hits here.
So, we're on day three. As we record, we're on day three. So, Elon's been on the stand for two days of the trial and several hours today. The first day was the direct examination. So, if you've ever watched a courtroom drama, you know this is when your own lawyer shows up and asks you questions about how cool you are, right? This is the time you're supposed to do very well and look very good, right? They're like, 'How did you... how are you such a genius?' And everyone's like, 'I can't possibly say.' And that's what happens, right? He's like, I'm just a benevolent guy who wants to make cars. And so that first day, he was really flat, because there's no conflict. Yesterday when cross-examination began, he just became a raging jerk. And cross is when you know, you get asked the leading question and the answer is yes or no.
And if you know that they can say, 'Your Honor, there's a hostile witness,' and you just have to say yes or no. Like that whole thing. That's cross-examination.
They say things like, 'Isn't that right?' over and over, and it's designed so that the witness is only supposed to say yes or no, right? And Elon hates this. It's like he's never watched a single courtroom movie in his life. So at many points he accuses OpenAI's lawyer of, quote, 'asking questions that were designed to trick me,' and then he answers the questions by saying things like, 'You mostly do unfair questions.' And then OpenAI's lawyer is saying things like, 'I'm trying to put the questions as fairly as I can. I am doing my best,' and Musk just goes, 'That's not true.' And this is just the entire... at one point Liz writes, he refuses to acknowledge the nature of linear time.
Oh yeah, that it was like, 'Before you were not a board member, you were a board member, yes,' and he just won't answer the question.
So he's just like fighting the basics. Like we're not even engaging on the basics of whether or not cutting off donations to OpenAI might create pressure on OpenAI, whether Andrej Karpathy, who he had recruited to OpenAI and then to Tesla, whether he asked him to stay at OpenAI, like just these very basic chronology questions. Elon's like, 'You're mostly unfair. You're mostly trying to trick me.'
Yeah. And I mean, there's another one of these that's just as good, and this goes to, 'Did you read the contract?' This is the one that really has stuck out in my brain over the last few days. Elon gets an email in 2018 outlining the bones of the for-profit structure. This is like one of the things actually that has come up in this trial and actually I think will end up being important is how long has it been a discussed possibility for OpenAI to be a for-profit corporation, right? Like there's this old clip that's going around of Sam and Elon talking and Sam refers to OpenAI as a company and Elon goes, 'OpenAI is structured as a 501c3 nonprofit.' And a bunch of people are like, 'But then...' It's increasingly clear that conversations about a for-profit OpenAI have come up many times and Elon has acknowledged that he was okay with a version of that. But there's a moment where they send him an email outlining the corporate structure they're imagining for a for-profit version of OpenAI. And he says on the stand that he only read the very first section of it. Which is, hey, he says, 'I read the highlighted box with important warning,' which is amazing. Just this is very... it sounds very much like the way the Trump administration handles Trump where it's like you have to draw a big red box around the thing you need him to read, otherwise you just have no idea.
By the way, the important warning was: contributors should consider their investments as donations with no return.
So, he's acknowledged that he's read that part, right? Which is important. And then Musk says, 'I didn't read the fine print.' We're going into the fine print of this document. And then OpenAI's lawyer goes, 'It's a four-page document.'
Brutal. You can just tell he doesn't like being on the stand in this way because he can't ignore your tweet. He can't downrank you.
His army of fanboys isn't there to distract with memes. He just has to do it in the lawsuit that he filed and insistently drove to trial.
It's a four-page document. It's just such a good...
Like those are one of those moments where you can just imagine like he goes back to his law office at the end of the day and it's just like high-fiving all the paralegals everywhere. It's good stuff.
He's got another line here where he says earlier, 'I don't lose my temper and I don't yell at people.' And OpenAI's lawyer takes this as an opportunity to bait him into losing his temper, which he does very successfully, immediately and successfully. So, in the same sequence, he says, 'I don't think I read this term sheet. I'm not sure I actually read this term sheet. I did not look closely at this term sheet.' And then he's been deposed already. So, there's already another version of his testimony. OpenAI's lawyer points out that in the previous deposition, Musk didn't say that he'd read the first paragraph and nothing else. And Musk starts yelling at him, 'I said I didn't look closely. I only read the headline.' And it's like, my guy, you can't be like, 'I'm very calm. I'm just here to save humanity. I never raise my voice,' and then start yelling at the lawyer.
And then at the next turn you start yelling at the lawyer.
There's another part of this. Again, this is if you set yourself up to be cross-examined, you do have to prepare. It doesn't feel like he prepared for this obvious eventuality.
And you can't trickster your way out of it. So, if anyone listening has ever done like mock trial in high school, this one to me as a person who did mock trial in high school is just so funny because you can see him being the smartest kid in class and it not working.
So, Elon writes an email about DeepMind and he says, 'DeepMind is moving very fast. I am concerned that OpenAI is not on a path to catch up. Setting it up as a nonprofit might in hindsight have been the wrong move. Sense of urgency is not as high.' So, OpenAI's lawyer produces this email. They show it to Elon. They say, 'Those are your words, yes or no.' Elon's response to this, and I'm telling you, if you did mock trial, you understand exactly where this came from in his brain.
This is a hypothetical. What... are these words? Yes or no? This is a hypothetical.
He's answering the question. Is that a thought you might have had at the time? And what they're saying is, did you write this email that you wrote?
And so OpenAI's lawyer says, 'So you thought it might have been a wrong move. Is this what you said?' And they had to keep going until Musk admitted that he had written the email, which again, if you've just done the high school mock trial, you know that you've now backed yourself into a fight in which the people judging you, the judge, the jury, the high school guidance counselors who are doing mock trial are like, 'Get over it. Just answer the question.' Because it's not so bad if you're like, 'I did write that email. Can I explain what it means?' It's very bad if you're like, 'It's hypothetical. I refuse to even acknowledge that you're asking about this email.' And he's just doing this over and over again throughout this trial. Liz was saying she was like, this room is so testy and so uncomfortable. I think her line was, 'I can't even imagine having this person as a co-founder.'
Yeah. I mean the funniest part about all this, and Liz starts her story with this and also a lot of our commenters and stuff have been saying it, that one of the most remarkable achievements of this testimony is that it makes Sam Altman look really good. Like Sam Altman is going to come off incredibly reasonable and kind in comparison.
I mean the thing about trials, and this has happened every time we've had a trial, like one side gets to go and then the other side gets to go.
And so this is the part where it just seems very obvious that, you know, Elon is going to get to make his case and it's not going well but he's going to get to make his case and then Sam's going to show up and it's all going to happen.
Yeah. And the pendulum will swing back and forth. But as of right now, Elon is just being argumentative and petty and, you know, Liz is in the room and she's like, 'It doesn't seem like a jury is buying this.' There's another part of this. This is the most Elon thing in the entire world. He's arguing with the lawyer. He's arguing with apparently his own lawyer. And so he says, 'I understand leading questions.' That's a leading answer.
I was just about to bring this one up. And the judge says he can lead. He can lead all he wants. Let's remind everyone you are not a lawyer and you've never taken a class in evidence, which is an insane thing for a judge to remind a witness.
And Musk's response is, 'I did take law 101 technically, but yes, I am not a lawyer.' This is a total AI pill man. Like he asked xAI how to do a direct examination.
And now he thinks he's a lawyer.
This is also like very clearly someone who is used to either being or at least being treated as the smartest person in every room he's in. And there's been a bunch of evidence in this trial so far of the other OpenAI founders like going way out of their way to tell Musk how terrific he is and that like every request begins and ends with unbelievable flattery of what a genius he is and what a pleasure it is to be able to just be in a room with Elon Musk talking about AI. Like Greg Brockman just makes an ass out of himself being kind to Elon Musk in every single email. And you can just see this is a person who meets somebody who doesn't want Elon Musk's money, actually is out to defeat Elon Musk in a very specific way that no one ever really is in Elon Musk's orbit, and he just self-destructs in the face.
You take him away from his money and his social media platform that he controls where he has built himself an algorithmic filter bubble of people who think he's great and you're like, 'Oh, this isn't going well for you.'
Yeah. At the very end of this, the judge says, 'He was at times difficult. Part of management from my perspective is to just get through the testimony. It's bad. This is all bad.'
Okay, this is actually where I wanted to end this one because what I can't figure out here is the stuff that Judge Gonzalez Rogers, good old YGR, is saying seems like stuff I've never heard a judge say in the course of a trial. And again, where so much of this is going to come down to Elon Musk convincing this group of people in the jury that he is a good person trying to save the world, boy does that all seem to bode very badly for how this is going for Elon Musk so far. The fact that... is it as out of pocket for a judge to have to say this stuff during a trial as it seems to me?
So they sometimes say it. They say it when they have like the sovereign citizen guys who come in and they're like, 'You don't have jurisdiction over me.' Like sure...
They say it in that situation.
But normally, we've all watched all these CEOs appear before Congress, right? They are way buttoned up because when the person with the power to take your stuff away shows up, you back it way down. You're right.
You're conciliatory. You say, you know, 'Your Honor, I'm going to have to check that out later.' Like he should be playing to the jury. And he's saying things like, 'They tried to steal a charity and this case is going to set a precedent that will let anyone steal a charity,' and this is his ex-workshopped applause line, right? And he's not allowed to say it. And OpenAI is like filing motions to strike his testimony because that's not true. And this is a real... I mean, you can have a lot of feelings about the legal system in the United States of America. You have a lot of feelings about the justice system in the United States of America. It is a structured system. It's designed the way it's designed to at least gesture at fairness, right? And you can't just steamroll it. Like that's why you have judges and procedure and rules and lawyers and all this stuff. And he's trying to steamroll it because he steamrolls everything. And he's just running into the reality of he's just a witness on the stand, right? Again, you know, it's going to flip and they're going to put Sam on the stand. And I'm sure OpenAI's lawyers are going to try to make Sam look great and Elon's lawyers are going to try to make Sam look horrible and that's just the way it goes. But right now, what you have is one of the wealthiest men in the world in front of a jury that has pretty clearly indicated statistically they don't like him.
Being a huge dick. And I just don't know what his plan is.
I don't know that he has the personality qualities to solve that problem on his own.
Like Liz kept noting on today, Thursday, that he was much more subdued than he has been in the past. Like clearly there is somebody giving him the 'this is not going well for you, buddy' note. And yet he can't help himself, right? Like this is the third day of this. It's not clear since we... this is happening while we're recording. So they've moved on to Jared Birchall, who runs Elon's family office, manages investments. We've already gotten to how many Teslas he has. It's good stuff all the way around.
Yeah, it is. Liz's big theory was that this is going to be a trial kind of about a bunch of things that don't make any sense, but there is going to be a lot of stuff that comes out. And I think that is already becoming true. There's just like there's a great moment earlier today where Elon Musk took a long time to acknowledge that his for-profit companies are for-profit. It's like...
It's just what are we doing here? Elon...
None of it makes sense. We have all kinds of coverage on the site. Obviously, Liz is in the courtroom. Hayden Field, our senior AI reporter, is taking all the documents that are coming out over the course of trial, looking at them, organizing. There's a long post, or if you just want to see the evidence that's being entered, you can go look at it. We're going to... I mean, it doesn't get bigger than this for us. So, we're covering this in every way we can.
Yeah. And there's a bunch of other OpenAI swirl happening right now. We should just talk about a couple of the things. I think the biggest one that happened this week is Microsoft and OpenAI essentially walked their deal all the way back, right? Like Microsoft and OpenAI for years have had this very complicated but very sort of important and lucrative deal by which they share revenue and technology with one another. This is how like Bing became what it was back in 2023.
They made Google dance. Nilay, listen. It was the greatest day of Bing's life. Give it its one day. And this was like a really important tie-up for both of these companies, right? Microsoft owns a huge portion of OpenAI because of all of this. This is how OpenAI got a lot of access to technology and compute for a long time. And their big thing was all of this gets sort of wound down and reconsidered once we get to AGI. And right, AGI was supposed to be this magical moment that if somebody got to AGI before OpenAI, OpenAI would shut down and would give all of its technology and resources to that company. All of this is nothing. AGI is nothing.
I can't believe you're not leading with this.
And the final acknowledgement of it is their deal is now just a normal contract for compute.
I can't believe you didn't lead with the AGI clause going away. This is the clause that started with if you make like a hundred billion dollars of economic value that's AGI and then it got reworked into we'll put up a panel that will evaluate AGI and they would never say who was on the panel and I still believe it should have been like the Pope and Mariah Carey and me.
Especially now that we have Chicago Pope, like that would have been a great panel. Like I just...
We could have sold tickets.
Is this digital god? Is that the question? Yeah.
Yep. We never even got to learn who they were thinking about for the panel.
So in the annals of tech history, who was on the AGI panel, it will just be a forever unsolved mystery.
But in tech circles, this is not because AGI is a fake concept that has no bearing on anything. It is completely immeasurable and nonsensical. It's because we're already there, right? Like we've been seeing this for a while that the Marc Andreessens of the world have been out there being like AI and AGI are already here. It's just not evenly distributed. And other people have decided that it's not about AGI, it's about superintelligence, which is apparently a different thing of equally nebulous definition. But yeah, this whole thing has just fallen apart. And very clearly what this is is OpenAI wanted to be able to go make other deals with other companies and not be tied up with Microsoft in this same way. Also, there have been lingering issues between Microsoft and OpenAI for a long time. Microsoft hired a whole team of people to go do Microsoft's own AI stuff. A lot of the reporting in the story about Sam Altman that was in The New Yorker that Ron and Farhad did was about the ways in which these companies have kind of if not lied to each other at least not been wholly good partners to one another. And so the whole thing is just very clearly like everybody wants out of this. We're moving on. We are now business contacts who have compute deals.
Yeah. They have a first-look exclusivity. So, OpenAI products will ship first on Microsoft Azure unless Microsoft cannot and chooses not to support the necessary capabilities. So, you've got to, if you're OpenAI, you've got to use Azure unless Microsoft is like, 'No, that seems boring.'
And then OpenAI turned around and immediately announced a gigantic deal with AWS and has started to say, 'We are an AWS-first product.' Like that's our focus now. So, as you and I have been talking about on the show for weeks now, it has just become very obvious to everyone that AI is an enterprise product.
This is a thing businesses are going to do to automate business processes. In particular, a lot of businesses are going to write a lot of code with AI in ways that they weren't able to before. And this is all great. And to make that all work and be good, they need access to lots of data. And the businesses already have their data in AWS.
It's where it is. AWS is a huge cloud provider. A lot of businesses use it. It's already there.
Amazon is winning that race and has been for a very long time. And they're very proud of it. And so Microsoft is like, if you want the coolest AI stuff you can migrate to Azure. And a lot of people were like, no, we'll just use... Amazon spent a lot of money on Anthropic. We'll just use that. And so I think as OpenAI needed to get to an IPO and make more revenue and compete with Anthropic, which is resolutely a B2B company, they needed to get to AWS. I think the weirdness between Microsoft and OpenAI exists in a lot of ways and now it's over and we're just never going to know if Mariah Carey thought the computer was alive and I'm personally very sad about that. As you can tell. But the decoupling of OpenAI from Microsoft has been in the works forever. This to me, it's just very obvious has happened. OpenAI needs more customers. Being tied to Microsoft does not get them more customers. And unless OpenAI was content to remain Microsoft's research division and let Microsoft go fight for customers, there was no point in being locked to Azure in the way that they were. And I think we're going to see OpenAI aggressive trying to get business customers now that they're on AWS. None of this matters to anyone normal for any reason whatsoever. This is the most boring enterprise cloud computing drama that has ever existed in the history of the world.
But it is happening for a specific and important reason which again you and I keep talking about which is that people, regular normal human people in their regular normal lives, are increasingly saying they don't want AI.
And they're saying it in ways big and small. Right. There was this company Sensor Tower, which does really interesting analytics data on the internet and on mobile apps in particular, came out with a study that says basically uninstall rates for AI apps are way up year-over-year. Just to give you an example, Claude has the best number. Claude is the best performer of all of the major chatbots with a 90% increase in uninstalls year-over-year. ChatGPT is up 257% and this was measured in particular around all the stuff with the Pentagon and the Department of Defense. And...
Wait, let me understand this rate. So uninstalls are up 257%.
Not 257% of people are uninstalling it.
Which is a very Trump-like percentage, but the rate of uninstalls is more than doubled. Right.
And next to that, the rate of downloads is also slowing down. It's going up, but the actual pace of adoption is slowing really fast. And we're just in this fascinating moment where we had a great story on the site this week digging more into this tension that young people feel where they feel obligated to use AI and so they do. And most people who are in high school and college report using these AI bots all the time. And they hate it and they wish they didn't have to. And they feel obligated to. They feel like they're being sold a bill of goods about how if they don't they'll be left out of the job market and their world will fall apart on them, but they don't want to use these tools because they understand it to be bad for them to do so. Yeah.
And there is just like this tension is real and it is serious. And the thing that ChatGPT has had going for it for a long time is that growth has been ridiculous. Like it is one of the fastest growing pieces of consumer technology in history. There's been a bunch of reporting recently about the fact that ChatGPT still hasn't announced that it has a billion users. Which is like a big important milestone that a lot of people have been waiting for and did not expect to be waiting this long for. And like there are just these things are huge. They're very successful. Lots of people use them. All that is true and yet there is just this mounting evidence that even the people who use it don't like it and more and more people are finding ways to not use it.
That billion weekly active user number is a big deal because people have been waiting for it. But The Information first reported and then The Wall Street Journal had the same bit that OpenAI CFO Sarah Friar is saying we can't withstand the scrutiny of going public right now.
That the kind of financial reporting we'd have to do we cannot do right now. And that as a result she is being in some ways frozen out of her executive roles at the company. Like the tension inside of OpenAI around this stuff appears to be pretty intense.
That's reporting. I will say they keep issuing statements signed by both Sam and Sarah saying we love each other. So...
Don't put in the newspaper that we're defunct. Nilay...
There's something here that we keep talking about that people keep arguing with us about because the numbers are broadly big. Google announced this quarter that Google search queries have hit an all-time high. Great. It is true that people are using the tools at high rates. It's also true that they're making up their minds. Like the polling data is super clear and everyone in tech has gotten so confused about what data is that they can't see the thing in front of their face, which is the experiences in the free consumer products are bad.
Just straightforwardly bad. The experience of free ChatGPT is bad. It is nothing but like BuzzFeed-style engagement prompts now. Like every query ends with, 'Have you thought about asking me another question?' Like it's like some leading way.
Yeah. The ads are starting to get more intense. You're hitting limits faster and faster on these tools. Like it is an increasingly bad experience.
And the reason it became such a fast product to begin with is that Google search at that time and to this day is so intensely commercialized and shitified. So as a competitor to that version of Google search which is 95% sponsored posts and an SEO web where the economic conditions of publishing a recipe require all the poor recipe bloggers to write 2,000 words of personal narrative before the actual recipe, and that was all bad. We've written about that at length. Of course the chatbot that we'll just talk to you was a better experience. Of course, that thing grew at infinite scale in the beginning, and now they have to make money. They're reverting to an enshittified version of the product, loading it with ads, loading it with more engagement prompts, trying to drive user engagement in all kinds of ways that we understand, and people are reacting to it.
It's the same with free Gemini. It's the same with free AI overviews. Yes, it's true they're doing more search queries because you can just ask the thing more open-ended questions and it will deliver you an answer. It's also true that overviews are generally bad.
Yeah. And you have to do another search to get a good answer. Like...
Right. And like no one... I just feel like the tech industry is lost in the data, right? They're looking at these numbers and saying something must be going right. The market is telling us that we're winning. And it's like also they're telling you that they hate the products. Tech companies love to talk about this phrase revealed preferences as a way out of the tension between people actually liking the product or not, right? That you say, 'Oh, people are having a...
bad time on Instagram and it makes them feel bad and it creates issues with mental health and body dysmorphia and all of this stuff. And someone at one of these companies points to a chart that says, "Well, they keep opening the app, so they must like it." And that idea needs to die. It just needs to die because it is not correct. Because the revealed preference of people using AI is, "I would like to have a job in the future," and the world has spent the last four years telling me that if I don't learn how to use Claude Code, I will never get a job and that the world will leave me behind. And so the revealed preference is not for these good products. It's to continue to be a functioning member of society. And we just have to get out of this idea that because people use your products, that means they like them.
This is the story of social media. We have spent a decade learning with social media that they don't have to be good products to be powerful and to be popular and to be massively used. Like screw you and your revealed preferences. People are saying out loud that they don't like your product and that they feel like they have to use it. And if you as a tech company think that's a victory, you should understand very quickly why everybody hates you. That's why everybody hates you.
Yeah. I mean, to the extent that you know, when Meta still had product managers, I think they're firing them all because Meta is firing everybody every week. You know, the joke was always that if Meta was having a bad quarter, a product manager could just turn a knob, increase the ad load, and make the number go up. Inside of a tech company, that sort of decision process where it's like, we can just make more money by making the user experience slightly worse for a few weeks, is pretty normal. Like there's a reason it's a joke.
By the way, Meta this week revealed that, to be precisely correct, Meta revealed that it lost 20 million users in its family of apps. Meta stopped breaking out which app had which many users because I think it would make Facebook in particular look bad. So it bundles all of its stuff together. And that whole number across all of them, which protects anything having kind of a down quarter, went down 20 million users. That's a big deal. Do you know what went way, way, way, way up? The amount of money that Meta made last quarter. Like one way, a record-setting jump.
Like then they just turned the knob and they will keep turning the knob.
You know, the way they did it is super interesting and I'm jealous of the story. So this is a New York Times story by friend of the Vergecast, Tripp Mickle, by the way, and Eli Tan. And I'm utterly fascinated with how the ads economy works. I think understanding the money of the internet helps you understand the internet. And so I'm always just fascinated by it. And there's this thing that's happening on Meta platforms and on Google platforms specifically where the old way of ad targeting, where you would show up with your ad, which is called creative in the industry parlance, is very good. You have to call it creative to make yourself feel good about it. So you'd show up with your ad creative, your video, your picture, whatever it is, and you would say, "I made a razor. I'd like to target this razor at men 18 to 35. Go find them for me." You know who your users are. Put this ad in front of them and they would do it because they have a lot of data. And then you could refine it even more. You'd be like, "I think my razor blades are going to sell really well to men 18 to 35 in metro areas who have previously looked at shaving cream." And they have a lot of data so they could target it and you, the advertiser, were trying to go find your audience, right? And you could run all these experiments or find different ways.
That's over. The big jump in revenue is that they have flipped the model upside down and they are using AI to look at your ad, to look at your product, and then they're going and finding you the customers that are most likely to want to look at your ad and buy your product.
Oh, interesting. So you just say, "I have shoes," and now you used to have to say, "I want to sell these shoes to XYZ kind of person." Now you just say, "I have shoes," and Meta does the rest for you.
Yeah. And not even "I have shoes." It's, "Here's my ad," right? I made a video. Like, here's the ad I made for shoes and go find me people who are likely to look at this ad and convert into shoe purchasers.
And so the line that the industry is using that I've heard over and over again is, "The creative is the targeting." So you just make different ads that you think will appeal to different groups and you make dozens of variations and you fire them off into the AI ad system and the AI ad system looks at your ad and says, "Okay, this video looks like these videos that this group of people is interested in. We're going to give this ad to them and they're going to go find you the customers." And this is, you can just imagine the next turn, which is, "Oh, we can edit the ad with AI to make it more effective to the customers that we're finding." Or, and Zuckerberg has said this out loud, don't even show up with the creative.
Show up with the product. Yeah.
And some money. Give us the money. We'll generate the ads for you and just deliver you results.
Right. Take one picture of the shoes. We'll do the rest.
And we'll do the rest. And by all accounts, this is a massive success for both Google and Meta.
Yep. Small companies in particular, who were never any good at advertising because they're small companies and who can't afford the big agencies that used to sell this kind of data analysis, they're just showing up and be like, "Find me some customers." And their sales are going through the roof, so they're just dumping more money into the system. And you look at that and you're like, "Well, that's actually great for a bunch of small companies." And then you also look at it and you're like, "This sucks so much."
Yeah. Really?
Like it's tearing the ad industry apart. It's certainly devaluing creative across the board. It's subjecting all of us to the creepiest kinds of targeted advertising that have ever existed in the entire history of the world. But then a bunch of small companies are finding more and more customers. I could not tell you, I've been thinking about this story since I read it, since I've been hearing all the ad people say the words, "The creative is the targeting," which is a real mindbender because at the end of the day, like they also want you to give up the creative. They want you to stop making the ads. I don't know, man. Like when I say it's enterprise software, when you say it's enterprise software, it's this.
They took the AI tools, they invested in all the AI tools, they built the tools for themselves to run their ad targeting system. And it is by all accounts more successful than any ad targeting system or any advertising platform has ever been in the history of the world. Even as the users go down, but it's not like great consumer experiences in any way, shape, or form. The digital god stuff is a marketing scheme to bring customers to your ad targeting system. Like that is sort of the story of the last 20 years of the internet. And it is just more nakedly that story than ever. But when we get into ad tech, that means it is time to pull the ripcord on whatever we're talking about.
Let's turn now to the hype desk. This is where our friends Ross Miller and Ashley Esqueda come on and tell us what's cool in the world. Sometimes it's sponsored. Today it is not. Ross, Ashley, welcome back.
Hello. Thank you for allowing us to return. I'm surprised you haven't changed the locks, if I'm being honest.
Nilay and I have been talking. We've decided that one of your most important jobs on this show is to tell us what TV shows are cool. This is literally just like a thing that Nilay and I both require in our lives at this particular moment. And this is a thing I think you guys can provide for us.
Ashley has just a thing for you.
I can provide. I can provide. I love television. I will binge-watch TV. I'm obsessed right now with Widow's Bay on Apple TV. This is so good. It's a horror comedy. Katie Dippold of Bridesmaids created this and Matthew Rhys is the star. He's the mayor of a small little island town in the northeast called Widow's Bay. And Steven Root is in it as like your resident weird guy. It's got all the things that you want. And there's some elements of like, you know, it reminds me so much of Twin Peaks. And I think that this generation needs a new, we needed Twin Peaks for this.
It is as weird. The other person who's kind of co-creating is Hiro Murai, who did all the Childish Gambino videos.
Yeah, he directed it. Yeah. Yeah.
He did all of Atlanta basically. So it's that kind of like modern weird. I will say a kudos to Apple for being the only streaming platform left willing to release things once a week and actually stick to it. Kudos.
And B, this suddenly the Twin Peaks comparison instantly makes it make sense why all of my TV nerd friends have been excited about this show for like months. Like this is the show of all of 2026. Every TV hipster I know, all of whom love Twin Peaks, have been extremely excited for this show.
It feels like it hits for me right now.
Fair enough. Like also again, kudos to Apple for, I miss the golden age of TV that was like Richard Plepler's HBO. I miss that era like a lot. And now I feel like Apple TV has been on such a tear over even just, I mean like what, five years now. I mean they're just crushing it. And if you're a connoisseur of fine television shows, a content person as it were, if you don't have an Apple TV subscription, like you're really missing out. There's so much good sci-fi, like so many good sci-fi shows on Apple.
Yeah. Do you think John Stankey knows them? Like I've read a lot of coverage from the Hollywood press. It's like, who's this new guy? Will he keep paying us money?
Does he know we're down here?
Yeah. Exactly. It's like his job is to just ignore it, I think, is what everyone wants to have happen here.
Yeah. I mean it is a really good value ad for I think the subscription stuff, which I, you know, I think that's the thing that they're wanting to sell is that, you know, all the subscription services where you have Apple One and it's that like what is it, 40 something dollars a month where you get like everything all packaged in there with like Apple News. And so I think that that is, I think Apple TV is the crown jewel of that package in my opinion.
All right, Ross, what do you have for us?
I've got Coyote versus Acme. I am stupidly excited for this. We weren't here last week, but right after the show went up, they dropped a trailer for this movie. And you all may not know about this movie. It has been so long coming. It got finished in 2022 and then immediately shelved by Warner Brothers for a $30 million tax write-off. It was one of the three movies, Batgirl and some Scooby thing.
This was a fully finished film.
Fully finished, ready to go.
Completely done. Not like, oh, it still needs post-production like visual effects.
Do I just not know how to do my taxes? It definitely feels like I should be like, I wrote this book I didn't publish. I'm shelving it and I'm not releasing it and it's worth $30 million.
No one will ever see it, but I could tell you it was going to be a million seller.
Mhm. Mhm. No. So it was one of those movies during kind of the early Zaslav years. They're trying to get the finances figured out. All these filmmakers who had seen it, they're like, "This is the next Who Framed Roger Rabbit." It's an animated kind of live-action hybrid. The plot essentially is Wile E. Coyote keeps going after the Road Runner with these Acme products, right? And they've broken every time. So he's fed up and he's finally decided to sue the Acme Corporation.
Will Forte as the down-on-his-luck lawyer versus John Cena, like the super fancy lawyer for Acme Corporation. Like this is the little guy David versus Goliath story. And of course, this is also now a metatextual read of the movie itself because it has been lingering for three or four years now. There's a couple of near misses. Netflix and Paramount tried to buy it. Warner Brothers said, "No, we want $80 million." Finally, I think they just gave up. This smaller independent distributor, Ketchup, they've done a lot of like little smaller mid-tier movies. They picked it up just recently, I think in March for $50 million, and it is finally going to see the light of day in August. The trailer just dropped. Is it funnier if this movie is great or if it's horrible? Like just from a pure, this is the weirdest story ever. Like is it a better ending if it's like a smash hit or if it sucks?
I will be equally satisfied either way.
Yeah, I think that's right. It's either a very funny story or like stick it to the man, Coyote versus Acme is awesome. Like it's kind of a win-win.
It's entertaining either way for us the viewer at home.
But yeah, I mean it could be like K-pop Demon Hunters but they just didn't realize what they had. Like Sony gave it to Netflix for dirt cheap and now it's like, shoot, now we got to figure out how to like monetize the back end.
I will say all of the people on the internet, I'm going to, because I am terminally online and I saw the outcry of people being like, this is terrible. We can't put the Looney Tunes in a vault. You better go see this movie. Like you better go see this movie. This movie needs to make like $500 million. Like it doesn't need to cross a billion. We don't need to, it doesn't need to be the Mario movie, but it does need to make a statement and that requires people to actually go see it. So go see it.
It's a lot to ask in 2026.
I know. I know. It's like $400 for a movie ticket now. I understand.
Yeah. I saw Project Hail Mary the other night alone and it was like $90 all in.
But wasn't it gorgeous? The popcorn was this big though. It made me very happy.
The popcorn supply chain is rough.
Big popcorn. That's for another day.
Ross, Ashley, thank you guys so much for being here. We will see you back here next week. We are going to take a break and then we're going to come back and we got some more stuff up. It's gadget time. We'll be right back.
Support for the show comes from Zapier. AI is becoming more and more a part of the real world, but just talking about AI doesn't help you be more efficient at work. For that, you need actual tools like Zapier. Zapier is a way for you to break the hype cycle and put AI to work across your company for real. With Zapier's AI orchestration platform, you can bring the power of AI to any workflow so you can do more of what matters. It lets you plug leading AI models like ChatGPT and Claude into the tools your team already uses so AI shows up exactly where it's most useful. Whether that's powering workflows, running an autonomous agent, supporting customers with a chatbot, or something entirely custom, you can orchestrate it all with Zapier. And it's built for everyone, not just technical teams. In fact, their data shows teams have already automated more than 300 million AI tasks using Zapier. Join the millions of businesses transforming how they work with Zapier and AI. Get started for free by visiting zapier.com/verge. That's z-a-p-i-e-r.com/verge.
Support for the show comes from Hostinger. Every business has its impact. And with AI changing the landscape, the barrier to entry has never been lower. Whether you're starting a side hustle or building the next big thing, Hostinger lets you go live in minutes, not weeks. Hostinger is an all-in-one platform that brings everything into one place. Your domain, website, email marketing, AI tools, and AI agents. You can create websites, online stores, and even custom apps without coding or design skills. Then use AI agents to automate tedious tasks and help grow your business. Turn your one day into day one. Go to hostinger.com/vergecast to bring your idea online for under $3 a month. Plus, get an extra 20% off with promo code Vergecast. That's less than the price of a cup of coffee per month. That's hostinger.com/vergecast. Promo code Vergecast for an extra 20% off.
All right, we're back. Nilay, we have a bunch of gadget news this week. All on a really fun spectrum for me of so a bunch of things that I think you might be really into and then a bunch of things that I think you may have absolutely no feelings about or interest for. So I've invented a new game and it's called Does Nilay Care About This?
Don't do this to me.
You love all gadgets in theory, right? Like you love that gadgets exist.
Sometimes you don't care very much about a gadget. They're like, "I'm so psyched that you're here. I want what's best for you and I hope you live a great life. Just don't talk to me anymore." That's how you feel about some gadgets.
So we have a bunch and I'm just going to name some gadgets and we're either going to talk a bunch about it or we're going to talk about...
You're going to be so mad at me by the end of this. All right, let's do it.
The one I want to start with is I think probably the gadget of the week. It's the Steam Controller. Do you care about the Steam Controller?
Okay, first of all, I need to say I do believe it's the gadget of the week. Without question, people were so into the Steam Controller on our site this week. I understand why. I intellectually understand why.
Okay. I am going to convince you that you should care very deeply about the Steam Controller. But first, let me just play you something that Jay Peters on our team made. He reviewed the Steam Controller and he made something for us explaining what this thing is and how it works. Let me just play it for you.
Here's the Steam Controller ringing like a telephone. I have in my hands Valve's new Steam Controller, which I've been using for about two weeks to play a bunch of games on Steam. It's a great controller for playing games, but it has all sorts of really cool small touches that make the experience even better. Like, if you lose the controller, you can just ping it. The controller has nice chimes when you turn it on or off. The included puck that gives you a low-latency connection also snaps onto the back of the controller so you can charge it magnetically. You can see a live indicator of that low-latency 2.4 GHz connection right from Steam settings. Here are the back buttons which open up all sorts of customizations for when I'm playing my games. And I love the quick access button that opens up a menu so I can connect wireless headphones or check performance in my games. You can even use the controller to type into text boxes on your computer. Or in a pinch, you can use it as a mouse and keyboard to navigate your BIOS. The Steam Controller will be available on May 4th for $99. I'm probably going to buy one.
