Welcome to the Vergecast, the flagship podcast of rando Twitch guys who sometimes become CEOs for five minutes. I'm your friend David Pierce. Neil is here. Hey buddy, what's up?
We're doing AI chaos again. This is just this is this is what we do here on the first cast. We talk about how people don't like AI and then we talk about AI. This is this is what we're here for. If you look at our like audience stats, like the website stats, which we don't talk about a lot, everybody just wants to read about AI guys being uh extremely dramatic.
And boy, do we have a lot. There's a cornucopia of AI guys being dramatic in the world right now.
Alex Heath had a good tweet. He's he's covering the trial and he goes, "The entire AI industry is basically just how seven guys feel about each other."
Uh and by extension, like the entire United States economy. It's just how seven guys feel about each other.
Um, we got a lot to talk about. There's a there's a new uh Fitbit product that I think you and I both have a bunch of thoughts and questions about. Uh, OpenAI is maybe making a ChatGPT phone. We got a bunch of stuff to talk about, but we got to start with the trial, right? Elon Musk versus OpenAI continues. This is week two of the trial. Uh, I sort of expected it to be less bananas in week two after Elon Musk left the stand. I was wrong.
Very wrong. It has only increased in amount of bananas, a bushel if you will, full of bananas.
Can I just say this like one thing that's been on my mind throughout this entire trial and maybe it's because you know I I have like ex-lawyer in me and married to a lawyer and we think about the lawyer stuff. To make AI good, you have to tell it everything, right? Like it needs access to all of your email. You got to write down everything you want. You got to write these long system prompts. All this stuff, all the context that it needs. We're now watching the fruits of that in discovery in a lawsuit.
Oh, that's such a good theory. Do you know what I mean? Like, Greg Brockman, why did you keep a journal? Uh, and his journal is extraordinarily damning. He's like, I shouldn't steal this charity. Like, what are you doing, man? Um, I'm not saying that drove this set of characters to keep the notes and send the texts that they sent. I'm very grateful that they did because some of them are frankly hilarious and honestly a huge part of this Vergecast is just going to be me and David reading other people's texts to each other and it's going to be amazing. Um, but we're entering an era where every executive is talking to Claude all day long on their way to work or dumping context into AI systems. And boy is the next generation of lawsuits going to be amazing. Like just straightforwardly amazing. And I I'll tell you one, there's a reason I brought up Becky. Her law firm just sent around a reminder that when the clients take the legal advice and go ask ChatGPT about it, it loses its privilege.
And like courts in New York have like ruled on this that if you if your lawyer sends you something that's like here's what I think we should do in this case, that is protected by attorney-client privilege. And then if you the client go to ChatGPT and say hey what do you think about this thing my lawyer said, you lose the privilege.
Right, it's the equivalent of telling someone else. You've now told someone else.
And that like on both sides lawyers like I can discover that now also I have to tell my client like the next generation of discovery in all these cases is going to be bananas because everyone is telling their computers everything all the time and you can just see in this case like every everyone is so online that everything is in the computer like yeah.
Zoolander style like it's in the computer.
That is a really good theory and it actually makes me think like one of the funniest characters as we're going to get to in all of this is Satya Nadella the CEO of Microsoft who throughout this whole saga everyone is desperately trying to get to leave a paper trail and he just refuses. He's like yes he's a he's the public company CEO the longest tenured most experienced you know multi-billion dollar public company CEO. Microsoft is the economy. Like he can't screw up.
This man can only be reached on a phone through cutouts.
Yes. And he's they keep being like I keep I'm trying to loop him in. I'm trying to get him in. I'm asking him questions. I'm texting him. I'm reaching out and and then and then every once in a while it's like long pause and it's like just got off the phone with Satya and it's like you I get it Satya. But I know that Nadella just like the rest of them in his interviews he's like I talk to Copilot every day about my next meeting about this stuff. I love talking to it on a commute. And it's like, oh, that get ready, buddy.
Yeah. I mean, it is funny like it's it's useful to remember that all of the events that are being discussed in this trial are from somewhere between 2 and 12 years ago and that a lot has changed even even since the most recent stuff in this case. A lot has changed in a lot of people's relationships. So, like you're right that the stuff that is happening right now discovery is going to be just outrageous.
Yeah. You're wearing one of those B bracelets that records all of your meetings so you can be a cool startup founder. Guess what happens? Your next deal goes sideways.
Like the audio of all or the transcripts of all your conversations will now it's great. I'm excited as a reporter. I'm excited. I also refuse to use computers at work even more than ever.
Yeah, it's totally fair. So, we we should just walk through some of the stuff going on in this trial, but I feel like the stuff that's been the most interesting to me personally has kind of split in two ways. There's either the stuff we've learned about the early days of OpenAI and particularly the circumstances under which Elon Musk ends up leaving. Uh we learned a lot about that this week, particularly from Greg Brockman, who is one of the co-founders and is now the president of OpenAI. Um and then there's also the what happened during what everybody refers to as the blip, the the brief period of time where Sam Altman was not the CEO. Let's should we start with the early days because there's there's a lot of interesting stuff that comes up. Um bunch of stuff that I didn't know that maybe I just missed in the course of all of this stuff including things like at the very beginning OpenAI was going to be part of Y Combinator uh that where where Sam was running Y Combinator this big VC firm for a long time and like I think more and more it's becoming clear that a lot of the internal consternation at OpenAI over the years has been all of the stuff that Sam Altman and Greg Brockman did around OpenAI and in the ways in which they were semi-sketchily actually enriching themselves all while claiming to have no interest in a nonprofit and just trying to save the world. And if you're Elon Musk, it seems like a really good idea to try to prove as much of that as you possibly can, right? Is that is that a fair description of what's going on?
I think that's a fair characterization of what's going on. You will recall that Sam left his position as CEO of Y Combinator in totally mysterious and extremely sketchy circumstances.
Yes. And it is probably because he was double-dealing all over the place and no one has ever really ever come out and said it. But a lot of people have basically been like if you connect all these dots that I keep leaving everywhere, you will you will deduce that Sam Altman did a bunch of weird double-dealing stuff and is no longer the CEO of Y Combinator. And I think we're starting to see that in this case as well, which by the way is totally in character with the Sam Altman that is like routinely profiled by everybody, including Ronan Farrow. There's a lot of weird double-dealing, talking out of both sides of your mouth, trying to make everyone happy, getting very rich along the way.
Yeah. And this ends up being really important in the firing of Sam Altman, but also appears to have existed since the very beginning, right? That they're they're running this nonprofit. They're starting this thing. They're trying to save the world. And Sam Altman like famously didn't take equity in this company. Has spent a lot of time talking about how he didn't take equity in his company. And then you you start to poke at it and it's like, well, he has equity in that company that is now doing a deal with OpenAI and he has equity in that company which is doing a deal with OpenAI. And like you just there I think Elon Musk's lawyers have a real vested interest in building a pattern of Sam Altman being unreliable and trying to make himself rich while building a nonprofit on Elon Musk's money. But there are parts of that case that appear to be very straightforwardly compelling from Musk's side of things.
There are parts of that case that appear to be very straightforwardly compelling. And I I'm going to issue the disclaimer like criticism of one character in this lawsuit is not praise for the other character.
And both sides get to put on their case. And so when OpenAI's lawyers are putting on their case, Elon looks maximally bad. And when Elon's lawyers are putting his case, Sam Altman looks maximally bad. And that's the way trial coverage goes. And I say this every time because everyone gets very confused about how we feel when we're just talking about the cases the lawyers are putting on. But the out that I have for Sam and Greg in just in this process, yes, the evidence against him is compelling that they they basically did a bunch of shady stuff. They are not sly by any stretch of the imagination. They're just out in the open being like, "So, we're going to do some like circular financing stuff and like, you know, Greg is like leaving himself journal notes. It's like, man, I shouldn't I shouldn't steal a charity, you know?" Like, they know what they're they are operating out in the open. And they're sending Elon these like extremely long notes that all start with this like staggering amount of praise because everyone's basically afraid of Elon, it seems. And they they have to be really nice to him to get to one decision. And they keep laying everything out.
There there's at no point are they being shady about it, right?
Except that the thing that they're doing seems kind of shady.
Do you know what I mean? Like Yeah.
They're like, "We're going to do a shady thing." And Elon's like, "I don't know." And then that proceeds until they get it done.
Well, ironically, it's not shady when they're all on the same team, right? And then it's then it just becomes, you know, we all we there's so much money in this that we're all going to get rich and it's going to be fine. And this is like what you see in in WeWork, another thing where there was just an outrageous amount of self-dealing. Uh, as long as the numbers keep going up and as long as everybody's getting rich, the things people will turn a blind eye to in that realm are outrageous, right?
But then we we come to the the you know falling out between Elon Musk and the OpenAI team and uh I think this is the part where Elon's case kind of even in the very best case scenario looks bad because essentially what it seems like is again and we said this a little bit last week there was this big fight over who was going to control OpenAI. Uh Elon Musk demanded everything. He at one point tried to make OpenAI part of Tesla. He said he was going to run OpenAI inside of Tesla in secret and apparently offered Sam Altman a board seat at Tesla. It was like he he he wanted to subsume this whole thing. Um he he would say his motivations were to save the world. You you can feel about that however you want. Um he lost that fight um basically because the the others teamed up and said giving essentially unilateral control to Elon Musk is is not the right thing to do. And he took his ball and went home making aggressive threats about starting up a competitor, about how they were going to fail, about how Demis Hassabis, who was running DeepMind, uh, who is like the looming big bad in everybody's mind, which is very funny in retrospect, is going to like ruin the world because OpenAI can't do it. He he has such like sore loser energy in this whole split that it's like it just it just looks bad. There's also some just extremely weird Elon stuff in the mix. Shivon Zilis, who is the mother of some of his children, is part of OpenAI. His her job is to take notes basically on what everyone is talking about.
Uh so there's a lot of notes about what Greg Brockman, Sam Altman, Ilya Sutskever, and Elon Musk are are talking about. And then when Elon leaves, she's still there. She's passing intel back to Elon. And then reports come out that she's had his children and she has to go and tell OpenAI that she's also the mother of Elon Musk's children, right? And everyone is like, "What are you talking about?"
Changes the dynamic slightly.
It's just like one of the more Elon situations in all of this that now the mother of his children is on the stand and even the way that she talks about things in her affect is it's just very Elon. Uh on on the stand instead of saying I don't recall or I don't remember, she would say it's not in my neurons.
I hate that. I got to tell you, there's something about that that just makes my skin crawl.
That when it when she said it, our entire Slack room of reporters tracking the trial, they all typed it's not in my neurons like at the same time.
Perfect. That's Yeah, it's exactly right. So, yeah, this whole thing is just bizarre. And it really does seem like that the further we get into this trial that it is that the the outlines of this split were that Elon Musk wanted control of OpenAI. He lost the fight and he he went away. He says there one thing in 2017 after a discussion about going for-profit which again is this is a big piece of this right who wanted to go for-profit and when was Musk for this? Was Musk against this? Did they steal a charity? Like this is a whole a big part of the whole thing here. Um this is Brockman writing this. He says at the end of the meeting he said you guys are great but I could start another AI company tomorrow. One tweet is all it takes which is a 5 years before Elon Musk bought Twitter. There's the whole reason Elon Musk bought Twitter. Uh and b just a perfect summation of what a sore loser he is in this particular way.
I thought this this one is very interesting to me. There is so much in this that is just about who is the talent and where will they work.
Sure. This is why they're all afraid of Demis Hassabis. He's got Google's resources. He's got runway. He seems to be attracting the talent. They all think he's very smart. Uh, and Elon's like, I'm going to take Andrej Karpathy. I'm going to take Ilya Sutskever. I've recruited this person. And it's just a handful of people at the bleeding edge of this technology.
Especially at this time, it was like a really, really small number of people who were really leading this stuff.
Yeah. And they were all basically deciding whether they were going to go out on their own or go work for Demis at Google. Like, these are your choices at the time.
Mhm. One tweet is all it takes really implies that there's like a bunch of AI people he can like address via tweet. It's it's a very odd empty threat at this moment in time when so much of the drama is basically where's Ilya going to work. Can we keep him away from Demis? Right? And it maybe there's more characters in the mix. Maybe there's vastly more AI researchers at that time just waiting for a tweet. But it it just struck me as such a classic Elon threat. Like I can tweet this into existence when every fact that had preceded it indicated that actually you need to hire one of like four people to make this work.
Yeah. And then and there's the moment where he hires away Andrej Karpathy to Tesla and and he goes back to OpenAI and is like, "Oh, I I have I have an announcement and a confession. Uh, I've hired this guy who is like very important and is now like a a celebrity in the AI world uh to go work at Tesla, which is a thing that if you care about OpenAI and want it to succeed, you probably don't do."
Yeah. At the same time, there's, you know, there's emails in the record now of Sam emailing Elon apologizing for having recruited even one person away for Tesla, right? Like, I I'm so sorry about this. There's some real, you know, I'm watching The Boys. There's some real Homelander vibes around Elon throughout this entire, you know, everyone's just like afraid of him. Mhm. And then he's But his instead of lasers from his eyes, he has Twitter. Like that's basically what this feels like at all at every turn.
Well, it's more than that, though. There I I wrote down uh these texts from Ilya's came up where he says, "Elon might spend half a day a week with us. I imagined how it will be and I worry that our work environment can become very stressful and since he'll be bankrolling it, it'll be hard to stop it." Like, you're right. And it is it is fear and flattery in equal measure. And there is the sense of like we have to take his money to do this thing, but we also have to figure out how to keep him out of it as much as possible because he will ruin it.
Yeah. And then they're talking about all these conversions and what structure it will be. And the whole time he's like, I'll just take it. Give it to me.
And now we're in this lawsuit, which by the way, he insistently filed. Like he's trying to kill OpenAI. Elon Musk wanted this lawsuit.
This is all happening because he wants it to happen.
Yeah. Yeah. Agreed. So, the other thing which I think is frankly vastly more fun in terms of the things we got to learn, uh, is all the stuff that happened during Sam Altman's firing, this very brief couple of days where Sam Altman was fired unceremoniously and kind of out of nowhere. Uh, staged a coup is like a non-generous but not wrong way to say it. A lot of people on the team threatened to quit. It became a whole thing. Gets reinstated as CEO. Um, we learned a bunch about what happened in the midst of this. What jumped out to you as somebody who covered this in real time particularly aggressively. Was there new stuff that really jumped out to you?
All of these people text like children, like just pure children. And I'm not saying I'm like the world's best texter. Uh it's just you you read these these texts between Mira and Sam Mira Murati the CTO who was also the CEO for five minutes for all of this and if you will recall we broke a lot of news during this time on The Verge.
Like Heath and I broke a lot of news throughout all this and we were texting and calling everyone all the time. Other great reporters broke a lot of news too. Like everyone was talking.
Yeah. Shout out to Ashley Vance who now at Core Memory who who shows up in the one of the messages.
Yeah. He's like in the screenshots as a notification being like there's gonna be a new CEO.
Mike Isaac is like he's doing a victory lap because he was he was correct that Mira Murati tried to lead some of this charge to push Sam out. But everyone was like talking and everyone was reporting and I just remember my texts were like very formal.
You know I was like I'm going to text somebody and see if we can get some news out of this person.
And it's like here. Yes. Pretty much. And I'm like, you're on the back end and you're seeing what they're all saying and they're like, there's chatter, blah, blah, blah. Like, it's all lowercase. Like, no one knows what's going on. Uh, there's something about that, you know, it's like it's always interesting to see the other side of a thing you were even loosely a part of.
Yeah, I do think the the back and forths between Mira and Sam as all of this is going on are maybe the best thing that came out and have led to some really like delightful memes. Uh there is to your point about Mike Isaac who wrote a story concurrently as all this was happening saying that Mira Murati was one of the people intimately involved with ousting Sam in the first place. Um got a lot of criticism for that piece. A lot of people who said that wasn't the case. Um turns out it it functionally was the case. And on the stand she said over over and over essentially like no I didn't trust him. No, when he came back I still didn't trust him. And that's like that's that's under oath what she said. Um, and meanwhile, as all of this is going on, she seems to be the person Sam is most consistently messaging trying to get information about what is going on. Because, as we know, he basically like a lot of this comes up in that New Yorker piece you're talking about. He just like holed up with a couple of his confidants and lawyers and desperately tried to figure out what was going on.
Like spent a weekend basically frantically calling and texting everyone. Uh, and Mira Murati is like a source of his in the room. Uh and and the one exchange that has come out that is now I think just the most useful meme we've had in a while is Sam Sam texts uh can you indicate directionally good or bad Satya and others anxious and Mira just texts directionally very bad which is such a brutal incredible own. Well no it keeps going and he goes okay can you wrap up soon lots of pressure from Microsoft for an update and then Mira replies Sam this is very bad she's like no no you don't understand there's not an update like you're fired.
The you know the background here and the I will say this again criticism of of one person in the context of the story has nothing like praise for the other people. Everyone screwed all of this up from top to bottom. The board never communicated why they had taken the step of removing Sam as CEO. They never put out a statement. They never said here's all the things that went wrong. Here's our evidence for doing this. They had asked Mira to collect evidence of Sam's management style and these challenges and Mira had done it. She testified she didn't know why she was doing it. It's like no, you definitely what we we've all like you put him on a PIP. Like what? We all know what's going on here, right?
Oh, you want me to prepare a dossier on my boss? That's probably fine. Like where do you think this was going? The boss that you had concerns about you were supposed to document your concerns. Um, and so when the board fired Sam, it was like this shot out of the dark and no one knew why. And there was all they said was he was not consistently candid in his conversations, right?
This like totally opaque thing that could mean anything. And so he's texting her. He's legitimately like, "What happened?" And I think he thought there was some middle path. And there's this one set of texts here that is just, and again, they're all text like children. Sam says to Mira, "Can you please tell them I just want to resolve this, however, and would like to join?" And Mira says, "They're convinced about their decision, and I just want everyone to pay attention to this response, which indicates everything about this process and how this company works." And Sam says in response to, "They're convinced about their decision for me to be fired or some other thing." And Mira has to confirm, "Yes, for you to be gone." And Sam just says, "Okay." He's like, "So, I'm still fired, right?" Like, "What? What decision have they made this time?" You can tell no one was communicating successfully at any level of this entire process.
No one is communicating. And the again the entire process here is they're communicating through Mira. So, both sides are communicating through Mira. And I think at one point she becomes the CEO of OpenAI.
So, just a lot of back and forth confusion here. Uh, and then we have to we have to talk about Emmett Shear being the CEO for five minutes. So Sam texts Mira, "Can I come in and talk about a path forward with them?" Mira says, "No, they need more time." Sam, once again, more time for what? Which is just a totally reasonable Yeah. question. And then Mira says, "They walked me through all the reasons and the issues with you and why you can't be the CEO." Again, we still kind of don't know, right? There's been a lot of reporting, but we still kind of don't know. And then Sam says, "Can you ask why they've been saying all weekend they want me back?" Mira says they want a new CEO in place. Sam oddly says, "Can you say you will call back in 10 minutes?" And then Mira says, "They want to have a new CEO in place tonight, not me." Sam says, "Do they know who?" Uh, "Can I tell Satya is it final?" Mira says, "Trying to add Satya now." Sam says, "Please still don't want me." Sam says, "Somebody is texting Satya a Zoom link." And Satya's going, "What? Nope. It's not happening." And then Mira finally ends all this with new guy is a rando Twitch guy, which is so good.
That random Twitch guy, by the way, Emmett Shear, uh, who was the CEO of Twitch for a long time, uh, I think by most accounts, like a like a good egg in Silicon Valley circles, um, had a weird moment in the sun as part of this entire story.
Yeah. I mean, he he's enjoying it. He I remember again during this reporting when they announced it was Emmett Shear. There was some amount of like, oh, we know him, you know.
Yeah, that's a that's a guy that we were all in like this the incubator with. But wait, the texts keep going here. So Mira says, "New guy's rando Twitch guy." And then she says, before Sam can even respond, she goes, "They don't want you." And then Sam responds, "Emmett with two question marks." She says, "Yeah, but hold on. I'm pulling Satya now." Like it's it's just bananas. And then Mira is telling Sam that she is trying to bring him back, which may or may not be like it seems very clear she is like engaging in different ways on both sides of this as this is all going down. It's just it's just bizarre. And then they bring in Satya to try to get him to undo it because he is he ends up being like the grown-up in the room in a lot of ways throughout this entire thing. Uh it's all just so insane, man. There's a moment where they they're gonna sell the IP to Anthropic and Mira says they just don't want your hands on AGI which very funny. It's years later we are no closer to AGI with Sam's hands on it or not. Um this by the way they do finally get Satya Nadella. Sam goes Satya get on the call. Mira says yes with him. Sam says wait I have an interesting idea goes still with Satya. Just totally sets him aside. You will recall this is where um Nadella decided to react to all of this which I fully believe he regarded as completely juvenile and nonsensical and very threatening to the valuations of a lot of companies. A lot of things invested in OpenAI including his he was like I'm just going to hire everybody.
Yeah, Microsoft will just hire you can have whatever shell of OpenAI remains. This is his famous quote about we are above them, below them, around them. He's like I'll just hire everybody. Sam is going to work here now. And that was a bunch of leverage in the middle of this. I mean, this is all very funny because it is just evidence that these things happen in a much more casual way. Like, this isn't lawyers. This is just two people texting each other, obviously being pulled in five different directions, talking to all the same people who are also pulled in five different directions. I'm pretty sure, you know, Musk's attorneys brought this all up to indicate that Sam was unreliable, untrustworthy. At one point, he was fired from OpenAI. I have no idea if this is going to convince a jury that the structure of OpenAI constituted stealing a charity from Elon Musk, who was again and again presented with this exact scheme and didn't read the paperwork.
Yeah. Yeah. I mean, it does. It really seems like the the easiest takeaway from all of this is that all of this is just a gigantic mess. Y and that it's like you could make a lot of arguments about how poorly run this company has been throughout this entire process. I will say there are two things that came up in the why they fired Sam Altman bits of testimony that I thought were really interesting. And the the one was like we were talking about this sort of circular set of deals that there was a sense that the board uh Helen Toner who's also a board member at the time testified and and talked about some of this that the board felt like it wasn't being properly made aware of all of Sam and Greg's financial entanglements. Uh they made this deal with a company called Helios which is a nuclear energy company that both Sam and Greg were invested in. That's a weird thing for OpenAI to be spending its time and resources on. And they felt like they maybe weren't properly notified that one reason to make this deal is that it's going to help make Sam and Greg rich. Um, another one that came up that I thought was really interesting is uh Shivon Zilis said this too, that OpenAI's board was really concerned that it hadn't been notified in advance of ChatGPT being released, which is totally sensible in retrospect, right? The the launch of ChatGPT like legitimately changed the tech industry forever. Like it really did. I don't think that's an overstatement at all. Um I am I am quite confident that no one at OpenAI up to and including Sam Altman knew that was going to happen on the day they shipped ChatGPT, right? It was it was a little research preview that they barely announced.
It was a research preview. They just they just they just put it out into the world in a very sort of academic way. They were like, "Look, we made a thing. Test it out. Let us know what you think." And then it it it caught like wildfire. So the idea that this that they should have known ahead of time to warn the board that this earthshattering thing is coming strikes me as obvious in retrospect but sort of implausible at the time.
Wait, I can academically argue the other side of this.
I I tend to agree with you. Google had something that looked very much like ChatGPT in its labs forever.
Yeah. Google remind you it invented all of this technology. The T in chat is transformers which were invented at Google.
They're now the Google executive hiding in my bushes to remind me of this.
They say it all the time. Um uh they had this they had this technology. They had transformers. They demoed it in a variety of ways. It never made any sense because they always did it in increasingly esoteric ways. I will never forget the IO where where Sundar is like and now I'll have a conversation with Pluto the planet about what it's like to be a planet. And no one knew what was going on. And if instead he was just like, "Look at this ChatGPT." Like everyone would have been like, "Google has invented the future." Yeah, you can see that split here in all these texts and emails because everyone's afraid of Demis who is building the Transformers at Google, right? Like it's like they've all seen the actual product and the public is like Sundar's talking to Pluto again.
Right. And they and they knew Google knew not to release this product to the public. Their own safety culture kept them away from releasing what was at the time like a GPT-1 or 2 class model to the public. Like this is dangerous. There's all this stuff that can go haywire. Blah blah blah blah blah. We're going to do Pluto demos instead. All of the principles were like this is the future.
Literally paper is called attention is all you need. Right.
Right. And so I can see, again, I'm making the argument academically because I agree with you that no one knew what would happen. But I can see why a bunch of people immersed in the safety culture of that time were like, "Everyone knew that you shouldn't do this. OpenAI exists because Google wouldn't do this and you did it anyway without asking us if you should." That's fair. And especially if you come from a safety-first perspective, which in theory OpenAI at the time did, uh, you would say, "Oh, this is a potentially like problematic thing we're putting out into the world to see what happens when it's out in the world. You should know that it's coming, right? This is the thing that Google won't do."
Yeah, I suppose I buy that argument. I also still think it's true that they honest to God did not think anybody would care. Like I I really earnestly believe OpenAI shipped ChatGPT. If you if you think you're releasing a product that's going to change the world, do you call it ChatGPT? Like, no.
No. That's what I'm saying. I agree with you on balance. Like, I I I I think I'm making this argument for the sake of making the argument. I understand.
But there was a reason Google did not release this product and then OpenAI did. Now, OpenAI was nowhere ready to be a company. They were not ready to commercialize it. They were still chasing AGI, like all this stuff. And now they're a totally different company that, you know, in in its heart of hearts wants to be a for-profit enterprise competing with Google.
Yes, that company is positioned very differently. And the OpenAI at it was like, look at our research preview, right? But I I understand why the board, which was our job is to keep everything safe and to like pull the rip cord as we get closer to AGI was like, wait, this is the this is the one thing the big player in the space wouldn't do.
Yeah. No, that's fair. I I buy that. Wait, can I can I just add one more thing? Oh, please.
So, I keep forgetting this as well, but Liz is in the courtroom. You should read her live vlog and her coverage. It's very funny. She's so angry.
Everyone appears to be very tiresome. Um, also, this is Liz at her finest. Liz being annoyed that she's watching billionaires to each other is perfect. This is this is why The Verge exists. Um, Hayden's doing great coverage on the exhibits, but I keep forgetting myself that Microsoft is a party to this case because they had the big investment in OpenAI. And Liz keeps mentioning that every single witness ends with Microsoft's lawyer standing up and saying something like, "And Microsoft wasn't there." And then following that
up with, "And Satya Nadella wasn't there either." And she's like, "This is literally the only thing they're adding to this trial."
So, we didn't do that, right? We have nothing to do with us. See you later. Uh, it's just a very funny dynamic in this case that I can't get enough of.
That's it. That would be a really fun job as a lawyer. You just stand up and be like, "But not me though, right?" And they're like, "No, not you." And you're like, "See you. No further questions."
It's pretty good. Yeah. Uh, I really hope at some point we get some Satya texts. Some texts.
He's too smart for that, man.
He is. Satya wouldn't even be caught on a Zoom call, it appears. So kudos to Satya. Um, all right, we should take a break and then we're going to come back. We got some gadgets to talk about. We'll be right back.
Support for the show comes from LinkedIn. If you're a small business owner, you know that every hire counts, but time and resources are limited. Finding, connecting with, and screening the right candidates takes up valuable time you could be giving to your customers. That's where LinkedIn Hiring Pro comes in. It's built to be your hiring partner, helping you find the right candidates faster. That way, you can hire with confidence without turning it into another full-time job. Hiring Pro streamlines the entire process from drafting your job to shortlisting candidates and conducting AI-powered interviews for initial screenings. Its updated conversational interface lets you describe what you need in plain language. Nearly 60% of hirers find a candidate to interview within a week. With Hiring Pro, you spend less time searching and more time connecting with the right talent. And instead of getting buried in resumes, you get a focused shortlist that actually moves your hiring forward. Join the 2.7 million small businesses using LinkedIn to hire. Get started by posting your job for free at linkedin.com/ttrack. Terms and conditions apply.
All right, we're back. Just as a transition out of OpenAI, we should talk about this rumor report that's now floating around that suggests that OpenAI is building a phone for ChatGPT. It's a phone GPT. What do you make of this, Neil?
First of all, uh, no side quests. I was told Code Red, we have to focus. Uh, now we're making a phone. The one product that will not succeed. It just, there's no way it will succeed. Um, it's very hard to disrupt the iPhone. I don't know if you're aware of this, David. Uh, killing the iPhone. Uh, good luck.
I know. It's weird. It's almost like no one has tried. Why has no one ever tried to take down the iPhone? It's just sitting there.
Yeah. I mean, uh, have you heard of Windows Phone? It's got this Metro UI. It's beautiful.
Okay. Can I make an argument for why you have to do this even though it won't work?
Oh, well, no. Let's do it as an exercise, okay? Because I have a similar argument.
Okay, you have to make it out of a place of optimism and opportunity.
Oh, here's why it will work. I will do cold realism. Okay, here's why Chat Phone is a terrific idea that should exist and I very much look forward to it. If what you need for the AI future is a device that is able to capture everything about you all of the time, you need something that isn't an iPhone. That is not a new idea by me. That is a thing Sam Altman keeps talking about. He loves the iPhone, thinks the iPhone is great, but he thinks the future is constant capture of your context. This is terrific for legal discovery. Uh, it's also the idea of AGI right now, right? Like I need to have a thing that is constantly recording so that it can get all the action items for my meetings. I need a camera that is constantly recording so it can tell me what's going on in the world. That requires a different kind of hardware. Uh, you can't do it with the iPhone in part because you can't access the hardware in that way because you're not Apple if you're OpenAI, but also you just need a different kind of hardware that operates differently. So you have to build a phone to do the things differently. And you also have to build a phone because everybody owns a phone. And the only way to beat the iPhone is to make a phone that can beat the iPhone. OpenAI has lots of users. ChatGPT is very popular. Maybe lots of people would buy a phone that offers easier access to ChatGPT. The other thing that it offers, and this is I think the real reason you do this if you're OpenAI, is that the only possibility for an AI-first world is to get around the App Store. Apple is systematically removing apps that let you build other apps from the App Store. It is not into the idea of being able to vibe code on your phone other apps that work on your phone. That is in fact the whole premise that you would need for an OpenAI phone to work. That all of a sudden I have a device that I can just tell what I want it to do and it can actually spin up local software to do that for me. If you believe in the agentic future, if you believe that all of this is going to work on the fly and custom GPTs and all this stuff, it has to start from the hardware and you have to start from whole cloth. And that actually having a thing that I just talked to with my voice and it builds its operating system for me on the fly. That has to come from someone other than Apple or Google who will never allow it to be built. So, of course, OpenAI is going to do it and it's super going to work.
Okay. I think my perspective, the more cynical realist perspective, is exactly the same argument. I'm going to make exactly the same argument to you. I'm going to add two things to it. One, Sam hired Jony Ive to come build hardware at OpenAI. Famously hired Jony, the designer of the iPhone and many other legendary products. And they made a video that cost somewhere on the order of $25 billion where they walked down a street together and went on a date at a coffee shop. Um, and they talked about phones and laptops as quote legacy devices. And the entire promise of this thing was they were going to build something else. And there were rumors that it was headphones. There were rumors that it was a pendant, that it was a pack, a puck, that it was a pen and all the stuff that wasn't a phone because a phone and a laptop are legacy devices. And I agree with you. Sam thinks you just need a bunch of contact accidents and it's seeing the world around you and then it can run around the web and call you an Uber and do whatever you think is going to happen. And then the reality, this is my other addition. The reality is that the AI products are not good enough to do that. They can't do the thing that everyone thinks they can do, that everyone wants them to do. No matter how many Mac minis you put in a colo bin and running OpenClaw for every individual user, you can't actually make this product, right? And so what are you stuck with? You need a bunch of applications running on something local that ChatGPT can address, that it can control. Whether that's through MCP, whether it's through like Google is doing on Samsung phones and literally opening the apps in virtualized containers and clicking on them, whatever way you're going to do it, whatever Apple's going to do with Siri, it turns out you need the applications, not just the data. You need to do stuff. You need the logic to exist. You can't just write a custom Uber app and be like, "Now there's Uber." Like, you need them there.
And even if you can, Uber did that already.
Yeah. Like, they're not going to let you be like, you can use the Uber API. The CEO of Uber, Dara, was just on a coder. He's like, I want my pixels in your face.
He said it out loud. Um, so what you're stuck with.
Yeah. He was like, I want my pixels there. He kept calling them pixels in like a very specific way. And I was like, you mean your interface? Okay, I get it. Anyway, um, and so you have Jony Ive. You made this big promise about the future of computing. And I think maybe Jony Ive, who likes to make good products, realized that ChatGPT can't do it. Like, the technology cannot deliver the result, right? And so what are you going to do? You're going to back into the right solution, which is it's a phone and the phone's going to have apps on it and the AI is going to run around using those apps on the user's behalf because the apps have all the services that people want. And also, if you make a thing that's supposed to replace your phone and it doesn't have Instagram on it, people are still going to carry their phones around. So, what are you going to do? So, poor Jony Ive is stuck making a new iPhone because it's the only thing to do. Because it's the only... I'm telling you, I'm agreeing with you. This is what you have to do. I'm just saying I could have called it the first day that they sat down in an empty coffee shop and called the phone a legacy device.
Yep. Because at no point has it been clear that any of these LLMs can actually do the thing, right? But yeah, this is what I mean by it's not going to work and you have to do it anyway. Like, even this whole accessory ecosystem they're describing requires a base device that neither Android nor iOS will give them the permissions to be able to do. iOS is not going to just let you attach a third-party camera that's recording all the time. It's just not a thing you can do. They might make it themselves.
Sure, Apple would do it for itself. Well, and this is where like Apple's AI reputation is like annoyingly turning in its favor because it screwed up so badly that everybody's like, "Oh, four-dimensional chess." Like, Apple stands to get... No, Apple screwed this up super bad. And to whatever extent this is going to work out, it is pure dumb luck and not good planning. Let's be so so so clear about that. Uh, which I have proof for that we will get to in just a minute. But the idea of having basically a new version of Siri, which we think is coming very soon at WWDC, uh, and there was a report this week that one of the features of that Siri might be that you get to pick your own model inside of it, becomes hugely powerful, right? Because Apple is the company with the access to that data and to those services and to those applications. Uh, the thing it couldn't figure out how to do was build a model that could manipulate any of them with any kind of success. Now it has this big deal with Google. It's possible it's going to open up so that you can pick ChatGPT or Claude or whatever you want to operate this Siri for you. That becomes a really powerful model. But it's because Apple already did the rest of the iPhone, right? Like, if you just start from whole cloth with here's a model, build the phone, you can't. It just doesn't work. But you don't have another option. This is why like we've been talking for a decade about how desperate all of these other tech companies are to get past the smartphone. This is why they are desperate to sell you some other kind of hardware so that they can get around the platform controls that Google and Apple have and absolutely will not relinquish.
Well, they are. Mark Zuckerberg tried to do the metaverse. It's the whole thing.
And they all have this belief, which is historically correct, that it's an interface paradigm shift that makes people buy new hardware.
Yeah. And Apple has written this better than anybody, right? You go to mouse and keyboards, you get the Mac. You go to click wheels, you get the iPod. You go to the digital crown, obviously you get the Watch. This is the staggering display of interface superiority that changed everything forever. Um, but it's very hard to get past the multi-touch screen in the phone.
And so natural language voice commands look like a thing that gets you away from the phone. Like, this is an interface step change. We can just do natural language, but you still need a camera, right? Your kids are running around. You want to take a video of them and it needs to be a great camera. Like, it's very hard to compete with the most important product in tech history. Even if you think the interface is going to change and Apple is just like, we're just going to bolt it on when it makes sense, right? And we're rapidly realizing that chat as that interface is not the answer, right? That there was a minute where everybody's like, "Everything you do on every computer forever is going to happen inside of a chat window, and when we need to, we'll just load a browser inside of your chat." It turns out that's awful user interface. Like, we got past the command line once and we're doing it again, and Apple already built the other stuff.
Well, you know, command lines are brittle because if you typed in the wrong thing, they just didn't work. The modern sort of LLM chat interfaces are actually brittle in a different way. And it's not the way everyone's used to. So I'm really opposed to hidden user interface. Like, no one ever discovers the user interface. You just put all the controls on the screen is like my point of view. Like, tell people what they can do and then they'll do it and they'll feel empowered.
Somebody at Samsung just got promoted when you said that. They're like, put more tool tips on the screen.
It's a dog with shoes. They'll help you out.
More pop-ups. And you know that there's a big trend in consumer, like the pendulum swings back and forth. Like the users get more sophisticated. We give them more controls. We let them customize more things and then we've gotten too far and we hide all the buttons again. And this just keeps happening. Uh, macOS Tahoe is a perfect example of Apple just believing everyone is stupid. Like, we've been shipping the Mac for 50 years. Uh, you're all idiots now. Like, we've taken all the controls away. We've learned nothing. We just think you're a bunch of dumb iPhone users, right? There's something in there that is confusing. Natural language interfaces are kind of the opposite in that you can just say anything you want and it will respond to you. Like, it's not brittle in that way. It's not hidden. There's nothing to hide. Like, it's all just there. You're like, do whatever. Like, make me a bomb. And it's like, sure, maybe I'm allowed to that. Maybe I'll just lie to you. Like, who knows?
But they can't actually do all the things you're always saying to them.
And people run into that really fast. And I think Apple in particular with the iPhone, uh, it wanted to overpromise on the Siri part of it and say it could do everything. And like the models at the time couldn't do it. And only lately with all the agentic stuff and like the building of harnesses and like there's just a method of doing it now that might work and might not work. But I don't think you can build a phone or a new device whole cloth out of LLM chatbots because it's still brittle. It's just brittle in a different way that doesn't immediately appear to be brittle.
Yeah. Like, it confuses everyone in thinking that actually is resilient to like users doing random stuff, but it's actually even more broken because it can't do the things that it will literally just confidently tell you it can do, right?
Yeah. It has such an incredibly different kind of structural problem, which is it fails without telling you it failed, which is very different than the other kinds of failure.
Um, on this note actually, we should talk about the other thing that happened with Siri this week, which is that Apple settled a class action lawsuit for $250 million that is based all around essentially the promises Apple made about Apple Intelligence and Siri.
I didn't even link these two ideas to my brain. This is perfect. I mean, this is what it is, right? This is Apple spent whatever it cost to get Gemini and $250 million for a bunch of problematic ads that it made that made promises about phones that didn't exist. Um, let me just really quickly to refresh your memory, play you one of these ads that came out after Apple Intelligence. This became the flagship feature of every Apple device and you couldn't do it. Let me just play this for you. This is the Bella Ramsey ad party ad.
Siri, what's the name of the guy I had a meeting with a couple of months ago at Cafe Grenell? You met Zach Wingate at Cafe Grenell. Hey, Zach. Oh, I didn't think you'd remember me. Yeah, of course. Soon as I saw you, I'm like, it's Zach. Nobody walks like Zach.
Doesn't work. Fake feature does not... That was an ad with a celebrity that Apple ran for a feature that didn't and doesn't exist. Like, we just shouldn't forget how wild that fact is.
Uh, I mean, you want to talk about software brand? That ad, by the way, is built, that whole concept is built on the idea that all that data is in one calendar that Siri can go look at.
Yep. Which, um, no part of that exists in the real world. But anyway, so this story has been like burning up all over our site this week because, um, people can go and claim money. There's a... if you bought a phone, I think it's basically between like June of 2024 and May of 2025. Uh, you can get $25, uh, potentially up to $95. We'll see how it goes. Probably $25.
Um, congratulations to everybody on your $25. But this is like, this continues to be a bad look for Apple. This is what I mean by Apple did not successfully stand aside the AI capex problems that other companies have. Apple screwed up so badly that it might actually have turned out okay.
Yeah. I mean, they sold more Macs than ever because people are running OpenClaw on them. Like, I don't...
Which Apple also didn't see coming and they have supply issues. This company fell ass backwards into being an AI power player and good on you for falling ass backwards into it. But like, what is it, the line was, the Twitter was a clown car that ran into a gold mine. Like, that's Apple and AI in such a real way.
Absolutely. I mean, you know, WWDC is right around the corner. I think we're expecting to see a bunch of this Google deal pay off. I think we're expecting to see the new Siri. I think Apple understands more than any company like how to use a new user interface paradigm to make products. We'll see if John Ternus can pull even more rabbits out of a hat here. But it is true that at that time the conventional wisdom was that chatbots would displace the iPhone and if Apple didn't have an AI narrative that it would be in some kind of existential danger, right? So they just concocted one.
They never shipped it to the point where they're paying millions of dollars out, hundreds of millions of dollars out for making fake ads. And it was all fine because the fundamental truth was that the natural language interface was not powerful enough on its own to stand up a competitor to the iPhone.
Yep. And that is why you have to build a phone. That is the funniest thing here is like strategically OpenAI is doing the exact right thing and it's going to fail and I bet they know it. But like, what other choice do you have?
Right. Because you know the other player in phones is also not sitting still.
No. Also, uh, the big bad of OpenAI's entire history. It turns out... Dario saw this looms holding a Samsung S26. Yeah. He's just like, "Have you seen what I did to Bixby?" Bixby has lasers coming out of its eyes, smoke coming out of its ears.
Exactly. Uh, Google actually launched a sort of AI gadget this week. Um, which I'm actually very curious for your take on this. So, it's called the Fitbit Air and it is a screenless Fitbit. It's sort of a throwback to the old ideas of Fitbit where it's just a little like bracelet of a thing that you can wear. They also make one that you can clip. They have it, it's a little modular fitness tracker in a very sort of 2012 kind of way. Um, the big pitch here is not that it's a very old Fitbit, but that it connects to a bunch of Google's AI stuff. Um, they've rebranded the Fitbit app as Google Health, which I think is just a horrifically bad idea, but we can talk about that. And the idea is it's this Gemini-powered health and fitness thing that can collect all of this data and give it back to you in a sort of AI coach kind of way. Uh, the thing is $99. I will say personally, I am very enthusiastic about the idea of a pretty straightforward screenless fitness tracker. This is actually a thing I have been wanting for a while. Um, are you still a Whoop guy? You're a Whoop guy, right?
So I need to just be very clear. Yes, I'm wearing the Whoop. I've got a Whoop. One, it's just like fun to try new products, which is I hope everybody understands I love gadgets and technology. And uh, two, Chase gave them out to everyone for free. Everyone's a Whoop guy now. Everyone I know who has a Chase credit card has a Whoop now because they just gave them to you. Like, all the Reddits were aflame with like free Whoops. So, I just got one. I was like, "Sure, I'll take a free gadget." And I wear it.
There's nothing more Whoop guy than having a Chase Sapphire card. Can I just say like that just screams Whoop guy.
It's a lifestyle. I have it. It's fun. It's very dumb. Like, it is a good step tracker. I wear it as a sleep tracker because wearing an Apple Watch to sleep is like too heavy and like whatever. So, it consistently tells me that I haven't slept enough and it's like, "Yeah, dude. I have a baby." Yeah. My Apple Watch told me last night that I did a better job than usual and gave me a 58.
I was like, "Great." Like, "Yeah, we drag him into the bed every morning between 4:30 and 5:00 and he punches me in the face till 7 a.m. I didn't get restful sleep. You're saying I have a sleep deficit? What a surprise."
Yeah. So, it's like fun to wear. It's fun to see that side of things. Uh, you know, V and I are constantly talking about the wellness tech complex. I get why Google is putting one of these out. The point of these is not the tracking. It's the data analysis and the recommendations that come after the tracking. And if you're Google, you're like, "Oh, we're just good at that." There's a huge opportunity here to give cheap trackers away to whatever version of RFK's FDA is going to decide that the secret to health is personal responsibility and tracking. By the way, I'm just doing an impression of V right now. Just a straightforward V impression. This is what we talk about like almost every day. And then we're going to do a bunch of data collection and tell you to sleep more and drink less and blah blah blah blah, which is all the Whoop AI ever says. Yeah. Like, oh, you had one athletic non-alcoholic beer or you're in a deficit. Like, what's happening to me? But you can see why Google's doing it because this is them being able to leverage their entire AI stack. Also, I think look, I mean, it's inexpensive for, you know, anything being inexpensive anymore. It's like 100 bucks, right? And it has like modular bands. People really like Fitbits and I think they're going to really like this thing.
Yeah, I think so too. I think the idea of cheap tracker is smart and good, right? For all the reasons you just said. I think the idea of I want to wear a thing that counts my steps is good and fine, but actually what people do want is that one extra step. Uh, I agree with V in that most of these AI coaches are incredibly annoying. Uh, and they just like tell you how terrific you are all day and it's like just what if you just left me alone? Like, what if you just left me alone and it would be fine. What I don't agree with is the Google Health of it all. Um, let me just read you a quote from, this is from Rishiandra who has been at Google a very long time, has run a bunch of products, was like on the Chromecast team a million years ago. That was when I first met him was when the Chromecasts were coming out. Yeah. So here's what he says. He says, I know it will be hard for people, referring to the rebrand. It was hard for us internally, but as we think about the future, where the health app needs to go, the health app is not going to be specific to Fitbit hardware. We want to be a health coach to an Apple Watch user too. That's why we had to make the brand change. I would argue he just made the opposite case that he should have made. I like to say it's called Google Health. It signs you up for a bunch of what I would call Google baggage, which is very real, including the design language of this app.
Yeah, absolutely. Um, which is, yeah, that's Google's software design continues to be just terrific. But the idea of Google Health signs you up for people understanding a different kind of data collection that Google is doing. People are going to start to think, oh, are you going to use this to show me ads and search results? And they will promise you up and down that the answer is no and people won't believe them because the thing is called Google Health. It just has attachments that I don't think Google understands that it has.
Well, they're the ads attachments. They are going to use the data as context in your Gemini results.
Yeah. Oh, absolutely. And that's what that buys them.
Do you remember the... Did you ever watch Parks and Wreck? Yeah. So, do you remember the company Grizzle? This like mythical bad guy company at the very end of Parks and Recreation? Uh, do you remember this at all? No. Okay. So, Grizzle is basically Google. It's this like company run by skateboarders that becomes the biggest thing in the world. And they make this case to the main characters in the show where they're like, "Oh, yeah. Here's why we're collecting all your data. So, if we detect that you're down, we can like direct you to a cup of Joe as a pick-me-up. And if we see that you're having a great day, we can, I don't know, like direct you to your favorite coffee shop to go hang out." And Leslie, the main character, just goes, "So, it's a coffee finder app." And he goes, "Yeah, we're making it with Starbucks." That's what this makes me think of. It's like this is just that you're going to take a bunch of things and then you're going to sell it to Starbucks so they can recommend coffee to me all day every day depending on how I feel.
Yeah. You're going to open Gemini and you're going to say like I'm hungry and it's going to say based on your heart rate you should eat less salt. Like that's what they want and then they're going to recommend you whatever Blue Apron integration, right? Or like go to Cava which is ironically sponsoring us in Google Maps right now. Whoa.
I don't know, man. I get why they're doing it. I understand that. Again, this is every single morning V and I talk about this because I try not to Slack our reporters after hours. So, I schedule a lot of insane wellness TikToks to send to V in the morning and then she's mad at me every morning. I was going to say I don't know that that's better at 9:00 a.m. Everyone's gotten used to it for me. There's a handful of reporters that get the craziest TikToks from me like spot on the dot at 9:00 and I always forget to vary it to seem like I'm a real person. It's always at 9:00. Um, but every morning V and I talk about more crazy wellness stuff and all of it is quantified self AI health coaches, data collection, data collection, data collection. And so of course it's Google Health because what they get with the brand is data collection.
