The Vergecast - How DeepSeek crashed the AI party

Episode Date: January 31, 2025

Nilay and David dig into the week's biggest story: the new Intel-powered Surface Pro. Kidding! They talk about DeepSeek, the out-of-nowhere AI company that sent both Silicon Valley and the stock market into uproar this week. Then, after the hosts debate what the real killer app for AI is — and whether we've even found one yet — we follow up on our question from last week about how people are actually using AI. We got so many good answers, and we talk through what to make of them all. Finally, in the lightning round, we talk about Brendan Carr being a dummy, the return of the Pebble, the continued rise of Bluesky and Threads, and Meta's $25 million check to Trump. Further reading: Why everyone is freaking out about DeepSeek DeepSeek says its newest AI model, Janus-Pro can outperform Stable Diffusion and DALL-E 3. Microsoft makes DeepSeek’s R1 model available on Azure AI and GitHub OpenAI has evidence that its models helped train China’s DeepSeek China’s DeepSeek AI is hitting Nvidia where it hurts DeepSeek’s AI app is restricting sign-ups due to ‘malicious attacks’  US Navy jumps the DeepSeek ship. DeepSeek wakes up Trump. OpenAI CEO Sam Altman on DeepSeek R1: “an impressive model.” Mark Zuckerberg tells Meta investors to not worry about DeepSeek The Pebble smartwatch is making a comeback, with some help from Google  Oracle and Microsoft are reportedly in talks to take over TikTok FCC chair says landlords can force bulk internet service on residents From NYT: F.C.C. Chair Orders Investigation Into NPR and PBS Sponsorships Meta agrees to pay $25 million to settle Trump account suspension suit Zuckerberg wants to Make Facebook Great Again Zuck wants to bring the “OG Facebook” back. Email us at [e-mail] or call us at 866-VERGE11, we love hearing from you. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Support for the show comes from Retool. Too many companies run critical operations on duct taped spreadsheets, Slack workflows, and whatever else they could cobble together. Not because they want to, but because building internal tools means weeks of waiting on someone else's backlog. That's where Retool comes in. Build custom internal tools just by describing what you need. Prompts something like,

Build Me a Revenue Dashboard on our Salesforce data. And Retool actually builds it on your company's data, in your cloud with enterprise security built in. Go to retool.com slash Verchcast. We all need to retool how we build software. What's up, y'all. I'm Skyler Diggins, seven-time WMBA All-Star, Olympic gold medalist, and mom. And I'm Cassidy Hubbard, host and reporter for nearly 20 years covering the biggest names and stories in sports and mom. And this is Am Mom, a community for athletes, game changers, and moms of all kinds.

dropping May 14th. Tap in with us. Do you ever wonder what's in your lotion? If you look at the back of the bottle, it could contain more than a dozen ingredients. And they may not all be regulated. The threshold is so high that only 11 cosmetic ingredients have been restricted by the FDA since 1938.

This week on Explain It to Me, the chemicals lurking in your cosmetics. New episodes, Sundays, wherever you get your podcasts. Hello and welcome to the Redcast. The flagship podcast stealing stuff Sam Altman has already stolen. You're on notice, Sam, you steal a car. I'm coming to steal that car. Look, I've been a lawyer in a long time.

Isn't that double jeopardy? Like, if somebody else has done a crime and then you do the crime to them, you're good. Yeah, if it's additive crime, I think it is fine. Yeah. I don't, look, I passed the bar exam a long time ago. I've drank a lot since then. Just in the past five days. the Trump administration.

They drank away one full bar exam. But I do believe, and correct me if I'm wrong, that if someone does a crime and then you do that same crime to them, the court's like, you're good. Listen, if we've learned anything in the last two weeks, it's that you can, in fact, just say things. Whatever you want. Yeah.

And sometimes those things turn out to have huge weight over how America works. So here we are. We're fully in the I declare bankruptcy stage as the American experiment. As I believe George Washington prophesied in his first inaugural, he's like, one day people are just going to start saying stuff. And you should row a boat away from here as fast as you can. Those are all the things I know about George Washington. All right. There's a lot going on this week.

I'm actually very excited about our second segment this week. Last week we asked a bunch of you to tell us what you were using AI for because there's a big gap, right, in how people feel about it, what the company say about it, the amount of money we're spending on it. how people are actually using it. And it turns out a lot of you are using AI. We got so much feedback, like so much feedback, including from a bunch of people who, like, were outraged at the idea that not everyone uses AI for everything all the time. And I really actually appreciate that. It was, it was very cool to go through and be like, oh, these are like real people who use it in their real lives and not just like somebody trying to convince me to give them billions of dollars. Are you sure that Sam Altman wasn't outraged and not everybody's using AI?

Do I think Sam Altman has a lot of burners and is in our inbox on occasion? I do think that. He does have access to chat, GPT. I will say that. So we're going to go through all those. I think that's going to be really fun. We got a lightning around. But we got to start, I think, with the tech news.

It's the tech news. And it's obviously that Microsoft is continuing to launch Intel powered surface devices. No? For $500 more, you can have a surface that has more words in the name and a chip you don't really want. And worse battery length. Congratulations. You did it. There's something there. They did announce that they're going to do this.

They also announced that Microsoft is running Deepseek now on co-pilot PCs that run Qualcomm chips. We're going to talk about Deepseek in a second. But it's just interesting that there's still, Microsoft has to still carry Intel in this specific way because that partnership is so long and so important to everyone. But it's like

everything about Microsoft releasing new service devices has changed. in the past 18 months to where two years ago, they're like, here's some service devices with arm chips. I think the reaction would be, eh. And now it's here's some service devices with Intel chips. And it's like, oh, man, you're still doing this? And that's a big change.

Yeah, and it really did happen. A, really fast and B, almost without anybody noticing. Like, I feel like every step along the way was also kind of meh. Right? Like that first round of co-pilot plus PCs. still the worst name in laptops, came out. They were fine, right?

Like, the overwhelming response was like, Qualcomm did pretty good work here, but the emulation stuff is still pretty bad. The performances and what we, it's like, it's just not, it's fine. And then somehow between there and here, we've flipped all the way to, like you said, I see Intel Surface pros and go,

that's not what anybody wants. Like, it's crazy. But I don't, and I, to be clear, I don't think the arm powered chips are doing everything yet. There's a reason to sell Intel powered laptops. Lots of people are still buying Intel powered Windows PCs, obviously.

But it's just interesting to see where the bleeding edge has gone to. Like, it's obvious that that thing has runway and Intel, and you know, along the way they did fire their CEO. They did do that. So maybe it's just vibes all the way down. That's not actually the biggest news. That was a joke and we ended up talking about it because it's a first. The big news is Deepseek.

Yeah. It has turned the tech world upside down in just a very, the only way to describe it is a very bubbly way. Like everyone was over-invested in over-hyping AI and then everyone overreacted to another AI tool. What do you think, David? So I actually think that's the right frame for it because the more I've learned about this

and the more I've talked to people about it, the more I've come to believe there are sort of three simultaneous things going on. On the one hand, there's the stock market freakout, which is just a stock market freak out. a bunch of people, I would say, like, half correctly understood what Deepseek is and made a bunch of weird, semi-thoughtful bets against Invidia as a result. And there's been some really interesting reporting over the last few days that a big part of the sell-off was triggered by this one super deep, super thoughtful blog post about somebody who was shorting Nvidia. And he, like, writes this big thing and lands on short Nvidia.

it has no moat, this thing is not, this thing is massively overvalued. Got shared by a bunch of people in tech and you can kind of argue that that's what caused a lot of that. So on the one hand, there is just a lot of money was lost very quickly. So a lot of people got in a like CNBC tizzy about that.

So that's like one thing happening over here. The other thing is very much a story about technology. And the actual question of like, okay, how do we think about what DeepSeek did here that is interesting and unique? And how do we think about open source AI being really at the bleeding edge

a lot of this stuff? And what did Deep Seek do that is interesting that's going to roll back to some of these companies? And what does it mean for Open AI? That's all interesting. I think we should spend most of our time talking about that. And then there's a third thing which is just China. It's just the China of it all. And I think everybody sort of twists themselves in knots about everything to do with China in part for good reasons and in part for bad reasons. And in part for bad reasons, but like coming off of all this stuff with TikTok and Red Note and the ideas about

what these companies are collecting about us and what it means that there is like technical leadership happening in those countries. And like everybody is just really spun up to have feelings about China. And so there was a lot of feelings about China in there altogether. So it just became this like perfect storm of I think under a couple of really small changes. This would have been like an interesting story and not like the only thing anyone in my group chats cared about all week. But because it was those three things together, it became everything. Like my high school friends are talking about deep seek. This has just been happening this week. Well, once you turn the entire economy into gambling, which is what we're doing. Yeah. Everyone cares a lot about the gambling.

And to like whatever, you mentioned the CNBC of it all, like most people's retirement savings are like in big index funds, which are all just big tech stocks. Right. And so... It's like eight companies now. Right. So when NVIDIA goes down or Apple goes down or META goes down, like actually the whole economy kind of goes with them, which is not great.

There's a lot of reading and reporting you can do about that. That's not what we care about here at the verge, but it's just notable that that's why it breaks through so often now. Yeah. Oh, I agree. And I think in NVIDIA's case, I mean, like what people have been saying about NVIDIA for months is that it has essentially propped up the stock market on its own for a very long time now, and it has gone through the roof.

Like it's grown like crazy. And so, yeah, everybody has been saying there is a bubble and just pointing at Nvidia. By the way, total side note, one of the funniest things about every time Nvidia is in the news is the number of different ways people pronounce the name of that company. There's a lot of no video happening, like NUH video, which I like. If Verge has, if you're paying attention to the verge, there's some of that floating around the birch lately. Listen, we're going to fix it. The thing is, I'm not sure anybody's wrong. You know, who knows?

It's nuclear and nuclear. Like, you know, it all gets the job. Wait, hold on. There's a, we can't slide this far. There's an answer. We're going to clean it up. That's all I'm saying. Okay.

So it's like, we agree that you would pronounce it like in Vidia. Like there's a sort of sign. There's an eye at the beginning, right? Yeah. Like a gentle eye. Their name is so stupid that they have pronunciation guidelines online. Do they really? Yeah.

They have for years, years and years and years. It doesn't matter because like the CNBC people will just say but also Nvidia might be the company that most violates my theory that you can't overcome a stupid company name and it's a stupid company name it's hard to pronounce it doesn't mean anything it has nothing it's just it's just like an Amazon brand

of letters that they happen to turn into Oh that's one of the biggest company's owner But it's good that's a rough chuckle That's really good I bought this Nvidia battery bank That's what I'm saying $39 with a 10% percent coupon code. Yeah.

Like in a parallel universe, U-Green makes all of the GPUs that people buy in the world. And this is where we are. But yeah, so the money of it is real. And I think to the extent that it's also tied up

in all of this other stuff, it's, that's important. But to me, that is actually like, that is the thing people got most sort of bent out of shape about and took it to mean

that there is all this other equally important stuff happening in these other ways, like with the tech itself. And the more I learn about it, the less I feel like that. actually the case. So I keep making the comparison to Bluetooth with AI. It's just, it's just my go-to. You know you've been doing that a lot recently? You compare a lot of things to Bluetooth recently.

Everything is Bluetooth. Right? Like we had to, you don't remember this. When we do it first cast coffee table book, by the way, that's the title. Just throwing out of it out of it. Everything is Bluetooth. Everything is Bluetooth. When Bluetooth first hit the scene, they were so proud of themselves. And they're like, everything is going to be Bluetooth. Bluetooth is going to change the economy. they were they were out here talking about personal area networks or pans oh yeah i wrote a lot of blog posts and gadget about personal area networks how's your pan doing it's awesome dude it's my apple watch and my hair pots yeah my pants this is a real thing that was going to happen right like the we were going to blow up the world with Bluetooth and then the actual products didn't match the

investment for years and so like it's just a good comparison 5G i'm actually is another great comparison. We were in a race for that, you will recall. Yeah. To what finish line, no one knew the answer. Robot surgery? Seems unlikely. Autonomous cars, there's some of them driving around. Sure.

Sure. But, like, that was the big input. And also with 5G, you will recall, we had to beat China. Because if China got 5G networks first, America was done. Right. You can't have it. Which is very much what people are saying about AI right now. It all just tracks.

Like these things rhyme, right? The difference is when AT&T and Verizon spent billions of dollars to build 5G networks against whatever hype and whatever grafts. They at least knew they were going to sell people cell phones and sell service. True. At numbers that people were willing to pay. Right? And the people who put Bluetooth everywhere, they didn't know that there would be AirPathes. But they could imagine AirPods.

And they were like, one day there will be AirPods. And then eventually Apple invented AirPods. Yeah, you could get all the things with wires and be like, what if no wires? And like that, that worked. You could see how you'd get there. But with AI, we just keep, we're like, what if it bangs you? Is that worth $500 billion? For some people, the answer is decidedly yes.

We're like, what if it's agents and it does some clicking around the web for you? What if the rabbit R1 is great? What if the humane pin is great? To quote David Pierce, is this a thing? And there's not the finish line. Right, where everyone's going to pay all of the money for the products that justifies all of the investment. And I think my version of the deep seek story is like it just showed that all of this big

upfront investment maybe isn't necessary. And so meta doing dick measuring by how many Nvidia GPUs it was buying kind of is like, are, wait, are you just wasting money? Like, is there, is there an end result of this? Or can, or can you just do it much more cheaply now? And I don't, I don't know the answer to these questions. Because I don't think anybody has really delivered a product that is going to pay it all back. In the context of Deep Seek, that conversation has been super interesting this week.

because like the case for all of this being good news for the Mark Zuckerberg, you know, talking about the size of his data center of the world is, okay, let me see if I can explain how the theory goes. The theory goes like this. If you make AI cheaper to make, two things will happen. One, you'll be able to make better AI for the same price. So you get the performance scale goes up, right? I can now make something 50% better for the same price as I was planning to do it before.

So, thus, good AI. Then all the stuff that I've already been doing gets cheaper, so it becomes more accessible to more people who use it for more things, thus also better. So it's like this flywheel of scaling is what everybody is very excited. Dario Amadee, the CEO of Anthropic, made essentially that argument, right, that now, because all this stuff is getting cheaper, and this has been happening for a while, Like, Deepseek didn't invent making AI cheaper to run. Like, this has been happening for a while.

One of the very first Google IOs and this whole AI tumult, like two years ago, literally Sindar Pichai was saying the thing that we're best at is making things cheaper. Right. And so now what you're thinking is like, okay, if you're Sam Altman, instead of spending $10 billion to do this thing, I can spend $10 billion to do the next thing, right? I can make the thing that I'm working on with all this money 50% better. But what is he making? This is what I'm saying.

So I actually buy that theory, right? Like it all makes sense, but the question continues to be, what are we laddering all of this up to? What is the pot of gold at the end of this rainbow? And my favorite thing that happened this week was, did you read, read Hoffman's New York Times opinion essay that everybody was dunking on all week? Yes. Can I just read out loud two paragraphs from it? Please. He says, imagine AI models that are trained on comprehensive collections of your own digital activities and behaviors.

This kind of AI could possess total recall of your Venmo transactions and Instagram likes and Google Calendar appointments. The more you choose to share, the more this AI would be able to identify patterns in your life and surface insights that you may find useful. Decades from now, as you try to remember exactly what sequence of events and life circumstances made you finally decide to go all in on Bitcoin, your AI could develop an informed hypothesis based on a detailed record of your status updates, invites, DMs, and other potentially enduring ephemera that we're often barely aware of, as we create them, much less days, months or years after the fact. This is Reid Hoffman writing about how AI is going to make humanity better and the best use case he can dream about is it's going to tell you which DM made you want to buy into Bitcoin. Like, what's the plan here? And now, and I've been thinking about this a lot with the questions of like, okay, well, then if all these companies don't need this much money, why are they raising all this much money?

And it's like, well, because money's good to have. And you should, if you can raise it, you might as well. And so Sam Altman is out there raising unheard of amounts of money and there are reports that they're about to raise anywhere from $15 to $25 billion more dollars now. Like, Sam Altman is as good a fundraiser as we have ever had in this industry. And so he's going to go out and raise money, but it turns out, I'm not sure anyone knows what we're building towards anymore. And I am less and less sure every day that people know what the answer is to what problem they are solving with AI. So we're in this place where like, Deepseek is very interesting in that it is going to, make, it is going to help make a lot of this stuff easier and cheaper to make. The fact that it's

open source, I actually think is on balance very cool. Uh, there are scary things about open source AI, but whatever. I think by and large, this stuff should be in the open and not in the hands of three companies. Uh, it's going to change the way that this stuff gets developed. People are going to like take the things that deep seek did, bring it in, make it better. The whole industry will get better. And to what end has not become any clearer at all. Yeah, it's funny. The reedhofen thing i read that too and you know like you do in trump administrations i i spend i've been spending a lot of time think about the inner lives of rich rich men sure like very rich men what do they what do they want for me because they seem to be in charge of everything um

and it's just like dude wants a friend like i don't i think a lot of these guys don't have all their relationships are just people who want things from right and so when they're like what what could this robot be for me they just imagine just a friend who's known them since kindergarten. But then do you think there's also a little part of them that's like, wants their AI friend to know how rich they are? Yeah. Oh, absolutely. They're like, why don't you think my ideas are good?

I don't want your ideas AI. I want you to like my ideas. I just think if I have a piece of life advice for everybody listening, it's like you should have some non-transactional relationships in your life with people who just like you and you like them and that's it. You're not trying to buy and sell everything all the time. That might be good for humanity. I always think about Steve Jobs.

You know, he would end all the keynotes with like apples at the intersection of liberal arts and technology. And like these dudes are like, fuck the liberal arts. It's like, no, that's the important thing. Like that's why that's why that company is important because they like care about culture. They're like the computers are to make art with. It's why they can't make AI for shit though. But that's that's so funny. Yeah.

I don't know what Steve Jobs was. I thought about AI. But it's like you look at the output and it's like, it's pretty bad. And it's like, is this good? Like, this is, do you want to make more stuff with this? I will say, deep seek, there's some copy going around that it had written about, like, the nature of tools, which is a very vergy thing to write about. Dieter made an entire video about whether or not computers or tools or instruments once in a while of time. It was pretty good.

You know, it's like, not great. But it had some bangers in there, you know? It didn't sound like chat GPT, which I mean is a compliment, right? There is this, like, I think we're all kind of learning how to sense when something is, written by AI, and it didn't sound like that. It was weird and kind of overdone and overwrought, and it was very much like an overly pretentious college sophomore writing a paper for their English class. But like, that's a lot better than chat GPT. So I'll take it. Yeah, it's not like a receptionist who's kind of annoyed that you're there, which is chat chip chip because he's default tone. Like,

I'll be so corporate that you will go away is kind of how chat GP writes. But I just think that's like very fascinating. Like, what are these, what is this tool for? And the first, answer with the image generators was art, right? And then the second answer was the chatbots will write me 10 more episodes of Seinfeld. And then it was like, they'll kill Hollywood. And then we just kind of like brought it all the way down to like, it'll be a friend that'll book an Uber for you and maybe buy a sandwich. And it's like, this is a really small vision at the end of the day. Just as we're talking, I think you just kind of mentioned it, David. Opened eyes like in talks to raise more money

at a $340 billion valuation. This is right on top of them announcing $500 billion to do Stargate, which is not real, right? That's the most, like, talk about Trump administration. Staying in the White House announcing a project that you're not going to build as like the heart of the Trump administration experience.

It's a giant novelty check. Like it really is. Yeah. That whole thing was a giant novelty check for $500 billion. Sure. But the reality check on this whole industry right now is you're spending all this money because no one was stopping you.

Everyone wanted crypto to be it and crypto wasn't it. So then maybe I'll be it. And because it can do something as opposed to crypto, like all the big companies are like, look at this new thing our phones can do. That's Apple intelligence. Like, great. But it hasn't done enough yet.

And then Deep Seat comes along and they're doing it for way cheaper. And I still don't think it's good enough. But it's so much cheaper that it's like wiping out this like investment thesis that you need all of the money to make a friend. Right. And again, it goes back to the thing of, uh, you have to start to explain to people what you're making. And I think like the this, this, this we are making God thing isn't going to work forever. It's already starting to come unwound. Like a thing I've really noticed is, uh, as Sam Altman and others have started to pull back on what they claim age. GI is. This idea that we are like marching towards a singularity where these things are going to be as good

as humans at everything is like getting more nebulous and weird. And so like if anything the goal posts are getting fuzzier for these companies, not more clear. And so it's wild that it continues to be this frothy when it feels less and less like anybody has any idea what they're going towards here. except it will reinvent the economy. And that's what you have to say, because that's the only way to raise $25 billion more dollars

is to claim you are going to change the way that people live their lives. And, like, that's how Adam Newman convinced Masa Stone to give him a lot of money, and that went super great for rework, and we're just doing it again at even more scale now. It is notable that SoftBank is investing in total, yeah, at large valuations. It is also notable that, you know, Sam Altman is, you know, making videos where he's saying AI will require us to renegotiate the social contract. And Mark Andreessen is like absolutely not.

And then in the next breath, Mark Andreessen is tweeting like, imagine a world in which wages go to zero and everything is free. And it's like, well, that's renegotiating the social contract mark. I don't, what are you talking like that? Also, none of that is going to. I cannot say clearly enough. None of that is going to happen. It's just not.

Like some, if that does happen, play this clip for me and I will be suitably embarrassed as I sit on the yacht that I got for free because of AI. This is not going to happen. And until then, it's like, okay, sure, let's work backwards from reshape the way humans live our lives. What's the first killer app that you're going to, like, what are you going to make for me that is great that I like and use all the time? It's Apple Intelligence, too.

You can summarize any email you want at the click of a button. Yeah. I saw an unnamed big tech executive and I just said the words Apple intelligence and they just started laughing. I think that's right. That was basically where they were at with it. There are two other funny parts of the Deep Seek story. Interesting but funny. One is that Deep Seek is the result of Biden administration policies to restrict chips in China.

And so it really feels like a bunch of. of US AI researchers got lazy, and they were not optimizing their code enough. And the Chinese researchers with H-800 chips, which are Nerfed H-100s, over-optimized those chips to get better results than the H-100s, which is just like a classic of computers story, right?

Yes. Like you started out with 64 kilobytes of RAM, and you just like worked your way up from there, and then the old head programmers are like, these kids today are wasting cycles on animations. Like it's all it's it's just there. It's the same story. Yeah.

I don't know if it's laziness. I think, I mean, there's there's certainly some laziness. They were counting on an infinite supply of processing power. Right. Right. But that's like it's,

you know, what does it constraints breed creativity? Like yeah, there is something to that. But I also think, I don't know that they were sitting around being like, well, we don't have to try hard because the chips will do it for us.

Actually, now that I say that I do, I know for sure they were kind of doing that. See what I'm saying? Yeah, no, you're right. Once you work it all the way. You're like, oh, they didn't.

They were just like, we'll just buy more chips. Like, Sam, get more money and buy more chips. Right. There was a brute forcing to this problem that everyone just assumed was going to be how we get better. And they've been making changes to the foundational technology. And like, Transformers are better than they were in 2017 when people started talking about this stuff. Like, they're doing that. But nobody at these companies spent a lot of time thinking about how do I do this on worse chips because they just had the good chips.

Right. And, you know, a month ago, we were talking about scaling laws. I'm like, can you just throw more horsepower and more data at this problem to make the models more efficient? And the answer was like, kind of no. Right. And that might have been when the bottom should have fallen out of all these stock prices because that answer was no. Then this moment was coming inevitably. Well, but then what they all did in response was say, okay, well, we're running into walls with our, you know, frontier models at these unbelievably huge things.

GPT5 is reportedly not what OpenAI wanted to. it to be like nobody can find kind of the next huge turn. And then they were like, but if we do even bigger data centers. And it's the same thing to your point. It wasn't like, let's go back down to the metal here and figure out like from first principles how to do this. It was what if we throw even more chips at it. Yeah. Which is how you get invidia. And very much, like, you know, Liz and I've been talking about parts of the invidia story for a while now. I think she's even written it up. Like there are startups that are financing their purchase of Nvidia chips with the Nvidia chips as a collateral on the loan.

Yeah. Because their assumption is that there will be so much demand for usage, they will be able to just like make it work. And it's like, yeah, that's a little, that's got some NFT vibes to it. You know what I'm saying? Seriously. And you just see, okay, well, they didn't, the Biden administration put restrictions on the chips we could sell in China. And the Chinese researchers at Deep Seek were able to ultra-optimized those. those chips in a way that U.S. engineers and researchers had no incentive to do.

And then this is, so that's funny, right? That's like an unintended consequence of a restriction. Yeah. That like, fine. Like, it's funny in the sense that like trying to nerf the Chinese AI market ended up tanking the U.S. stock market. Weird. It's pretty good. That's not what you want there.

And then the second extremely funny part is this idea of distillation, where part of DeepSeek was trained by just ripping off OpenAIs models in various ways. That's pretty good. And like, you know, Sam Hallman is like a little bit outraged by it. And it's like, Sam, you stole the entire internet.

Everyone's so mad at you. Yeah, I mean, Deepseek is, apparently was built on top of Lama, meta's model, and used vast quantities of output from chat GPT to train itself, which is just, it's just perfect. Like, it's just, if every single stereotype you want, to believe about the AI industry, every single stereotype you want to believe about China. Like, it's all right there. It's great. Yeah. And at least meta, to its credit, is like, yeah,

we made the model open source. That's what it's here for. Like, you know, uh, open, I'm making noises that like, to have evidence that this was. And it's like, yeah, who cares? Yeah. Like, I don't know they broke your terms of service. How awful. People have a lot of evidence about you, my friends. Like, you know what else should take the bottom out of the AI industry, losing the number of copyright lawsuits that you are currently engaged in. Yep. I feel like I need to disclose at this time that Vox Media has signed some kind of licensing deal with OpenAI to make our stuff show up, show up and search TPT. And honestly, no one uses it. So it's fine. Seems right. That's way above our heads. But there it is. There's a disclosure.

They just turned Gemini on for our Google workspace. That's a thing that happened this week. Disclosure. We had to have a meeting trying to turn it off and keeping our data inside our own house. Even that, right? We're going to take a big bite out of the search market and knock Google off it. Like, It hasn't played out. And that's the biggest identifiable market so far, right? You will ask the row out questions and it will deliver your answers. Right. We're going to make Google dance.

All of that doesn't seem to have played out. No. And I think that's why everybody's betting so big on agents, right? It's like, okay, somewhere between we're going to take a cut of your DoorDash order and we are going to be the payments processor is like that's the idea everybody has had. Like, I think it was you that said. to me at one point on the show that everybody's business model ultimately is just payments processing. Like, it's really, it's really extremely true. Everybody is just desperately trying to do payments

processing, including now X this week, which we can talk about. But these ideas just aren't working. Like, it is true, and we're going to talk about a lot of the stuff that people are doing with these tools. It is true that people like using these tools. And that is a meaningful thing. but the question of what are they for and maybe even more importantly what is the business here super duper unknown like it's important to keep reminding you that the bigger open AI gets the more money it loses and that continues to be true and like there was like the Uber thing that was like you know when when we win we'll take over and we'll raise our prices and like they they did win they did raise their prices they had to reach

trench in like a huge way and now they're making this much money. Yeah. That's not coming for any of these AI companies anytime soon, no matter how many billions of dollars you raise. It's just not coming. Right, because their product isn't as useful as when you push a button, a Toyota Highlander shows up. Right. Right. And it's, it's being commoditized even faster than that. Like, the thing everybody was worried about about Uber 12 years ago was robotaxies are going to come and just throw all this into chaos. That's happening in the AI world. Like, all this stuff, whatever you build as an AI model company is going to be copied and bested, like, immediately by six other companies.

And this is just what we're seeing. So let me ask you this. Microsoft had earnings this week. They're doing fine. But, you know, it seems like the Microsoft Open AI relationship has been rocky for a minute. Yep. You know, they were like part of the starting announcement, not really part of the starting announcement.

That was mostly funded by Oracle and SoftBank. Nadella was asked about $500 billion, and he's like, we've already committed $80 billion to Azure buildouts, and I'm good for my $80 billion, and just like wouldn't talk about Stargate. And then this week, Altman and Nadella, Altman posted a selfie of the two of them,

so the next chapter of this relationship is going to be great. But then the companies had also announced that, like, Microsoft's deal basically goes until 2030. I can't tell what's going on with these two companies right now. And especially he posted that after Deep Seek, and after Nadella is trying to calm down, the market by talking about Jevin's Paradox, which is what you're describing, where you make the thing cheaper so people use it more. What is going on with these two companies?

I think it's as simple as they are, they're sort of in a mutually assured destruction thing where like Microsoft benefits so much from OpenAI being a huge success because it has these agreements for all the cloud stuff that's happening because it's such a huge investor in the company. Like, as Open AI gets bigger, it uses more Azure, which is more money for Microsoft. Like Microsoft literally wins coming and going when it comes to Open AI. So if I'm Satina Della, it makes sense to be doing both of these things. You both want to shore this stuff up in-house in case this company implodes in some meaningful way, which A, it has already done once, and B, could just happen again.

Like, the people a lot this week have been talking about the fact that most, Most of Open AIs co-founders aren't there anymore. There are many litigious people circling around this company. There are lawsuits circling. Like, Open A.I. is both absolutely chockful of money and, like, pretty fragile in a lot of ways. So I think if you're Satsia Nadella, you're like, okay, I need to not be fully reliant on it. But I also am going to make so much money from Open AI if it hits that it's worth a selfie and, like, a nice tweet a couple of times a year. Like that's a pretty low risk way to at least keep the piece enough that open.

Nobody gets super nervous about open. Like one selfie. Like he's got a chart in his office. It's like options. One selfie. Invest $200 billion. Yeah. Leak phone call to Sam.

Like, yeah. Be caught having lunch with Sam at Sweet Green is like $200 billion. Like it's I really, I think these companies hate each other. And I think they know it's better to be stuck together for a while and to root for each other because it will be successful for everybody. But I think they will be in, they, I think Sam and Satya both go home at the end of day and like throw darts at a, at a, at the other person's face. The really interesting thing is, you know, Microsoft's core business is Azure, like in the realest way. Azure's doing great, you know, their earnings are out. there are parts of the Azure business that are like different than the others.

Like their new customers are not signing up for AI. They're just signing up to be cloud customers. And they kind of announced for the earning call. Like, yeah, we stop trying to do that. Like we're back to just trying to get cloud customers. Yeah. I mean, that's what that that has happened in a real way. Like that's actually been a big moment that both Microsoft and Google were like,

okay, we're going to charge $20 a month extra on top of the normal like workspace or office subscription. and you're going to get all the AI stuff, and people didn't. Like, $20 per month per person for nebulous AI features turned out to be an awful lot to ask. And these companies, like you said, they need more cloud customers, and they need people to use the AI in order for the AI to get any good. And so they have just walked it all the way back. And what they want is they want you to use their tools that make the money, and then they want to shove AI at you so that you use it and maybe someday we'll pay for it. Yeah. We'll see how that goes. But that's like a really interesting piece that Microsoft is at least reselling whatever capability it has.

Google's reselling it. Amazon is reselling it. InVIDIA sells the chips. And then they, you know, like Jensen flashed on the world on his jet made of money. Meta is really interesting in here because they're the only ones who don't resell the capacity. They are buying all the GPUs. They're building a huge data center the size of Manhattan, like crazy. But they sell none of it. It's all for meta. And either that's a lot of risk, right? They have to get the value out of it. Or they figured it out. And in their earnings, this isn't great.

I don't love it. But they're like, we have increased ad targeting yield by sorting through all of the ads we can show people faster. And that's what the GPUs are for. And it's like, oh, they actually figured it out. They know why they're spending the money. They're not just trying to resell capacity to somebody who's going to make a killer app. or assuming that they can make digital cod. Yeah.

They're like, oh, there's a thing we do with this that actually goes faster because we're doing it. Yeah, I mean, the killer app for AI might be ad targeting. Yeah, which is terrifying. I want to be clear. That sucks. Yeah. And meta would love for you to not think about that

and instead think about smart glasses. But like meta is building those data centers to serve you better ads, not to like let you wear glasses on your face. If they can do all that stuff, terrific. But the business case for that Manhattan size data center is advertising. And I think that's what is happening across the world. It's true for Google. It's true for a lot of these companies. Like, ads are going to be how a lot of these things support themselves in a way that I think people are not yet ready for. In the case of meta and Google,

specifically, they have already started down this road where the end result is your some advertiser and you just type in who you want to reach and you put in the first version of your ad and then it makes infinite variations of that ad targeted at specific people. So everyone is seeing their own ad. Right. And that's what the generative AI is for.

And this is either good or bad, you know, like, it's bad. But like, you know, from those companies' perspective, they might spend less money in advertising, but meta might make more because they're showing more ads that convert more clicks and purchases. You can just see how you get to this is all good. And then you can also see how the user experience of the platforms will go to hell because they're going to be just choke to death with AI slop and extremely weird AI ads. And it's already happening.

Like I don't want to pretend. I don't want to be like this is some like vaporware idea. Like TikTok has some of this capability in their ad sac today. Meta and Google are already saying to creators, let people chat with AI avatars of yourself. They are training the user base to see AI content and think of it as the regular part of the user experience. And then it's just one turn to, hey, Ford,

give us a bunch of money, and we'll have infinite ad for explorers that we know everything about this person, and we'll be like, hey, we saw that you bought a lot of stuff at the store today. that would fit better in a Ford Explorer. Like, down to that level of targeting. Or it'll show me the color of Ford Explorer that I'm most likely to like

in the setting that is most likely to make me want to buy it because that is data that these companies have. Like, it's going to get really weird. And I think the, like, uncanny value of targeted advertising is about to be explored in a really fascinating way. I will say,

one of the funniest things about this is that it is going to make Amazon look even stupider because Amazon still does the ad where you're like, I buy a hat on Amazon. And then for six months, Amazon's ads, are just like, do you want to buy that exact hat that you already bought? Yeah. How about a hat? Do you like hats? Here's a hat. And I'm like, and it's just like Amazon, if AI can fix that, I'm in. Give me AI. Just boil the oceans. Take this hat away from me. That's weird. Look, okay, we've done enough negativity around AI. I do think that this thing is a bubble and it has

pop to some extent and it might bubble again. But there's just, I don't, it's not just deep seek. It's the products. It's, I'm going to compare it to, uh, to sell networks again because it, this is just, this is just the world I was raised in. The money we spent on LTE was good. Like if you, if you, we have a lot of newer listeners, but if you are with, here with us from the beginning, you will remember that the iPhone getting LTE was a big deal.

Yeah. Right. And the iPhone launching without 3G was like a legitimate criticism of the product at that time. It had edge. It had 2G networking. It was too slow. And putting Wi-Fi in it was the big, like the big innovation. This is all impossible to think about it right now. But when the iPhone got 3G, that was a big, big deal because it meant that it could access the internet faster and was usable on the go as an internet device. When it got LTE, that meant it could do video. And that's when a bunch of mobile video like took off and like became a thing. When I got 5G, everyone was like, that's going to happen again. Right. And it just didn't. It didn't. It's super didn't.

What is this for as a question people should ask much more often about technology? I will say, I just want to say one nice thing about deep seek and then we should take a break and talk about cool things people do with AI. Learning about deep seek sent me down a deep rabbit hole on chain of thought. reasoning. Yeah. Which is one of the technologies that Deepseek did a really good job with, basically like these reasoning models, all these companies are coming out with, uh,

that kind of show you the way that they think. Truthfully, I thought of that is kind of a gimmick that was like less doing something different and more just like, pretending to be human so that you trust it more. Uh, but what it actually is is deep seek and other companies have found ways to basically, instead of generating one token that is like the answer to your question,

It's able to generate a bunch at a time and actually go step by step and try to think through a problem. And what DeepSeek figured out how to do is how to use a technique called reinforcement learning, which basically rewards and punishes a model when it does things right and wrong, and teach it how to do that kind of step-by-step reasoning inside of the model. And so there's a great moment in one of the technical papers where the model starts to answer a question, goes step by step, gets it wrong, realizes that it is going down. the wrong road and is going to make a mistake, backtracks and starts over. And that is like a powerful moment in the history of AI. Because all this stuff is like, if you just think about it as

like it's generating the next word every time. It doesn't actually have the capability to go back and be like, wait, where did I go off track? It just keeps going off track. And with something like this, like this is the kind of thing you will start to see in every other model very quickly, I think, because the reasoning model is, A, they have, there are pretty strong studies to suggest that, It's vastly more accurate when you ask these things to go step by step, because instead of trying to answer the question all at once, it answers it step by step, which is like the correct way to do things. But that also, the more you let the model do that, like the longer you let it take and the more steps you let it take and the more you let it think and try and poke around and do stuff, the more accurate it actually is. So there is a path toward we can do this better. And it looks like these reasoning models.

and there are huge compute challenges there, there are huge latency problems there. That stuff is going to be hard to do in a way that is good user experience. But the idea of you can teach a model to be wrong less seems to actually be plausible. And I think that's very cool.

And DeepSeek is way out in front of that idea. Fine, you've convinced me. Let's give Sam $500 billion. Fine. I believe you. Did you download the Deep Seek app? No. I didn't either.

I got so freaked out by like just their terms of service are just like, we'll give the data to anyone in the government who asks. And they've already leaked some. Yeah. Yeah, there was a whole database that got leaked. Like, don't download the app people. And definitely don't log in with like your Google account and your email. Like, be, be careful. Please.

Please. Also, when I try to download it because none of that was my actual concern, they had shut down registrations because they were overloaded. I see. So you tried and were foiled. Yeah, I tried to get in that club. You said, listen, China and said, no, no. I should have that club drunk. and they would not let me in.

They were like, no shoes, no shirt, no service, sir. All right, let's take a break. That's been enough. We've been enough negative, Nancy. Let's see what the people are doing and whether that justifies $500 billion. Sam Altman. We'll be right back.

Support for this show comes from Shopify. Every thriving, successful business has to start somewhere. A good place to start is a relatively simple question. What if, given the right tools, I really put my all into this. One tool that can help grow your sprouting business to new heights is Shopify. Millions of businesses around the world rely on Shopify for e-commerce. They offer a host of helpful tools you can take advantage of,

from payment processing to analytics to website design. Their design studio includes hundreds of templates to help you create the exact website you've been envisioning for your business. If you're wondering, what if I need help? Then no worries, because you're never left to fend for yourself. Shopify's award-winning customer support is available 24-7. It's time to turn those what-ifs into a thriving business with Shopify today. Sign up for your $1 per month trial today at Shopify.com slash vergecast.

Go to Shopify.com slash vergecast. That's Shopify.com slash vergecast. Support for the show comes from Upwork. The days of doing it all, all by yourself, are over. There's no romance in burning out while you're trying to scale. Instead, you can check out Upwork. Upwork helps grow your business by giving you fast access to specialize talent across more than 125 categories so you can fill skill gaps, launch projects faster, and scale without committing to full-time headcount. And finding the right talent is easy. You can browse profiles,

review pass work, and get help scoping the role so you can get started quickly. Seriously, you could connect with the right freelancer in just a few hours, especially when you sign up with Business Plus. Their AI-powered shortlisting pairs you with the top 1% of talent in under six hours. No endless searcher required. You can visit upwork.com right now to post your job for free. That's upwork.com to connect with top talent ready to help your business grow. That's UP-W-R-K.com. Upwork.com.

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 slash track. Terms and conditions apply.

So last week, we pointed out a gap that I think we see in all of our AI coverage. So you might have even felt in the last segment that there's how people feel about AI, which in our comments is bad. straightforwardly. There is a real like principled stance against it as a whole that a lot of people are on the chain. And maybe it is. Like I said, the other,

the other leg of the stool that he kicked out is the New York Times winning its copyright case against, uh, open act or Sarah Silverman or any of these. Wasn't it Mark Andreessen who was like, if you don't let us steal from the whole internet, this all falls apart? Yeah, he wrote it in a letter to the copyright office. Perfect. Where he was like, I need you to, I need you to do me a favor.

Then he bought the government. So we'll see what happens. There's that. By the Copyright Office this week announced a clarification in its rules, saying if you make art with AI, that is copyrightable, but art made entirely by AI is not. No gray area there. No, that one would be confusing to anyone. Super straightforward. Anyhow, there's a gap.

The gap that I'm saying is people are mad about it, especially in our comments, across the Internet and various social media platforms. It's movie studio puts up a poster with AI. They get bullied right off the Internet. Then there's what the companies tell us, which is people are using this stuff like crazy. Deep Seek is at the top of the app store. chat GPT, most popular app by growth in the history of all applications. I think it's a little bit less than that, but you get what I'm saying. And then there's just the stats that we hear from the companies themselves that are sort of like less invested in the hype.

Like the one I always call out, I called out last week, Adobe CEO, Shantanyi Ryan told me on decoder that people use gender to fill at the same rate as layers in Photoshop, which means they just have opened Photoshop. Right. Like, that's just an identity. Yeah. There's this big gap in how people are saying they feel about AI. The hype we're going to make digital God. And then the sort of like mundane reality of a lot of people are just using this stuff. And a lot of people like it, but they don't perceive themselves as like using AI.

So we asked, listeners, send us what you're doing. And David, it sounds like everybody responded. Yeah, we got a lot. We heard from a bunch of folks on social. We got like dozens of calls to the Vergecast hotline. We got a whole bunch of emails. Everybody, it turns out, uses AI. I. And there are at least

a lot of people out there who had a use case that they wanted to share. So I figured we should just go through a bunch of them. And I think I tried to pull ones that are reflective of trends that we saw in their responses. So if you're

down, I'm going to read you some emails and I'm going to play you some hotlines and we're just going to talk about it because they all made me feel a lot of feelings and we need to sort through them together. What is the show but sorting through feelings? I'm here for it. So I'm going to not say the emailers names because some of this stuff is very personal. And also most people in their emails don't explicitly give us instructions whether to say their names.

I think a few of the folks on the hotline say their names. But I feel like if you say it into the voicemail, knowing we're going to play it on the show, that's fair game. So I'm so sorry if I'm outing you here on the first cast. This first email is from Sam A. All right. Let's start with. Ilya S has a note for us about how good AI is. So here's one email we got.

It says, I was recently diagnosed with a stomach issue, so it's been difficult for me to find and know things that trigger and don't trigger a flare-up. So when it's really acting up, it's super easy for me to just take a picture of a menu at a restaurant or bar and just ask, what can I have? And from what I've told ChatGBT from the doctors, it can just tell me what's safest to eat and drink, even telling me the exact ingredients that are safe in those dishes. That's awesome. like unambiguously cool use case in my that's a great use case it's just like I again I mean just being

negative I don't know how the restaurant actually made the food I don't know if things hallucinate like that it's just this is what I mean like I'm just prone to see the risk and I think that's an email where people see the benefit so this is you just brought up the exact thing that is going to come up an awful lot in these which is uh you are going to see a trend in which there is a question of how much should I trust this? And one of the things that consistently surprised me is a lot of people saying it doesn't really matter. That like I understand it has flaws. It might not be perfect. It's something. Like here, let me, let me read you one. I've often felt intimidated by the professional and technical demands of public service.

However, chat GPT has given me the ability to communicate more effectively, present my ideas professionally, and quickly understand complex topics. From rebuilding water treatment facilities and developing billing systems to ensuring compliance with the Environmental Protection Agency regulations. I've been able to handle responsibilities I never imagined I could take on. Again, very cool. To me, I immediately read that and go, how many of the EPA regulations do you think chat GPT just flatly gets wrong? Like, just fully hallucinates some made-up thing.

And I feel with a lot of these things, I keep coming back to like, it's going to work for you until it blows up spectacularly in your face, right? And it's like it's the same with the eating in the restaurant. Like as long as it's right. And let's say it's right 80% of the time, which is probably kind to most of these systems. But let's say it is. You take a picture of your food or you upload a PDF of a regulatory document. Ask it questions.

It's often going to be right. But when it's wrong, if you trust it, really bad things can happen. And so I feel like where I fall on the like, but what if it doesn't? work scale might just be different than some other people. But that just is the thing I keep thinking about with all of these. Yeah. I don't trust computers. I have a weird job in that regard. But like as David knows, when I record this podcast at home, I insist on recording locally into another recorder and not letting my Mac do the recording because I just don't trust the Mac, which is weird because I should. But I will say on the side. But I just like,

that's the frame that I'm in. Right. Whereas it's like what I hear that email is not about trust. It's about I worked out this idea so that I would have the confidence to go have, like, have the idea. And that's really different and interesting. How so? Say more about that.

I mean, that person sounds like they have a very complicated job. They have to know how to do everything. And they're like designing a water plant, right? Yeah. I hear that and I think, well, you already, you already did know how, right? Like, you had the capability of doing it. And what you needed was a conversation or more easily to adjusted reference material or whatever to get you over the hump. Like, you just solved your imposter syndrome.

The AI didn't do it for you, right? It just helped you get to where you were inevitably going to be. I think there's something to that And I think there's also A recurring theme in a lot of these That is basically like I just needed Something I didn't have a tool to do this thing

And now I have a tool to do this thing And even if it doesn't necessarily do it well The fact that it does it at all Is something right And it's like I think that I started with the food one Because that is one that there isn't A clean other answer to Right like I see on Reddit all the time

the people who go to a restaurant with like those huge placards of like, here's all the food I can't eat and they hand them to the chef. And then the chef posted it on Reddit being like, this asshole came to my restaurant. And to have something that is like, I can take a picture of this and it is going to give me even reasonably reliable information about what this is and whether I can safely eat it. Reasonably reliable and easy is probably a huge step up from the alternative in this case.

And I think that's something really meaningful. And so there is this sense that like something is better than nothing is so pervasive through a lot of this, and I think is really interesting in a way that I didn't give it credit for. All right, let me play you a couple of voicemails we got. Hi, this is Alex, and I wanted to call in to talk about

how I use LLMs in my daily life. I use chat GPT as an on-demand advisor of sorts. You know, how can I improve this recipe? How much cat litter should my cats be using per month? What are some good ways to save time at the gym if these are my priorities? A lot of these things are things I could use Google for, but it's just quicker to ask chat chabit.

Another thing I use it for is as a way to organize my thoughts. So if I'm ruminating about something, I'll just talk to chaty pt. It's no replacement for my therapist, but it helps reorient my thinking when I needed to. I know I could also get this for my friends, and sometimes I do. but chat TPP is nearly always available instantly. I find that very compelling, I have to say. Yeah. I mean, again, it's hard to tell anyone they're like doing it wrong.

Yeah. That's great. I'm like very happy that that's working in that way. The thing that grabs me out, that one in particular is it is the thing that Google is afraid of, right? But it's also the thing that these products have shown themselves to be the worst at. Like how much cat litter should my cat use every month is well first of all it depends on the cat But second of all it's like that's the one where it's like I you should you should feed your cat rocks like it's like that's what will happen

Put glue in the cat litter is like it's just sitting there. Yeah, and I think this is just really eye-opening for me in that Yeah, you're to your point. It's just good enough. Yeah, it's just something. Yeah, and right in and as we have covered exhaustively typing that sort of thing into Google right now is like getting mugged and now Seriously. Yeah, no, there is a real thing that I keep hearing on a bunch of these that is just how aggressively Google fumbled the bag of search. That like, what people want is better Google, not chat GPT. And Google just threw all of that away. But I think one of the things that you say all the time is that the internet is about to be flooded with C plus stuff. And I think what we might be underrating is how valuable a C plus can be sometimes. Like maybe a C plus, delivered to me easily and quickly at all times is actually something. Is that bleak as hell? Probably. But it's something.

All right, let me keep going. That's like the, that's like how they, why they invented legacy admissions at Harvard. Like that's what you're describing. Like, what are we going to just accept a flood of C plus applicants? Yeah. And that goes fine. What are you saying about? All right.

Here's the next one. Hey, this is Jake. All right. I'm listening to your recent podcast where you've mentioned, and you want people to follow in and say whether you use Chad, GPD, for in the rule. I use it weekly to make my shopping list.

I ask you to make the recipes based off of what I have in my house and what I'm going to buy. Try to make it as macro-friendly as possible. One of the features that I've started using more recently is whenever I make a shopping list, I ask it to make an optimized shopping pattern for a standard location of where I'm going. So if you're going to Costco or Aldi, I can make sure that I'm shopping and no suspicion. and then I ask it that find some substitutes for recipes.

And that's not I use it for pretty much anything that it comes to for like vacation, trips, they either of a planning. That's some of the main use cases that I use it for

outside of my work. All right. So first of all, I just, I, that's all great. Man, Google blew it.

You just described everything Google has ever demoed at Google I. I know. But you're doing it in chat cheap. I know. Yeah. Yeah, like there are a hundred Google Keep features that you just described that no one uses. Yeah, no, this to me was the one that I was like the most sort of down the middle simple use case is like, I'm going to write a list of stuff that I have in my fridge.

You're going to spit back some recipes and then I'm going to say, can you optimize this list to go to Costco? Like chat GPT is great at that. Do it. Love it. Yeah. And then the next turn is go on Instacart with operator. order it for me. Right. Like, you can see how you connect all those thoughts. Next time I go to Costco, I'm going to tell Chad GPT to give me directions around Costco, and I'm only going to follow ChatGBT.

We're going to see how it goes. This is actually the part where he said that, I'm dying to know more here. It really implies that ChatGPT knows how Costco's organized, and just based on my Costco in, it's a war zone. I'm not sure Costco knows how Costco is organized. That's what I mean. It's like, I've asked Costco employees where things are, and they're like, No, no, no, no, no. Good luck to you, sir.

Yeah, it's a secret. You're not allowed to know. But yeah, that kind of thing, like, I am always talking to people about, like, what I, what do I think are good use cases for chat GPT? And the one I always come up with is movie recommendations because it's like super low stakes, but actually a thing AI is unusually able to be good at where I'm like, I like these three movies. What else should I watch? And it's like, that's a thing LLMs do very well. And this kind of like recipe remixing is another a good example of that. So I just thought that one was great. All right, let me read you another email. Real tone shift here. You ready for this? Yeah. We got an email from somebody who's going through

a divorce and said, Chad GBT has been a surprising source of support. Let me be clear. I don't use it as a talk therapist, parentheses, gross, but rather as a tool to help me communicate effectively and with kindness. My wife is very sensitive to perceive criticism, so I draft emails to her and then use chat GPT to help me soften the tone, make them warmer, and remove anything that could be interpreted as critical while still ensuring my requests and boundaries are not diluted. Filtering emails this way has made a difficult process a little less hard. It's also helped me curb any instincts to be petty, sarcastic, or passive-aggressive. It helps me keep my side of the street clean.

Wow. Right? So I'm married to a divorce lawyer. I have heard about an infinite variety of passive-aggressive emails between partners for the past several years. I feel like I should just go tell her about this. It's really interesting, right? Yeah. This one, I was like, I don't know how to feel.

On the one hand, there's like a piece of this that kind of makes me sad that it's like, it's a bummer that there is a web app that has to be your tool for this. But on the other hand, like, again, what's your alternative? Where else would someone go to do something like this? If you took this out of the bigger context of AI and all of the noise and all of the hype and you were like, there's a new feature in Grammarly, right? Five years ago. Yeah.

Or you're like, Google Docs has a new feature that just, helps you, like, de-passive aggressive in email. I don't know that anyone would have any reaction to this other than that's actually pretty good. Yeah. Yeah, I think that's right. And it's only in the context of all the AI hype and, you know, that horrible Apple commercial where it's just like, bad employee sends professional email and like, this is

what Apple intelligence is for. This is maybe the most human of the notes that we've gotten. Yeah. Right? And that, I mean, not to say everyone else isn't human, but like, this is a very emotional task that is being performed. Totally. And I really do think if you took this one out of the AI bubble and all of baggage that

comes with it, you put this in any other product five years ago and everyone's like, oh, this rules. Yeah. Yeah. I mean, and I think it goes to show that A, our vocabulary about AI is still so primitive. Like, we call everything AI. And actually, AI is lots of wildly different things. And we need to get to the point where we're better at talking about them as different

things. but it also to some extent goes to show how versatile something like chat gpt is like all this stuff is happening in the same text box right like it's a pretty remarkable thing also very notable everyone has said chat chbtee every single person you haven't gotten one dolly no no apple indolid little bit of clod love in a few of these coming okay a little bit of claude love um i don't think i pulled any of the ones that people use claude but um sorry claude sorry claude you're just calling it sorry claude you're sorry claude you're users. Claude was mentioned. I don't think Jim and I came up one time in the emails that we got. Brutal. But Claude, people were like, oh, I mostly use chat GPT and like some Claude. There was a lot of some Claude.

All right. Let me play you another voicemail. Here's a, again, we're just going to tone shift all over the place here because that's what AI does. Here we go. Hey, this is Leo from Atlanta, Georgia. Neil, I asked what everyone's using their AI assistance for. And I use mine almost exclusively to solve.

hard questions in the word puzzle game word salad. Also to note, I recently discovered that something about David's voice makes me super tired and helps lull me to sleep. So I want to give a quick shout out and thank you to Mr. David Pierce for improving my sleep quality and helping fight my insomnia. Thanks, guys. Love the show. So I mean, David's voice makes me super tired.

First question, was that a compliment or an insult? Yeah, that's really, that's a tough one. You make me sleepy. It's like a really. Yeah, he said thanks. I don't think you meant it. But that one, like another in the line of like movie recommendations, right? Like pretty low stakes, pretty straightforward.

This thing has a lot of information and it can spit some information back at me. Yeah. It works. You're cheating, but it works. You are cheating. To be so clear, you are cheating. It's okay. I've Googled what's the word will answer before.

It's fine. It's bedtime for you, sir. We all have bad days. Yeah. Okay, let me find the next one. I've been fascinated by my mom's usage of AI. She's 49, and while she uses it a lot for her small business, writing, posts, etc., I've noticed that over the past year she's also started to use it as a general Google replacement in her personal life as well.

She wants to know a fact, she asks chat GPT. She has to write a WhatsApp message that requires minimum thinking. She asks chat GPT. Last week, she found out one of her close friends is struggling with her mental health and having suicidal ideations. And I shit you not. She asked chat GPT, what are the best ways to support someone who's depressed. Again, just Google. fumbled the bag. Like, people want

better Google. Yeah, but this is the one, again, these are real people, everyone use a computer however you want. I'm happy it's working. But this is the one that's the blinking red danger sign. Right? It's like, oh, everyone

just needs a friend. Sure. Right. There's just an element here where this is, to me, like, this is where the robots hallucinate all day long. And what you're getting is better Google with worse results, presumably, but in a better interface or more reactive

or just someone to talk to you in a way that's natural. And it's like, this is the one where it's like, if you tell me this is the use case, I will confidently tell you this is not support $500 billion data center built out. Oh, 100%. Because it doesn't scale. Like the thing will happen that happened to Alexa in Siri,

which is you'll do this until it fails you, and then you will stop thinking you can do anything. more than this right like you're like Siri tell me a joke and it's like I don't I give up right like fine timers or music it is and I'm not saying the ceiling is the ceiling for a chat should be so much higher obviously but there's still a ceiling and the ceiling is when it lies to you right when it makes something up when you ask it for movies and it gives you three real movies and one fake one and then people are literally showing up at public libraries asking for books that don't exist

and that that's what I mean it's like when lawyers show up in court citing cases that aren't real and they get disciplined by the bar association. That's the problem. They can't overcome. They don't know how to overcome that. And so like this one to me is like, yeah, every experience is valid. Like great. And it's really interesting to see how different people are approaching it.

But this is the one that hits the wall the fastest. Interesting. Yeah. I mean, I think I agree. But I also think there is a more charitable way to look at that that says the human is still kind of in charge. right? Like, I think it's, it's one thing to have these things do things on your behalf, right? And this is where I think things like operator get really messy. Like, would you let ChatGPT text your wife for you? Like, even something relatively innocuous. Again, David, I don't, I don't think I've made this clear enough. I'm married to a divorce lawyer. No, but even something very simple where you're like, that woman can have ChatGPT divorce me. It's just a macro on our computer. It's a work issued windowed laptop with a button. that says D and then my life is over. Fair. Yeah. But like,

okay, fine. Helen Havelack, our boss. Would you, would you like voice mode chat GPT and be like, hey, will you text Helen and tell her I'm running late, but I'll be like 10 minutes late to the meeting? Why wouldn't I just text for that? No, no, but just go with thought experiment with me here for a second. Okay. Would you let it take that act, like do that on your behalf?

Not say, here's a message you might send, but be like, send this message for me, chat, GPT. Would you have it do that? Yeah, abstractly, yes. Really? See, I wouldn't. Like, just the, there is a certain thing where I'm like, okay, if you want to write the text and then I'll look at it and send it, that's something. Yeah, I guess that's what I'm imagining because that's how I use Siri in the car. No, but I even get freaked out in the car because you say it and it reads it back to you. And I'm like, what if it, like, put weird punctuation and a period at the end and my wife is going to think I'm mad.

There's just middle finger emojis between every word and you just don't know. Yeah, okay. And I still like, I truly get nervous about doing a text that I, I can't look at the words before I send them. And I think that's where all of this agenic stuff gets really messy for me. Because I'm expecting you to do things. But again, just to finish this out, in this case, the idea that, like, I'm asking ChatGBT to give me ideas about how to support a friend, I think is less terrifying than the idea of, like,

send my friend a supportive text message, which would be horrifying. I scrolled down and I'm reading the rest of the same email, about the mom. And the second part of the email is the, the most important part of the email. Okay. Here's a quote. Most people I'm friends with or that listen to the show are probably concerned with accuracy, environmental impact, development of critical thinking, et cetera.

She, the mom, could frankly not give less of a shit. And I suspect people like her are likely to be the majority, not because she's dumb. She's brilliant other ways because she's just not wired that way. And this is your point about good enough. Yeah. Right? It's just the computer's talking to you and it sounds fine. And that's good enough.

And I, I'm just very convinced that we live in an age where just being confident is like, all you need. Like just blustering through it is very effective in a variety of domains. And maybe that collapses on itself. We'll find out. But that's like the whole population is primed to just be like, confident talking.

That must be right. Totally. Wired ran a headline the other day. Something to the effect of the less people. know about AI the more they like it. And based on a study that I think it was like a marketing firm did basically to find that if you don't understand what's happening,

AI seems great. And that makes perfect sense to me. That feels exactly right. And I think most people don't understand what's happening. And maybe that's terrifying. But I think that's what's coming. Based on this and that second piece, which is super interesting to me, I understand why a bunch of companies are so excited about this. because it's the first time using a computer has felt different for people. Right?

The jump from the command line to the graphical user interface felt different, and you built an entire paradigm of computing around it. And then the jump from desktop computers to touchscreen mobile phones was an entire paradigm shift, and obviously phones could go with you and all the other stuff. But that created an entire economy. And they've been searching for something else that makes computers feel different to people.

Are they going to fold? You know, like, what's something? It's got to be something. And this feels different to people because they're just talking to them. But there's not an application layer behind that that builds the whole economy or supports all this. Like, and also I just come back to the serious example. Like eventually mom's going to hit a wall, right? It's going to lie to her.

She's going to make a mistake. It will embarrass her in some way. And then we'll stop because she can't trust it. And that trust once it's gone is gone. Yeah, I definitely agree with that. And I think the question of, A, when that happens and B, what the stakes are, is the stuff that makes me really nervous about all of this. Yeah.

And either I'm just a chicken or everyone else should be more afraid or both, I don't know. All right, let's do one more. And then we'll get out of here. Yeah. Here's one more. Hey, The Verge and or the Slack channel. My name's Florian. I'm calling from New Jersey taking up your request on what I use chat GPT for.

I would say primarily I use it as a more reliable resource than Google searches for informed decisions. So whether that's in-depth settings menus, I can find Bluetooth. But, you know, the other ones, I'm a recent college grad, so I've used a lot to help friends with their career growth or updating resumes, kind of like the bad code, good code arguments. I come back for practical advice, namely curious topics or other things like that to gain some information on very quickly, but I would say primarily, it's for things like when I can't find the answer on AV forums for how to set up the Blu-ray player through ARC with the soundbar and making sure that Dolby True HD plays from my disc.

Neil, I don't know for you, but thank you for the great pod and hopefully someone in the Slack sees this. Have a great day. I love you. I love this man. First of all, just call me. You're going to talk to the robot? I'm around.

Yeah. This is why the Virdcast. We will solve these problems for you. That's great. The in-depth technical problem, the issue that I've always run into with that is the answer is buried in some forum posts from six years ago, and it's just been muddled. And I actually want to find the old forum post where the person actually has the answer, not the, do you know what I mean? Yeah. Like, not the synthesis over time.

This is where I think like Reddit answers, the Reddit AI search is actually super exciting because it is geared toward that exact thing. Yeah. That is like, what if we just did a smarter job of searching this stuff, but we're still interested in finding actual people's actual answers? It's like, that's a cool mix if we can figure out how to do it right. I lied. I have one more to play for you and then we should go to break. I just want to say to my guy, you're my best friend. We don't know each other, but we're in it now. You are one of us.

We're family. Welcome to the verge. All right. I got one more. True HD for life. This is Matt from Springfield, Missouri. I was just answering your question about using AI daily. All these tools came out when I started dating and I used it to help write my profile and also compose my messages because I'm a machinist.

I have good ideas, but I'm not good at words. And AI helped me fill that gap. It must have worked because now I'm in a relationship. I'm going to have a son this year and I still use AI today to compose thoughtful messages to send her throughout the week. Oh, this is great. AI is good, Nilai. I got to say, I mean, that's, we just got it there. We should shut down the verge.

Negative Nancy is good out of here. That's fast. That one made me really happy. Chatsby only hit two years ago. We went from zero to sun. Killing it. Pretty good. Yeah.

Pretty good. Pretty good. You know, most people move in together for, I'm just saying, anyway. Not anymore. This is an AI world we live in. No time for that. Okay, I'll give you, I'll give mine. And it's very different than that, but it's in the same zone.

I recently sold something on eBay, and I let eBay's AI just write the description for me. Really? Just gave it. It was so dumb. It was like we bought a new stove and we bought different color handles for the stove. So I sold the old ones. And I just took a picture and I figured out what it was. And I was like, do you want to write a description? And it's like, this is the lowest stakes shit in the world. Like, anybody who's searching for a placement stove handles

and sees a picture of this, like, they know. Like, you don't have to describe them very much. And I was like, fine, here's some copy. And it did it. Poshmark today, actually, announced that they're letting people do. descriptions. I think there's this world in which like weird low stakes fields get filled in. Obviously dating is a higher stakes field, but I can see why it's very tempting to people. Yeah. By the way, that's a great story. Yeah, Matt, congrats. We're very happy. I hope you name your son Al. Think about it.

All right, we should take a break and we got a lightning round. That was great. Send us more of those. Yeah, thank you to everybody who wrote in. It was super fun. I learned a ton. This has been really enlightening and interesting. And I do want to make sure our coverages round it out. Yeah. All right. We're going to take a break. We'll be right back.

Support for the show comes from Anthropic. Not every question has an easy answer. And the ones that are really worth asking usually come with a healthy mix of inspiration and backpedaling. Aha moments and quiet meditation. When you're working through one of those problems, you want a partner to bounce ideas off of and figure out where the deeper issue lies. That's where Claude can have.

help. Claude is the AI for minds that don't stop at good enough. It's the collaborator that actually understands your entire workflow and thinks with you, whether you're debugging code at midnight or strategizing your next business move. Cloud extends your thinking to tackle the problems that matter. Plus, Claude's research capabilities go deeper than basic search. It can have comprehensive, reliable analysis with proper citations, turning hours of research into minutes. Ready to tackle bigger problems? Get started with Claude today at cloud.aI slash Vergecast. That's Claude.aI slash Vergecast and check out Claude Pro, which includes access to all the features mentioned in today's episode. Claude.aI. slash Vergecast. 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.

It's 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 focus 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. dot com slash track. Terms and conditions apply.

Buzzwords like progressive and affordability are thrown around all the time in politics. But what do they actually mean? For me, being a progressive means at least two things. One, being willing to unite lots and lots of people, all of the folks that are getting screwed over against the powers that be that are making your life worse. And then second, being progressive. is essentially a hopeful enterprise, that you think, I think that the world can be much better,

that we don't have to settle for crumbs or settle for the status quo. And is there a difference between what it means to the elected officials and what it means to the people? So money is essentially the root of everything. I don't care if you're gay. I don't care if you have all that. That's like secondary. Third, like that doesn't, that's not a priority. That's this week on America Actually.

Let's begin. All right, we're back. unsponsored lightning round. That's how you know it's wild. It's crazy times out there. Unfiltered. Uncensored. The First Amendment's back, people.

And by back, we mean Trump will arrest you if you criticize. All right, there's actually a lot in this lightning round. Let's start with some very exciting, very niche Vergecast news. Google is open sourcing the Pebble source code, this

original smartwatch. And Eric Mijikovsky, the founder of Pebble is going to make another watch. David, you like talk to him. What's going on here? Yeah. So I saw Eric at CES. We had lunch and he had a pebble on his wrist. And I said, uh, we were just talking about, you know, life in the future and gadgets. And I was like, I was like, so you, you've left automatic. He sold his company Beeper, the messaging app that we talked about a bunch to automatic last year. He left the company looking for something to do. I was like, so are you going to, are you going to just like do pebble again? Are you going to re pebble? Uh, and little did I know. He asked me again.

and we talked this past week, and he was like, did you know? And I was like, what are you talking about? And he was like, we're calling it re-pebble. How, like, why did you say it? So I, I'm feeling very proud of myself, except that I completely let him off the hook and just moved on to something else.

But anyway, so Eric, I guess over the last year, has been working with Google to open source all of Pebble OS. So Pebble sold to Fitbit, which sold to Google. And a lot of the, like, pebble ideas sort of percolated around those two companies. And you can see some DNA in like the Pixel Watch.

But the Pebble software itself was just not being used. And so Google, to its credit, I cannot report that Deeter Bone was involved, but I also believe in my heart that Deeter Bone was involved. Dieter, famous Pebble fan was, just works for Google now. You know, there's just sometimes there are pieces there. But anyway. Look, we don't do conspiracy theories here except for that one. Except for that one. And Google agreed to do it.

So Google open-sourced all of Pebble OS, which is not everything. There are like third-party drivers and stuff that are occasionally complicated to open-source. The name Pebble still belongs to Google. But functionally, all of the software is now available. And so Eric is starting a company that will not be called Pebble, but the goal is just to make Pebbles again. He's like, we got it right. His pebbles now still work. He said he had to replace the battery on all of them, but they still work and they still do the things he wanted.

And so he's like, I'm going to keep making these like long-lasting, e-ink, gadget-y gadgets to put on people's wrists. And the plan is hopefully to start making stuff quickly, like even this year. And there are a lot of pebble nerds out there who are very excited about this. So the pebble story, as I recall it, is that they just ran into API problems with the iPhone. Because the original pebble could get notifications. Right. Like Apple supported sending notifications out that way.

But you couldn't reply to a message. they were just limited in very real ways. You could just like look at a text message but you're going to get your phone. Apple's never going to open that up. No. So probably Android phones,

the pebble app on the Android phones they could do some of that stuff. It was always much better, yeah. There's more to do on Android. But also I think for Eric in particular, I think the idea of it being like a messaging machine seems to be less compelling. He's like, he's very into it as like,

it's an easy way to get kind of glanceable information, but it's also like, simple and the battery lasts a long time and it has buttons for music controls. Like, buttons for music controls is one of the specific things he named as like a core competency of pebble. And it's like, it does not have to be a messaging system, which I actually kind of agree with. Like, I would like lots of features to exist, but I find that I only use a very small number of them on my wrist. And I would trade a lot of them for like five day battery life. And I think it's going to be interesting to see what they do this time. I'm excited. I love a good open sort.

hardware hacky project. It's going to be cool because there's like there's been a pretty big community of people keeping this stuff alive for a long time. And the fact that now they have honest to God software to play with instead of like hacking around the edges of existing Pebbles, I think we're going to see some pretty interesting stuff here. Like that that community is really engaged and I think is going to get up to some stuff. Okay. So that's Pebble. Next one. Oracle and Microsoft reportedly in talks to take over TikTok. This story makes me so tired. Although Trump denied Oracle, right, in some press conference, but he doesn't know. I mean, like, when I say he doesn't know, I don't just mean be dismissive.

I mean, literally, he gave them a 75-day illegal extension, which Apple and Google do not trust enough to put TikTok back in the store. And then at a press conference, he said Congress has given me 90 days, which is 15 more days than he gave himself. Yeah. Congress has not given him that. Congress didn't give anybody anything. The time's up. Like what he gave TikTok was a 75 day pause on enforcement of breaking the law, which again, Apple and Google did not trust because that's billions of dollars. Anyway, so I'm just saying like the facts are confusing on the ground.

Like, well, does Trump even believe is going on here? Because he's just saying different numbers. And then, but he did say, I don't think of Oracle is involved, which is weird. But Microsoft was a bidder the first time. Nadella famously said it was like the weirdest deal he's ever been a part of. Yeah, a bidder might give. Microsoft too much agency in that. It seems a lot like Microsoft was like dragged into a room and yelled at about buying TikTok for a while.

No one knows what's going to happen. Like truly no one knows what's going to happen. And then this other plan that Trump keeps floating where someone does a joint venture and they get 50% of it to the United States, our own big tech companies are extremely nervous about that plan. Oh, interesting. Because if you just think about it, that means India could shut down YouTube and demand it gets sold and 50% of YouTube be given to India. Like, that's a precedent you do not want to play with. Interesting. Yeah.

Because other governments are way less skittish about nationalizing companies or doing speech regulations. Everyone's like, don't play this game. Right. We'll see if that. India is actually a particularly interesting one in that it is a government that I think would be. excited to do that kind of thing and is such a crucial market as a growth engine for so many of these companies. Oh yeah, I picked it. India has banned TikTok. Straight up and it is very reliant on YouTube, but it has very draconian speech policies. The Modi government does not enjoy

dissent in very specific ways. You could see Brazil and other country like that Russia exists. There's a million countries that would play these games. And so like our own big tech companies He's like, do not, you do not want to open this door. Like, you're already in 50 different trade wars. Like, don't do this. Is it possible that the door is opened now that just someone has said it out loud? Like, do you wonder if there are a bunch of people sitting around in, like, government boardrooms being like, how did we never think of this? They do.

But, like, if you want to roll up on the United States that way, the United States would be like, no. Fair. Yeah. But then the United States starts doing it. And they're like, okay, tit for time. Like, here we go. Like, we're off to the races. Like, the norm has been shattered.

And, like, because we're doing trade wars, you can already see it. You can see people be like, we don't know what's going to happen, but don't do that one. We'll see if any of that goes through. We'll see if there's a buyer. We'll see what happens at 75 days. Like, 75 days is not a long time to close a deal of this magnitude. No. Even if you had agreed on it already, it would be a short amount of time.

Right. And so far as you know, there's no lead buyer. And then there's the big question of, are you selling TikTok U.S.? or are you selling bite dance? And there's a lot of chatter that the answer is they're going to sell bite dance. And so that's a much bigger deal. Right.

With a much different set of investors, one of whom is Jeff, yes, Trump donor, you know, like a lot of his money is set up in bite dance, not TikTok US. And then there's the question of like,

what does control mean? Because the law specifies it can't be under the control of a foreign adversary power. And then Trump is supposed to run, an interagency process to blah, blah, blah, blah. Is he going to do that? So there's just, I don't know, man, I think we're, I think we're going to run to the end of that 75 days, and then we're all just going to look at each other, and it's, who knows?

It really does seem like at the end of day 74, we're still going to look around and everybody's going to be like, what happens now? And no one is going to know. Like that, it feels like that's where we're headed. And that is going to, you know, not that Trump isn't running into the same. separation of powers problem every single day, but you get to day 74 and he's like, I'm extending it again, a bunch of Republican senators are going to say, no, no, no, no, no. Tom Cotton is going to be like, no, here's what we're going to do. We're going to shoot TikTok

in the street. That's where he's already at. Yeah. And so, yeah, we're just barreling towards a full sale of bite dance, which is nuts to contemplate with the permission of the Chinese government. Byte dance, probably either the first. second or third most valuable startup on Earth right now. Yeah. By April. And by the time you're listening to this,

it will be very close to February. Like, this is just not going to happen. Yeah. It's crazy to me. Yeah, I really, like, we were on the show for a couple of weeks trying to, like, handicap possibilities, and I have given up even trying to guess.

Like, I couldn't possibly begin to make an educated guess about what's going to happen here, which is so interesting and also so insane. I do think there's some reasonable likelihood that Microsoft actually goes through with this and tries to do it. They have the money to do it. They have the ability to run the infrastructure. They have Azure. They have a consumer market with Xbox users.

Like they have a thesis for why they would run it. They have an advertising business. It's fine. Not great. Like, sure. Meta can't do it for sure. Can't do it for all the reasons we can talk about. But, like, you know, the FCC is still pursuing a breakup.

Like, they're not going to do it. Would buying ByteDance be the biggest acquisition in history? Company is worth, like, upwards of $300 billion is the number that keeps being fluttered around. I don't know if the number is. But then I've heard numbers for TikTok that are higher than $300 billion. So, like, ByteDance, by the way, also kind of in the deep seek race of putting out much cheaper AI models that can be right. Like, there's a lot happening in that company. There's not happening in that company.

Also, just the chatter that I'm not. I've heard is the number changes based on who buys it. Right? So, like, if meta or Google buys it, for example, they have gigantic monetization engines that are really, really good. So they don't have to spend any money, like, getting their money back out. Sure. Right?

They just have to deliver ads inside of TikTok. And they already have the clients and they already have the tech and they already have the, yeah, like, whatever it is that you think they need. They already have it. It's just really expensive scale, essentially. Yeah, they just bought more scale. If you are Amazon even, you got to go build an entire monetization system. So you're going to pay a lower price because you've got to allocate some money to that stuff.

So how much you're willing to pay, how much is worth is all based on what you can do with it. And there's only so many companies. And the two companies that could do the most of the most are not going to do it because the government's trying to break up. Yeah. So literally they're trying to break up Google. Google has to sell Chrome but can have TikTok. It's like a very funny outcome. It's nuts.

All right. On the next one, this is a new segment that you're going to have to get used to in the Vertcast for next four years. I'm just calling it Brendan Carr is a dummy. It's a podcast within a podcast. He's a moron. He is a total political hack job with no conviction and no morals. That's my belief. Brendan, if you're listening, you can come on to Coder and you can try to prove me wrong, but good luck, buddy.

He doesn't even have a cool mug. Like, if you're going to be terrible at this job, at least like have a funny mug. You know, Ajit Pai, he was a guy. He was also a character in the first Trump administration. He had a governor. Like, he, there was a line of depravity that he would not cross. Fair. Like, you could be like, I disagree with this irritatingly Verizon captured conservative

Republican commissioner. Brendan is just a political animal. Yeah. Here, two examples this week, and they're just going to happen every week because the man loves attention. He is rolling back a rule that allowed people who live in apartment buildings to go get their own internet service.

So, like, a lot of apartment buildings, your landlord would be like, I'm just going to sign everybody up for whatever ISP, Verizon, AT&T, Comcast, whatever. And all of you have to use it, and you can't go get your own. And because I have collected us all up together, I'm going to get a deal.

Right? We have some scaling, gosh. And the ISPs, if you've experienced this, you're like, I don't, what? Like, it's just the same number as everything else, right? You're not getting a deal. You're just being lazy or you're getting a kickback from the ISP, and you're not passing any of that on, like, this was my experience in an apartment in Chicago, was my experience in an apartment in New York.

Same. Yeah, a lot of people have had this experience. Yep. Biden, FCC, passes a rule, says you cannot do bulk building. You have to increase competition and let people sign up for their own ISPs and apartments. Brendan Carr has revoked this rule and his argument is this will lower prices. Which just on its face doesn't make any sense.

It's one of those things that like I can't even make the bad faith argument that that makes sense. You know what I mean? Here's a quote. I have ended the FCC's consideration of a Bryden era proposal that could have increased costs by 50%

the price that some Americans living in apartments faith for internet service. This is regulatory overreach and it would have artificially raised the cost of internet service. There is not a world in which letting people choose their service provider raises their prices. There might be a world in which you live in an apartment building that's so big. Then you have a landlord that's so good that maybe they were negotiating a terrific rate.

I don't know anybody who lives in an apartment. No. There's one of those in America, and that's the one when Nagar is talking about. Yeah. I'm like, what is doing by a political animal. Like, he's straight up saying reducing your experience of. Broadband competition will lower prices. Yeah.

It's ridiculous. All right. What was the other one? You said there were two. Pure double speak. All right. Well, there's two, actually. There's two other ones.

The second one is he's reopened investigations into NBC, ABC, ABC, and CBS's broadcast licenses. It was very complicated because the stations have them, not the networks, but so it goes. And in particular, he opened up the NBC one because he says they violated the equal time rule by having Kamala Harris and Tim Kane on Saturday. live. Oh, right. Do you remember this? This was like last fall that he got all fussy about it. Yep. And so on her way at the door, Biden's FCC chair, Jessica Rosenworsel just like wipes out these investigations. She's like, these are stupid. I know they're politically motivated. We're not doing this. He's reopened them. Whatever. You can say whatever about two of them. They're dumb. But in particular, the NBC one, he knows that NBC complied with the rule because he has emailed people

and said that they have. This is such a brazen lie. Right? He's like, they had her on. They complied with the rule. Here's their notice. And then immediately had Trump on a NASCAR race the next day because they gave the proper notice

and the campaign took advantage of it. I know he knows it because I've seen the email. And he's reopened it to be like, I'm investigating. Like this is just total political theater for an audience of one. Yeah. And this is dangerous, first of me.

Amendment territory. I don't like this speech, so I will cost you money, is not what you want. Any government, it doesn't matter if you like Kamala Harris, you love Donald Trump, you hate one, it doesn't matter. Unelected bureaucrats saying, I will cost this money, I will make this painful because I don't like your political speech, is red flashing warning light of free speech problems. Yeah. And then on top of that, just as we sat down to record, Carr announced investigations into NPR and PBS. Okay. Because he says they are illegally advertising, and what he specifically means is that they do

fundraisers. And so his line is, taxpayers should not pay for commercial activity and then, you know, the usual garbage about woke Sesame Street or woke NPR. He's attacking NPR in particular because the C of NPR is Catherine Marr, who used to run Wikipedia and the conservatives hate Wikipedia right now. And so this is just Brendan Carr trying to chill speech, right? He's like, I will defund PBS and NPR because I've constructed a legal reality in which them doing fundraising drives or saying we're supported by viewers like you who call us to give us money is advertising for an agenda. Right. And all of that is transparently nonsense. But it maybe doesn't matter because he's in charge of the FBI.

It doesn't matter in like the alternate reality of X where this man is trying to build a profile. And I'm telling like if he, Brendan, I know you pay attention to us. You have for years. You want to come on Dakota and explain this like craven political bullshit when you are playing with the absolute fire of the First Amendment and claiming you'll do it.

Come on. I'll talk to you. But it's not going to be easy. Because I think this stuff is incredibly transparent. Yeah. I also, this is not important in the scheme of things. But like, if you find yourself against Sesame Street, like in any way,

business, politics, like, if you're canceling Sesame Street or you're mad at it, like, you've done something wrong. Like, just, it's not good. You shouldn't be against Sesame Street. That's it. That's all I have to say. That's it. Sesame Street is great. It's pretty straightforward. Leave Sesame Street alone. I will stop. It's lightning around. We've got to move on. Yeah. Oh, but I have another super fun, deeply craven political speech thing. Do you want to talk about it real fast? Love it. My favorite. Meta agreed to pay $25 million to settle the, the, the, suit over suspending his account after January 6th, which is just like in a sea of Mark Zuckerberg

nakedly sucking up to Trump for business reasons. This is the simplest one yet. Like, here's a bunch of money. Please be nice. And then there was some reporting that Trump was basically like, I think what he said is like, we have to settle this lawsuit before I can like bring you into the into the fold. I think he called it the nest, into the nest. And it's just like, this was the price of admission, right? Like, this is what happened at the inauguration. They all spent a bunch of money to hang out with him and go to parties and he would say nice things. And now META is just spending another $25 million to make Donald Trump like them because that is worth it.

Can I just say why this one is like particularly galling to me? Meta would have definitely won, right? The theory of this case is that meta, when it banned Donald Trump, was somehow the government and that violated Trump's First Amendment rights. Right. Which is nonsense. Which is nonsense for the very simple reason that the president at the time met up banned Trump from its platform was Trump.

Yeah. He was the government. It turns out. He was super the government. Like, you don't even have to get to the like pedantic forum stuff of like the First Amendment is about the government, not slashed up. Like, I can do that all day. I was raised in that fire.

I'm just saying this one is particularly stupid because you're like meta, acting on orders from the government who was me illegally banned. Yeah. It's so stupid. He's lost other versions of this case. But this is everyone is doing. There's other cases that he would have, ABC settled one, Disney ABC settled one. They're just trying to get out of the line of fire because he controls the regulatory agencies. All these companies have decided it's worth the money.

And the thing is, they're right, right? Like, I would say pretty transparently, you can make the case that this is just a good use of $25 million for META, which is... Yeah, this is a sneeze. Has a lot of complicated cases going on there. There are antitrust questions about META. Trump has threatened to throw Mark Zuckerberg in jail a bunch of times. Like, I can see why you write this check. It just sucks.

Like, it just sucks so much that this is the state of things right now. I mean, yeah, the FTC is going to try... They're in a trust case to break up meta is coming. The various teen health cases about Instagram causing mental health crises across the country. They're all coming up. You can, if you have an angry Trump who's like, break him up and you can get out of that for $25 million. Yes. There's a ruthless cynicism here.

Yeah. But man, we are, we are just playing with fire with the First Amendment right now. Like, left and right. We are doing weird speech regulations and claiming. to do them in the name of free speech. Weird. Super weird. Yeah.

All right, we got to end with a happy one. Blue Sky. Yeah. I'm happy Blue Sky's working. I have, I'm all in on Blue Sky. I was, wait, just the concept of Blue Sky is your last happy? Blue Sky now has 30 million users.

It's like there's news, right? Not just knowledge like I logged into Blue Sky. It's growing. Threads is much bigger. Threads continues to add a million people a day. It has 320 million is the number. Mark Zuckerberg. gave on the earnings call this week.

That's a gigantic number. But Blue Sky, Blue Sky is the fun one now. I think we can say pretty certainly that like Blue Sky has replaced Twitter as the place that is fun to hang out. And it is true, it's a bunch of like liberals being annoying to each other. Super annoying. But in addition to that. This is a group of people.

It's also a lot of. It does not love jokes. No. Yeah. At least that's my experience. Yeah. But like, A, I'm psyched that there are multiple versions of this thing that are succeeding. I think it would have been a bummer if we had just replaced Twitter with one thing.

The fact that there are a few different cuts at this that are going to work is really exciting. Mark Zuckerberg also made noise about wanting to make the OG Facebook experience come back, which that's not going to happen. Sorry, Mark, you ruined that 15 years ago. But like, Blue Sky, man. Blue Sky, I think, has hit like Escape Velocity in that this thing is going to work and it's here. People are building stuff on the protocol. People like using the app.

It's coming. Yeah. I don't remember what, but they're doing it. I would say I'm a blue sky person. My harshest criticism, blue sky is for all the promises of that protocol and how open it is. There's not one other server. No, there's not. But there are people building somebody built a cool blue skyy app that just looks like Instagram and it just takes photo posts. Like that's cool. That's the Fediverse, baby. Give me weird ideas about how to look at blue sky. I'm in. Yeah. But that's all just like AP. front-end views. Listen, I want one other place to log in.

One thing at a time, man. I want to post, maybe that's going to be the verge. You said good news. Let's end on good news. Okay, good news. We're happy. The Fediverse is around.

I like it. I do like it. I do love an open source, weird pebble hack, and I do love a weird band of misfit, decentralized social media hackers. That's the good stuff. Hell yeah.

That's the rebel stuff. Well, you know, the bad thing happened. That's what we want. Send us more, this is your, send us more weird, rebel technology. Right? That's tough.

That's your homework for this week and we'll play the voice smells next week. Love it. All right. That's it. We're way over time. That's the Vergecast. Bye, Carol.

And that's it for the Vergecast this week. And hey, we'd love to hear from you. Give us a call at 866 Verge. The Vergecast is a production of the Verge and the Vox Media Podcast Network. Our show is produced by Will Por, Eric Gomez, and Brandon Kiefer.

And that's it. We'll see you next week.