Silicon Valley ICE Debate, Anthropic CEO’s AI Essay, Social Media Addiction Trials | Jan 27, 2026

0:13Welcome, everyone, to the Information's TI-TV. My name is Akash Pasricha. It is Tuesday, January 27th. First up today, our editor-in-chief, Jessica Lesson, joins the show to break down Silicon Valley's reaction to what is happening in Minnesota right now. And we'll also separately discuss Jessica's latest essay that she published in The Information. We'll then get some more analysis on NVIDIA's deepening ties to CoreWeave. And we'll take a look at some of the upcoming trials that will put social media companies under the microscope. And finally, Pinterest is cutting up to 15 % of its staff as part of restructuring to put more resources towards AI.

0:52We will look back at some recent comments that Pinterest CEO Bill Reddy said on this show about his AI strategy at large. It is a busy show, so let's get right on into it. A number of leaders across Silicon Valley have publicly begun speaking out against the violence in Minneapolis. That has also sparked public debate between leaders at Coastal Adventures, and it has also led to billionaires clashing with each other on X. To help us unpack all of this, I want to bring on our editor-in-chief, Jessica Lesson. Jessica, welcome back to the show. Hey, Akash. I was going to say good morning, but people watch this anytime, so hello.

1:29Hello. Jessica, I want to understand how you are thinking about Silicon Valley's reaction right now to what is happening in Minneapolis. You know, Akasha, and obviously the events of what's happening are horrific. So I'm going to comment on what I comment on, which is how tech leaders are reacting and what it means for their businesses. And, you know, it seems like the liberals of Silicon Valley have found their keyboards. You know, with the dawn of the Trump administration, we've seen a lot of people move both to the right and then the people who weren't to the right have basically kept quiet.

2:08And I don't think there's been a sea change yet, but we are certainly now seeing a variety of people, including sort of very senior executives at companies like Google with Jeff Dean, a major AI leader there, you know, speaking out. And we're seeing debate within firms as well. Over at Coastal Adventures, we have Keith Ruboy and Vinod Kosan, very different publicly sides of this issue. So it raises some questions about whether those disagreements will erupt and cause more drama. Or maybe I'm hopeful, maybe we can, you know, just have some level of civil political debate. I probably not, actually, because we are in the world of the Internet and big egos.

2:55But I think the sea change, especially over the last 24 hours, has been people who don't share their views of the administration saying so publicly, even those who work at big tech companies. And as sort of basic as that sounds, we really have not seen that for a long time. I want to ask you about recent times this has happened. This is not the first time that people within companies have disagreed on issues and that has spilled out on social media. If you think about what has happened in the past, how do you think this plays out moving forward? You know, it's early to tell, Akash, and obviously this is on the heels of tech CEOs, you know, attending the Melania documentary premiere or screener or whatever it was, right?

3:38So I think the macro trend, certainly in the technology industry and business, I just got back from Davos where we saw the same thing, is businesses sort of very much wanting to appear in lockstep with the administration. So I think that is still the big trend now. But if we start to see more individuals speak out and, you know, there were reports this morning that Sam Altman has said that ICE and internal communications at OpenAI said ICE has gone too far. If we start to see leaders saying that, it's going to be really interesting just to navigate internally the mood inside these companies. You know, there have been many moments, intense political moments, where corporate America has had to kind of set a tone.

4:22And in tech, the tone has been, we're not political entities. You know, leave your political discourse for the dinner table. You know, come here and do your job. That's what tech companies have been trying to set the tone. But obviously, that is very difficult in these really challenging times. So I think it kind of portends a lot more in turmoil internally, more calls from employees that question doing business with departments of the government, which we've already seen. We've seen those in the Valley time and time before, but I think we're due for another wave of those as well. And then just to finish us off here, I mean, if you think about the connections between Silicon Valley and then the government specifically, and then, as you mentioned, the turmoil that this could cause internally with employees, what are the possible business implications of all this, do you think?

5:19Well, one, I guess the questions of contracts. I mean, personally, I don't see these large tech companies pulling away from doing business with the United States government. We are in a, I guess, choice of words, but we are in a battle to and all these companies to get their AI, their software, their technology into governments. And I just don't see that changing. But I think there's going to be more unrest internally at these companies about those business decisions. Now, you mentioned Sam Altman. Dario Amadei from Anthropic was one of the leaders to also speak out against what's happening right now in Minnesota.

5:59He also, you know, it was part of a larger essay that he wrote. It was, I think it was 20 ,000 words on the current state of AI. 30 pages. I needed Claude to get me through it, Akash. Right. Isn't that kind of ironic that we use his product to digest his own words? Although maybe that's why he made it in the first place. What do you make of a 30-page essay and the key takeaways from that? I mean, it's hard to get through something like this even. What did you think? Yeah, well, look, on one hand, I admire Dario for being outspoken about the implications he sees about AI. He's always done this.

6:39He's a man of essays. Apparently he wrote this in three days over the December break. And I guess as a writer and someone who also finds Dario to be really thoughtful, I kind of like welcome these. I'm interested in reading them. I think they're sort of net positive. But they're really principles, Akash, and not a playbook for solving all of the problems that he outlines, which by the way are really scary. I mean, at one point he talks about the fact that we have trained AI on science fiction. And, you know, does that mean that the AI is going to adopt the mentalities of all the sort of worst AIs of science fiction, right?

7:16I mean, he's putting that as one of many hypotheticals to make his point. And it's sort of intentional fear-mongering, right? Like, Dario does want to light a fire around what he sees as a major inaction and regulation. But, you know, even his points at the end of what to do about it, to me, fall short. They're talking about, you know, constitutional government. I mean, the idea of getting Elon Musk with what we've seen over at Grok in recent weeks to like adopt some constitution with these other AI companies just does not seem plausible to me. So in the end of the day, I'm glad this essay is out there.

7:55It's thought provoking. People should read it. The bigger question is what is going to happen to mitigate some of these risks? What is the will in Washington around that? You know, one of my takeaways from Davos was actually a very split opinion on what's coming for the next wave of tech regulation. On one hand, you saw, you know, a lot of murmurs that there seems to be, especially around AI, a lot of concern, a lot of political concern. And on the other hand, the notion that anyone was going to take a sort of aggressive regulatory stance right now made other people laugh. So that's where my brain goes on this topic.

8:38But the essay is certainly worth a read. And so let's talk about then an essay that you wrote yourself, which was, as you just sort of alluded to, You went back and looked at your own reporting notes that you took at Davos. And Claude Cowork is the tool, the toy of the day that everyone just can't get enough of. And the premise of your essay, I'll let you tell everyone, but I mean, you basically use Claude Cowork to see if it can make sense of your reporting notes from the trip. And the results were decidedly mixed. Tell us about your process and what you learned from all of this. Absolutely, Akash.

9:19And, you know, it was actually sparked by an interviewing Brett Taylor, the CEO of Sierra, who we've had on the show and is sort of an AI native customer service company. And he was saying, look, I'm an engineer by trade and I am most vulnerable to AI disruption. And this is sort of what I'm doing about it. And he went on and sort of described how he's this orchestra of these AI tools. So, you know, kind of very bleary-eyed at the end of this trip and in my way, I had an hour or so in the lounge before. I said, okay, let me, inspired by Brett, I'm a reporter. Some people think I'm also deeply vulnerable to destruction.

9:59Obviously, information is and how that's transferred. I'm going to embrace co-work, finally had time to download it and such. And I had recorded all of my on-the-record interviews, put the transcripts in, give me the takeaways. And, you know, on one hand, it was accurate, right? I mean, like the takeaway minus one number that it got out of context, it was accurate, you know. And Claude identified and made a PowerPoint presentation, you know, big themes like the AI transformation is just beginning. business adoption of ai is a focus of you know topic of conversation right and it was a little bit a little bit vanilla in other words a little bit vanilla but but you know what it did for me and and i think i think this is important right i think everyone where all professionals are sitting here thinking are these tools you know where do i fit how do i turbo charge my powers with these tools and how do i look around corners for where i should focus given the march of these tools.

11:04And for me, my takeaway was, you know, if you weren't following the technology sector and you weren't reading the information or any other sources, you'd probably learn something from this stack, right? And by the way, you wouldn't be manipulated. It wasn't fake news. It wasn't slop. It was coherent. The slides looked good. But what it made me realize is what I try to do and provide some unique value to a specific audience whose sort of interests I know, whose questions I know, and obviously do the interviews to get that original information before. And so it was sort of a lesson in knowing your audience, which is such a simple principle, but I think business professionals across fields should really, really embrace it as we enter this next AI age.

11:53Because at the end of the day, you know, my takeaways were much more around, you know, which companies were changing their talking points, who executives were meeting with and what that indicated for their strategies, what wasn't said, who wasn't there. You know, that's based on 20 years as a technology reporter and editor, and I think will be more valuable to our audience. But it made me want to like really tap into the brain of our audience even more, which I think is a great takeaway to have from, you know, where we are with this tools. And by the way, I did just sort of jot down those takeaways.

12:31And then I asked Claude Cowork to make me some nice slides. And, you know, slides are not my forte, probably saved at least 45 minutes to half an hour. So I could go back to thinking about the next interview. So yeah, overall, I think it was a valuable experience. And I really made me think about knowing our audience. And I think that's a lesson that's much broader than just for the media business today. Great. Well, Jessica, I want to thank you for coming on. That is Jessica Lesson, our editor-in-chief and founder here at The Information. Okay. We have been thinking a lot this week about what NVIDIA's latest investment in CoreWeave announced yesterday will mean for the broader AI data center and NeoCloud story.

13:15I want to bring on someone who has been following that story very closely. David Levy is a general partner at Porch Capital. He previously spent nearly a decade with AWS. David, welcome to the show. It's great to have you here. Thanks for having me. So why do you think NVIDIA did this deal? Walk me through what you think their calculus is here. The super obvious is that they and Coreweave are highly committed to making sure that GPU Cloud AI infrastructure build-outs happen as quickly as possible and that capital does not become one of the constraints. That's super obvious. The slightly less obvious but still obvious is that NVIDIA feels a little bit of pressure from Google TPUs and they want to make sure they are still able to influence the direction of the hardware that's used for inferencing.

14:11But I think the real story here is that inference workloads and prompting and tokens are about to explode because of agentic inference, which is going to have models prompting models rather than just humans prompting models. I'll get to the inference part of this in a second because you've written about that topic extensively. But I do want to ask about NVIDIA's role as an investor, a supplier, a customer of CoreWeave. I mean, when you have one company that has so many different ways of connecting to another company, I mean, is this a good idea to you? Is it inevitable? What do you think of it?

14:49I think what's happening is that there is so much demand and so little supply. And the way to finance these types of build-outs can't be done by traditional mechanisms, bank debt, etc. So we're seeing unique types of financial structure. For example, when Corby first built out, it used the private credit markets. But what we're seeing is that instead of going to just private credit markets or traditional means of financing, they're going to the large corporates who have the best interest of themselves and of getting these build-outs done. So that includes Microsoft, that includes Google, that includes NVIDIA, of course.

15:30And so the capital's got to come from somewhere. It's not a demand problem at all. Now, you worked at Amazon Web Services for a long time, and we've had executives from neoclouds on the show before. And the question I've asked them is, hey, how do you think about competing with the larger cloud providers as you provide a smaller but more specific offering? I wonder, when you were at AWS, how does AWS think about defending itself against these these NeoClouds, these AI specialized clouds. I mean, is it really a threat? Do they have more specific offerings that they hope to target that segment? Well, so when I was at AWS, I'd say I was there right up until the NeoClouds started to appear.

16:18And my first look at them was like, how can they possibly compete with Amazon? The reality was, is that the economics of building the types of data centers that the neoclads are building are superior to the economics that the hyperscalers have. And if you look at pricing, for instances, that's what you see. I don't know what the exact pricing is today, but in general, especially when this first started, the neoclads were coming in 50 % or more cheaper and more effective at getting these workloads up and running. And still today, you're seeing some of that, and it's very difficult for the legacy cloud providers to do that overnight.

17:02How did they get that price efficiency? What makes them able to offer it so cheaply? So their data centers are designed to handle GPU workloads. And their data centers are designed to be able to more effectively incorporate technology like NVIDIA's H100s, B200s, etc. And so they're fitted with the types of racks, the types of design, the types of networking, and especially the types of cooling to do this. the legacy data centers, it's not simply a matter of like, oh, you're just going to swap out one rack of CPUs or older GPUs for these new things. And I don't blame the hyperscalers. They have, you know, between the three major hyperscalers, they have$200 billion a year in revenue to serve.

17:50And so it's not like I can shut that down and start to serve the GPU cloud from that. They have to build their own new stuff. And I think the initial deal that Microsoft did with CoreWeave is a proof point right there. Microsoft was like, well, it doesn't seem like we can do this. Let's work with CoreWeave to do it. I want to pivot to talking about something that you wrote about earlier in this year. You were talking about themes that you think will dominate the conversation in 2026. And this was on your blog. And you wrote about edge inference, and you wrote about that becoming a topic that you think really will dominate much of the conversation.

18:32Why do you think that will be the case? And I should also say, for those of us who don't know what edge inference is specifically, explain to us what it is even. So this means, what I mean by edge inference is basically running inference workloads anywhere that is not a centralized data center. So that could be on device, that could be at co-location facilities, that could be at any places in between those. And so when I started to look at that, I first thought, I was like, well, this is a natural progression. The same thing happened with cloud and hosting as we went to content distribution networks.

19:06But that was from latency. Latency will become an issue, but the real issue here is what we saw at the end of last year. There's no power. There's not enough power to build these data centers right now. And so there's any number of solutions to this, whether it's getting more power, whether it's using gas turbines or more efficient ships. But another way to reduce the strain on the power supply is to have some of these inference workloads deployed in less, I don't want to call them less capable, but less expensive data centers and facilities to run. And so the core leaves and hyperscalers of the world will continue to build these megacenters as they should.

19:49But for some workloads, that's going to be overkill and the economics just won't work. And you can have these other locations that have much less power constraints than these larger data centers do. So the idea here is basically being able to subdivide the workloads between what needs to happen at a big data center versus what inference can actually happen on my iPhone sort of thing. I mean, that's sort of the prediction that you're making. Yeah, I mean, like iPhone is one spot. I mean, I think it's proven pretty difficult to get inference on a device. Apple's been trying that for a while. We haven't quite gotten there.

20:25I think we'll get closer. But whether it's on-device, on-prem, or in content distribution networks, that's where I see some of the power under-supply being mitigated. Right. And last question for you then. I mean, what's not there yet about the technology? Why hasn't this happened yet? Well, so first of all, you can't put H100s and B200s and other types of devices into these smaller parts of the network. But you also need to be able to right-size these. And it hasn't been the predominant issue. But as we move to, as inference becomes more and more significant portion of workloads, it's already more than 50%, and it's going to explode with Agentec AI, with models prompting models, this is going to become a much, much, much greater problem.

21:21And so we're going to need to deploy this infrastructure in other places. Great. Well, David, I want to thank you for coming on. That is David Levy, general partner of Porch Capital here on TITV. Okay. Today in Los Angeles, jury selection will start for the first of what will be thousands of trials involving state and federal groups suing the big social media companies. The lawsuits allege these companies have designed their platforms to be addictive, and that in turn has adversely affected young people's mental health. My colleague Erin Wu wrote a deep dive around these trials in last night's briefing newsletter.

21:59I want to bring her on to talk all about it. Erin, welcome back to the show. It's great to have you here. Thanks for having me. So walk us through this. Who exactly is suing the social media companies right now? Right. So there are thousands of cases going on right now against the social media companies. This first case centers on a 19-year-old girl known only by her initials KGM, who's in Chico, California. She is claiming that her social media addiction caused her depression, suicidal thoughts. Originally, she was suing all four of the social media companies, Meta, TikTok, Snap, and YouTube.

22:35As of this morning, actually, TikTok also settled. So both TikTok and Snap have settled this first case. But essentially, there's thousands of cases split up into state and federal cases. The state cases are all families, like KGM's cases. The federal cases are families, but also school districts, state attorneys general, and Native American tribes. And so what are all these groups asking for specifically? Yeah. So the state cases are personal injury cases. So essentially the plaintiffs are asking for damages to compensate them for their own medical expenses, but also pain and suffering, and then punitive damages to encourage the platforms not to do this again.

23:15In the federal cases, they're also asking for court orders, essentially forcing the platforms to change their behavior. So that's in the school district cases and the state attorney's general cases. So it's not just money, in other words. I mean, they're actually asking for changes to the platforms. Right. So what the plaintiffs are arguing is that certain platform features are defective and are causing addiction and these harm to young people. So it's not actually about the content on the platforms. It's around how the platforms themselves have designed it. And so they're calling out things like infinite scroll, autoplay, targeted notifications, features designed to Keep people on the platform longer and keep them coming back and back.

23:56And so what Accord could order is for the platforms that have to change some of these features. Now, I'm just thinking about this case. And I'm someone with no legal background whatsoever. So I really know nothing here. But I am sort of thinking about how it should be relatively easy to prove maybe that social media is addictive. but maybe the challenge here is really showing that that specifically is tied to young people's mental health. What sort of evidence are the plaintiffs thinking about bringing forward here? Yeah, so essentially the plaintiffs are going to have to prove two things. Like, A, they're going to have to prove that the platforms are addictive, and B, they're going to have to prove that there's a causal link between that addiction and then the harm to the kids' mental health.

24:45So I was chatting with a psychologist yesterday. there is some scientific evidence to support the idea of social media addiction. There have been some studies that have broadly shown a correlative link between time spent on social media platforms and negative mental health outcomes. And so some of that research does exist. What the platforms are obviously going to be arguing is that, oh, there's a lot of factors going into this. These kids were bullied off of the platforms, and so they're going to be trying to muddy the waters as much as possible. Okay. And so basically, it's not a slam dunk. I mean, what seems obvious to us is not a legal slam dunk is what you're saying.

25:27Yeah, so I mean, lawyers that I talked to are saying that it's really hard to know how these cases are going to go. And one thing that's interesting there is that these are jury cases. So basically, what starts today is jury selection in the case. And so these cases will be tried by people like you and I, who have grown up maybe using social media or at the very least have been steep in this world for the past decade plus. And so it's going to be a lot of people with thoughts on social media and kind of like coming in with preconceived notions. So it's going to be really interesting to see how jury selection goes.

26:04How much is at stake here, do you think, maybe even money-wise, for the social media companies? Right. So the first group of cases is about money. So the plaintiffs are seeking compensatory damages and they're seeking punitive damages. And so it's kind of hard to know how much money exactly is at stake at this point. So the big wild card is how much money the jury awards for pain and suffering, which is a more fuzzy amount than just simply medical expenses. And what one of the plaintiffs lawyers was telling me was, for example, in a case where a family's child died by suicide and they're blaming the platforms on it, you're essentially asking the jury to put a number value on a life.

26:50And so it becomes kind of hard to guess at the outset. So TikTok and Snap have both settled this first case for some amount. We don't know how much that is, but that likely sets the floor for them in terms of how much they would be able to settle the other cases for. Marc Lanier, the plaintiff's attorney, is likely to give a number that he's asking for an opening statement next week, which will also give us an example. But the very largest global settlement, which is how these cases often were solved, the tobacco industry was$206 billion over 25 years. So I mean, like, this can be a lot of money.

27:25I wonder what you think this tells us about lawsuits that may inevitably come up with the AI platforms in the coming years. And some of them are here right now, of course. You know, I sort of think about social media being a way that we get information and a way we consume content. and the AI platforms are certainly going to be, if not, they already are, the new way of getting information. I mean, did your reporting around these trials, did it tell you anything about how you think that situation could unfold? Yeah, so these lawsuits are going to be really important for how the generative AI lawsuits play out.

28:06Some of those have already settled. Character AI and Google, there are other lawsuits about ChatGPT. So essentially, what the plaintiffs in this case are trying to prove is essentially novel legal theory that's treating these platforms as products, as products with, in some cases, what they're arguing is defective product designs or product liability essentially. And so rather than go after them for the content on the platforms, which has generally been held to be protected by Section 230, they're going after the platform design itself. And so this is the same legal theory, and in some cases, the same plaintiffs prosecuting the cases against the generative AI chatbot companies.