Anthropic的世代级狂奔、OpenAI陷入恐慌、AI护城河、Meta连输重大官司

All right, everybody. Welcome back to the number one podcast in the world: the Fantastic Four, the original. Oh, the cast is back. The cast is back. Brothers in arms. Here we go, good boys. We've got a big news week.

David Sacks is back, and he's in the great state of Texas. How's it been, Sacks? How's Texas been for you so far?

It's been great, although I just got back from D.C. I got about 3 hours of sleep last night, but we had a lot of news this past week.

Yes, and we'll be talking about PCAST and your role going forward in the Trump administration. Big news that we'll be talking about today also relates to you.

Oh, Sultan of Science, David Friedberg, with your background from the iconic film—for those not watching, it looks like the iconic Thelma & Louise. I wonder if that has something to do with the budget of California, which you've been outspoken about recently. Great rant, which I retweeted with Molly.

If only you could be allowed the time and space to do those kinds of rants on this pod.

Yeah, thank you very much. If you keep talking, Jake, and let him know. Here he is; he's going again. He's going again. It's Dr. Doom. Dr. Doom, your mayor and your new governor. Would you consider it, Friedberg, after a holo[?], running for governor?

There is no after a holo[?].

Oh, please do it. Oh, please do it. Wow, that'd be so great.

I'm tempted to buy a holo[?] for $5 or $6 billion so he just doesn't.

It's a dirty game, pal. California politics are dirty, man.

I don't even know what it does. I'll just have somebody else deal with it.

The oppo research on Friedberg?

No, no, no, no. We get him elected. He would do an incredible job. He would save the fourth-largest economy in the world. It would be incredible.

It'd be amazing.

Here's the oppo research, Sacks: David Friedberg went to a rave in 1999 and stayed up until 10:00 a.m. We have witnesses. Once he got tilted at the poker game, stole a bunch of pistachios and Lactaid, and ran home.

We open-sourced it to the fans, and they've just gone crazy with it.

Anthropic is on a generational run, and OpenAI is crashing out a bit, boys. Let's chop it up here.

Just looking at Anthropic, it's been a pretty major heater this year. In January, they launched Claude Cowork for business users. You know what that does? Cron jobs. You can connect to your Gmail, your Notion, whatever it is.

And then Opus 4.6, which, consensus-wise, everybody thought was a major step function. Jensen, Michael Dell—everybody's called it out. Jensen actually called it, back in November, an inflection point and the first agentic model. And Opus 4.6 has basically hit a threshold that we haven't seen before in terms of real productivity in teams, as Dell said.

In February, they dropped a bunch of Claude Code plug-ins that caused the SaaS-pocalypse—the SaaS, software as a service.

Well, it was a SaaS-pocalypse, too.

Yeah, there was a little bit of that back and forth as well.

No, I mean, as a SaaS investor, it was a SaaS-pocalypse. My exit comps were affected, that's all.

Yes. It seems like you may have divested at exactly the right time.

All right, $6 billion in annual run rate was added in February alone, as Brad referenced a couple of weeks ago here on the pod. Earlier this week, they announced computer use, a new agentic system for enterprise-grade, kind of, OpenClaw functionality. Now you can use the Claude app from your phone to control your desktop computer. Really slick feature.

Here's the calendar release over the past 2 months for the team at Anthropic. Dario, come on the pod anytime.

Sacks, you've had a couple of flare-ups, and obviously the administration and the Department of War had their kerfuffle. But just looking at it objectively, what's your take on the surging Anthropic generational run, as I've described it here?

Well, I've never been a critic of Anthropic's products. I've always been an admirer of their products. I think last year I gave them credit for MCP. I agree that they seem to be performing very well now.

The company made a big bet on coding as the big breakout use case. Whether that was done for business reasons or ideological reasons, I'm not sure. Anthropic is sort of the most AGI-pilled of all the frontier labs, and I think they made this bet on coding as their way to get to recursive self-improvement.

As it turns out, it was a very good business move as well, because code is the gateway into enterprise and enterprise IT budgets. And so they were able to grow revenue pretty quickly as a result of getting into enterprise.

Also, coding seems to be the basis for these other product extensions. So, like you said, they went from Claude Code to Claude Cowork. The idea being that, well, if you can generate code, you can also generate PowerPoints or spreadsheets, and you do that by generating the code to create that output. So that was the first extension.

Now they are extending into agents. This computer-use product is kind of like an OpenClaw knockoff. So it looks like the generational run for Mac mini is just about over.

Look, I think they're firing on all cylinders. My issues with them in the past were related to what I've called the regulatory-capture strategy. They do want a permissioning regime in Washington for chips and models, meaning you have to go to Washington to get permission to release new models or to sell GPUs anywhere in the world.

I think that's excessively heavy-handed. Their motives for doing that may be pure. It may not be regulatory capture; it may be ideologically motivated. Regardless, I do think it is a form of regulatory capture because it plays into the hands of the big companies and creates moats that new entrants will not be able to overcome.

So I have, let's say, a philosophical objection to that part of it. But again, I'm not a detractor of their products by any means.

With respect to what happened between them and the Pentagon, I'm not involved in that. I've stayed out of military procurement. In general, I don't get involved in what are called party matters. I just focus on policy matters that affect the whole space.

I saw Emil Michael making this point a couple of weeks ago on our podcast: If you, as a company, don't want your products to be used in war, don't sell to the Department of War. It's in the name. But if you do decide to sell to the Department of War, you should expect it to be used for all lawful uses.

So I think that was a very pragmatic observation. Again, I just have to underscore this: I'm not involved in that dispute. It's the basis of a lawsuit right now, and I don't want someone trying to draw lines between dots that aren't there. So I'm staying out of that one.

Yeah, and in fairness, objectively, they've been treated the same as any other large language model, even though they're not fans of the administration, they're not donors to the administration, and they have specifically been critical of the administration as a company.

Perhaps cynically, Friedberg, as a strategy to get—you know, one of the conspiracy theories here in Silicon Valley is that Dario's taking the position of being anti-this administration, anti-President Trump, in order to get all the PhDs. There's like 3,000 or 4,000 of these highly sought-after PhDs, and it's a way to have them vote with their presence to come work at Anthropic. Your thoughts on that, and then just generally their generational run?

I believe it, and I think they've actually created and fostered a culture of that since the beginning. I think that they're representing it as a branding exercise at this point, but I don't think it's made up. I think it's a direct representation of the people who work there and what they believe.

Yeah, and it's a strategic advantage because probably of those 3,000 PhDs, 90% of them are left-leaning and wouldn't want to work necessarily with Anthropic.

Like most things we see in the world today—in economics, in markets, in business—everything seems to be politicized, and you have a left and a right version of everything. You have a left and a right version of media, you have a left and a right version of what food to buy, and you have a left and a right version of what AI tool to use.

So this effectively may just be the natural manifestation in the AI market of what's going on elsewhere in society, as we all fracture and hustle over to our side.

All right, Chamath, before I go to OpenAI and their recent moves, any thoughts on Anthropic and Dario's positioning of the company?

Look, I think both are incredible businesses. We're in the part of the cycle where we're trying to create drama where I don't think drama exists, because they're still fundamentally in very different go-to-market motions. Now, they may converge and compete over time, but I think it's important to separate where each of them are good.

From an enterprise lens, which is where I see most of the action, particularly through 80/90[?], it's all Anthropic all the time. And I agree with Sacks. My philosophical issues with the management aside, around their ideology and sometimes how they use some of the capital for things other than tech and R&D, I have issues with those things.

But in terms of the quality of that technical team and what they create, it's head and shoulders above anything else. It allows us to build a vibrant business now. Do I have issues with how much it costs? Yes. Do I have issues with how fast we're consuming tokens? Also, yes. But I think those will get sorted out, and those are really tactical issues.

So the reason why I think we're all breathlessly trying to pit OpenAI against Anthropic is because we want some drama.

But the reality is these are very different businesses, and Nick found this tweet, which I thought was really interesting. Even at the absolute highest level, these things are presented in an apples-to-oranges way. There are very basic issues of revenue recognition that are fundamentally different.

You may say, “Well, who cares about revenue recognition?” The people trying to write the headlines that say one is overtaking the other miss the fact that they’re in completely different businesses, which has guided how they even think about growth. If you normalize these 2 businesses, what you would see is that OpenAI is still the overwhelming revenue generator in this space, and over time, Anthropic is catching up.

This is a little diagram that tries to explain this. OpenAI is 3/4 consumer subscriptions and 1/4 API. Anthropic is almost the exact opposite. OpenAI is used by consumers overwhelmingly. Anthropic is used either directly or through things like GitHub and Cursor.

As a result, OpenAI has a very conservative way of recognizing revenue. Anthropic, they sort of recognize gross tonnage as their revenue. When you start to hear things like, “Oh, this thing is at $20 billion and OpenAI is at $N billion,” they’re 2 totally different conversations.

I think right now it’s more about the press cycle of trying to create clicks than it actually is about the underlying quality of each business. Both are incredible businesses, as this demonstrates. By the time they go public, both of these businesses will have a very clean and, I suspect, normalized way of telling the story so that you can actually compare them. What I would tell people right now is that everybody is running with numbers to try to create a narrative that I don’t think makes sense or applies to either business.

Yeah, and there have been a lot of strategy changes. Some people are saying OpenAI is crashing out in panic mode. Obviously, they own the consumer with ChatGPT. They are the verb, like taking an Uber or Googling something. Consumers always just say, “Hey, did you check ChatGPT?” But obviously, other large language models are catching up.

Can we say something to that, Jason? I mentor tons of startups. Sacks has done it. Friedberg has done it. I do it. What is the one thing we tell folks? Focus, focus, focus, focus.

100%.

Do 1, maybe 1.5 things, but do them incredibly, incredibly well. Everything else, you start to bleed and smear. What was that Brad Garlinghouse term? Peanut butter. Yeah, you smear the peanut butter too far out.

This is a good moment, by the way, if either of these 2 companies are in the smearing phase, to recalibrate and reset because you just can’t do everything.

Speaking of smears, I couldn’t help but notice that Emil Michael was smeared by an article—was it in The Lever or something like that?—accusing him of having a conflict in the Anthropic dispute. Did you guys see that?

I didn’t. No.

Yes. I saw the—there was an article that said he was an investor in Perplexity and therefore conflicted in his negotiation with Anthropic. That’s what the article said, pretty much.

Perplexity is LLM-agnostic. That’s a stupid claim.

Right. It’s obviously written by people who don’t understand anything about AI, really.

Okay.

Perplexity is a wrapper. You’re saying it uses multiple AI models. I don’t think they sell to the Pentagon. They’re not a competitor to Anthropic. Moreover, as I understand it, Emil’s ownership of shares in that company was blessed by the Office of Government Ethics.

Nonetheless, I think the timing of this is very suspicious. It reminds me of what happened to me when I started opposing Anthropic, and all of a sudden there was that hit piece in The New York Times accusing me of having conflicts.

I’ll just say that Anthropic may pose as this company that’s on the side of the angels, but they’ve hired a number of very seasoned, brass-knuckle political operatives in Washington, including members of the Biden administration. Laura Loomer actually just had a piece today on one of them. I’m not going to rehash that.

The bottom line is that this is, frankly, a political operation that’s willing to get down and dirty. They’re not always on the side of the angels. I think they can be quite ruthless.

Sacks, remember I told you: You get 1 Biden mention a month in 2026, so you just used it up. I don’t want any more Biden, Biden, Biden. He’s retired.

That was not a Biden mention.

No, but the truth is, whoever wrote the story missed the actual best feature, or among the best features, of Perplexity, which is actually got a really great co-work competitor called Computer I've been playing with. I’m not a shareholder, to be clear. There’s no book being pumped here. The Model Council is the greatest feature—

Yeah.

—that they have. What’s really brilliant about it is that you ask it a question, Sacks, and it will go to all 3 different major models. You can pick which ones, including open-source models. Then it tells you where they differ and tries to figure out why they differ.

This is one of the great features of the product. I think Perplexity could be a really great company as well, even without a language model. But let’s talk a little bit more about OpenAI here and its market share, because I think you’re correct, Chamath, but they are getting off their game.

Here’s what’s going on. Take a quick look at the consumer market. Obviously, they started with 100% market share. They created the category in 2023. It dropped to 85% market share in 2024 and 75% market share in 2025.

But by how much has the market grown?

Precisely. The market is still growing. In terms of the number of searches and queries, they’re obviously growing tremendously. But they have major, major competitors, and their market share is going down.

I had my team over at This Week in AI do a more thoughtful analysis of where this is going. If you take a look at this, there are 3 players that really haven’t shown up yet: Apple, Meta, and obviously Windows. All 3 of those are underrepresented.

If you give them credit for just getting half a point of market share here and starting to intercept, which I think those 3 players were here and will be here, they’re going to be well under 50% market share. I think ChatGPT is going to have some big challenges on the consumer side.

While they’re doing this stuff in consumer, they’re cutting back on all their side projects. You probably heard about the Sora video app. That’s been shut down. This is major news because Disney was going to put $1 billion into OpenAI as part of it, and they had done a licensing deal. They were going to integrate Sora, this short-video product, into Disney+. All of that has now been canceled.

The $1 billion is not going in.

The $1 billion is not going in—the licensing deal, all of that. In addition, there’s supposedly a newfound focus at OpenAI on chasing Anthropic down the enterprise path.

So, they’re getting off their game, getting a little discombobulated, perhaps—or maybe getting on what matters, which is enterprise. Apparently, in terms of revenue, OpenAI also offered private equity investors a guaranteed minimum return of 17.5% as part of a joint venture that would help PE firms deploy AI and ease the high upfront cost of that.

There are lots of questions here, Chamath. I don’t know if you’ve tracked this PE model, but obviously a lot of people are doing roll-ups in services, accounting, and legal. Josh Kushner has a big effort here. A bunch of private equity firms are trying to essentially, I guess, end-run the transition process. Arguably, that’s what you’re doing with the software factory at 8090. Your thoughts on OpenAI, this pivot, and this private equity—

I think it makes a lot of sense for OpenAI to focus on a few things and do them exceptionally well. I disagree slightly with your first part, which is that I think people like to make new decisions about new experiences, and I think OpenAI has incredible consumer mind share.

I just see how my kids use it. They started there, and it’s very hard to get them to switch, even when I say, “Hey, have you tried Gemini?” They use Gemini, and to your point, the reason is because they stumble into it more.

Yeah.

But if you give them a cold navigation experience, they rely on ChatGPT. It’s the same, by the way, on the other side in the enterprise. If you give us a cold problem, my default reaction would be to use Anthropic.

I actually think that’s quite healthy because you’re going to segregate the market. If you go into the Wayback Machine, when we first started talking about this thing, this is sort of how we all postulated this would work: Even if OpenAI just won the consumer business, it’s a multitrillion-dollar company with enormous scale and value. I think that’s okay.

I think what they probably need to do is say, “Where are we the strongest? Where is there the most obvious traction? Can traction in another market, like enterprise, bleed into consumer usage?” If it’s true, then you have to win the enterprise.

I think winning the enterprise, though, is a very different game than winning the consumer. It’s a very different set of features and a very different set of expectations. People either have to decide they’re going to compete everywhere, or pick 1 thing and just nail it.

If I were OpenAI and you had to pick 1 thing, you would pick consumer because they’re the juggernaut and the clear leader, and they have an enormous brand.

Chamath, let me pull you into this because my base case here is that all consumer queries are going to be free.

Apple's going to make them free. They're already free for Google. I think Meta's going to make them free and actually have a decent product soon. Microsoft, same thing. ChatGPT has decided to push off advertising. They were going to put advertising in it. You remember they got mocked by Anthropic with their Super Bowl ads.

So, what do you think's going to happen on the consumer side? Consumers generally don't pay for services. Usually, 5% to 20% of the market pays for services, and everything else is free and ad-supported. But it looks like Apple and Google are going to just let it rip. So, that could take the revenue oxygen away from ChatGPT. What are your thoughts on who wins consumer?

I don't think that it's going to be free. I think there are 290 million subscribers to Spotify. They're paying—what are they paying?—20 bucks a month or something?

Probably less on average because it's a global number, but yeah.

Netflix has 325 million paid subscribers. AI that can book your travel, answer questions for you, track your calendar, do your email, et cetera, is likely going to be the most valuable—call it—meta-service that consumers have ever seen.

Mm-hmm.

And I think it's very likely that we're going to end up seeing many more consumers subscribe to a consumer AI service than we've seen even with cable television. Think about your cell phone. Everyone's paying 50 or 60 bucks a month for a cell phone. Why not pay 80 bucks a month?

100 bucks?

You pay 100 bucks. And by the way, in the pandemic, remember what we saw? The 2 things that people refused to cancel were mortgage payments and car payments. You were willing to go into arrears and into default. The 2 things that people would always keep were the cell phone, number 1, and electricity, number 2. ChatGPT will be there.

So, I think that's going to be the case with these consumer AI services. They're ultra-valuable, and they're going to layer in services on top of them. For example, do you want to watch video embedded in your consumer AI app? Do you want your consumer AI app to do your finances for you?

It could be that the consumer AI app becomes the new platform, much like the iPhone was for the app economy. There could almost be, whether it's through connectors or embedded tools, an incredible ecosystem where traditionally advertisers actually pay to be embedded and show up inside of the AI app. The consumer can either pay for it, or the advertiser can pay for it. So, I think there's going to be a very different economic model, and it's still very early days.

Sacks, the numbers right now would, in my estimation, be more like 50 million people subscribing to ChatGPT. They have 1 billion users, or they're trending to 1 billion. I think they'll probably hit it in the next month or 2. Certainly, they had 900 million 2 or 3 months ago. So, it's about 5%. Where do you think this winds up? Do you think it becomes 300, 400, 500 million consumers willing to pay 20 bucks a month for this, or do you think it's more free, with the data and the ad-supported nature of it going the Meta and Google route?

I think it's possible that you could get a few hundred million subscribers for the premium tier. Look, I think most consumers will take the free service in exchange for advertising—some ad-supported model—which, by the way, I think could be quite successful.

When ChatGPT started displacing Google for search, a lot of people were predicting the death of Google's model because who'd want to look at 10 blue links? I think that's true, but I think you can do something much more compelling in AI chat compared to just the list of links. So, in any event, I think ad-supported models might make a comeback here in addition to premium models.

All that being said, as an investor, I always liked B2B businesses better than B2C because it is hard to monetize consumers. Their willingness to pay is not that high, and they tend to have high churn rates. Whereas businesses tend to be very sticky, you can upsell them, and you can get more than 100% net dollar retention year over year. So, if you can make an enterprise business work, it's always been a model I've liked.

That being said, obviously some of the most valuable companies in the world are consumer companies: Meta, Google, and Apple. These are all consumer-first companies. Ultimately, both models can work.

Obviously, both can work. The question is which one? I guess it really comes down to how motivated we think Google and Facebook will be to build that bridge from their ad networks to their AI offerings. Obviously, Facebook is kind of MIA in all of this, but Google is not. I think that will be the determinant here.

Google is going to compete very vigorously for the consumer because it is existential to them. It's very clear that search and AI chat are kind of merging into one space. That means that ad links will kind of merge into in-chat advertising. So, they have to adapt to that and compete for the consumer.

I also think that Google is in an outstanding position to do the whole OpenClaw thing because they already have access to your calendar, your documents, and your email. The agent doesn't really have to earn your trust because you already trust Google with all of your stuff.

Right. So, I'm kind of waiting for the Google version of OpenClaw because I don't really want to share all my documents with some new service.

They're the only one that has so much free cash flow that they can almost view it as 2 separate companies, which it effectively is. GCP over here runs the enterprise play, and Google consumer over here runs the consumer chatbot play, and they can keep them segregated.

That's so much harder for a startup to do because, on top of just keeping everybody organized, you have the financing problem of constantly having to raise more money because you don't yet have a profit engine that spits out cash. They're probably the only one. You can see it in the valuations, actually, which I'm going to get to in a second, but people believe the durability of Google more than they believe the durability of anything else.

And Sacks, I think you weren't on the pod last week, or maybe 2 or 3 weeks ago, but Google announced Google Workspace Studio to do AI automation. It's online, and people are playing with it already. So, they have joined the OpenClaw party.

By the way, Jason Calacanis asked a different question earlier as well, which is around the PE story, and I think the PE story is a window into the rest of the broader market.

The real open question is: What are these companies worth? There's a threshold question—kind of a very important fork in the road right at the outset—which is, do you believe that we're on a path to superintelligence where everything is incredible, where there's infinite abundance, where you can magically describe things and beautiful things appear, complex things appear, groundbreaking things appear? Or do you believe that it's good next-generation software? The answer to that question is really important because we're financing things like it's the former.

And GC has a big move in this. Chamath was on the program in January when we did the interview show, and he's got GC buying up and rolling up accounting firms, health care firms, hospitals, et cetera. This seems to be part of the future of venture capital: taking AI—Sacks—and actually buying out, or I should say big VC kind of starting to look like private equity, buying out hospitals, accounting firms, and business-processing firms in India, putting them all together, and then running them with AI.

Any final thoughts on that, Sacks, as a business strategy?

Well, I think it's interesting. They're kind of betting on the idea that they can own the change management around AI. Everyone just kind of assumes that you throw AI over a wall, and a business automatically knows how to use it and drive efficiency from it. What we're seeing is that it's pretty difficult. Chamath, you've seen that. There are McKinsey studies showing that 95% of enterprise pilots aren't successful.

There's tremendous latent value in AI, but it's hard to know exactly how to deploy it at this point in time. What these private equity firms are saying is that we know how to drive value from this, and if we own the business and then own the change management, we'll be able to create value that way.

Their business model makes solving this problem existential. This is sort of along the lines of this essay that I wrote. Let me just give you the thought experiment, and you guys react.

Today, we live in a world where the whole market is trying to debate what P/E ratio you'd be willing to pay. Facebook is durable, so I'm willing to pay 30 times. Nvidia's really durable, so I'm willing to pay 40 times. Tesla's incredibly asymmetric to the upside, so I'm willing to pay 200 times. Caterpillar or John Deere, I'm willing to pay 15 times. I'm just using these as examples.

What it's effectively signaling to you is how durable all of these cash flows are. All we do in the public markets when we make an investment is guess when the cash flows run out. We try to say, “Here's how much it's worth, and here's how much I'd be willing to pay for it today.”

But if you go back to this example and say, “What if there's this superintelligence on the horizon?” I think it's fair to ask the question: What is anything worth? What is anything worth in year 10, or year 15, or year 20?

Because if you have infinite abundance and you have all this creativity, won't all companies be disrupted? Won't we be in this constant churn of everything getting disrupted all the time? If you were faced with that problem in the public markets, how would you react? I think the canary in the coal mine are the SaaS stocks.

Yes, we jokingly call it the SaaS apocalypse, but I think it's much more important. I think it's a big societal question. How do you view capital markets? How do you view the health of a company in a world where we've been told there's a superintelligence on the horizon that makes everything much more fragile than it was before?

The market reaction is to put all these companies on a spectrum. They started here in software, and they're re-rating everything down. They're changing the way that things are being framed, from price-to-earnings to a multiple of the cash that you have on hand.

I think that has huge implications, mostly to Silicon Valley and largely to employees, because we all sell the dream. We start a company and we're like, "Okay, small salary, big equity upside." But that's implicitly saying, in 15 or 20 years, this thing is going to be worth some gigantic number.

But if instead every business gets disrupted every 5 or 6 years, all you're going to end up with is just the cash. So what should employees do? The rational reaction from employees will be, "You know what? I don't want your equity. Give me more money." And if all of a sudden you do that, the valuation multiples and the complexity change again.

I had my team put this chart together. What is this? We took a handful of SaaS companies and took the Mag 6. I just said, "Okay, if you take the market cap and divide it by the annual free cash flow, what that tells you is how many years it takes to get back if you bought a share of stock?"

How many years does it take for the free cash flow to come back so that you've earned back the cost of 1 share? Snowflake, in 2023, would have taken you almost 100 years. Where is it now? It's been cut in half. ServiceNow, Atlassian, Workday—you see it.

I think what this speaks to is the beginning of this re-rationalization in the public markets, saying, "If superintelligence is coming, we have to be very careful about what we're willing to pay for these things." But if you look on the right-hand side at the Mag 7, what's so interesting is that Apple, Microsoft, Meta, and Alphabet, the market has completely flipped the other way.

They're saying, "We believe that these cash flows are essentially monopolistically durable forever." That's the only reason why you would walk them up like this, except Nvidia, which is the most unbelievably accretive, well-run company, with the highest margins, making $200 billion. They're treating it like they're treating ServiceNow and Snowflake.

I just think it's so interesting what's happening. I can't explain this, but this data sort of shows the reset that we're going through—a very complicated reset in the company.

Sacks, what's your take on this reset as Chamath describes it? Do you think this is just a flight to the quality of the free cash flow of the Mag 6 and how much cash they print? And then maybe the other ones are smaller footprints and they're just more disruptable.

We had this discussion many years ago that Google, Apple, Microsoft, and Facebook would all be disrupted at some point, and that simply hasn't happened. They've gotten much more nimble at copying products or incorporating features and products into their core offerings. So, your thoughts?

It's probably generally correct that there will be a decline, but there's also the selective opportunity. Do you guys see the LP slides that went on the internet from Thoma Bravo's LP conference?

Yeah, those are great. Nick, maybe you could find that.

They kind of highlight that within the broad market landscape, there are companies that are not just going to sit idly by and let AI delete their business value. They're integrating AI themselves, they've got high-quality people to do so, and they're reinventing their products themselves.

They've already got a beachhead, they've already got customer access, and they've already got enterprise users. In fact, if they can integrate AI into their products and tools, there's almost this selective dispersion that happens in the market.

The winners are going to win, I think, truly in every market—not just in software, but across every market, including industrial supply chains—based on who is going to implement and utilize AI tools and agents to do work. It's going to expand the work productivity of that organization, not just create new features, which is what we focus on when we talk about software companies.

But really imagine a complex business being able to do 10 times the output it can do today with the same capital equipment and the same labor force.

Friedberg, what do you pay for in a world of superintelligence versus in a world of non-superintelligence, in terms of the durability of a business?

Yeah, it's hard to say, man. We don't know, right? And that's why all the discount rates are going through the roof. That's why the valuations are collapsing, because we just don't know what multiple or what discount rate you apply, or what terminal growth rate you use, which effectively implies the value.

What do you do in your P.A.? Do you say—

I don't think Disneyland is going anywhere. I think there's some stuff that you could say is the counter-AI portfolio. The counter-AI portfolio, I think, is physical experiences.

HALO. They call it HALO: high asset, low obsolescence.

Okay, that's a great example. I'm not a big investor or trader like you guys, but that makes a lot of sense. Intuitively to me, that seems to be an area where people are going to be spending a lot of time, and they're going to have durability in those businesses.

I think businesses like natural gas production have durability. I just bought LNG, that Cheniere company.

Cheniere? Yeah.

Cheniere.

You've got Cheniere.

I bought Cheniere given all the craziness in the Middle East. I visited there with Doug Burgum, so it was the first time I'd ever been exposed to this business, and I checked it—

That's a great business. It's a really well-run business.

I think that's got durability, obviously, unless it gets blown up by some enemy. That would be a problem.

So, Adam's thesis that T.K. shared?

Adam's real life mining is a great one. Obviously, I think the space industry is going to be bigger than we recognize. I actually think there's probably a $15 trillion to $30 trillion a year economic opportunity on the moon.

That kind of business, like you're going to see with SpaceX's IPO, is going to have an insane multiple. So, to your point, Chamath, there's a lot of stuff getting steamrolled here with crazy-high discount rates where you just don't know.

There's a bunch of stuff where the pathway over the next 15 to 30 years is maybe independent or unlocked because of AI. Then there's a bunch of stuff that's just unaffected. That starts to get a higher multiple because that's where capital starts to flow.

Sacks, there are probably 3 things that we would agree are great moats for businesses: brands, network effects, and the management team. Those come into play here as well, yeah?

I don't know if I would consider the management team to be a moat. Warren Buffett says that you want businesses that are so strong that they could be run by a bunch of monkeys, because one day they probably will be.

I was thinking more like, obviously, Elon and Tesla are going to just relentlessly innovate.

It's a great point, by the way, and it's true, especially in the age of agentic AI.

I think, though, that does up-level things a bit. You are right that the key question is moats, because I do think that there are still strong moats in a lot of different kinds of businesses, and a lot of them are very subtle.

Like you said, some of them are network effects. Some of them are the difficulty of producing physical-world products, things like that. So, there are a lot of different types of moats out there, and that is the key question as we enter a world of, let's call it, digital abundance.

Yeah, the network effect of Apple's ecosystem, and they have hardware, right? They've been on that path of making their own silicon. That's incredibly defensive. And they have brand, right? So, that is pretty strong for Apple.

Then you take Tesla: you have Elon relentlessly innovating, and it's hard hardware stuff. If you look at Meta and Google, these are incredible brands with great management teams and constantly innovating.

If I had to bet, I'm going to bet that brands go to 0.

Really?

Yeah, because I think that when you can make things that are as good or better, and you can make them in a cheaper, faster, better way, people want that abundance more than they want affiliation to a brand.

Example?

The perfect example is actually what Tesla did to BMW, what Tesla did to Mercedes, and what BYD and Geely have done to the car manufacturing cycle in China.

This is a fundamentally cheaper, faster, better product. Yes, it's got a great brand, but nobody's going to pay a premium for these products. The reason why the Model Y has outsold everything else is because it's priced better and it's superior on every operational dimension of comparison.

That's also true for the cars in China. So, I think it's the opposite. I think brands and the pricing power of brands go away, other than maybe premium luxury goods. But even that's eroding.

Look at the stock—I don't know, Nick, show the stock chart of LVMH or Ferrari. This is not a commentary on the quality of the actual product, but what this shows is an erosion of pricing power.

All of these things are being eroded away.

Yeah, this would be the value prop. Most people in brands would just say “value propositioning.” So JetBlue is a value brand. The Tesla Model Y is a perfect example of a value brand. And if you look at Apple’s recent cohort, what did they focus on? The Apple MacBook Neo, which is a value laptop—$600, $700. So they’re even going downmarket to try to capture that value.

And to your point, maybe the right word is abundance. The brands that bring abundance, that bring more to the table than their competitors, and are able to bring more at the same unit cost or less, capture share. That’s probably true.

You know what’s interesting about that, Chamath? As we open up the aperture of this, one of the theses we talked about here a couple of years ago was, “What happens with AI disruption, job disruption, et cetera?” Cost coming down on cars, with the Model Y getting cheaper, the Cybercab coming in, BYD—obviously, if you go to any foreign country, BYDs are everywhere and they cost $15,000. Then you look at Apple making the MacBook Neo—that’s a $600 laptop. Everything getting cheaper seems to be happening.

I just think it’s very hard to know which of these companies are going to be disrupted. A year ago on this pod, we were saying that Google was going to be toast, or some of us were saying it, because it looked like—

Yeah.

Okay, fine, but it looked to us like ChatGPT was taking massive share from Google Search and the AdWords model was becoming obsolete. Now, because of the success of Gemini and the potential for personal digital assistants and personal agents, I think we’re probably pretty bullish on that company. Look at their stock chart; it reflects that. I think it’s doubled in the last year.

You’re onto something huge, because what you’re saying is, I can confirm this to you: 80/90, we sell enterprise software. I’ll tell you, three conversations with huge enterprises asked exactly what you just said, and we jokingly called it “strangulation as a service.” They all say the same thing: “Get all this complicated UI out of the way. Get all of these products out of the way. Find a way to create a shim where I can just write what I need and tell it what I need it to do. It deals with all this complexity in the background. I never want to see these things ever again.”

That’s what people want, to your point. They want to be able to say, “Okay, pay this with my Venmo, or use my Amex in this situation, or get me that flight in some way,” and have all these wonderfully smart agents do all the work behind the scenes.

And that actually tracks with exactly what I’ve seen with OpenClaw, Perplexity Computer, and Claude Cowork. Instead of going to your Notion, instead of going into your Gmail, instead of pulling up your calendar, you ask it, “Hey, what’s on my schedule this week?” It brings it to you. So you could have a dumb, flat terminal—a chat interface—on a $100 device, and it would do just as well.

Just to make the counterargument against myself, even though I think that you’ll increasingly tell your agent what to do instead of clicking around or touching the screen, you still need a dashboard or a user interface to check on it and see readouts of information. You still need to be able to visualize it. Where is the Uber that I’ve asked to be sent to me? How far away is it? I still want to see it on a map.

So I could imagine that even if, let’s say, Siri++ becomes the dominant way of interacting with your iPhone, you’ll still want that Apple user interface. Hard to say. I see a lot of people out there tweeting that Apple is brilliant for missing the whole AI wave and not spending a lot of money on data centers or CapEx. Then there are other people who say, “Well, wait a second. They’re missing critical capabilities.” Then there’s a question of, “Will they be able to make deals for an AI-powered Siri?” I don’t know. I think this is very hard to know at this point in time.

All right, and we will be discussing all these hard topics at Liquidity. May 31st through June 3rd. Chamath, some big announcements here from you. You have taken control. This is what happens in the Game of Thrones known as the All-In Corporation LLC.

The partnership.

The chaotic partnership.

No mids.

There literally is chaos. But wait, wait, wait, why are we doing this in California? I don't even think I can go back to the state at this point in time.

You can come for 48 hours. Make an occurrence.

By the way, there are some people taxed that have zero days in state at this point in time. You can do 48 hours. It's okay. Well, Gavin Newsom might meet you at the airport. You can get picked up at the airport.

Let me announce the next two speakers.

Oh my god, I'm so excited. But just for background, Chamath comes into this thing. Liquidity was this little conference I did. It's now part of All-In and I start setting up all the speakers and then Chamath goes, 'Not good enough. I'm not showing up unless I pick all speakers.'

What did you call me?

Melon farmer. So this melon farmer took unilateral control of the 10 speakers. 10 speaker slots, very coveted.

Okay, so we've announced Dan Loeb. We've announced Sarah Friar. And I'm very, very, very honored and excited to announce that Bill Ackman and Andrej Karpathy will also be speaking at Liquidity.

Four heavy hitters, six to go. Two more goats on the roster of goats.

And we have a handful of other—

Yeah, there might be some surprise guests and unannounced guests show up. TBD. Let's get Dario to show up. Dario, come to the show. Come hang out with the All-In boys.

I'm really excited. You know what Andrej's going to do? What Andrej agreed to do is he's going to do like five or 10 minutes of slides on the future of the world with AI and then we'll do a fireside.

Love it. That's going to be tremendous. And you know, he's been on a heater himself with his recursive GitHub.

Autoresearch is incredible. And Bill Ackman has a ton of points of view on all of this stuff. So we'll get a lot of his thoughts. I want to give a shout-out to Friedberg, because when we did the interview with Jensen, he said, “Guys, I had a glass of wine on a Sunday. I was like, ‘I wrote a replacement for my HRIS system.’” I asked all my team to do it, and then I was like, “Oh my God, this is incredible.”

So I asked our folks, “I don’t like the website; it’s 80/90.” The next day, they were like, “Yeah, we vibe-coded a new one. We’ll have it up.” Then I asked, “Well, do you like the CTAs and how it’s doing?” They were like, “No, no, no, no. We put it into Autoresearch, and we doubled the click-through rate.”

To Friedberg’s point, this would have been many man-months and tens of people. Instead, all these recipes—these playbooks—are available. The person who runs growth at OpenAI publishes his recipes.

Yeah, you’ve been Claude-pilled. That’s it. You’ve been one-shotted.

It’s really incredible. This is why—I mean, you commented on my post yesterday. I wake up every day and my head spins. I’m like, “What is going on in the world?”

It’s crazy. Every day feels like a new era right now.

Yeah.

And it’s disorienting.

Because I think what’s really interesting is, you pull what would normally have been something that takes a period of time—say, a year out or 2 years out—and you can pull it in and say, “I can get that done in 3 days, and I don’t need to hire people.” All the sequencing and staging that would normally go into accomplishing something has been reduced, and then the time rushes in. So all your ideas rush into you, and they’re all immediately accessible. That’s why every day is like, “Oh my God.”

It’s disorienting.

Yeah, it is. I bought the domain name Annotated.com about 15 years ago for $4,000, and I wanted to create a service like a bookmark service, Chamath, where you highlight a paragraph from The New York Times and then you write your comments on it, save it, and you basically have the service.

I was talking to a developer about making it, and I literally vibe-coded it as a Chrome extension this past week. I was like, “Okay, I’ve been sitting on this domain and project for 15 years, and I did it in a weekend.”

Isn’t it crazy? It’s very weird what’s happening. It makes me feel like I’m in a simulation, that everything can just manifest itself. It’s the Star Trek version of the world. Remember the replicator, Sacks, where you just say, “Earl Grey tea at this temperature?” That’s happening in business now. You’re like, “CRM system, build Annotated.com. Build this new website for $80 or $90.” Boop—it’s a replicator that just gives it to you. It’s very strange.

I’ve only felt this feeling 2 times. Once, I was on the outside looking in. I was a derivatives trader in Toronto, looking at the 1st dot-com wave, and I was like, “I’ve got to be a part of this. How do I get to be a part of this?” I got a job at Winamp, and the rest was history.

Winamp, the original iTunes.

But I missed the wave. Financially, it didn’t do anything for me, but I was in the right place.

You got to the beach with a surfboard. That’s all that matters.

Then, in the 2nd wave, I crushed it in the move to mobile and social. But this wave feels like 100 times bigger than that.

It’s a tsunami by comparison. I had 3 of these. It’s probably what it’s like to be at Nazaré in Portugal. You guys ever see those clips of that crazy place where there are 100-foot waves, and you’re just towed in and you’re like, “Okay, let’s just—this could end 1 of 2 ways. Let it go, or see you, goodbye.”

No, it was literally when I first saw a PC, when I first got on the internet and then saw Mosaic—those 2 moments in the early ’90s—and then seeing the iPhone. I think those 3 moments were very special.

Very special.

All right, rough week for Zuck. Two verdicts went against Meta in 2 days. They were first found liable for allowing child predators to access minors on Facebook and Instagram. A New Mexico jury ordered Meta to pay $375 million in damages. The attorney general’s office there ran an undercover investigation. They created fake child profiles, and those accounts on Facebook and Instagram were contacted by predators. People showed up. Yada yada yada.

A whistleblower and former Meta engineer testified that his own 14-year-old daughter received sexual solicitations on Instagram. Then, on Wednesday, a Los Angeles jury found Meta and YouTube negligent for designing addictive platforms that harmed a young user’s mental health. Basically, the plaintiff in this case said they started using Facebook at age 6 and Instagram at age 9. She testified that features like notifications and algorithms made the apps so addictive that they caused depression and anxiety through compulsive use.

You have some thoughts on this. We had Jonathan Haidt, right? Didn’t we do an interview together? Was that 1 of the first joint interviews? Great book.

I did. You and I did, yeah. Incredible book. Things are obviously addictive. He had a very important point, which is to try and keep kids off cell phones and social media until they’re 16, and he was kind of cheerleading this verdict.

But I’ll take a little bit of a contrarian view to the popular sentiment on this. I’ll just talk broadly about this idea of tort litigation. Tort litigation costs our economy $900 billion a year in the United States. $900 billion a year. That’s how much is spent on litigation costs, settlements, and judgments. It’s 3% of GDP, and it’s growing roughly 10% per year.

These civil penalties decided by juries are going against big companies like Meta and YouTube, but it’s also food companies, restaurants, and everything else. Anytime there’s a window to sue someone and extract value from them, tort firms are all over it. It’s called the tort tax now in America.

It’s not just losses paid by the companies, because fundamentally, when a big company pays out these tort taxes, they’re going to invest less. There’s less R&D, less product development, prices stay high, there are fewer new product launches, and there are all these crazy restrictions on stuff.

Look, I agree that social media causes immense harm. It particularly causes harm for kids. Kids should not be on social media until they’re 16. I absolutely agree. Maybe adults shouldn’t either, but, you know, we’re adults. Fundamentally, I think there’s an important question that we often ignore, which is: Who is fundamentally responsible for that harm?

Should the sugar beet and sugarcane farmers be responsible for diabetes in America? Should the soda companies be responsible? The retailer selling the soda? The FDA for not stopping it all? Fundamentally, I think we have to ask the question: What role does individual choice and individual responsibility play in this equation?

If everything is a liability, what do you think?

I think we have to take personal responsibility. I think the parents who are absent from taking care of their children are responsible for harm to their kids. You shouldn’t let your kid play with a gun. You shouldn’t let your kid go to some sketchy neighborhood after hours by themselves. You shouldn’t let your kid play video games 100 hours a week. You shouldn’t let your kids eat nothing but soda and potato chips. You have responsibilities as a parent.

I think parents should keep kids off screens and keep kids off social media. Once the harms of excess use or exposure to that sort of thing are known, I think there’s a responsibility that sits with the parents.

What about things like tobacco or processed food?

Yeah, this is the key. The same is true of alcohol. Alcohol is terrible for you. There’s nothing good about alcohol. But I think I should have a choice on whether or not I want to consume alcohol, tobacco, or processed goods. The recognition that it’s bad for you should be publicized by the government.

If I had to summarize, what you would say is that product liability law makes no sense. There should be human liability and human responsibility expectations in society.

We never talk about responsibility. We always talk about where the government failed us and where these companies failed us, and we never talk about what we individually did wrong. How did I individually choose to eat 100 sodas a week? How did I individually choose to get my kids addicted to social media? Where was I as a parent? We don’t talk about our responsibility.

By the way, this fundamentally addresses the point about human agency, which I think is more critical in this era than ever, because AI is going to flood us with everything all the time, nonstop. What we choose to do in a world where we’re already getting everything, and how we choose not to take everything that’s being offered to us, is going to be a critical part of what distinguishes human success from human failure. It’s going to become more apparent in the future.

Not everything is about liability, and not everything is about the government failing us. It’s about people making choices, and we don’t talk about it.

What I’ll counter—I agree that personal choice is super important. What you’re probably leaving out here, which you’re definitely leaving out here, is when these companies know they’re doing something damaging and do it anyway. That was the key to RJR Nabisco—

The whiskey company? What about a whiskey company selling whiskey to an alcoholic? What about a potato chip company selling potato chips to an obese person?

I’ll put those 2 aside because I don’t think we’ve seen major cases about that, but I will say the auto industry knew for a long time about airbags. You remember that? They didn’t deploy them.

RJR Nabisco knew that these were addictive. They designed the cigarettes to become more addictive, and they didn’t tell people about the health risks. Asbestos, same thing. Lead paint, same thing. This has happened over and over again, where corporations subvert the release of information to make additional profit.

So the question here with Facebook is: Did they know how addictive these were? Did they know kids were being sexually assaulted and that they could have done something about it, or didn’t they? I agree that there’s too much litigation.

Absolutely true. The kids being assaulted—absolutely true. Not releasing information about the level of addiction, if they had it, is certainly bad. But should the product be legal, number 1? And if the product is legal, who’s responsible for using it? Where do we draw the line? If we don’t want people to make a choice, then we shouldn’t put it out there.

But if you, as a corporation, know it’s dangerous and then you lean into that—if you know, as in the case with Facebook, that this is super addictive and super damaging to young girls, and then you lean into making it more addictive and don’t put safeguards in place, and you can prove that, like RJR, asbestos, and many of these other products—

What safeguards does a whiskey company put in place? What safeguards does a casino put in place?

Hold on. You asked the question. Age is 1.

It’s not about age.

Second, after age—which is what we’re really talking about with kids—Jonathan Haidt would agree that they shouldn’t be using this until they’re 16. I think that’s a perfect analogy: age-gating and then labeling. If they know of something that’s really damaging, they should release that information.

So, there was a whole thing about alcohol and pregnancy, and they covered up in the alcohol industry or didn't disclose exactly how damaging it was to drink alcohol while pregnant to a fetus or a developing fetus. And then remember all those signs that went up in bars in the '70s and '80s? That was directly because of that.

So, labeling, information, and age-gating would be the logical things to do for social media, and that's what's happened to PSAs. So, that's the answer to your question.

I don't know about age-gating, but I think informing parents about the risks is fine. That should be a responsibility. But the tort lawyers are one of the largest donors in political elections in the United States. They donate largely to Democrats in local elections and Republicans in national elections. And then they are the largest donor class to elections of judges.

It's a business, and I think we don't talk enough about the business of tort law, the business of litigation in this country, and we often ignore this question about choice. And I don't want to live in a world, J Cal, where the government and companies are telling me what to do and what not to do, how to live my life, et cetera.

Let me jump in here. Personal freedom and personal responsibility versus corporations maybe knowingly doing things.

Responsibility, J Cal.

Parental responsibility in there as well. Yeah. What do you think, Sacks?

Well, look, there's no question that the trial lawyers want to turn Meta into RJR Nabisco and the cigarette companies and try to fit their fact pattern around that. I just think that the activity is fundamentally different than smoking. Smoking is manifestly harmful to you regardless of what your age is. The only reason we allow it is because of assumption of risk. It's a free country.

I think in the case of social networking, it's much more unclear what the harms are and what the benefits are. I think it's much more subjective and it's much more of a personal choice for adults and also for parents. Friedberg, one area where I disagree with you a little bit is you said the harms of this are immense and well-known and understood. If that's the case, then let's just ban it for under 13 or under 16 or whatever it is. But I don't think that's the case, and because it's unclear, I think it's up to parents to decide what's for their families.

At the end of the day, I think the right way to deal with this is parental empowerment. You give parents the controls to set screen-time limits or to decide what apps their kids install.

Now, the debate has moved over to AI apps, and there are a lot of parent groups that want to ban kids or teenagers from being able to use AI chat apps because there were a couple of cases of self-harm. My reaction to that is, look, I've got a 10-year-old, and if he starts using ChatGPT to get answers to questions, I would consider that to be a good thing. I'll keep an eye on the usage, but I want him to be an AI native. I want him to be able to do research. I want him to know how to use these tools. I want him to get the right skills to be successful in the 21st century.

In China, they're incorporating AI into K–12 education. Are we going to ban it for our kids and teenagers? I think that'd be a terrible mistake. So, I think when it comes to AI and AI chat apps, it has to be up to the parents because there's too much manifest good that can come out of kids learning how to use these apps. Maybe social networking is in a different category, but I do think that the harms have been exaggerated because the trial lawyers have an incentive.

And just to give you some facts about that LA case, in this LA case, they were sued by a 20-year-old woman who claimed that she became depressed because of using social media. She apparently suffered body-image issues because of social media, and she was able to win this judgment for millions of dollars. Look, I think there are big causation problems with that case.

In the case of that plaintiff, the evidence showed that she came from an abusive home. Her father had abandoned her. Her own mother had body-shamed her. So, it was very unclear where her body-image issues were coming from. I think it's possible that social media contributed to them, or it could just be that social media was a scapegoat.

I also got to say that I do think it's a dangerous precedent. Are we going to allow plaintiffs' lawyers to sue Spotify because you created a playlist of sad music and that music contributed to your emotional distress? That's kind of what we're saying. You got to remember, these are free services that people have a choice whether to use them or not. If social media is making you feel bad, if listening to the wrong playlist is making you feel bad, then stop doing it.

But what's going on here, I think, is the trial lawyers are trying to create the next Big Tobacco, and their goal is to try and sue these companies into oblivion. I don't think that's the right answer either.

The consensus, Chamath, is that kids who use this 2 or 3 hours a day—this has been across many research studies—have massive correlations with depression and anxiety and eating disorders, specifically in young girls, if you get to 2 or 3 hours a day of this. So, that correlation's pretty well established.

You pretty famously said—and you worked at Facebook, so you saw this coming—but I think at Stanford, in the 2017–2018 timeframe, you actually had some comments on this. We could either play the clip, or you could just describe it, I guess.

I said what was obvious to me then. I lost a lot of friends at Facebook when I said it, unfortunately, but essentially what I said was, I don't let my kids use it, and I don't think that this is a constructive part of a developing child's diet.

I do think that Sacks is right that I view AI chat differently. There are different guardrails that are required if you go down some sort of dark corner, but that's possible to understand, and the product is architected in a way to create cul-de-sacs if you're thinking about self-harm or these other things.

Social media is very different. It's an incredibly fast-switching algorithm, and that's the optimization. Until the incentive for that optimization changes, these outcomes will continue to compound.

I think the interesting thing is that the LA lawsuit was an individual lawsuit. The young woman, I think, was awarded $3 or $6 million. Yeah. The other one, though, that they lost this week was in New Mexico, and that was for $375 million. That was more around, I think, child exploitation. Yeah.

The thing that I'll note is that these trial lawyers, which I'm not a fan of either, have been trying, as Sacks said and as Friedberg said, to make these folks a target because there's so much money on the line. And they've been batted back pretty successfully. But this was the first time where they were able to navigate the Section 230 protections that Facebook and Google have typically used to protect themselves, because Google was a part of the LA lawsuit. And they were able to go down the pathway of product-liability language.

Now, to Friedberg's point, I think he's generally right. I think it's my responsibility as a parent to take care of myself and my children. But to the extent we don't change these product-liability laws that are on the books, I think the door has been opened and a map has been drawn, which is: This is how you navigate around Section 230, and you can get a decisive lawsuit in your favor against these large companies with enormous cash flows.

And so I expect that this will be a death-by-a-thousand-cuts kind of scenario where folks are just going to rally around this. I don't think it's right, but I do think that's the rational reaction to what just happened.

I mean, look, $6 million to an individual 20-year-old girl or $375 million. These are huge numbers. And the reality is that I think we've opened the floodgates. I do think there should be a better response. And I do think that parents should take a lot more responsibility, but I do hope that these products allow us a kill switch for when our kids are under the age of 16. Frankly, I would love a kill switch under the age of 18.

And there have to be simpler ways to age-validate. We used to work around this thing called COPPA compliance. COPPA's bull—it's a nothing burger. A 6-year-old can vibe-code their way around COPPA, so it doesn't do anything. And so age verification is completely broken. It's harder to get age-verified for a porn site than it is to get un-age-verified for Facebook.

Chamath, how did you get around that?

Yeah, how do you know that?

How do you know that?

That's just for uploading. You were uploading at the time, Chamath. This is proof of age.

Not your OnlyFans.

Proof of proof of age.

We actually deal with this in our new AI framework that the administration released last week because the number-one issue is around online child safety, which I think, again, is mostly about social networking, but it touches on AI, and so it's kind of gotten lumped together.

And I'm referring now to the White House, which has said it's willing to support some form of age-assurance technology and parental controls. So, look, I think there needs to be a conversation around how true it is that social networking is just a fountain of ills for young people. If that's true, then why wouldn't you just disallow it, right? But if it's unclear—and I think it's more on the side of unclear, that's just me—then it's up to the parents, and what you want to do is, again, have the age-assurance technology and the parental controls and let the parents decide.

I don't think that social media is net bad, broadly speaking.

But I will tell you that if my child uses it for 2 or 3 hours a day for multiple days in a row, they become…

Yes, they act weird.

And so, I don't know what to tell you, except I would love a kill switch until these kids turn 18 on all these products.

Well, you can. You should.

This has to do with dopamine, by the way. What I do instead is use Apple's Family Sharing, but it's hard, and I'll tell you why it's hard. Our schools give our kids Chromebooks and say, “Use these Chromebooks, because that's the way you're going to access your LMS, your information, your content, your homework.” And you know what's an integral part of that? YouTube. And then it becomes a backdoor.

So even when we take the devices—I take their devices, we lock them down, they can't bring them to their room, the whole thing—if they need to do homework and I catch them during a little break, I'm like, “What are you doing on YouTube?” They're looking at Shorts.

Of course. That's what they do.

Yes.

And so it is a whack-a-mole problem for parents.

I think I'll just give myself the final word here on behalf of the group. It's obviously a dick thing. Obviously, the industry has not policed itself, nor would it police itself, because they want to get people addicted now so they're addicted when they're adults.

The rest of the world has realized this. Australia now has a minimum age of 16 in force; Malaysia has 16; and other countries are right behind them—Spain, Germany, and the UK. What should happen here in the United States is that this should be done with the handset manufacturers. If Android and Apple showed leadership with Facebook and said, “By default, when you buy a phone, you're going to have to age-verify kids under a certain age”—which you kind of do already when you create a family plan, like you did—we could solve this problem, and then parents would opt in to giving their kids access to this.

No good comes out of kids under 16 using social media, and Jonathan Haidt, in our interview, said you should be putting phone lockers in. Our school is doing that; other schools are doing it. It is the greatest thing ever. Kids complain, and then they love it.

I don't think it's that great. I'll tell you why. Our school does the phone condoms—I don't know what they're called—but you put them in the bags and whatever. The real problem is that there's enormous social pressure when you're in high school to use these products.

I'll give you a specific example. Our rule is you cannot get Instagram or TikTok until you're 16. I would love it to be 18, but we all agreed on 16, and we're able to maintain that rule. But then high school comes around, and I have 2 kids in high school now.

“Oh, we need Snapchat.”

“No, you don't.”

“Yes, we do.”

“No, you don't.”

“Okay, great. We'll have no friends. Thanks a lot. We'll just sit here in our rooms, dark in there.”

There's an enormous amount of social pressure. So when you talk to the other parents, it's like, “Hey, guys, can we all agree that we don't need Snapchat?” You just can't get uniform agreement across all parents because everybody views this problem differently.

And so I had to change the rule. Now, when you get to high school, you're allowed Snapchat, but not Instagram. When you're 16, you get Instagram and TikTok and all that stuff. That's the best we could do. But it's an impossible task for a parent.

I find it indispensable to be able to use iMessage with my kids, and I also like the location awareness. You can track their location. You compare iMessage and those features to Snap—is it really that different? I don't think it's that different. So you're really going to ban Snap? You're going to ban Snap?

No, I can't. My point is I had to give them Snap, but you're right. I was like, “Why can't you use iMessage?” They're like, “Dad, don't be a loser.”

Yeah. No, I signed up for a Snapchat account just so I could see where they are and where they're messaging. I prefer they were just on iMessage, but anyway, that's the reality.

So you talk about banning all these things, but again, I think you're on a slippery slope. And I have to point that out.

As a parent, ban them. No, I'm saying, as a parent—

Yeah, you're setting guidelines.

In the house, these are the rules. When you become a freshman, I'll give you Snap. When you're 16, I'll give you Instagram and TikTok. Otherwise, shut the fuck up, put your head down.

By the way, I also have a tip for parents. You know this issue with kids in headphones, where they want to be listening to stuff all the time and you have kids walking around like zombies with headphones? I replaced their headphones with over-the-ear headphones from this brand called Shokz. They go over your ears, but your ears are open, so you can listen to an audiobook or music and still be able to talk. Since we did that, there's been much less drama around the over-the-ear headphone issue.

Man, managing people's headphones.

Well, again, it sounds like you're exercising parental supervision. Does it really make sense that later, when your kid turns 20, they could sue these companies for emotional distress? It doesn't make sense.

Right, because to your point, that girl said she was using one product from the age of 6 and the other from 10. That's crazy.

Yeah, it's ridiculous. Crazy. I agree with you, Sacks. That's crazy. Where were the parents?

All right. We've got time for, I think, one more topic. President Donald Trump, the 47th President of the United States, announced his Council of Advisors on Science and Technology. It's called PCAST. Sacks, you have now—am I correct in saying—moved from the czar of crypto and AI to now the leader of PCAST? Is that the correct way to frame this?

Well, the President has appointed me to be a member of his Council of Advisors on Science and Technology and to co-chair it along with Michael Kratsios, who's the director of OSTP. I'm still an AI advisor, but I do it on behalf of PCAST now.

You remember last year I was in SGE. We got up to 130 days. I used that time up, and the President appointed me to this new role. It allows me to continue being a technology advisor—in fact, on a wider range of issues. Before, it was AI and crypto. Now it's whatever PCAST wants to study, talk about, or make recommendations on.

I think, in addition to AI, other areas that are interesting are nuclear power, quantum computing, advanced semiconductors—all these different areas. I think we've got some biotech.

Thank you, Friedberg.

And we have some incredible people who are now on PCAST to run with this.

Yeah, I'm looking at it: Marc Andreessen, Sergey Brin, Michael Dell, Larry Ellison, David Friedberg, Jensen Huang, Lisa Su, Mark Zuckerberg, and some other folks who maybe aren't as recognizable to the audience. How were they selected, Sacks? The only criticism I've seen of this is: lots of business leaders, great; lots of technology, great; but maybe a little light on the scientists.

Well, I don't know. We have people who've won a Nobel Prize in physics on there. We're talking about people who are experts in, like I mentioned, quantum computing, fusion, nuclear, biotech—pretty much everything across the board.

I would say that one difference between this PCAST and previous ones is that you have more doers, more builders—people who've actually created products or companies. We think that's a good thing. Why would it be a bad thing?

I mean, is it a bad thing that Marc Andreessen invented the first Internet browser or that Jensen invented the GPU? If you're going to make recommendations about advanced semiconductors, don't you want to have someone who actually invented some of the key products in the space?

The big question, of course, Sacks, is—as we've seen here with Science Corner—how do you plan on staying awake for these meetings if it's going to be all science? Usually, when you take your bio break or your—

Well, it's science and technology, so we're—

You'll be awake for the science.

On the tech stuff.

Got it.

And then we've got Friedberg to focus on the science stuff.

Got it. So, Friedberg, this is incredible. You've joined President Trump's administration now.

I will say I'm honored to be invited and appointed by the President, and I appreciate Sacks and Michael Kratsios.

You look back at PCAST, it's kind of rooted in FDR, when he formed this Council of Advisors on Science, when nuclear physics and quantum mechanics were starting to reinvent what was possible in the world. We're sort of at a similar era today, because arguably AI is reinventing what is possible in the world.

I think there's this acute moment that we find ourselves in, in this extraordinary race against China. I'll give you a statistic: 10 years ago, China published 50% as many scientific research papers in peer-reviewed journals as the United States. Last year, they published 50% more than the United States. This is across all disciplines and domains, including physics, materials science, chemistry, biochemistry, and the broad life sciences.

There's this moment that we're in right now where the world is being reinvented by AI, but there's also this extraordinary race with China—not just in fundamental research and discovery, but in the industrialization of new discoveries and new technologies.

You could feel it in D.C. this week. I was at the Hill and Valley Forum, but literally everyone in Silicon Valley, everyone in D.C., is absolutely honed in and focused on what is going on in China. It used to be, in biotechnology, for example, that China was kind of a copycat and a me-too, or they were really good, for example, in manufacturing.

But it is now the case that in many subdomains, China is becoming the scientific leader in biotechnology and in life sciences. And that is a scary thought because ultimately China could end up engulfing the entire pharmaceutical industry and becoming the leader in things like medicine, but most importantly, foundational things like AI.

I want to defend and speak for a moment about the choice to put what I would say are industrial science and technology leaders on this commission, because this is a moment where there's an industrial race, not just a discovery race, underway. That's why this moment is so critical. Anyway, I'm very appreciative to have an opportunity to serve and thankful to David for his leadership.

In related news, Chamath will be joining the President's Advisory Council on bomb pots and wagering. I'll be joining that. And when I saw this PCAST, I was like, “Oh, is it podcasting? I'm in. Finally, I've been invited to the President's Council on podcasting.” So, big news coming. I don't want to tip anybody's cards.

We'll let you know when that happens.

Yeah, absolutely. When the President's Council on podcasting emerges, Lex Fridman and I will make our way to Washington, DC.

I want to thank the President as well, and that's a great honor to be named to co-chair this. It is, like Friedberg was saying, a fairly august body that goes back a long time. I think the modern incarnation was created in 1990 by George Herbert Walker Bush, so this has been around for over 30 years in its current formulation.

I think we have a slightly different take on it, which is that we're going to have these builders and doers on there. Like Friedberg said, we have some catching up to do on our industrial policy. We have named 15 people, and there's still 9 more slots. We can have up to 24.

I think it's possible that at some point there'll be a second round, and if we're missing some expertise, it can be filled in. Obviously, that's up to the President. He can decide later who else might join this.