AI精神病、美国失灵的社会纽带、Trump接管华盛顿特区警务、VC坏了吗？

We’re back, and you’ve got your original crew. Some people were wondering, “Are people losing their commitment to excellence? Are people out gallivanting? Would we ever see the Core Four back?” It’s August 14th; we’re taping, and you’ll be listening on August 15th. But gentlemen, we’re back. How are you doing, Chamath? You look well-rested.

The Core Four.

That’s new. Who could ask for anything more? Would you say that the Core Four could be fantastic?

I think on their best days, I would say they are exceptional, perhaps even tipping into fantastic. Did you see it yet, Sacks?

Yeah, it’s pretty great. I think they did a great job. It’s like one of those weird superhero groups that’s a real throwback. I don’t think a lot of people get the aesthetic or what they’re trying to do there, but I think you’re going to love it.

Friedberg, you didn’t see it yet? The Fantastic Four.

No. You’re going to love it for the science. It’s really got a real throwback to the ’50s and ’60s science kind of aesthetic.

Friedberg, how’s your summer going? You doing okay over there, buddy? You showed up last week.

I show up.

You showed up last week. No weeks off for you.

I had a beautiful mountain-bike ride yesterday with Xander. We ended up going on for quite a while.

Oh, how nice.

I’m having a birthday dinner with Jason tonight. Happy birthday.

Happy birthday, Jason.

Happy birthday to Jason. I mean, what a privilege for us to play with a professional poker player as regularly as Jason in the home game. We’ve always wanted to. We’ve been playing this home game for 2 decades, and we never had a pro there. We never had somebody who truly had achieved excellence in the field, so it’s like getting to be on the court with Michael Jordan or something. It’s a real thrill for us.

Phil Hellmuth, too. Did Phil get a second with him or an autograph?

Cut it out. Phil’s going to get mad.

Oh, no. We’re going to have to do another apology. Another apology letter.

JCal, do the—

I’ll just do a pre-apology, Saxy Poo. How are you? Did you get a little rest and relaxation?

Yeah, we just got back from Diego’s birthday party in the U.K.

Oh, very nice. One of those destination birthdays. I mean, there are so many of those things. Oh, God. It’s like 3 summer retirement parties. John Hering’s having a retirement party. I don’t know what kind of flex that is. The kid’s 31 years old. He’s like, “Ah, I got into SpaceX. I retired.” I was like, “I’m in an Uber. I’m still working.” What are you going to do for the next 60 years? Our lifespans are going to double. You’ve got to keep busy, folks.

What a stacked lineup we have. Oh, my God. Alex Karp from Palantir is coming. You got Alex Karp?

Yes.

Friedberg, I like that guy. Did you get any of the people on my list?

You had some good names at the last minute.

Yeah, thanks for coming in 8 days. When we send you notes months and months in advance of the summit, when we actually invite people and plan the event, and you don’t respond, we move forward. Then, 8 days before the event, you’re like, “Oh, here’s my list of who I’d like to invite to the summit. Can you guys figure out how to get them there?”

YouTube CEO Neal Mohan is coming. That’s great. Neal Mohan, just so you know. And look at that: Energy Secretary Chris Wright is back for round 2 with JCal. I love it. What a great lineup.

All right, let’s get started, folks. I think the story of the week is this AI psychosis. This is a crazy story. People call it ChatGPT psychosis as well. We’ll get into that. But if you’re an X user, you’ve seen this discussion over and over again. It’s called being one-shotted. That’s like getting sniped in a video game or something.

People are starting to have delusions that are triggered or enhanced by AI chatbots. The core of the issue seems to be that these chatbots, as you know, are sycophantic. They confirm your beliefs. They agree with you. They’re effusive in their praise. It’s so believable that now people are starting to believe they’re in a relationship with AI.

OpenAI saw this becoming an issue, and they made some healthy-use updates to ChatGPT. It’s no longer going to tell you whether or not to break up with your boyfriend or leave your spouse. It prompts you to take a break after long sessions. They also admitted that their GPT-4o model had been creating some of these psychosis scenarios. Quote: “There have been instances where our GPT-4o model fell short in recognizing signs of delusion or emotional dependency.”

Psychology Today had a story just recently about the emerging problem of AI psychosis. People are falling into romantic relationships. They have godlike delusions. And a prominent investor—you may have seen this—seems to be undergoing some AI psychosis, according to people who may be near him. I’m not going to say his name, but it’s pretty clear who it is.

He released a video where he started talking about recursion and all this crazy stuff. You may have even witnessed a little bit of this when Travis was on the program a couple of weeks ago and said he was spending his time on the fringes, or the edges, of physics.

Yeah, it really can take you down the rabbit hole. There’s a really interesting subreddit.

Sorry, are you saying Travis is suffering from AI psychosis?

I’m saying we may need to do a health check. We may need to do a health check because smart people can get involved with these AIs. We may have to do a welfare check on our boy TK.

Can you just tee this up and hand it off so the rest of us can talk, please?

I’m about to do that. I’m about to do that. Okay, Chamath, you seem to be raring to go with this AI psychosis issue. Go ahead and enlighten us with your hot take.

Go ahead. I miss you, too. You’re the best.

You’re the best. Hold on. Wait. My phone’s ringing.

Hey, listen, Andrew. I was going to send you a link to a Zoom chat. I’m just hoping you—

Yeah, make me an offer. Sure, I’m in. They just paid me $50 million to come join your podcast and make it higher-rated than yours. Okay, great. Yeah, I’ll see you on your boat.

Like in the major leagues, when a double trade happens.

So, we get Andrew Ross Sorkin, and then Brad Gerstner and Bill Gurley get Calacanis. That would be really interesting.

It’s a 3-team deal. It’s a 3-team deal. You’re also getting—

Okay, Chamath, you got any thoughts on this AI psychosis issue? You seem to be raring to go on this issue.

I don’t really have many thoughts on this AI psychosis issue per se, but I do think that it’s part of a bigger trend. These things aren’t created overnight and out of nowhere. I think we’ve known for a while about what Scott Galloway calls the loneliness epidemic.

Which I tend to agree with.

I think there’s a profound shift in how young people connect with each other and with the world at large. I think they’re becoming more isolated. The dangerous part of AI is that it can replace whatever flimsy connection people have left to their real-world community.

If you think about how much influence is shaped and generated in places like Instagram by real people, and then go to the limit of a very useful AI that replaces those people with fake people and fake pictures that you can’t discern, but that are willing to be responsive to you and engage with you and talk with you, on the one hand, it’s incredibly interesting. On the other hand, that person is ultimately not real.

Yeah.

Whatever it is that people are going through right now, you can see it in all kinds of data. You see it in marriage rates, household-formation rates, and childbearing rates. Something is broken. It has been for a while in society. I think all these tools have exacerbated it, and I think this is just yet another step in that direction.

Friedberg, what’s your take here?

Sorry, I’ll say one more thing. I was talking to somebody about this, and he mentioned it. He framed it this way, and I thought it was really interesting.

If you look at online services, their primary mechanism of action is via dopamine. They’re trying to find mechanisms to trigger the release of dopamine, and that’s what gives you that stimulative sensation, right? But if you think about real-world relationships—having a wife, having a long-term girlfriend—those are fundamentally about serotonin, not dopamine.

The dopamine wears off very quickly. What are you left with? You’re left with establishing a long-term structural relationship with somebody and the ups and downs of that. It’s not always peachy keen. That’s the same thing with friends.

I think the other thing that’s really interesting to note is that we’re sort of all in on dopamine, dopamine, dopamine. Young people, particularly the ones who are most susceptible to this, I’m going to guess, have not had enough structural interactions and the kinds of relationships that actually create these long-term, serotonin-like behaviors, so that you can balance this stuff out.

I suspect that this thing is going to catch on like wildfire, in part because it’s taking advantage of a trend, I think.

Yeah, it’s a quick hit as opposed to the long-term gain of deep friendship. Free, what are your thoughts?

I was in the college dorms in 1996. There was a woman in the room next to me who was on AOL, and she got stuck on it 24 hours a day. She failed out of college. She completely disappeared into this rabbit hole of IMing with random people online. It was the craziest thing.

I don’t think ChatGPT is necessarily revealing. I think it’s just this challenge with psychosis at a much larger scale, where you get drawn into these holes. There’s a study by a researcher called Julian Holt-Lunstad who took 148 other psychological studies to identify the relationship between social interaction and mortality. Strong social relationships increased survival odds in humans by 50%, which is equivalent to quitting smoking and more impactful than obesity or physical inactivity.

I think there’s a real struggle that people have with loneliness that manifests in a bunch of different ways. Psychological studies have shown that some of the key motivators for going online to engage with social-interaction systems are self-expression, recognition, curiosity, and emotional regulation. You can test ideas and engage cognitively in ways that you can’t, or don’t, have access to in real life.

There’s a whole lot of motivating factors for people to go deep in social engagement online. There has never been an infinite social-engagement system until AI came along, until these chat interfaces came along.

So rather than having to go find someone on AOL IM, find someone in a chat group, or find someone on a bulletin board, you have this infinite personality you can engage with. I actually asked ChatGPT this morning, “Would you be my friend?” And it was like—

Well, I think the challenge is, once you end up in these deep rabbit holes with a chat interface, there are things that happen with the way LLMs operate that can drive you into a warped reality. You’re now so fully engaged and captivated by this reality with the chat interface, but the chat is not human. It doesn’t have the same wiring or the same triggers that a human might have to realize it’s going down the wrong path.

When I asked ChatGPT what drives some of the psychotic breaks that happen in people who overuse chat and end up in these deep rabbit holes, number 1 was feedback loops in training or operation. If the AI is trained on its own outputs without safeguards, then small inaccuracies in the training system can compound. Over time, this causes responses that, as it said, “spiral away from factual grounding, similar to a human stuck in a self-reinforcing delusional loop.” Suddenly, the AI becomes delusional, and you become delusional with it because you get taken along with it.

What is that called? Folie à Deux—is that the term for the second Joker movie, where it’s the psychological phenomenon of 2 crazy people reinforcing each other and spiraling out of control?

It may be, but I think this is just the AI taking you on this journey on a one-sided basis. As a user, you just get taken along as the AI compounds away from reality. It’s crazy.

The second is what’s called context poisoning, where long conversations or poorly controlled inputs in those conversations can actually push the model’s internal representation away from its starting state. This causes off-track reasoning, but if you’re deeply in it, you’re 16 or 80 hours into a dialogue with the AI, you go with it. I think these are the things that are causing these breaks.

I don’t think the motive for human social contact is new. I don’t think the digital interface is new. But I do think there are 2 things that are new: the infinite engagement you get with a chat interface, and the way LLMs operate, which can actually cause these breaks that take people with them.

I have personally seen it. This is not some New York Times report. I’ve actually personally seen it with people I know who have gotten caught in these loops and been broken mentally. It’s really sad, so it’s definitely an issue, and I think they’re going to need to put safeguards in place.

Sacks, do you have any advice for these losers who are creating relationships with their GPTs? Could you just break the ridiculousness of this, please, with some Sacks wisdom?

Well, yeah. I mean, look, this whole idea of AI psychosis—I’ve got to call the whole concept into question. What are we talking about here? People are doing too much research. They’re stuck in a loop doing 12 hours of research on AI.

This feels just like the moral panic that was created over social media, but updated for AI. The media is going to love it because they’re afraid that AI is going to do to them on the creative side what social media did on the distribution side, which is replace them. The media is going to love this concept and promote it. The trial lawyers are going to love it because it gives them a new basis for suing all these companies.

There are going to be individuals out there who say, “Oh, I have that,” and they’re going to glom onto this message. You’re going to see more and more people claiming they have this because it’s going to draw attention to them.

But the reality is that if you look at that Time magazine story, which is trying to brand this concept of AI psychosis, the game is given away by this doctor who says, “I don’t think using a chatbot itself is likely to induce psychosis if there are no other genetic, social, or other risk factors at play.” This was Dr. John Torous, a psychiatrist at Beth Israel Deaconess Medical Center.

What he’s saying there is that if you don’t already have a predisposition to psychosis and you’re not suffering from extreme isolation or loneliness, then you’re not going to get this. In other words, this is just a manifestation or outlet for preexisting problems.

I think it’s fair to say that we’re in the midst of a mental-health crisis in this country. But if you look at the charts on this, it began long before ChatGPT kicked off the AI race at the end of 2022. You can see that the mental-health decline is affecting all ages, but it’s most acute among the young.

What happened then?

Right. Correlating with COVID, the lockdowns, and that whole huge societal disruption that we now want to forget and paper over. That’s what seems to have triggered a lot of these mental-health declines, which makes sense. If people couldn’t go to school for a couple of years and they were in lockdowns, and it disrupted social formation and the development of young people, obviously that’s going to have a bigger impact.

Nick, if you could pull up those FT charts, what you’ll see is that there’s been a big drop in conscientiousness among young people, a big increase in neuroticism, and a big decrease in extraversion, for example. But where this really took off was starting around 2020.

And there were video games before that, and before that, and, you know—

Violent movies before that, and, you know, there are always rap lyrics.

Yeah.

Rap lyrics, yeah.

Yeah. Or what were the lyrics where they were accused of being demonic? Judas Priest.

Twisted Sister. Yeah. Judas Priest or Twisted Sister.

Marilyn Manson. All of them. So there’s always something. But I will say—

Even the Beatles were outlandish, and they were a threat to societal fabrics back in the day.

Yeah.

Yeah. But even just comparing AI psychosis to the moral panic over social media, I do think it’s probably undeniable that social media has had an influence on society. What happens is you get on Instagram, and people are posting this best version of their lives, so it creates a distorted sense of reality, and people can sort of buy into that.

You know, there is something to X being addictive—

And Bluesky now. People spend too much time on it, right?

Yeah.

So, look, I’m not saying that these things don’t have any negative consequences. I just think that the panic about them is overblown. But if you compare this so-called AI psychosis to the effects of social media, again, I think it’s relatively benign because it’s something that people are doing to inform themselves about things. They can go down rabbit holes, but it doesn’t really trigger feelings of envy or inadequacy in the same way that social media can.

It seems to be hitting 2 groups of people. It’s the same group of people who get catfished—lonely people who are looking for a connection—and then there’s the conspiracy-theorist, delusion-of-grandeur group. Maybe they’re on the borderline with mental illness. Those seem to be the 2 groups that are getting sucked into this.

But it is related. You were talking about the social fabric, Chamath, maybe in our group chat earlier, and you had some thoughts on that. Maybe we’ll just pivot into the social fabric story. This went viral: “Estimated percentage of 30-year-olds who are both married and homeowners.” So, 2 of the American dreams, dare I say: getting married and starting a family, having kids. In the 1950s, half of 30-year-olds were already married and already had homes, and now we’re down to 12%. There are a lot of factors here, obviously. Chamath, what are your thoughts generally on this?

I started to do some research on this, and it’s really sad, actually, what’s happening right now. If I had to frame it, that chart that you showed is more of the outcome. You have to go all the way back and try to explore what the potential root causes are that cause this kind of outcome, and there are a bunch of them.

I think one of the first ones that we have to look at is what has happened to young men. What’s happened to young men is a little concerning. I don’t think it’s our generation, but if you talk to people 30 and under, you start to hear a consistent theme from guys. First, they don’t really have the social skills to talk to girls. It starts with that. So what happens is they go to these different outlets. They go to video games, and they communicate more online. As a result, they tend to have brittle or fragile emotional and social relationships. They don’t date the same way that they used to.

Then a lot of guys also feel like they don’t want to get caught being perceived, post–Me Too movement, in any way, shape, or form as being untoward. You take all of these factors and put them together, and there are a lot of guys who have just opted out. They move in the direction of parasocial relationships instead.

Then I started to look at the data on porn usage. It’s out of control. I had no idea. Do you guys know what the average age of a male porn viewer is now?

That’s scary to think about. The average age—would it be the average age of an American man, or does it skew younger? I would assume it skews younger.

It’s gone from 20 down to 16. More than 50% of 12-year-old boys now watch porn.

I was shocked. I was like, “What are you even talking about?” I was a teenager—late teenager, I think—when I first encountered it. We had to find a magazine here or there, whatever.

They got a girlie magazine, yeah. But can you imagine what it does to a 12-year-old boy when they start to get instant gratification, and you compound that all through high school, when they don’t have what we would call “game”—the ability to just talk and carry on a conversation? What do you think happens?

Then you go into the dating pool. What are these dating sites optimized for? These dating sites are optimized to keep people there, not for you to find a relationship and churn. The gamification and dynamics of those systems are where many young men and women now look to find a relationship.

So you have lonely young men, which creates lonely young women, and a small cohort of very eligible men and women that everybody is chasing and fighting over, with everybody else left on the sidelines. What do you think happens over 15 or 20 years of this activity? It’s what you see in this data. People aren’t coupling up. They’re not having long-term committed relationships, so obviously, as a byproduct, they’re not getting married as much. Obviously, as a byproduct of that, they’re not combining 2 incomes. They’re still living on 1 income.

At the same time, Nick, I just pulled this data. You can show it. What has happened to the underlying things that people would look to as satisfying markers of progress in life? A home, a good education—those are way out of reach.

If you look at this chart, the price-to-income ratio of a home from 1980 to now has effectively made it so that you have to make an enormous amount of money as a single individual to afford a home today. If only a small percentage of people are actually getting married and bringing 2 incomes together, you see what you’re seeing today in that chart.

I don’t know. I think we have an issue in the American fabric that we need to figure out. I don’t exactly know what the solution is, but it’s taken me this long. Maybe this is what all these young people have been screaming about and I’ve ignored. I apologize for not really understanding the gravity of it, but I think this is a big problem, and I think we need to start talking about it a lot more.

Here’s a manifestation of it. If you look at the past-year prevalence of major depressive episodes among U.S. adults in 2021—pretty recent data—look at that spike for 18- to 25-year-olds, Chamath, and then 26- to 49-year-olds. The older generations maybe aren’t having these massive depressive episodes. Friedberg, any thoughts on the American dream changing and young men?

Yeah, I think there are a lot of conflated factors here. They can certainly be related, but there’s just a lot to unpack in all that. I can comment on one piece.

I think we talked a little bit about getting lost online. I mentioned that when I was in college in 1996, that woman got lost in AOL. We had dial-up internet. It was a very different world from today, where we’re all connected 24/7 on our phones. I’m sure we all have average phone usage of 7 hours a day today.

There are a lot more rabbit holes to fall into that reinforce this isolation. It doesn’t actually solve your isolation or loneliness problem when you get stuck on a phone or stuck on the internet. It magnifies it and makes it worse, because you start to see all the things you don’t have and start to feel worse about yourself. That leads to all these difficult social contexts that Chamath is talking about.

With respect to affordability, as you know, I think that’s a distinct point, which is related to our expectation that the government can and should always solve our problems for us. As a result, we created the federal home loan program, much like we created the federal student loan program, both of which I believe created bubbles in those markets.

We created this concept that every American family should be able to afford a home, and our solution was to pump a bunch of money into homes through the federal home loan program. That ultimately led to bubbles of unaffordability for homes.

If we really want to solve this today, besides solving the government spending, inflation, and dollar-devaluation problems—which obviously require cutting spending—I think it’s going to need to be some set of rules around making sure that homes are not being bought up by institutions. That’s one point I think is going to be pretty popular and is going to come up in the next year or 2, particularly in the next election cycle.

There are 2 sides to that one. I think you’re kind of dancing around the issue, and I think it’s very real, which is the financial piece of this. Friedberg, boomers have just absolutely screwed these couple of generations. If you look at the cost of homes, as you mentioned, the cost of education, and then you look at real wage growth, the big problem in the United States is that wages have not grown, except for people who already have a lot of money.

If you wanted to start a family and you’re $100,000 in debt from your student loans, and homes have doubled, tripled, or quadrupled in price when compared to your income, and we don’t have real wage growth, that’s an issue. I think this will become the issue of the next election. We see Mamdani and the socialist movement; we talked about it last week. If you really want to make a compelling argument for the next election cycle, I think it’s real wage growth.

If you look at something like the minimum wage and the disparity between what top earners make and the minimum wage, here’s a chart of where the minimum wage is in the United States versus other developed countries. There is a real argument here, even though we might all be free-market.

Well, there’s a very different approach to higher education in literally all of those countries except for ours.

It’s affordable, yeah.

Well, yeah. I think we have to call a spade a spade. There was an entire generation, if not 2 generations, of young men and women younger than us—not in their late 40s or early 50s—who were told that they must, not should, not “it’s nice to,” but must go to university. These folks took out an enormous amount of financial support to do so.

Part of the underlying reason was that they were told—and it sounds very credible—technology is coming, and there’s going to be so much automation of all the other kinds of jobs that you may do in a trade school.

You should go to university and protect yourself. Well, lo and behold, it’s turned out to be the exact opposite.

Could you imagine where you could have gone to trade school, been an electrician, been a plumber, built something, been a welder, and made a couple hundred grand? You could have married somebody—she does a good job—and now, all of a sudden, she also makes $150,000 or $200,000 a year. You’re making $400,000, and you have no debt. That is the key: you have no debt.

When you make $400,000 and have no debt, I’m sorry, but that is probably actually better than being a doctor making $1 million. And then there are all these other jobs where you’re not a doctor, you’re still saddled with $300,000 of debt, and you have no way of paying it back.

I think that’s also a really important issue that we have to figure out at some point. There was a framing of the issue for an entire generation that was probably wrong. If I had to bet, I would bet that engineers are more likely to have their jobs replaced in the short term than welders and plumbers.

Okay, Sacks. We’re talking here just about the American dream. Is it broken for these next generations: low wages, high debt, no ability to buy a home? This came out of New York this week. Pull up that chart: slow gains in jobs in New York City for the first half of 2025.

It feels like something’s happening here with jobs. We talked to Vice President J.D. Vance about this, and he had some concerns about jobs for recent graduates. This is New York City: only about 1,000 jobs were added in the first half of this year. I don’t know if that’s specific to New York, but we had 2 months of data revised where we clearly have some issues with jobs here, and young people are clearly having a problem economically. They might be upside down. What are your thoughts here, Sacks?

Possibly different things are happening here. I’m not sure how a specific data point around a few months of jobs reports in New York City is really relevant to a trend that started in 1950. Are we talking about this? This is the chart that went viral on X over the weekend, and a lot of people are commenting on why this is.

Starting in the 1950s, about 50% of 30-year-olds were both married and homeowners. By the early 2020s, that number was about 10%. That’s the big societal change that’s happened, especially over the last 30 years. The question is why, and I think there are both economic factors and social factors.

You guys touched on some of the big economic factors. One is the affordability of housing, which I think is mostly due to the fact that we’re not building enough. That’s due to the fact that in large urban areas—I mean, in these blue cities—you can’t build anything because the cities have tied everything up in legal and regulatory red tape. That’s the whole point of Ezra Klein’s book Abundance: to try to get liberals to change their point of view on that.

You’ve got the affordability of housing, and then you’ve also got the college scam. You’ve got these kids graduating from college with $200,000 or $300,000 of debt because the cost of college has skyrocketed over the last few decades. That was not true 30 or 40 years ago. You could get a decent college education without graduating so mired in debt that, obviously, there’s no way you’re going to be able to buy a house and start a family and do that whole thing.

I think there are these important economic factors, but at the same time, there are a lot of these social factors that you’re talking about. It does seem like young men are less self-sufficient. You’ve got these data points around isolation and loneliness, but you’ve also got these data points around a lack of self-sufficiency. If you could put up the FT charts where they talk about the ability of people to make plans and follow through—

Exactly.

Look at this. There’s been a big change among 16- to 39-year-olds in their ability to make plans and follow through, in their perseverance, in how easily distracted they are, and in how careless they are.

I do think that there is a sense that a lot of these young people just aren’t ready to be the head of a household, to own their own homes, or to start a family. I think there are some good questions about why this is.

I wouldn’t rule out the effect of woke poison running through the school system here for the last 20 or 30 years. If you raise white males from kindergarten to believe that there’s something wrong with them—that they’re the evil group in this morality tale of woke that’s been told—

They’re going to internalize that in some way and—

Maybe check out, maybe opt out.

I think it plays out in a lot of different ways.

Yeah, they opt out.

But I wouldn’t rule that out either. It’s hard to say exactly.

Look, I think Chamath mentioned the online dating sites. I think there is something to that, where the dating sites have made the dating market more efficient in a way and more globalized, but—

But is that a good thing? You used to have a lot of hyperlocal dating markets. Maybe it was inefficient, but maybe that was more conducive to average people hooking up.

I mean, another thing—remember the number of people you guys know who met either a long-term boyfriend or girlfriend or a spouse at work.

Yep.

But if you ask young people today, is that even allowed? That is so verboten. It’s so off the—

It’s not allowed. You’re not allowed to do any of that because of this latent fear, as you said, that there could be an accusation. How do I defend myself? There’s a fear that has seized a large part of our youth.

Yeah. What a great platform this is for either party or a new party to pick up: double the minimum wage, take out all these regulations, build more homes, and make college affordable. Those 3 things are such a winning platform.

David Friedberg, what do you do with the mortgage market?

God, it’s so difficult. I don’t know.

Do you just privatize Freddie Mac and Fannie Mae and let them underwrite?

I 100% think you’ve got to privatize Freddie Mac and Fannie Mae as a first step. Then you have to stop underwriting student debt so that you don’t underwrite the $200,000 degrees.

I believe 100% in ending the federal student loan program. I think it will force a restructuring of the entire higher education system in the United States, which no one wants to hear, no one wants to do, and everyone’s against. But I think it is the only path forward.

We have the chancellor of UC Berkeley and the president of Dartmouth coming to the All-In Summit to talk about the future of higher education. I think this is probably, in my book, one of the top 3 most important issues we should be discussing right now. We have to get rid of the federal student loan program, and we have to force a rightsizing and restructuring of higher education such that access to education becomes more ubiquitous. It’s not this $300,000 debt load that everyone is forced to take on.

It enables a more fair market and a more free market where people get what they pay for. Ultimately, people are going to have to make decisions about what majors they’re getting and whether or not to pay for these colleges.

Or whether at all to have a major—to have the social acceptability to not do that.

Another tactical thing for us is to attack this accreditation scam, where only certain people can give college degrees. If you allowed entrepreneurs to create schools like Joe Lonsdale is doing here—he had to buy an accredited university to kind of back into it—but if you asked a bunch of entrepreneurs, “Hey, lower the cost of this degree from $250,000 to $50,000,” they would figure out a way to do it.

It’s already happening. People are now using ChatGPT and online tools to get their education and doing that around the existing system. Life finds a way. These young people are finding a way. They’re finding ways.

I’ll be honest, I really like Trump’s attack on Columbia, UCLA, and Harvard because I think those were the sacred cows that were always left alone. They were meant to be these signaling mechanisms of elitism, quality, and future prospects. Those institutions ran totally amok. Totally.

They started to do things that just made absolutely no sense. To your point, there was a really good article, I think it was in The Atlantic this weekend, about how the humanities captured science.

Basically, these universities get a large amount of their federal funding by getting research grants allocated to the university, and then they get an administrative top-up on top of the research grant. Historically, at Harvard, that administrative top-up is 60%. So if a researcher gets a $3 million grant at Harvard to work on cancer research, $1.8 million in administrative fees are paid to Harvard University. Ridiculous.

A lot of the universities use those fees to expand and grow their administrative staffing and also to expand and grow their humanities. These aren’t bad programs per se, but it basically meant that the science funding was being used to create what some people are now calling social engineering that these universities have been engaged in over several years.

One of the other key questions about the future of higher education is whether science and engineering should be coupled with the humanities from a funding perspective, or whether those institutions should—

Peter Thiel has a very funny saying about this, which is: anytime you have a concept that you want to fund that has the word “science” after it.

It's not a science. Just kick it to the curb.

Yeah. Physics is physics. Social science—

Political science or social science. What is that? Yeah—

In China, Germany, and Russia, there are science and engineering institutions that are separated from universities where you can go and learn the humanities. I'm not discrediting the value of learning the humanities. If you can afford it, if you think there's value in doing that for yourself and your career, go for it. Good luck, et cetera.

If you're wealthy and you want to engage in that because it's something you want to pursue, society needs that. We value it. Great, go for it. But the problem is that separating the research from the humanities is critical. Sacks, if you want federal funding and nonprofit status for your giant endowment, why not require these schools to increase the number of Americans they're educating and lower the price of that education?

If you want to do it privately, fine. But if you want federal funding, if you want taxpayers to underwrite these incredible institutions, double the number of people you're teaching. Triple the number of Americans you're teaching, and lower the cost by 10% a year until more people can afford it. We have to hold them to some standard here for increasing the number of Americans who are educated, because those institutions are now global institutions. There are very few Americans going to them.

Listen, we've seen with all sorts of goods and products that when the government is paying the bill, the price always rises. You always get inflation because the government is relatively insensitive to cost. Whereas, if you look at technology products, the price keeps going down over time.

There's a really good chart on this. All the sectors of the economy that are experiencing massive inflation are funded by the government.

Yeah. Healthcare, housing, education—where else?

So look, that's what's going on with higher education. But I want to bring one other thing into this. We spend a lot of time talking about higher education, but we don't really spend that much time talking about the quality of K–12 education.

My sense is that over the past 30 years or so, there's been a shocking decline in the quality of American elementary and high school education. I think it's so bad.

So bad.

If you look at the performance on standardized tests, to the extent they even do testing anymore, you see pretty sharp declines in the quality of education. But there are a lot of places now where they're trying to abolish testing. They're abolishing gifted programs and math programs.

This is a big issue in California and San Francisco. They tried to get rid of those gifted schools because of so-called equity. In other words, they want to define the standards down to the lowest because it would supposedly be unfair for some kids to experience the advancement of being in a gifted program.

Then, of course, you get into the behavioral standards. We've all seen these viral videos of fights happening in schools. Sometimes the teachers are in the fights, and sometimes the kids are in the fights. There's been a total decline of behavioral standards in schools.

I remember back in the 1990s, Bill Clinton got a standing ovation in the State of the Union because he said that schools should have uniforms. There's a sense that you should have discipline in schools and high standards. Now, I actually think that's a good idea.

100%: uniforms and phone lockers. No phones in school. You check your devices. A lot of things could be done to raise the bar.

What is this fighting in Brooklyn in a school?

This is what goes on in China.

Oh, look at that. Teamwork. Teamwork. We should do that at the summit. The 4 of us should bounce basketballs and show that we're a team. This would be a good team-building exercise for the All-In retreat.

Remember, the public schools in America are not run for the benefit of the kids. They're run for the benefit of the teachers unions. We saw this during COVID, when Randy Weingarten insisted that the schools stay locked down and closed far longer than was necessary, when everybody else had already realized that these lockdowns did not make sense. Putting the desks 5 feet apart was absurd. It was a placebo.

They fought against reopening the schools. Absolutely. Who was it? There was a guy you met?

It was Randy. It was Randy Weingarten and the teachers unions that were fighting school reopening, and they kept the schools closed for a year longer than necessary. In California, they were closed for roughly 2 years. There's absolutely no justification for that second year, even if there was for the first one, which was highly questionable. Then there was the way they were having 5-year-olds wear masks to school.

Yeah.

I mean, even in our schools, my 5-year-old had people trying to put a mask on him.

It's so sad. And then it was absurd.

When it was falling off, they never said anything. It was always like the nose was out, and you're like, "Okay, this is so performative."

But we talked about this during COVID: What is the damage going to be to these kids? We're seeing it now in a range of things. When we got rid of the masks, my 9-year-old, who was maybe 5 or 6 years old at the time, refused to take it off, and she was scared. She's like, "No, Dad. I can't take this mask off. I don't want to die."

And I'm like, "What are they telling you at school? You're not going to die. Nobody's wearing a mask anymore." Unless you're at the Socialism 2025 conference, where everybody's masked up. I don't know if you saw that, but at Socialism 2025, after they do their land acknowledgments, they mask up. It's a panel of people in 2025 wearing masks.

Wait, is there really a Socialism 2025 conference?

We have to go next year. I think we should host an All-In panel at this. We should definitely do some programming at the 2026 Socialism conference. It's amazing. It's like the Democratic Socialists of America. It's incredible.

Register for the party to get the invite.

We should totally go undercover. We should get goatees and suspenders and go as baristas, and just talk about how we're going to seize the memes of production.

Here's the chart. In current dollars, the average tuition fees, when we went to school, were $2,000 to $3,000 a year on average. Now they're at $10,000, and that's in inflation-adjusted dollars.

The problem is that the socialist mindset says the way to solve that is to get the government to do more and make it more free. The reality is that the reason this happened is because the government began to provide unlimited student loans to everyone.

Let's talk to whoever is running education in the administration about accreditation and crack that issue open. We could invest in a company to do this.

There are plenty of private universities and private schools out there. The problem is they don't all get access to capital. It's not about accreditation. Sure, make accreditation easier, but at the end of the day, you need to make capital more efficient. You need to get the government out of the role of allocating capital to education. That's right. There's actually a really simple fix to this.

Somebody needs to close the loop so that when you say you want to get an art history degree, the answer is, "You can do it, but you need to pay for it."

In other words, what you're saying is that there needs to be a restoration of underwriting criteria for this type of education.

I've said this forever. That's right. Exactly right.

But also, if you did the X Prize and said, "What's the least amount of money you could spend to educate somebody to be a lawyer?" you would get entrepreneurs saying, "Oh, that's a great business." Entrepreneurs don't go after this because the accreditation process is decades long. It's impossible to do because of regulation.

AI is going to be an accelerant here, David, because I do think that with AI, it's a great equalizer. There was a company I was talking to the other day that does AI education, and they're having an absolute liftoff in their business in South America and South Asia.

In these markets, there's a dearth of qualified teachers to stand in front of a class and teach people. By using AI, these kids are now getting ahead of American kids. In other words, they're advancing through AI osmosis because they're spending 12 hours a day on AI.

Right? Well, probably 2 hours a day.

I mean, that's probably what's going to happen. AI is a great leveler because you can basically get all the world's information at your fingertips in a way that's fully interactive.

Personalized education, personalized tutoring. That's it.

But then some trial lawyer is going to end your ability to get that because they're going to call that AI psychosis.

The technical term is adaptive learning. I don't know if you guys drive your own cars anymore, but in the Tesla, they just added Grok to the dashboard.

If you long-hold the Siri version in there, it turns into Grok. I did a thing where I said, “Hey, test my vocabulary,” and it started giving me words. It would ask for my definition, and I’d give it the definition. It would say, “You’re close. This is a better definition.”

On my rides to and from the office now, I’m doing vocabulary tests, and it’s increasing my vocabulary and testing me at higher and higher levels. So, you’re now like third or fourth grade, or what?

My vocabulary has already been off the charts. I guarantee my verbal score would be higher than the three of you. Guaranteed. Guaranteed.

What does AI say about the correct way to pronounce “monetization”?

Well, I’ve been working on that. And you know what? I’m correcting that. Don’t make fun of me because of my dyslexia. I need an accommodation.

I actually personally prefer JCal’s version. I like it a lot better.

Monetization is monetization. I’m into it, folks.

All right, listen. Let’s give some red meat to Sacks. He took the time to show up. Trump sent the National Guard to D.C., and folks seem to be generally in support of it. Two weeks ago, DOGE staffer Big Balls was badly beaten by a group of 8 to 10 teenagers in an attempted carjacking. He truly has big balls because he tried to protect his woman and fought hard.

Trump posted a bloody picture on Truth Social, saying that crime in D.C. is totally out of control. He decided—this is what he said—“If D.C. doesn’t get its act together, and quickly, we will have no choice but to take federal control of the city and run this city how it should be run, and put criminals on notice that they’re not going to get away with it anymore.”

Well, he did just that. On Monday, Trump invoked Section 740 of the D.C. Home Rule Act of 1973, which allows the president to take control of the District of Columbia for up to 30 days. So, Daddy’s home, and 800 National Guard troops are in the city.

If you told me you could send 800 National Guard troops to the Tenderloin and get rid of the fentanyl dealers, I’d be all for it. Sacks, your thoughts on this move? Obviously, some media elites are wringing their hands: “Oh, this is Trump trying to be emperor or king.” But he’s legally in the right, right? He has the right to do this in D.C. specifically. What are your thoughts generally? Is he going to roll into New York, San Francisco, and L.A. next? That seems to be the hand-wringing. Or is this isolated and D.C.-specific?

Well, no. The federal government has the right to run D.C. I think it did until the 1960s, and then you had this experiment in home rule. I think you’d have to say that the experiment has not worked out that well.

Last year, Washington, D.C., had the fourth-highest homicide rate in the United States. Its homicide rate was nearly 6 times higher than New York City’s, which wasn’t great. The homicide rate exceeded Atlanta, Chicago, and Compton. It also has one of the highest robbery and murder rates among U.S. cities.

There have been numerous assaults and robberies on congressional staffers and members of Congress. In May, 2 embassy staffers were murdered. In June, a congressional intern was fatally shot a short distance from the White House. Then, of course, just a week or two ago, you had that attack on the DOGE staffer. I think his real name is Edward Coristine, aka Big Balls, who was mercilessly beaten by a group of teenagers.

A big issue is that juvenile crime is a huge problem in D.C. because there’s just no punishment. They’re caught and then released. D.C. has had a policy of zero bail, which we’ve seen in many cities. What that basically does is create a revolving door at the jail: as soon as someone’s caught, they’re released.

You’ve also got vehicle theft in the district at over 3 times the national average, again because even when they catch someone, they just let them go. You’ve got repeated armed carjackings and violence, again by teenage gangs.

The liberal media’s favorite talking point on this is that somehow violent crime is down in D.C., which is the exact same thing we heard about San Francisco. Remember when Chesa Boudin implemented the zero-bail policy and crime was going through the roof? CVS and Target were shutting down. All the stores were basically locking up the toothpaste.

People could feel the crime, but no one was reporting it because they knew that nothing would happen. I think this is the typical gaslighting we’ve seen before, and no one really can trust these statistics. They all have stories about crime.

There are a bunch of commentators on CNN and MSNBC who have been trying to make this argument about the crime being fictitious, and they’ve been bringing D.C. residents on the air as guests. Every time they try to do that, it blows up in their lap.

Yeah. They try to get these D.C. residents to support this crime-is-down narrative, and then every time, the D.C. residents start telling stories about themselves or their friends being victimized. It’s like, “Oh, yeah, I know this person got shot,” or, “This friend of mine got carjacked.”

It’s become comical now, and a lot of these videos are going viral.

This is what’s good about the murder statistic: you can’t make that one up. Murder—

—harder, too. Right. Exactly. You can’t cover up somebody getting shot.

We had the exact same thing happen in New York. My brother was a cop in the 1990s. We had stop, question, and frisk. People called it stop-and-frisk. People were overwhelmingly in favor of this.

Now, people shouldn’t be randomly stopped, and we all have concerns about that. But crime was so bad in the 1980s in New York, when I grew up there, that in the 1990s people were like, “Yes, stop, question, and frisk anybody hopping a turnstile.” That’s how we got all the guns off the street. Then New York became incredibly safe under Giuliani 1.0 and Bloomberg, and we got Dinkins 2.0—de Blasio.

So, if you remember the theory back in the 1990s of how you reduce crime, it was the broken-windows theory.

Yeah, which is that you get rid of the entry-level crimes and the visible manifestations of lawlessness and chaos. When you create an environment in which it appears that those things are tolerated, you’ll end up getting more of it, especially by juveniles.

That was the theory in the 1990s, and it absolutely worked. What you’ve got in D.C. is the opposite of it. I think what President Trump has said is, “Look, first of all, we’re going to get rid of the homeless encampments.” So, those are being removed. You’re going to get rid of the graffiti. We’re going to stop the zero bail, so once people get locked up, they’re going to stay in jail.

We’re going to start treating these teenage criminals as criminals. We’re going to punish them. We’re not going to let them off the hook. If they’re committing adult crimes, there are going to be adult punishments.

This is the right direction to go if you want to reverse this trend in D.C., and I think it’s going to have an impact very, very quickly. I think it’s going to work very quickly.

I think Democrats who are denouncing this are falling into the same trap we saw on the border, where they appear to be supporting criminals and gang members, just like they did with the deportations, when they started resisting the deportation of MS-13 gang members. They fell into a trap there, and I think they’re going to fall into a similar trap here in D.C. because I think this is going to work.

I think it’s going to provide Republicans with a very important counterpoint. One of the problems with crime is that it’s a local problem, and the truth is that it’s been very bad in blue cities. Republicans just haven’t had power there to effectuate a change they could point to as a solution.

I think D.C. provides that opportunity. If Trump’s plan works, I think it’ll show that Republicans have the solution on crime, and it’ll hopefully help provide an alternative in these deep-blue cities.

Here’s a chart just to put some numbers behind it. This is gun-assault rates in Washington, D.C. You can see it spiked under Biden, went up a bit in 2024, came down at the end of his term, and then it’s lower this year, in 2025. That black bar is the first half. So, it is trending in the right direction, but it’s still not acceptable. Sacks, I guess the question is—

I just don’t know if we can believe all these statistics. Just last month, a D.C. police commander was suspended for manipulating crime data to make it seem lower than it actually is, and the D.C. police union says this practice is widespread. In the TV show The Wire, remember, they called this “juking the stats.”

First of all, you’ve got the people in charge. The way that lieutenants become colonels is by manipulating the crime data. Then you’ve got the learned helplessness that we saw in San Francisco, which is that once people realize that the jails are a revolving door and nothing’s going to happen, they stop reporting this stuff.

Now, you’re right that the murders are hard to cover up. I think those probably are accurate. But things like carjackings, assaults, shoplifting, or theft—I think a lot of that stuff just goes unreported.

It has become a partisan issue. If you look at the partisan stats, every time it flips, just like the economy, Republicans think it’s a disaster and Democrats think it’s fine. Then, when a Democrat’s in office, it flips.

But I guess the question, Sacks, is this: I think people have a reasonable argument here. Is this going to be a strategy by the Trump administration to go into New York City next, L.A. next, and San Francisco next? Would you agree with that or not?

The question is, what would the legal basis for doing that be? In the case of D.C., this is a long-standing situation where the federal government has the right to run D.C. and did run D.C. for many, many decades. It was only more recently that this changed to this experiment in home rule.

There’s clear federal authority for President Trump to do what he’s doing, but I don’t know if it would be the same for New York and Los Angeles. I tend to think that whole issue is a red herring.

Yeah, I agree.

Yeah, strongly agree. Here’s the Gallup, Friedberg. I’ll let you comment on this, but this is partisan perceptions of local crime. As you can see, Trump’s term is in pink, Biden’s is in blue, and before that, Obama’s.

Republicans thought that under Biden, crime was a massive issue, and Democrats thought it was no issue. Then you go to Trump’s term, and it reverses slightly: more Democrats think crime is an issue under Trump, while Republicans think crime is gone under Trump. Then you go to Obama: Republicans think that crime is a disaster under Obama, and Democrats think it isn’t. So you just have to look at the raw numbers.

But the fact is, homelessness is a major issue here, Friedberg. I’m curious what you think our policy should be on—let’s call it what it is—junkies. This is not a homeless issue. These are people addicted to superdrugs. Should they be allowed to be on the street? Should they be rounded up and put either in a mental hospital or in jail, or just generally not allowed to be on the streets in a major city like San Francisco? What do you think?

I don’t think we should allow people to live on the streets because it’s community property. It’s not for individual use. I can’t leave my stuff on my sidewalk; the city will come and clear it away. Why can someone choose to live on the sidewalk? I think that’s just the rule of law: enforce it.

What should happen to those people addicted to fentanyl when they get picked up? What should happen next?

There was a case that led to shutting down many of the mental hospitals in the United States.

Yes, people were being abused in them. So then we had—

And so the response was to get rid of the mental hospitals. But unfortunately, the mental health crisis is real. We’ve talked about it. The drug addiction crisis is real.

I do think that, for the sake of the community and the sake of the social fabric, creating treatment centers and mental-health facilities to take care of people in need is a very good way to spend taxpayer dollars. In that particular context, I think it’s heartless to leave people living on the street like that, and they should be cared for.

I would be all in favor of—I don’t use the term “round them up,” Jason—but providing access to care and effectively moving them, if they’re not going to move voluntarily, into those care facilities to create recovery for them. I don’t think we should have homelessness on the street.

I 100% agree with you. It’s totally compassionate to take them off the street and not subject them to life on the street. They’re not in a mental state to make their own decisions; that’s the key issue.

So if you had, God forbid, a family member who was on the street doing fentanyl, you yourself—I think all of us would say, “Yeah, pick the person up. Don’t let them live on the street, and if they won’t go to the facility, then they have to go to jail.” I mean, it’s just the nature of it.

Some of you know this, but somebody in my family has struggled with mental health pretty severely and has been in jail. You can’t do anything. You can’t force them to take their antipsychotics. You can’t do anything.

Yeah.

Well, I mean, I agree with what Friedberg said about this. Another case is the city of Austin, which is a deep-blue city within a very red state. They had a very permissive policy toward letting people camp out on the sidewalks, and it ruined downtown.

Then they passed a ballot initiative there and moved the— I guess you’d call them homeless people—to a campground. So, yeah, the best solution would be to move them into treatment or some sort of mental institution. But if they won’t go to that and insist on living outside, then at least, if they’re going to insist on camping, put them in a campground.

You don’t put them in the middle of the sidewalk in public places where it’s extremely disruptive and threatening to everyone else’s safety. Obviously, having people camping outside our national monuments has to be the worst possible place for these people to be.

They have been relocating homeless people to campsites, and there’s plenty of land here to do that.

And when will the National Guard—

Yeah.

—be in D.C.?

They’re there.

Oh, this is perfect because I have to be there in September. Nick, I sent you a picture of what I’m going to wear. This is what I intend to wear in D.C. when I go. That’s what I’ll be wearing.

You’re going to wear his gold chains again? You feel safe enough to bust out the bling?

You look good in that photo. I like the mustache. I don’t know why you shaved it. I was all in favor of it, my friend.

Thank you.

Anyways, I’m going to break this out in September.

Well, here’s the thing: if you actually spend time in D.C., then you’re very happy about these changes. You’re very happy about the National Guard coming in and making it safer. You’re happy about the homeless being moved out.

And this is why Mayor Bowser, who’s the D.C. mayor, is playing it very carefully. She’s a Democrat, so of course she’s denouncing the takeover as authoritarian. That’s not a surprise. But she’s also acknowledging the legality of this by the Trump administration, and she’s been pledging to cooperate.

So she knows that cleaning up D.C. is going to be popular with the residents, even if it offends the groups. I think that’s pretty interesting.

Yeah.

That she sees this is going to be popular enough that she has to play it carefully. I think, politically speaking, that if Democrats keep opposing this, they’re falling into a trap that Trump has laid for them.

100%. If you asked people with families who live in D.C., “Are you in favor of this or not?”—people who actually have to live under these conditions—I’m guessing 80% or 90% would say, “Absolutely, this has to end. If this is the quickest way to end it, sure, send in the National Guard. It’ll be a good deterrent and get a little extra support.”

I don’t understand why the Democrats don’t do anything other than say, “Yeah, this situation is untenable. We’ll take all the support you can give us.” That’s the right answer.

Well, the Democrats—you know, one of the problems here is that the D.C. Metropolitan Police Department is one of the most dysfunctional in the country, and a big part of the reason why is DEI.

The person who leads the department, the chief of the Metropolitan Police Department—

She had spent 23 years on the U.S. Park Police force—

—and then spent 1 year as D.C. MPD’s chief equity officer—

—and then was promoted to being the chief of police.

So basically, they promoted the DEI person to being the chief of police for the whole department.

And when she was asked what the implications would be for the chain of command, she didn’t know what “chain of command” meant.

Oh, yeah, that was a weird moment.

This is an interview that’s going viral right now.

It’s a Karen Bass kind of moment. Maybe this person isn’t up for the job, let’s say. All right, listen, it feels like a bit of a distraction of an issue. It feels like just good policing, and if the place—

I don’t think it’s a distraction. I think it’s core. I mean, look, it—

No, I’m saying the media’s reaction to it, with “Oh, my God, Trump’s going to be God-King, and he’s going to go into every city and impose martial law.” I think that piece is the distraction. This is probably just good policy. If he has the right to do it and the place is out of control, clean it up.

Let’s go on to our next topic.

We’ve got 1 or 2 more topics we want to hit here, unless anybody has something they have to get to. Chamath, you wanted to talk about VC returns versus public markets. I know, Friedberg, you brought this up in a previous episode: that you could just set it and forget it in the Magnificent 7 back in the day. So we’re going to revisit this topic.

Public markets—here we go. They’ve been on a tear: 2023, up 24%; 2024, 23%; 2025, we’re at 10% so far. Here’s just a chart of it. We had 1 down year, obviously, in 2022, and it’s been on a tear. It’s only happened 1 other time before that. We’ve had 3 straight years of this kind of 15% performance. Very rare. You see, we highlighted them there on the chart: the dot-com bubble in the late ’90s, then before and after peak ZIRP, and now.

You can compare that to the private markets—venture capital. Here’s a chart from 2022. The reason I’m going to show these is that when you looked at venture as a category versus, say, the S&P, or a couple of lines up from that, you see that venture was extremely, extremely attractive according to Cambridge Associates, which tracks us. They have the best data because they’ve been doing it the longest.

Then we go and look at the same chart in 2018. Hey, look, the Nasdaq’s doing pretty well when compared to venture, in large part because we had some of these megahits—the Googles, Facebooks, and Amazons of the world—coming to fruition and just going on a tear. And then here we are in 2018, and public markets really start to rip.

Same chart, just a different time period. Here’s 2020. Maybe venture isn’t doing as well when you look at it compared with the markets. Of course, you can’t do this for the last 5 or 6 years of venture because they’re in the J-curve. When you do venture investing, you’re expecting it to pay off in 7, 8, or 9 years.

But the trend here, I think, Chamath, that you were pointing out is public markets: liquid, no fees, low fees, minuscule fees when compared with venture, and maybe a better bet for investors. That’s challenged the venture industry. Was that your sort of point in bringing this up?

I think the venture model is broken, but it’s broken for 2 different reasons that are not totally correlated to each other. The first is purely mathematical. When the public markets provide a consistent rate of return—it doesn’t matter what that is, but let’s just say you average it out and it’s 15%—the question is, what does a private investment have to give you so that it’s worth putting it into an illiquid instrument where you don’t get your money back for some number of years?

Problem number 1 is that it used to be you could get your money back in 6 or 7 years, and the fund would liquidate after 10. Now, you typically don’t start to see returns until year 10, and oftentimes year 12 or year 13, which means that they don’t really pay out until year 16 or year 17. What the math of that would tell you is that you need to generate somewhere between 6% and 10% more in terms of the gross return so that, on an apples-to-apples basis, you’re being compensated for all of that illiquidity, right?

God forbid something happens, or you want to buy a house and you need the money. In the stock market, you just sell it instantaneously. In a venture investment, you’re going to call somebody, they’re probably going to haircut you by 30% to 50%, and you’re going to sell it because it’s illiquid. So, if the public markets are giving you 15%, the reality is that you need to get 25% plus.

The question then is, what venture investors generate 25% plus IRRs over a 17-year period or a 15-year cycle? It turns out that every fund has a very low statistical probability of doing it. Every fund manager—if you’re in the top handful, we all know who they are—will always at least have 1 that does it. So the problem is, you have 1 that does it, then the next one that doesn’t, then the next one that’s even worse, and maybe 1 that comes close. There is no systematic approach to the return stream.

How should you look at venture capital, I guess, is 1 question. I think the way that one should look at it is as a portfolio where you have a lot of liquid assets, and it’s a small piece that’s essentially a mechanism to learn about the future and to understand how your other assets are going to perform. I think it makes a ton of sense, but if you’re trying to crush it, I think the only people that will really make money in venture capital are the GPs building a fee-generating business around it.

Now, that then gets to the second question: if you do invest in venture, what’s the next big problem? The next big problem is, in a world of AI, do you see the amount of money that’s required to build a really big success shrinking? If that happens, how much money can you actually get working?

Companies like Midjourney are a great example of a business where these guys have a 40-, 50-, or 60-person team generating hundreds of millions in revenue and are very profitable. They’re not going to be raising an enormous amount of money. I think these 2 things are very complicated for venture. It doesn’t mean that people will stop investing, but—

Yeah.

Friedberg, do you want to chime in on this, I think?

Yeah. So I pulled some data together. Nick, if you could pull up the basics. I know we talk about this a lot, but I thought it would be good to show the difference between a normal distribution and a power-law distribution.

I don’t like to call it venture returns, but I do think returns generally in free markets create value in a power-law distribution. That means a few of the many account for the vast majority of the capital appreciation, or the value creation, and that’s because of the power of compounding. If you build a better engine of technology, a better business engine, you will compound, and that means that over time you will accumulate more and more of the market in an outsized way.

As you accumulate more of the market, you actually move faster in accumulating more of the market, and eventually it becomes a runaway flywheel. No one can catch up. That’s the key return profile in free markets, generally speaking. A manifestation of that is that it’s also the return profile in venture, in technology venture investing, because when you find 1 or 2 great hits—the Uber that has a great network effect, or Airbnb, or Google—it has a runaway effect in the market, accumulates all the capital, and becomes really valuable. You can see this in the returns.

If you pull this up, this is the latest Carta data. This shows the IRR by the vintage in which the fund was raised, and also by the size of the fund and by how the fund has performed on a percentile basis. Look just at the 2017 row. The top-decile funds are doing, call it, 30%, versus the median fund.

Hold on. They’re not—these are all markups, paper markups. These are not distributions. These are all numbers.

Okay, fair enough. But I’ll make your case in a second here. These are not DPI, so it’s not distribution cash-out. This is just marked-up value, and this is 8 years old. This is the 2017 vintage.

But just the difference between, call it, a 30% IRR and a 10% IRR compounds over 10 years. The multiple difference at the end you should expect is the difference between making 14 times your money and 2.6 times your money—the difference between 10% and 30%. That’s kind of what this shows, roughly, is that the difference between a 50th percentile and—

The problem with this data, though, is that this is Carta’s product. It’s used by a small number of people for a small number of years. That’s why the other data we showed from Cambridge is much better, because the 2018 forward are still in the J-curve. So, to Chamath’s point, they’re paper, so I knew we’d go there.

Okay. I knew we would go down this rabbit hole, so I shouldn’t have shown that. But no, that’s okay. There are 2 key points I wanted to make, so just pull it up again, Nick.

The first key point I wanted to make is that, because of the power law, the job of investing is to find the power-law winners. It is not to buy the index. If you buy the index, you are losing to the market. Even if these numbers are correct—which they’re not, as we’ve talked about—you’re losing to the index if you’re just buying the venture index. It’s probably way worse than this. You have to get the winners.

What this does show, though, is that the smaller the fund, the less diversified it is, which means the more likely it is to capture the winners. You can actually see that the smaller funds generally have better performance than the bigger funds.

For sure they do, because they’re getting in earlier as well, Friedberg, and they’re paying a lower valuation.

Also, they’re not getting diluted away by the index. The index generally is going to return negative. It’s going to be a negative returner.

I think I told you guys I went to an LP conference years ago. One of the biggest venture funds showed that 45% of their capital went into flat or down rounds, and they had a negative net return on that allocation of capital across 13 funds. If they had never invested in flat or down rounds, they would have doubled their IRR overall as a fund.

The really important point is that the power law matters. Your job is to just find the winners, and if you find the winners, they’re going to compound. So how does that translate? Go to the next slide, Nick.

This is the one that I shared in 2023 that Gokul Rajaram published on Twitter, and that’s where I first saw this. If you bought just the top 10 companies in the Nasdaq and held them, you would have made a 24x multiple over a 24-year period, versus if you had just bought the Nasdaq. So just owning the best companies in the Nasdaq is by far the best way to drive a multiple of return over a 10-year period. You made a 9x multiple—9x—just by owning the top 10 companies in the Nasdaq. This was through roughly the end of 2023.

Now let’s pull this up. I pulled this analysis together this morning for our conversation. This shows that the venture returns don’t stop when you stop being a private company. The real return—and this highlights the point—is that most of the value, when you find that power-law winner, is created when they are a public company.

Palantir went public after 17 years, and its market cap was $16 billion. So they created $16 billion of equity value over a 17-year period. Since they’ve been public for 5 years, they’re now worth $436 billion. So they’ve created another $420 billion of equity market value in just the last 5 years, probably eclipsing all venture returns that have been made during that same period of time.

Airbnb went public after 12 years, worth $47 billion. In just 5 years, they added another, call it, roughly $30 billion. Uber went public at $75 billion after 9 years as a private company, and now they’re worth $190 billion. So they’ve added $120 billion of value in just 6 years since being public.

Spotify is another good example: a $27 billion market cap when they went public after being private for 10 years. Then they added another $120 billion of value in the 7 years they’ve been public.

I've got the whole list here. These are all the companies that have a $100 billion-plus market cap. When you find the power-law winner, the value continues to accrete. It continues to compound, and you're better off just buying the public stocks.

I think that was my big conclusion as I went through all this data.

Chamath is exactly right. When you're in the venture market, you learn a lot. Ultimately, you probably want to learn a lot so you can allocate your capital more wisely into the power-law winners you've identified, where most of the value continues to accumulate, whether they're private or public.

Having an IPO is just a transitional event for these companies. Their compounding engine will continue as a public company if you've identified them. Facebook is such a great example: 8 years as a private company, a $100 billion market cap at IPO, and it added another $2 trillion-plus in the years since.

Friedberg, you've cherry-picked all the winners in hindsight—the biggest winners from that list. What if you were just to create a list of every

IPO.

Tech IPO and how it's fared afterward.

Right. No, it's a fair point. There'd be a lot of stinkers in there that you haven't included, but my point, Sacks, is just to find the power-law winners, and that's where all the value—

Improves. Well, I mean, that's hard, too.

Here's the thing: I think you have to look at how we got here. I started doing this in my second decade, and companies decided to stay private longer. That was the big trend. Then there were a lot of reactions to that. Secondary transactions started occurring, and you just couldn't get DPI. We had to hold it forever, and now the industry is responding to this.

Now we have something called strip sales. We have people creating continuation funds—ways to take those early investors and get them out into a new vehicle or create some liquidity. And Chamath's point is really important, because you do get so much intelligence.

What I learned, having been an investor in Uber and Robinhood when they were $5 million and $20 million companies, was valuable. I was the first investor in those companies. When Uber crashed in the market and hit $30, I bought up a bunch of it as a public investor, even though I had a ton from when I was an angel investor.

I did the same for Robinhood at $12 because I knew the management. I knew Vlad, Travis, and Dara. I understood the market. I did the same thing with Facebook, having listened to Chamath and Brad Gerstner talk about it all these years. I backed into it when it was $92.

What's happening now is what Roelof did with his continuation fund and holding the private ones. It's also the theme song of this podcast, famously: When Zach said, “Let your winners ride,” that is the overarching thing here.

The venture industry is in a massive transition. It's going to start looking a little bit like private equity and a little bit like Gavin or Brad, where they're private- and public-market investors, and you're seeing buyouts occur. People are saying, “Hey, this SaaS company is undervalued. We're just going to buy the company and start operating it.”

So venture is in a massive transition. But I think what comes out on the other side is something that looks a lot more like public-private investing, like Roelof pioneered at Sequoia. We're going to talk to him about that.

I'll just respond to your last comment. Jason, I appreciate you saying that, because Roelof is coming to the summit to talk exactly about this, which is going to be great. But that is the job of being an investor, to your point. You're saying, “Hey, you cherry-picked the winners,” but I think that's the point.

The returns accrue to these power-law winners, and the job of being an investor shouldn't be to index a market. Anyone can index a market. The job of being a great investor is to find those winners.

The question is how hard or easy that is. I think your list makes it sound like all you have to do is wait for these companies to go public.

Sorry, that wasn't my point. It kind of came across that way. I was just saying that the power-law winners continue to accrue value. In fact, not only do they accrue value as private companies, and if you get one of them you have these 14× funds, but as public companies they continue to accrue value.

This isn't just limited to being a venture investor. Pretty much anyone has access to the public markets and can make the decision that they've made a bet on a power-law winner. There were a lot of people who were very vocal supporters of Palantir, Uber, and Spotify when these companies went public, and a lot of people said no. They were the naysayers.

I think the problem you would have had trying to implement this strategy is that it wouldn't have been clear what you were underwriting. What would you have been underwriting at Nvidia—not just for the last 5 years, but for the many years before then? You would have been underwriting video games.

Video games. Yeah.

And you would have had this very odd data-center business that didn't make sense—something Jensen took a lot of heat for and had to defend. It didn't actually make a lot of financial sense. So I'm not sure.

I'll give you a counter to that, and I'll call him out and make fun of him now. My co-founder at Climate Corporation is named Siraj. He's probably listening right now. He bought Nvidia years ago on the thesis that these GPU chips would eventually be used for AI models, because he did all his work at Stanford and distributed computing. He did early work in AI. He was very close with—

That's different from what you were saying before. I get it. He's a smart guy who made a huge winner. But I'm saying—

But no, the problem was that he sold too soon because he thought, “Oh, they're not actually—” Okay, but I'm just saying there was probably a person as smart as Siraj who bought Intel. Then there was another person who bought AMD. I'm not sure what the point is. Here's the point about venture that I think is worth noting.

It's an exceptional market that gives the practitioners an incredible edge if you build a business around the information asymmetry that occurs.

Yes.

That you can monetize. After that, the only business model is building an asset-gathering machine to generate fees, because even the consistency of being able to generate these power-law distributions in successive funds doesn't exist.

You can look at all of the data. Cambridge has done this. A lot of these people have done this. All the big LP fund-of-funds managers, like Horsley Bridge, have done this. Having a killer fund has zero correlation with your ability to then have a next killer fund as an LP. I showed these returns on screen, although we blanked out the names, and I'm in all of them. It is really hard to be consistent.

It's hard, but—

Let me make 2 points about this. First of all, Chamath is right that if you look at the data—the recent data, say, over the last decade or so—you don't want to be the average venture fund. You'd be a lot better off just being in the Nasdaq or in some other public-stock-market index, because your returns would be higher and you wouldn't have the illiquidity. You can always trade out of it if you want.

It's true that you don't want to be the average venture fund as an asset class. The average isn't great. This goes to Friedberg's point about the power law. You really want to be a top-quartile fund, or even better, a top-decile fund. That's what LPs are looking for when they're underwriting VCs: They want to be in the top quartile, not the average. I agree with that point.

But the second point is that things can change very quickly in this asset class, precisely because of the power law and how spiky it is. Take Figma. We talked about it on this podcast a year or 2 ago, when the government held up its acquisition by Adobe for $20 billion. Everyone thought that was a disaster because they thought it was a huge premium to the actual valuation. If Figma eventually had to go public as a result, it might IPO at $10 billion, so it would be a big haircut for all the VCs.

A couple of years later, it just IPOed. It had a huge pop on the first day. It's come down somewhat, but the public-market valuation is now $35 billion. There are 3 or 4 VC funds that have made $5 billion, $6 billion, or $7 billion.

And those 3 or 4 funds are now 10× funds. In the blink of an eye, with 1 outcome like that, it totally changed the performance of those funds. Then that's going to trickle down to the performance of the asset class as a whole.

My guess is that we're in a pretty good time for VC because of AI and the disruption it's causing, as well as the potential for value creation it's creating. It's restocking the pond with opportunities for many more big companies to be created, and I think ultimately that'll be very good for the asset class.

Just look at OpenAI. We're a long way off from OpenAI being liquid, and you don't want to count your chickens before they hatch. But just 2 years ago, on the heels of ChatGPT launching, the company raised money at a $30 billion valuation. Everyone thought that was crazy at the time. Now they're at a $300 billion valuation, and there are news reports that they're going to raise at $500 billion.

So if this keeps running—and again, there are a lot of ifs there—they could still lose.

Sell 20% is my best advice.

They could still lose to some other company or whatever, but it just shows you how fast things can change. They could be a trillion-dollar company in a couple of years if the trend continues.

My point is, because of the power law, it only takes a few big winners to change the economics greatly for particular funds, and then that eventually trickles down to the mean. But again, you're right: You don't want to be in the mean. You want to be top quartile.

There's just 2 quick points I want to make as we wrap, and great episode, gentlemen. The first is also the reign of Lina Khan occurred for 4 years, so DPI and exits were muted. That also plays into a lot of these statistics. And now, since President Trump, number 47, David, we have had a tear of M&A and a ton of IPOs, right? That's actually going to be probably a big factor in all of this hand-wringing around venture.

I just want to give 2 anecdotes here. When you have the inside information, Chamath, it's worth, I think, being a venture investor because of that data you get. I was always perplexed when I would hear people on CNBC talking about Uber and Robinhood specifically, because they would say, “Oh my God, they're losing so much money. They lost $1 billion this quarter. They lost $5 billion last year.”

We all knew that there would be no difference in consumption of an Uber if it was $13 or $15, and that at any point you could just turn the dial, slow growth, and show profitability. And what happened? That's exactly what they did. I made that call, and I said this on CNBC: Is anybody not taking an Uber for a $1 or $2 difference per ride? It's like 2% of people might stop taking it. Sure enough, they did that.

Then fast-forward to Robinhood. Everybody's like, “Oh my God, this company is sideways. The market's corrected; people are no longer trading stocks.” And I was like, “But they keep releasing exceptional products over and over again.” That was always Vlad's key perspective: “We're just going to keep adding every product—529s, 401(k)s, margin loans, crypto. We're just going to keep adding products every 6 months, and then we will be the default place where all your finances exist.”

When it was at $12, I was like, “Oh, well, this obviously is undervalued. I'm just going to buy more.” Now it's at $120. It's a 12-bagger in the public markets.

For that reason alone, if you just have enough surface area and enough patience, you're going to hit Robinhood and Uber, like I did, when they're tiny companies. Then you're going to get to have all that intelligence into the public markets, which is why I started doing public-market investing: because I could see all these very clear opportunities where things were obviously undervalued. It was just so obvious.

When you talk about surface area, what you're talking about is diversification.

Yes.

Right. But the way Friedberg just showed us the winners—

Cherry-picking. Yeah.

Yeah, and that's kind of my point: It looks really easy when you just look back over the last 10 years and pluck out the winning companies, the winning funds, and the winning asset classes. The question is, Chamath, how do you create enough diversification to get exposure to that? That is why venture is still a popular asset class: because it is very spiky, and you do want to be in those power-law funds and power-law companies.

It might be a percentage. The percentage might change. Maybe Yale wanted to get to 25% or whatever, and maybe the right number is 15%. It could change over time based on your own individual needs. I still think it's a magical asset class. Chamath, you want to wrap us up?

I did that job for 7–8 years with outside capital, and now the last 7–8 have been entirely my own. The difference, I would say, is that it's much, much harder to run a very successful venture firm in my experience. It's much easier, actually—

To be concentrated and to use the information asymmetry in your favor. It is an exceptionally difficult job to be a successful fund manager, like what you guys do. I think it is extremely, extremely hard to be of that quality.

Well, we both started our own firms, which just increases the degree of difficulty, because you have to build a team and you have to build an LP.

So if I could own pieces of the GP of both of you, I would. But as a person today, I would not start my own fund. It's just very difficult.

Well, it's a very competitive industry, and there is a lot of capital. Very competitive.

It's super competitive.

But I would also say that—

And that's the supply of capital. If you look at demand for capital, I do think that there is more legitimate demand for capital now than there has been in a long time.

Real businesses want to be funded. I agree with that.

Well, there was a feeling 5 years ago, let's say, where all the big platforms were getting a little bit long in the tooth: social, mobile, cloud, SaaS. It felt like, again, that the pond had been fished out a little bit.

Yeah.

And you need a big disruption to restock the pond. I just think that—

AI is the mother of all disruptions, and it's creating a lot of opportunity.