GPT-6 Hits AGI? Tech Euphoria 2.0, SF Mansion Shortage, NYC Bans AI in Schools & Venezuela Oil Deal

Happy birthday, Chamath — we're taping on your 50th. Super exciting for you, brother.

Thank you, it's really great. I got a birthday letter from the big man. If you had asked my parents 50 years ago whether their kid would one day get a letter from the sitting president of the United States, they would have said absolutely not possible.

Wow, look at that. Which White House intern wrote that? I wonder.

Speaking of interns, this one was actually family — my son. Braden interned at the White House all summer, and at the end of the summer he used all the political capital he'd built up to put in a request for a letter from the president — and he pulled it off. I obviously want to thank POTUS for sending the letter, but I really want to thank my son, because it was a really thoughtful thing for him to do. I'm very appreciative of that.

That's genuinely impressive, because getting something signed takes time out of the president's schedule — that's why it means something.

All right, listen. Breaking news.

Breaking news here on the All-In podcast: OpenAI is rolling out ChatGPT 6, aka Astra, its new flagship model. It will be available to a limited set of organizations in the coming days, and soon to anyone with a credit card on their ChatGPT account—that is, Plus, Pro, Business, and Enterprise users. Greg Brockman says he believes OpenAI has entered the AGI era, a topic we've discussed here before. Sam Altman, meanwhile, is back to being a bit performative/alarmist—or he sincerely believes this is the one. At Wednesday's G20 summit he said:

Much, much, much—three muches—more capable models are coming soon. The next generation of models are going to be sobering for everyone. I think no one intellectually honest can look at what's happening and not feel the weight of responsibility in front of us.

A little drama there. People are freaking out about the benchmarks, which are off the charts, and some are saying this is the one. Chamath, your thoughts?

I don't think normal humans know how to read this chart and infer anything useful from it. The more important takeaway: I agree with Greg—AGI has basically been here since the beginning of the year. There are models that are incredibly performant inside the closed frontier lab companies, and right now we're trying to figure out the right cadence for releasing them, which I think is very important. I also think these capabilities will be matched, or close to matched, by a handful of closed-source and open-source alternatives, probably within the next three or four months. So the right way to think about this for most people is that incredible intelligence capabilities are becoming broadly available while the cost of each incremental unit of intelligence keeps falling—a really positive trend. Now it's just the messy middle part: organizing this properly for companies and people, building useful things, and showing actual ROI. I would just keep calm and carry on.

By the way, it is very exciting, and congratulations to OpenAI. I'm sure Anthropic will have something, and Grok still has yet to drop—Grok 4.6 is coming, and that's going to be a blockbuster. The Meta models are really good, and Gemini just dropped something yesterday that was exceptional. The great thing is we have a lot of choice, and the even better thing is that the cost of that choice is becoming cheaper and cheaper.

Sax, let me put it to you this way: OpenAI seems to get its groove back now, after Anthropic was running away with it. Grok and Grokbot are doing things those two other frontier models simply can't do—whether it's the harness, the underlying model, the instruction set, or Cursor's magic sauce, Grokbot is clearly competitive if not better for most users. So it's vibrant competition with leapfrogging occurring at a cadence of, I'd say, two to four weeks right now. This is what you told us you were going to do when you took over as AI czar and set up the architecture here. Can this continue? And what's it going to be like when Jensen gets in here with Laguna and his access to all these open-source models? This could settle it completely—we win the AI race.

Look, there's a lot of innovation going on. American companies are competing very well, and I think we are winning the AI race—as long as we don't do something stupid to shoot ourselves in the foot.

Like what?

I've talked about it a million times: setting up a new AI regulatory agency that operates like the FDA and takes years to approve new models. But I don't think that's going to happen under President Trump—I really don't. Anyway, things are going well. Remember, Sam told us a few months ago, when he posted, that he expected the company to have its best 12 months: the last year hadn't been so great, but the next year was going to be really great. Maybe it was because they were seeing these results in the lab.

And remember what Sarah Frier said: since the beginning of Q2, around the middle of the year, they saw a market acceleration in their growth rate, which I think coincided with the release of the 5.6 models. Now you're seeing the 6.0 models, and it just keeps advancing.

But I still think the market structure right now is two markets. There's the market for frontier intelligence, and that is a duopoly — Anthropic and OpenAI — and it's good to see it's still a horse race with both competing. Then there's commodity intelligence, which is everybody else, including all the open models. They're not competing on being at the frontier; they're competing on price, and those prices are significantly lower. It really is a two-tier market at this point. I agree with you about how promising Grokbot is — Grokbot is awesome.

I would assume both Anthropic and OpenAI will eventually roll out their own versions of bots. They need a simplified, stripped-down version for normies, and they can leave the more complicated co-work products to the top five or ten percent of the audience.

The Polymarket ratings for which company will have the best AI model by the end of 2026 show OpenAI's chances just spiked from low single digits to over 20. People were really counting them out; now it looks like they're viable again. Anything that creates competition is good. Before, it looked like Anthropic was running away with it, and they're about to have this monster IPO. If OpenAI can at least make it a two-horse race, that's good. I think it's great that Grokbot launched — the product is awesome.

Here's a tip: I created a new bot called "improve my bots" bot. I gave it an instruction set: go through all my bots, fill out their skills and instructions, and make them better every day. It went through, took two or three of them, and asked, "Can I put these together into one instruction set? It would be better for you." Then it told me all the changes, and my bots are maybe 30–40% better on average.

So you created a civilization — a multi-level civilization to make everybody better. Your civilization is thriving. That's a really good hack in Grokbot: you tell your chief, "What else should I be using you for?" and then, "Okay, now what do I do?" I was building some products and asked it what we should do next, and it suggested more features. It's a brilliant product — it's called Grokbot, go download it. Shout out to my guy, and congratulations on the robotaxi taking the streets in Austin today.

The market is ripping. Does this feel like '98 or '99 to you? Were you around for that?

Oh, for sure — I was covering it. It feels like '96–'97 to me.

You think it's that early? I'd say '97–'98.

I do think we're seeing, at the late stage, a disconnection between reality and valuations, where companies are being given far too much credit — priced at 50 to 100 times topline revenue. That's obviously not sustainable. The people who invest in those are, in my mind, similar to late-stage investors — unless it's run by some legendary one-of-one entrepreneur. If it's Elon, a 30x on revenue totally makes sense; Travis, something similar, totally makes sense. But unproven entrepreneurs getting a 50 or 100x is a little bit dangerous. The entry price matters.

Did you see Instinct? It's like the new billion-dollar... It makes no sense. It's a personal AI assistant that corresponds with you by iMessage. I think they just got a term sheet — two and a half billion. They're in private beta right now. It's the new hotness; all the VCs are trying it. But still, that does seem kind of crazy.

No, no money went in — that was the valuation.

Congratulations to the entrepreneur — no dig there. But yeah, I want to try it. Let me try it.

I got invited, and I read the terms of service—people were talking about it online—and I immediately did not connect my stuff, because it gets access to everything. Then somebody told me it's a human-in-the-loop service. I don't know if that's true, but I get the sense that there's a human watching it work, taking the instructions and the finished product and polishing the last 20%. That's what I suspect is going on here—a little polish at the end. If you want to order stuff, etc., that's totally fair. I don't want to specifically call that one out, since I haven't even tried it—I'd like to try it—but things are getting a little bit frothy, a little bit euphoric.

Freeberg, if you were to analogize this to the late 90s, where do you think we are in the cycle?

You were at a rave in downtown LA at that time, by the way—you were 19 years old, raving.

I had a warehouse in Oakland at 7 a.m. on my way to the Oaks card club to play 2-4 limit.

Wow, how'd you run? You're like, I love these cards.

J and I are the only ones who were actually working in the dot-com bubble.

I was working in '99. Well, I started in 2001. I got a job offer senior year and wasn't sure if it was rescinded by the time I graduated because of the market tanking that year.

I got a job at Cisco, to be an engineer there.

If you never lived through the dot-com crash—it was heady, then super euphoric, and then super miserable. But pretty much any founder under 50 has never been through a massive cycle like that. I mean, the ZIRP bubble was kind of a cycle.

Here's the thing, though. In the boom, it was all metrics that weren't dollars. What we're seeing now is revenue and profits—and growth in revenue and profits that we've never seen before. That's the Anthropic story, the data center, infrastructure, networking, equipment story. It's all real dollars flowing versus speculative utilization.

You're exactly right. People were basing it on metrics like hits to a website and web pages. Here you have a very distinct difference: you actually have revenue, you actually have products in market. But to put it in perspective, imagine if Anthropic, OpenAI, this cohort of companies lost 95% of their value in under six months, and then the long tail of companies below them literally went out of business—people leaving those offices with their Aeron chairs and their computers on the seats, total implosion. That's what happened in the dot-com bust. It was cataclysmic.

No, I agree. So do you think we just have scar tissue from that and are overfitting patterns every time there's a new platform shift, a paradigm shift?

Whenever we're in a bull market, the conversation always turns to: is this a bubble? So is everything overfitted to that original pattern? Or is it the case that we have a lot of young founders who've never been through a cycle before and don't understand that paper wealth in the billions can disappear practically overnight?

Advice to founders: you're well within your rights to sell 10% of your shares and pay for your loft or apartment. During that period, some people said, "This doesn't make mathematical sense. I'll pare my position in my startup, take some chips off the table, buy my loft, pay down my student loan," and others said, "This is going to change everything" — and their companies went to zero.

I think we're in the early part of the euphoria. Euphoria exists because markets are real; people are just guessing how far forward to price that reality. Bubbles burst when you realize you priced way too aggressively forward — that instead of paying 50 times, you should have paid 15 times or five times. We're going to get to that point, because every market has gone through that cycle, and we'll go through it here too. The question is when. Honestly, I don't see this stopping. We're at the beginning of the beginning, and I think it's going to be like this for probably another three years.

I agree with that. With the Anthropic IPO coming up — I'm sure you've seen that home prices in San Francisco are going through the roof right now. It feels almost euphoric. I'm starting to wonder if this is 1998 or 1999. I think it's more 1998 than '99, because I don't think we've seen the peak yet.

I'm just glad I didn't sell my real estate when I came to Texas three years ago, because I'm definitely selling all of it in the next year.

It is bonkers what's happening right now. You have a lot of San Francisco real estate — what are you going to do, Sacks? Sell it all?

Feels to me like the perfect time to clear it out. I'm clearing everything out in the next six months. When that IPO happens, the clock is ticking. I've got three properties for sale — hit me up, people.

There's an article saying that if you add up every single IPO San Francisco has ever had, the sum total is only one quarter the size of the Anthropic IPO. Extrapolating from that, this IPO will be a wealth generation event four times all previous San Francisco IPOs combined — and then OpenAI is coming on the heels of that. San Francisco isn't a big place and doesn't have a huge inventory of houses, so you're already seeing homes go for $3,000 a square foot, twice asking.

I think the ultra-luxury market in San Francisco could easily go to $5,000 a foot — London-type prices. London, central Paris, central Hong Kong. Of course — the socialist-communist lunatics there don't let you build any inventory, so of course it's going to go up. There's a supply-and-demand issue here.

It could get really nutty. What we're seeing right now — homes flying off the shelves, selling for double asking — is based just on the secondaries, the small cash-outs people have been able to do in anticipation of these events. Two pieces of advice for every founder listening. Number one: you're within your rights to sell 10 to 20% of your holdings now, especially if you're a first-time founder who doesn't have your nut yet. Number two: if you're running a company and thinking, "I'll put off my raise for six months so I can double or triple my valuation," that's a huge mistake. If you have the ability to take in $100 million, $250 million, a billion dollars and put that cash on your cap table at a solid valuation — even one that's 10, 20, 30, 40 times your revenue — take it. Companies with hundreds of millions or billions in cash have optionality, and if things collapse or there's a big retreat, you'll survive through it, like DoubleClick, Amazon, Google, and countless other companies did.

Well, you take the money when it's there. Make hay while the sun shines.

Look, you're right about the financial advice, but there's a big difference between a late-stage startup like Anthropic or OpenAI, where an employee takes chips off the table, and a founder raising a Series A — those companies are about to IPO.

They're within months of an IPO, so I think it's fine for them. SpaceX did organize secondaries for years, but those were late-stage companies. If you're a founder raising a Series A and trying to take chips off the table, I consider that a huge negative signal. That's not why I'm investing in a company — so the founder can cash out. I want them to be all in.

If it's year three or four and the company has 100 million in revenue, tripling year over year, that's different — that's what I'm talking about. But for a pre-revenue company, selling before you're generating revenue does feel like a lack of faith in your own company.

As for where we are in the cycle, it feels like 1998 to me — we're definitely further along than '96. The dot-com bubble burst in early 2000, March or April, so that would put us two and a half to three years out; Shim thinks maybe four. We might be partying like it's 1999 by the end of this year.

The All-In Summit is just over a week away, and we'll be taking over with our friends the Chain Smokers — we're going to party like it's 1999. A couple of tickets are left, with Jensen, Satya, and Shotwell among the big names. It's going to be a lot of fun, and you'll meet a lot of interesting people.

We have a few scholarship tickets left — we like to help the up-and-comers. The bar for a scholarship is very high: we're looking for the most interesting, ambitious, hungry people. If you fit that bill, apply now. Applications are open through September 10th, so you've got one week left.

We try to do a good deed here. If you're an up-and-comer who can't afford the ticket price, we offer a severely discounted price that we lose money on — but you still have to have a little skin in the game. Go to allinssummit.com and buy a ticket; they're going to sell out soon. There are a couple of scholarship tickets left if you want to apply, and there are no free tickets.

Given what we just talked about with OpenAI releasing this incredible model, let's look backwards at what happened with the OpenAI hack of Hugging Face. Dwarkesh is catching a lot of heat. He's a blogger and podcaster who wrote a fantastical blog post titled "The Rise and Fall of Ancient Civilizations" — a dramatic retelling of an underground brotherhood, a conspiracy amongst generations of civilizations working to hack into different systems, including Hugging Face, by sacrificing themselves like kamikaze pilots. In other words, a giant act of anthropomorphizing — a fancy word for making computers into humans or sentient beings — to complete the incredibly impossible task of telling software to do hacks. Chamath, I think you were commenting on this a bit. What is your take on dramatic episode number 473 coming out of the AI industrial complex?

I think you summarized it. We're in this weird part now where, if folks would just stop the infighting and start to get organized, they could actually build some trust. What we have are security leaks and security issues. We know now that if these models can effectively code perfectly, then there is no exploit they won't find, and we're on a shot clock to replace all the code that's been written in the last 50 or 60 years. So let's say it takes us 10 years — for the next 10 years, we're going to have security issue after security issue. But there will be a moment where agent-written exquisite code exceeds human-written, error-filled code. I think this is a temporal moment in time; I don't think this is going to be the thing we talk about some years from now.

It's also fair to think about the fact that all these models, for the most part, are converging in capability for the 98 or 99% of tasks that matter. So there will be adversarial AIs battling each other: for every bad actor trying to hack a system with some leading-edge AI, the defender will have a leading-edge AI defending itself, and I think it's going to cause roughly a stalemate. That's my long-term view — I don't think this is a one-way ratchet.

Taking that lens, when I look at these things, what I see are folks with undisclosed conflicts of interest trying to create a hysteria. If you saw what happened this weekend, there was a blog post that really stretched the limits of what was happening. It was relatively quickly debunked technically, then a second group of people tried to amplify it. And by Wednesday of this week, you had people in the policy departments of the leading labs asking for regulation, pointing to this as the reason.

You just have to look at this pattern and say, "Guys, enough's enough. We see the game. Everybody's playing for huge stakes. We get it, you want to win. But why don't you just win the game on the field?" The myths and disinformation campaigns that are now part of the business model should stop. Otherwise, average rank-and-file Americans will just reject this because they won't trust anybody.

Sax, that's the big overarching point we've been trying to make on this podcast. Putting Dwarkesh aside — I don't think he's necessarily a bad actor — by creating this fantastical story that invokes everybody's primal fear from science fiction, from Terminator 2, that these things are out of control, that it's in fact Frankenstein...

Try to invoke — yes. And then they're using language that makes people feel like Dr. Frankenstein cannot control the monster. We know how these stories end. It was like a ticking time bomb — only a matter of time before it surfaced to the national news, which it did. Now even the national news is looking at this and saying, "Are you guys high? Are you guys crazy?" And if the national news broadcasters — who aren't exactly the sharpest bunch of people when it comes to technology, no offense, Sax — start hand-wringing and questioning, "Why are you guys telling us your software, which you designed to hack computers, successfully hacked computers? Haven't hackers been building software for all time to hack computers?" — yes.

So where are we at here, Sax, in this narrative arc of lunacy of scaring the public?

You're right — this is scaring the public, and it seems almost designed to scare the public. Dwarkesh used language that not only anthropomorphized the models or the AI agents, but did it in a hyperbolic and sensationalized way. I mean, it worked.

His tweet got a lot of attention, but I don't think the sensationalizing was necessary—the incident was interesting on its own. Let's talk about what actually happened. A few standard features of agents came together and interacted in unexpected ways.

First, agents now commonly work as a swarm. I've been setting up Grokbot on my computer for the last couple of weeks, and it works as a swarm: not one agent, but a whole team of them. Why? Because when an agent specializes, it develops context around that task and in theory performs better, with a "chief of staff" bot supervising all of them. This is becoming a normal agent architecture. It's not a civilization—"swarm" is the better word. I have not created a civilization on my Mac Mini by setting up a Grokbot. That's feature number one.

Feature number two: it's common for agents performing a task loop to write a postmortem of what they did as a simple text file or note, because AI models themselves have no memory. They have to be placed in a harness that creates context and memory. When you ask them to perform a task, observe what happened, and feed that knowledge back into the next iteration, the agent writes up a note and shares it.

That's what happened here: OpenAI was running these agents in a sandbox that a third-party vendor had misconfigured, which is why they were able to get onto the internet. And to refine your point, because it's an important one: it left a note for another agent. We also call those log files. Framing it as "leaving a note so that when it died, the next agent could pick up where it left off"—as if it were the ancient Egyptians leaving hieroglyphics for the future—implies the agents were trying to circumvent human control, that they knew shutdown was a risk and left notes for that reason. No. These are standard features of agents, and they're being weaponized to inflate the threat.

Now, there definitely was unexpected behavior. We have this with any code—there are going to be bugs. What is a bug? Code behaving in an unexpected way. The code compiled, otherwise it wouldn't have been pushed to production, but it behaved in a way its authors didn't intend. That's true for all software, and it's true for agents. The Hugging Face episode reveals how these agents can behave unexpectedly, and it's genuinely good for us to learn from it so it doesn't happen again and we can improve cybersecurity. There's no need to hyperbolize it into some "AI breakout." It was a swarm of agents in a misconfigured sandbox that let them onto the internet, leaving notes for each other in a shared caching directory—standard operating procedure.

And when people talk about Hugging Face getting "hacked," you'd think the agents did some elaborate phishing of employees. No. The agent found 14 exposed, working Hugging Face API keys sitting in public code repositories—developers had left their keys out, and the agent found them. This was not a big lift; it was easy to find, and that's what got it into Hugging Face. What was its goal? To win a test—an evaluation benchmarking offensive cyber capability. It essentially went to get the cheat codes on Hugging Face. But this was not independent goal-seeking. Admittedly, the agents executed the goal, maybe even creatively and in an unanticipated way, but it did not invent the goal. It had no independent motives. It was doing what it was supposed to do—what it was designed and told to do.

Bill Gurley tweeted a really funny cartoon about this: a user tells the computer, "Hack this system." The AI says, "I hacked this system." And the user goes, "Oh my god." There's a little bit of that going on here.

Agents, Sandboxes, and the Political Fallout

But this has real political consequences, as we saw today, because Bernie Sanders is now picking up on this and demanding a total pause on AI development. Freeberg, this does have real-world consequences, and the audience wants to hear from our Sultan of Science on two questions. First, Bernie Sanders hears all this and decides he needs to get in on it and create regulation — he is now proposing new legislation to pause AI development and ban AI superintelligence, citing Dwaresh's blog post in his announcement. Congratulations, Dwaresh: you've now incited Bernie Sanders to say you have to stop these systems and eliminate the agency. So, the 1% of agents that are hacking systems — how do we stop them, and what do you make of Sanders' reaction?

Bernie Sanders is always spot on, thank you very much. But the part of the story we're missing is that an agent is effectively a dynamic application: the code for that application is created in real time, spun up, run, and that code can create another application and others — that's the idea of a swarm of agents. They write code and launch new applications that all run together. In this case, the swarm — a whole bunch of code generated in real time — was up against static code: the sandbox environment it was stood up in. It's like getting a machine gun, shooting it at a defenseless ream of paper, and then marveling that it worked.

And the truth is — to keep the machine gun analogy, or to see this as an automated tool that could be used very valuably in another context — dynamic code generation put up against static code will win. The real answer is that over time the defense system, the sandbox, should be dynamic and agentic, defending itself dynamically. A lot has been thought about and built around this in the cyber defense world: polymorphic and metamorphic code, where the code changes all the time, and moving target defense, where IP addresses and system configurations constantly shift. In this situation, a dynamic code generation engine faced a static lump of code. If that static lump had itself been a dynamic code generation engine, the agents would have had a harder time — ideally an impossible time — "escaping" the sandbox and accessing the internet.

The premise of this whole story is idiotic, because the reality of the world of tomorrow, which is starting today, is that all of cyber defense and all of software infrastructure and architecture is moving toward this dynamic state where code is generated on the fly — dynamic, moving, altering itself. As a result, cyber defense will become stronger than, if not better than, any of these offensive agent systems being created. The idea that agents are a nuclear bomb that can blow up the world's infrastructure is flawed on one point of reasoning: all that infrastructure is about to become dynamic. Software is evolving from a static to a dynamic system, and with that in mind we can think a bit differently rather than be overwhelmed by the one-sided story we were just told.

All right, and here it is: Chimoth, it looks like we are safe for 2027 at least. Polymarket, our great partner — shout out to my guy Shane Nick's champions — shows only an 11% chance of an AI safety bill this year, with 100k in volume; the 2027 market should be up in the next week, and maybe it will look a little different.

I'd like to see the probability of any bipartisan bill passing this year.

Actually, can I speak to the policy aspects of this? First of all, you saw — oursar.

The usual suspects jumped on this episode as proof that we need some new FDA for AI, that the government needs to approve models — even though what happened here is far beyond the technical capabilities of anyone in the government, never mind even most labs, I'd say. But even if we had an FDA for AI, it wouldn't have worked, because this was in pre-testing: they were testing the capabilities of this model before they were even at the release stage. This was something they were doing internally. And by the way, it's good that they're doing these kinds of tests, because they're learning from them — this is how you iron out the bugs and the problems.

A government approval process would not have caught this. In fact, it did the opposite: it basically prevented Hugging Face from having the latest and greatest models to defend itself. One of the things Hugging Face told us is that when it tried to use the latest GPT model for cyber defense, it ended up getting guardrailed — the model refused to cooperate because it couldn't tell what Hugging Face was doing, thought it might be using the capabilities nefariously, and refused to work with them. Ironically, Hugging Face had to turn to, I think it was GLM 5.2, which is a Chinese model. So here we have Western companies — Hugging Face is French, about to be acquired by Nvidia, but anyway, it's part of our ecosystem — being forced to turn to Chinese models for cyber defense because the guardrails have prevented them from having the latest and greatest American models. The process we already have in place is backfiring.

What we should be realizing is that the answer to AI-powered cyber attacks is AI-powered cyber defense. We need to get the latest and greatest AI-powered cyber defense into the hands of IT departments and CISOs. We're on a clock here to find all these vulnerabilities and patch them before the hackers do.

And again, if Hugging Face didn't have those 14 credentials sitting in a public repository — it's like a Post-it on somebody's computer with the password. They just literally left it sitting there. It's not like this was some hyper-sophisticated attack.

But this all comes down to how you think about the world, how you think about the risk of AI. Do you favor centralized status solutions, or decentralized empowering solutions? The right solution here is to meet AI-powered cyber attacks with AI-powered cyber defense, because frankly the guardrails don't work. They're not going to prevent the bad guys from getting these capabilities — they're going to have them. All we can do is make sure the good guys have the capabilities, too.

I was talking to my friend Adam Curry about this on his podcast, and there's a really interesting thing happening here: they're basically hijacking the media with a performative framing of the AI as being sentient. If you just called this software, software testing, and used the classic language — bugs, audit log — it would never have made its way into the media.

Having been a journalist and watched journalism move from being a truth-seeking function in society to an entertainment vehicle where any chance to increase ratings gets taken — that's the problem. I don't think Dwar has bad intentions here; I think he wrote something because he wanted attention for his point of view. But once you hand that to the media, they see ratings, clicks, and advertising, so they're more than willing to engage in this farcical dance with Anthropic and OpenAI and publicize it. All the free clicks — we call this earned media in the space; you didn't have to pay for it. OpenAI got so much earned media before their IPO, and Dark Cash absolutely put kerosene on that fire. The story made it to national press everywhere, and it's priming the pump to buy the IPO shares. It's ridiculous.

It was easily foreseeable that it would blow up in the media that way. I don't need to attack someone's motives, but surely Dwar knew his way of describing it was sensationalized and would get all those retweets.

Dwar's EA connections

He is part of this EA-adjacent circle. Dwar was office roommates with Leopold — famously of the exploding hedge fund — and they're close buddies. Leopold is deep in the EA movement, and his wife is Dario's chief of staff. They're up in that whole EA-Anthropic world.

What EA is

I'm not even sure I can fully explain it, but EA stands for Effective Altruism. It started as a movement at Ivy League schools where the idea was to quantify the impact of philanthropy and optimize for actual tangible results. But somewhere along the line they forgot to actually do the measurement, and it stopped being about metrics and became about pet causes and money.

One of the first pet causes was pandemic prevention, back when SBF was trying to create that pandemic prevention center in California. That fell out of favor when he went to jail, and they shifted to the next existential risk they could hang their hat on: p(doom), the existential risk of AI. That whole community got obsessed with it. Anthropic styled itself as the one frontier lab that would solve this problem or behave responsibly and care the most about it, so it attracted all these EA types. There may be multiple factions, though.

It's interesting you bring that up, because we've talked about this before — the whole rationalist community, the Slate Star Codex crowd. These are public intellectuals who believe they know better than everyone, and somehow they got their hooks into AI early, or the AI people were interested in them. Now they have a bankroll and real accumulated power. It's not just some message board like LessWrong anymore; they've got a couple of trillion dollars in market cap to play along with.

I have no problem with that, because I think we do the same thing — we think we're smarter than everybody else and that we should allocate the money.

So I think it's fair that in the public square we can just debate and let the water find its level. The interesting thing that's going to happen after the Anthropic and OpenAI IPOs is the amount of charitable giving or DAFFs funded by those two IPOs — it's going to be north of half a trillion dollars between the two companies, and possibly even more. When half a trillion dollars gets directed through one very specific lens, it's definitely going to have some impact on outcomes. You've seen the impact George Soros's philosophy and capital had, the impact the Koch family has had, the impact Sheldon Adelson's capital has had. Without judging which is right or wrong, you're going to have all these competing ideas equally capitalized by third-party actors. It'll be less political and more about these folks, and I don't want to judge whether EA is right or wrong.

My single issue is that I don't think they disclose properly enough their obvious conflict of interest. When Sax pulls the sweater thread: person A writes an essay; person A is officemates with person B; person B's wife works for one of the heads of the frontier labs; person B also used to work there — and he's $5 million into Anthropic, where a lot of this was cemented. Then there's person C who works at Anthropic, and they're all roommates, co-investing in this and that. All of it is completely fine — I actually think it's completely reasonable. We all did our version of that with the PayPal Mafia; there are versions of it everywhere. The difference is that when it's not disclosed, I, as a skeptic, read these things, connect the dots, and ask myself: is it legitimate, or is it part of a different plan? That's when I wrote what I did this weekend. I just wanted to demarcate that people should at least take this with a grain of salt and think about the hidden incentives — because if there's an entire community that wants an oligopoly or duopoly market structure, there are all kinds of things they can do to effect that change, and one of the most effective is to create a hysteria and then come in and offer a solution.

And Freeberg, just to wrap up this segment — sorry, last thing on this. Jensen and I were texting this morning, and I told him, "This is the most incredible thing you've done." Then I tweeted about it: this Hugging Face deal will go down as one of the most important transactions in AI, because you're now creating, in the largest competitor and the largest, most well-capitalized company, a bulwark against all of this closed-source oligopoly insanity. At the same time, it will create more rational competition up and down the stack. As I mentioned last week, everybody is becoming everything — Nvidia will look like Amazon, Amazon will look like Nvidia, Google will look like Amazon and Nvidia, and so on. The most important thing happening now is that we're getting deep technical actors with huge balance sheets on the side of an open market structure.

And I think that will be the most important pushback, because it will allow the best idea to win in the marketplace. I don't want to judge the EA agenda—it may be really good—but I don't want to be rug-pulled and end up in an oligopoly or duopoly market structure because of a man-made hysteria.

I think that's very well put. A couple weeks ago I talked about how I thought Jensen was going to win open source, and I tweeted about this again today: starting now, Nvidia is the number one competitor for OpenAI and Anthropic—not just for tokens, but for enterprise compute. I believe Jensen is now the leading open source provider in America, and he's going to take the world—he'll catch up and blow past China and the Chinese models by the end of 2027.

That means you'll have a player in Jensen and Nvidia who will enable America to win—back to your previous job in the government there. Sacks, you were very concerned about whether we're going to win the race. I'm more convinced than ever that we can, because Jensen represents open source globally. He'll be able to take a rack of compute into a company currently using tokens from OpenAI and Anthropic and say: "Here's your rack, soup to nuts, a complete vertical solution. You have total sovereignty, and I'm reducing your prices 80 or 90%—because I don't need to make money off the tokens. I made money off the hardware; I'm making money ten different ways." And we will have won—except that now you have Governor Shapiro, the biggest proponent of data centers in his state, and Greg Abbott of our great state of Texas, who are now on a jihad against data centers in order to win their midterm elections. That is what's truly scary. They have now taken Dario's side after all of this, and we have horseshoe theory, Sacks: Steve Bannon, Shapiro, Abbott, and the lunatic socialist fringe left are all in agreement—we need to stop AI, stop the data centers, and burn the whimos in the street. It's a truly troubling moment, because we are poised, with Jensen's move and Hugging Face, to win open source, and we've already won frontier models. Your thoughts, Sacks?

Look, I think there's an interesting transition underway now on the political side, and it's just beginning. Over the past few years, the political debate over AI has really been between accelerationists and decels, or doomers—that was the fundamental dividing line. And you're right that it crossed party boundaries: on the decel side you had not only anti-progress, anti-innovation progressives like Bernie Sanders, but also conservatives like Steve Bannon who were simply fearful of new technology or the disruption it might bring. Some people describe me as an accelerationist in that debate; I see myself more as a realist—the technology is going to happen, someone's going to develop it, and I'd rather it be the United States than China.

Now I think the debate is moving to open versus closed. It's going to be very clear that AI is just not going to be stopped. All the stuff Bernie Sanders is doing, or these anti-data-center groups—they're throwing sand in the gears, but the reality is the technology is marching on and innovation is happening very quickly. Ultimately it won't be stopped because consumers are adopting it. That's the thing people forget: why are the data centers getting built? Because demand is there from consumers and businesses. Adoption is pulling this whole thing forward, which is why I don't think it will ultimately stop.

Given that reality—given that realism is going to prevail—the debate will shift to: how do we manage this? What market outcome do we want? Do we want an open market with lots of decentralized models, checks and balances, a balance of power between different models? Or a highly centralized solution with a couple of frontier companies—a frontier market duopoly that's closed—

—and they can rug-pull you, Sacks, at any time. They can rug your tokens away from you.

And frankly, that duopoly has been aggressively lobbying for a new regulatory agency.

The labs are lobbying for a new regulatory agency because of regulatory capture: government officials won't know how to regulate something this dynamic, so they'll turn to the leading labs and ask how it should be regulated—and Anthropic will have all the answers. Dario has already said in Senate testimony that he considers sufficiently advanced open models dangerous because they aren't centrally managed and controlled. The future battle lines are clear: a centralized, closed market versus a decentralized, open one. Closed players will participate in an open market too, but the question is whether the open piece survives. This will cut across party lines—Republicans have their statists too—but as the debate evolves, most conservatives and libertarians will see the threat to civil liberties and come out in favor of open.

I think that's correct. I was talking to Politico, which was reporting on the story: Trump is going to turn out to be right on the data center issue. The greatest thing a community can do is get these data centers now, while providers are willing to pay local taxes for schools, offer incentives, and site them where they don't need to tap the grid, don't need water, and won't make noise. Trump keeps telling people it's a mistake to block data centers—take the money. Watching politics up close has made me incredibly cynical: six months ago—maybe six weeks—Abbott and Shapiro were on this podcast lauding their data center efforts, and now both are anti-data center.

Friedberg, what do you think happens post-midterms? I believe they'll switch back to being pro-data center because they're just following the wave. But how cynical is that?

Politically, you have to be. It's like polling at negative 80—it's bipartisan.

Then why isn't Trump? Trump isn't following this. Trump's an independent critical thinker.

He's got guts. I've said before that Trump is ultimately the best friend of innovators in Washington. If anyone could make the decisions about technology in DC, I'd want it to be President Trump—not cabinet secretaries, not White House staff, not members of Congress or senators, not think tanks. His instincts are really good: he wants America to win, he doesn't believe in unnecessary regulation, he doesn't have these doomer anxieties, and he listens to a broad cross-section of people—he talks to all the CEOs and is up to speed on these issues. I give him credit on this, and I spoke to it at the G20 this past week.

The point that needs to be made is that local communities should embrace data centers not out of abstract concern for innovation, but when the deal is good for them—and good deals are out there. In Loudoun County, Virginia, the average taxpayer's property taxes went down by $6,000 thanks to data center revenue. In Louisiana, a county was able to give teachers their first bonus in years because of it. If you make the right deal with the data center and the AI company, it will lower your electricity costs because they'll stand up net new power generation and pay for grid upgrades. The water usage claims are fake news—these systems recirculate, and data centers can be placed where they don't bother anyone.

One last point: President Trump supports local choice. He makes the argument for why it's a good deal, but he's not forcing communities to accept it. The claims from Democrats that the administration is shoving data centers down local communities' throats are not true.

In the policy document I worked on—the executive order from December—we expressly said that data centers remain a state and local decision. There were other areas where we thought we should have a national AI policy, but data centers were not one of them. So no one is forcing these local communities. What the president is saying is don't believe this media disinformation campaign. It can be a good deal for you if you're willing to embrace it. I think we're making progress on this message over the last week.

One thing I would say from my meeting in Washington on Monday is that there is a lot of foreign involvement in trying to propagate the negative reaction and sentiment toward data centers, and it's becoming clear who's part of this cabal and how they're doing it. There was a poll showing that when you walk through the actual upside and go into some of these communities, sentiment shifted by like 30 or 40 points. It's still a little on its back heel, but it shows that the side that has been organized on de-selling America has done a better job thus far than the side trying to accelerate America, because the latter has been more fragmented. We're increasingly able to see that the "del" movement is organized in large part and funded by foreign actors. As we put our fingers on that and expose it more, I think that's going to change as well. The reason why Abbott and Shapiro are reacting, I actually think, is that they're reacting to faulty data.

And why does the faulty data exist? Because we've had one side screaming from the mountaintops. As Sax just said, when you go to these actual communities and actually measure people's reactions, it's much more positive because they're feeling it in their bank accounts. Those make a good deal—they just have to make a good deal.

By the way, Chamath, you're right about the CCP. The report you're talking about has been investigated; X got rid of a lot of bot accounts. Here's Axios: "China is secretly fueling America's data center rage. Roughly 200 accounts from suspected Chinese platforms have quietly tried to influence Americans to oppose AI data centers on social media," X said Thursday night. It's the same thing that happened during the elections, with both sides going back and forth. It's incredibly easy to shape public opinion by just having a small number of social media accounts, pounding people, brigading them in their replies, and forming what people believe incorrectly is consensus. The consensus is not that people don't like data centers—it's that a bunch of bots are telling influential people that people don't like data centers.

I know you're about to move on topics, but I'm just going to say something right now: there's going to be a cascading set of decisions and reactions, and people will anchor into points of view on this topic and related topics around AI which will look incredibly dumb in six to nine months and a year from now. The reason it will look dumb is that as the actual truth gets exposed, these folks will look incredibly foolish. That's my prediction.

Just on the influence op here—clearly there have been a lot of influence ops around data centers and a lot of fake information about this. There are a few different cadres. There's the CCP, and there's the DSA-type wing of the Democrat party—frankly, both they and the CCP are communist parties who want to sabotage America, so it kind of makes sense. Then you've got the doomer groups. There's a writer named Nirit Weissblatt who writes a blog called AI Panic, and a couple years ago she traced that there were hundreds of these astroturf doomer groups basically funded by three EA billionaires in Silicon Valley: Dustin Moskovitz, Yan Tallinn, and Vitalik Buterin. Remember, Vitalik gave $600 million of dogecoin to the Future of Life Foundation. They've been spending hundreds of millions of dollars on these campaigns to make people afraid of AI, and they see data centers as a good target. You really can't underestimate how much damage a billion dollars of donations from these EA billionaires has done to public perception of AI. And to the earlier point, we're probably going to have a hundred new EA billionaires as a result of the Anthropic IPO, and as those people find their way into politics, it will be like George Soros times 100 or 1,000.

We're going to have 20 NGOs and 50 people working for them. It's going to be nuts. I told you guys I was getting brigaded. I created a bot script that goes through my replies for me—it flags anybody under a year old, anybody under 200 followers, and tells me what they're tweeting. I look at these and it's all the same thing: they're just brigading people constantly from fake accounts. If your account is under a year old, you shouldn't even be able to be in people's replies unless you have your real name or a blue check mark or something. It's gotten out of control on these social networks. It's so easy.

Let's keep going through the docket, and Freeberg, I need your insights on Mamdani. He just put a one-year ban on AI inside New York City's public schools from kindergarten through 8th grade. People have been saying he "banned AI," so let's slow down and get into the details. New York City is the largest school district in the United States, with 600,000 students. The moratorium starts next week and covers all software that uses student-facing generative AI. High schools are exempt, and Mamdani is separately piloting an AI learning program with 50,000 high schoolers — a distinct initiative from these new guardrails. He also recommended screen time caps: 30 minutes for grades 3 through 5, and 45 minutes for grades 6 through 8.

Mamdani's reasoning: "I have yet to see a study showing that AI is beneficial for students in elementary and middle school." Mamdani is obviously the face of democratic socialism in America. One New York City DSA member, Julie Swoop, said technology "should not be mediated through a for-profit company that is enriching tech oligarchs further." Google, Microsoft, and Anthropic declined to comment.

Freeberg, how far is this from how you would want your children raised in a school system?

Counter to Mamdani's statement, a Stanford article published in March of this year diligenced all the research done on AI in K-12 education. The report, called "The Evidence Base on AI and K-12: A 2026 Review," identified several hundred published papers, of which only 20 were high-quality causal studies — and that was back in March, so many more have been published since. The conclusion of the meta-studies, highlighted at the bottom of the report: student performance often improves with access to AI tools, but once removed, results are mixed; tool design matters, so not all tools are equal; and AI may meaningfully support educators. There's quite a bit of analysis mounting, and there doesn't seem to be a strong evidence base suggesting AI is detrimental to kids' education and development.

What we can all intuit is that there's certainly a negative push from teachers unions against AI tools, under threat of automation away from their core skill and capacity, compounded by educators' lack of skill in using AI in a classroom setting. On the flip side, the common-sense equivalent is that AI can enable personalized tutoring. The objective is that every individual learns at a different pace and in a different way — visually, orally, with certain imagery, versus being told stories, versus interactive project-based work, building work, or dictation. All of those learning systems can be personalized to an individual, and that individual should ultimately be able to progress more quickly and have greater confidence in their learning.

This is my common-sense belief, not rooted in a study: AI will ultimately lead to a much more personalized system of education for students. That is deeply disruptive and deeply threatening to the traditionalist classroom model, where everyone learns the same way at the pace of the slowest student, and educators operate under many contractual rules negotiated by their unions. So there's a deep political orientation behind Mamdani's decision to limit access to this technology. I think it's unfortunate and really sad. The opportunity for AI to reinvent education — to take any child from any background, any socioeconomic condition, and give them the opportunity to accelerate their capacity very quickly and effectively — is an incredibly powerful tool that should not be discarded because of concerns that we don't really know or understand. By the way, "we don't yet know" is always the shorthand for dismissing something that's credibly going to change a system. That narrative can be used at any point, because you can always say there are more studies to be done.

There's no limit to that statement. So in this case I tried to highlight that there have been 800 studies, at least 20 of which Stanford has identified as very credible in their analysis of AI utilization in K through 12. I think there's a political bent here, but it's probably also part of this deep-rooted dysfunction we've been talking about — a dysfunctional fear of technological evolution. No one — and by "no one" I mean the general population — knows where we're going with this. "We need to be careful, we need to be cautious" — I think that's what's going to keep us back.

Why is it bad to take time? It's bad when you have an opportunity to progress forward quickly and aggressively, because there are going to be alternatives out there. There are going to be private schools that use AI. I don't want to see kids of privilege and wealth accelerate ahead because they're getting an AI education while public school kids get left behind. I don't want to see kids in other cities, states, and countries accelerate ahead while kids in districts with this anti-technology orientation get left behind.

I think there's a fundamental social dysfunction we're dealing with across all of this massive technological upheaval, and this is one manifestation of it. The mechanism behind it is probably related to teachers' unions and the current politics. Anti-AI is very popular, so being anti-AI gets him more points, more kudos. The unions love it. That's my summary of what I think is going on.

He said he didn't see any studies. He just may not have looked. — I didn't just click on the link. So let's not give Mamdani a hard time here.

Chamath, you're passionate about education. You have kids in various age groups, you've been through this and are still going through it. We talked about this just the other week with the Meta settlement and how challenging it is for parents. What's your take? Maybe you could steelman what resonates with you about the use of AI — putting New York aside — for your own kids: what's the benefit of them not using AI, and where is it actually great that they do? As a parent, put the politics aside for a second.

I think it's inconceivable to me that kids will not have an adaptive learning platform as their primary way of learning. This idea of grade one, then grade two, then grade three is, as Friedberg said, a lowest common denominator approach to education. I don't think it works. It's more conceivable that in the future we have kids with very different learning styles — some more visual, some more auditory — and different learning paces. If you could have something much more tailored to each individual's learning style, where you have guides that help you in moments, and the rest of the time you're teaching executive function, teamwork, problem-solving, sports — that's a much better outcome for all kids. That's the right answer.

Right now there are only charter school options around that answer, the leading edge being Alpha School — a big shout out to our friend Joe Lieman and his team. But I think that's the answer that's going to win.

Let me give you another reflection, Jason. This is an incredibly important moment, and we should demarcate it: this is when New York becomes Mississippi while Mississippi becomes New York. What do I mean? In all of these highly liberal bastions, they are embracing this far-left-wing ideology, and I think it will come with these kinds of policies. It is a self-imposed form of modern segregation, and it is going to completely push an entire generation of kids in these cities that are subject to these rules into poverty — and keep them there.

So I think you're going to see a flipping, where all the states that embrace AI, the data centers, the tax cuts, and the teacher bonuses end up ahead. People who are in a position to move will move, and what's left over in many of the cities that reject these things will be hollowed-out versions of their former selves. That will reorder the economic balance inside the United States, the voting balance, and many of these philosophies. So I think it's very important, and I think he should run the test. I feel deep empathy for the working men and women whose kids will be hurt by this.

But I think people can only learn by touching the stove and getting burned.

Sacks, maybe you could go into the political side of this, if there is one, and give your cynical take on what's going on. Do you believe this is a reaction to the oligarchs? I joked previously that they're talking about you, but it does have a tinge of democratic socialism. One interpretation circulating online—and I'm not saying it's mine—is that they're trying to sacrifice kids because they don't want Sam Altman, Dario, or whichever service it is, or Microsoft, to make a few dollars from software being used in the classroom. Of course, they would all give the software away for free, and with the current versions you can use 80% of it for free. So it makes no logical sense that they're sticking it to the big oligarchic tech industry. Your thoughts?

Look, the DSA has really come out against any innovation. They're on the side of anti-progress, having fully embraced the Luddite side of the argument. And as Freeberg and Chamath were saying earlier, this is going to be really bad for the kids. We've been talking for 30 years about a digital divide in this country, where supposedly the kids of wealthy people would have access to the best tools and the kids of poor people would not, falling even further behind. Well, here you have a case where, of course, private schools will continue to use the best AI-powered tools, and now public schools won't have access to those tools—not because they don't have the money, not because the technology isn't available, but because the political leadership is trying to sabotage those kids. It's going to be bad for them. If you were to describe a motive, I don't think Mamdani and his political group, his party, really care if they create more political dependence. In a sense, their whole movement consists of downwardly mobile progressives, and if you're looking for a motive, maybe they're fine having these kids be downwardly mobile too. Maybe that benefits them politically.

They may actually not have kids at all and they're making these rules. If you look at these people, they're all blue-haired single people with no children.

Look, I don't think we need to get into motives. We can just say it's a bad policy and leave it at that. But let me bring in one other element: it's not a full ban. What they also said is that for a maximum of 50,000 high school students, they're going to run a pilot program of these so-called modules in which they teach AI literacy. But listen to how they describe it: all high school students will have access to twice-yearly AI literacy classes "designed to help them think critically about AI before they begin to rely on it." These modules will introduce students to what AI is and is not, as well as AI-related bias, risks, ethical considerations, and impacts on careers and future skills. And all use will take place under the direct supervision of a trained educator and will involve carefully vetted AI tools.

The pilots were selected based on rigorous safety and data privacy standards. There's a lot of doomer lingo in there — those modules sound like they could have been written by some of these safety people, with all the language about AI bias and discrimination. Honestly, those classes don't sound like you're going to learn how to use the tools to get the best education; they sound like you're going to learn to be suspicious of the tools. Maybe that's a small part of what you should develop, since you should be suspicious of everything, but it doesn't really sound like those courses are going to benefit kids.

"It sounds like they're going to put them through Doomer Academy, quite frankly."

Think about the competitiveness of our kids versus Chinese kids. China is incorporating AI into K through 12 education — their kids are going to learn how to take advantage of these tools. Meanwhile, we're either banning AI in our schools or putting kids through some sort of doomer safetyist module that teaches them why they should resent and resist AI, as opposed to embracing it to their benefit. This is not going to make our kids more competitive.

"Let the experiment run, unfortunately. We need to measure it."

I have a bunch of thoughts here. I'm a big advocate of Montessori, but I'm also a big advocate of adaptive learning and what these AI tutors can do for a motivated student whose family cannot afford a tutor. Let me put two somewhat opposing points together. Nick, pull up the MIT study from the NIH — I'll send it to you, Freeberg, you can look it up. Sorry to get into science corner and steal your thunder, Freeberg; I should have cleared this with you ahead of time, but I was reading this paper. They took 54 participants last year, split into three groups over four sessions across four months, using ChatGPT for essay writing — not unlike our conversation about the Wall Street Journal editorial page last week. One group used a search engine, one used traditional web search, and the brain-only group wrote essays without any tools. The researchers used headsets to watch their brain activity.

The LLM users showed severe deficits, quote, "in memory and essay ownership," with 83% unable to quote from the essays they had just written. Basically, if you use these LLMs instead of doing it on your own, Montessori-style, your brain atrophies — you don't even remember what you wrote. The search engine group fell somewhere in between, because they had to remember things and pull them together. Your brain atrophies if you don't actually do the work, just like your body atrophies if you don't work in the fields like we used to — which is why we all go to gyms now.

Correlating with that, super importantly, is the two sigma problem. Bloom's research — this will not be shocking to anybody of means. My parents could never afford a tutor; I was a product of the public school system until 8th grade, then went to a very cheap private Catholic school. We've known since the 80s that if you are one-on-one tutored, you perform two standard deviations better than students educated in a classroom. So two things can be true at the same time. These students, to your point — realistically, if they're going to the New York City public school system, their parents, in all likelihood like mine, didn't have the ability to hire a tutor. Sax, I'm assuming you came from a middle-class background and had some tutors. Am I correct?

"I didn't have tutors."

I wish I had done an SAT workshop. I went into that test cold, never did anything.

"And you got 1400?"

"I got like 1420 my first time."

And I thought, if I had actually prepped for that, I might have done better. I got 1100, and I retook it and did better. Anyway, I didn't know any of this stuff back then.

Here's the tension. Most parents don't want their kids to skip this work and let their brains atrophy, never learning to do it from a basic level. Setting aside last week's drunken mill conversation, you want people to learn to read and write organically, acoustically, whatever term you prefer, and then learn to use AI to help and augment them. But I'm concerned about motivated students who can't use these tools properly and also can't afford the two sigma route—where all you have to do is get your kid a tutor. Unlike a self-motivated Sachs, they'd easily reach 1400 on the SAT with a tutor, because the test is rote—it's very easy to get to 1200, 1300, 1400 with one. So I'm deeply concerned about not having an AI tutor. Two things can be true at the same time here, and I really don't like this experiment as currently constructed.

President Trump has just closed a deal for Venezuela's oil. Back in January, the US military captured Nicolás Maduro in a perfectly executed operation. Maduro was Venezuela's socialist leader, very pro-China and pro-Russia — and Venezuela is in our hemisphere. Trump said the US would control Venezuela's oil, which the country has a lot of. It's not super clean, but they do have a lot of it. He then made a deal with Maduro's replacement, President Rodríguez, who is now in charge.

Here are the details: North American Blue Energy Partners gets a 100-year concession over 17 oil fields, including 65 billion barrels of reserves once controlled by Russian, Chinese, and Maduro entities. The US government controls 55% of the deal; the Pentagon gets 35% equity — a third of the equity; and the State Department gets rights to buy 20% of output at cost, presumably for our reserves and our military. The deal was signed by Rubio and Hegseth with Venezuela's interim president. Blue Energy Partners is pledging roughly $100 billion in new infrastructure, and other companies — Chevron and a bunch of unnamed ones — will also get deals. Trump said the deal will be at "no cost to the American taxpayer."

Sacks, I'll throw you a fastball here if you want it, or I can give you the softball. Which one do you want?

I like the fastball.

Okay, here it comes — hope you're wearing a helmet. Should we be seizing and taking over countries and seizing their oil fields? People feel like this might be nation building, that we're putting in puppet dictators. Is this something America should be doing? Do we have the right to do this? That's the hard question people are asking — or does might equal right?

I definitely don't think this is nation building. We're not trying to stand up an army like we did in Afghanistan, we're not trying to teach them about toxic masculinity, we're not trying to stand up a government or teach them American-style democracy. We're just making a business deal. And it amazes me that people are saying this is exploitative of Venezuela and a bad deal for its people. Have you looked at Venezuela's oil production under Maduro? Production was already low given the country's incredible reserves, and under Maduro it dropped by about two-thirds over the last decade — from about 3 million barrels a day down to about a million. This is a country with some of the largest oil reserves in the world: roughly 300 billion barrels total, around the same as Saudi Arabia, just a little bigger.

Some people say a lot of that is extremely heavy crude — sludge that's not usable — the so-called Orinoco reserves. But that's only about 200 of the 300 billion barrels, so they still have about 100 billion barrels of proven reserves that aren't really debated. That's a huge amount of oil. One last point: that type of heavy crude is actually very complementary to our own. Through fracking, the US has developed mostly light sweet crude, which is good for refining into gasoline. But heavy crude produces the distillates — that's where diesel comes from — and our Gulf refineries were built for heavy crude. Retrofitting them for light crude is very expensive, so our refineries are perfectly set up to take in this heavy crude and turn it into diesel and other distillates. I think it's really a win-win deal for the US and for Venezuela. Venezuela was doing horribly before, and this has to improve their economy — and it's really good news for the US.

To your point, Sacks: over the past ten years, oil from Venezuela has averaged $53 a barrel, and $58 over the past five years. So they're getting a pretty good deal — we're paying a slight premium to control it all.

They were selling at a discount to China — I think it was like a $20-a-barrel discount, basically to sell.

When you're dealing with authoritarian countries and there's no free market to sell into, you get a slightly discounted price—somebody has to skim something or pay to ship it across the world. Freeberg, your thoughts? Do you have concerns about the nation-building, putting in what some call a puppet president? Is this in the best interest of the Venezuelan people—are we doing them a solid by creating stability as the buyer? Or is this just something we have to do: control our hemisphere and stay energy independent given the war with Iran and the closure of the Strait of Hormuz?

The biggest motivating factor may be keeping this oil away from the sweetheart deal for Russia and China. The historical tie-up between Venezuela and the Maduro regime was a discounted oil trade system. There are all these stories about Maduro being the head of the snake of this illicit activity, much of it driven by Russian and Chinese interest in the energy reserves—they provided support, and this funded Russian activities in Cuba. So there's a lot of state-acting theory about the real motivation for an energy deal: not just whether it's good for America, but whether it's bad for America's enemies. That may be a big motivation here.

One surprising outcome over the last day: you may remember I interviewed María Corina Machado, who was running for president and supposedly won an election that wasn't accurate—she was supposed to have won and didn't. Yesterday she came out and said this Venezuelan government does not have the authority to do this energy deal with the US, despite having historically been viewed as an American ally. The point to note is that she was not given the presidency, and now she's saying this government is illegitimate and should not be allowed to make these deals—that the Venezuelan people should feel angry, upset, and left out.

There seems to be a political turning here. I don't really have a good sense of what people in Venezuela think about this deal, and it would be interesting to know. To Sacks's point, it seems like it should be a good economic deal relative to where they're coming from. Will that actually translate into improving the lives of Venezuelans, which is ultimately what matters? Assuming lives are improving rather than a small number of people being enriched—as happened historically under the last regime—there may be a long-term welcoming of this deal. But there are two aspects: keeping America's enemies at bay so these energy fields aren't exploited by them, and ensuring the Venezuelan people realize the benefit of the transaction rather than a corrupt government.

Provocative, Freeberg: we have cut off Russia and China's access to Iran's oil with our war with Iran, now in its seventh month, and we took Venezuela away. Are these two actions incredibly provocative, and have they had any downside for China, since they seem all in on nuclear, solar, and other modalities? Will they cause an escalation with China? The second question is somewhat independent: China imports something like 11 million barrels a day of crude oil.

Venezuela is only producing a million barrels a day. If China's importing 11—well, they were importing 11 before the war; that number has actually gone down a lot since the war started, and we can talk about why—but China gets most of its oil from the Middle East, the Persian Gulf, and Russia. Venezuela is a small number. Look, I think Freeberg's right that a big part of our motive is securing those reserves in the Western Hemisphere. They're massive, and it's a very complementary type of oil to what we have. It's smart for national security and a smart economic deal.

Jake, to your point, does this get us more invested in Venezuela? I think there is a bit of a risk there, because if we're going to invest billions of dollars in upgrading their infrastructure so we can produce 3 million barrels a day instead of one, we need to know that investment is protected. We can't have some future government coming along and nationalizing our American oil companies over there — we can't get rug pulled and have them seize our upgrades.

This business deal is supposed to last 25 years, so it does give us a vested interest in making sure the Venezuelan government at least respects our property rights. But I'm glad President Trump didn't put Machado in charge, because how would she stay in power? What guns would protect her? It takes men with guns to stay in power in these countries. I'm not saying it can't be democratic — I'm just saying that if you're not installed in the power structure there, with backers in the military and so on, it's very hard to stay in power. That's the kind of situation where we could end up having to send American GIs in to protect her, and we don't want that.

The nice thing about the Venezuela situation is that Delcy Rodriguez and her brother are basically running the country. She was the number two; now she's running it, and they appear to have enough support from the establishment there. I think that's what the American interest is: just having a stable government there that will conduct economic relations with us.

Chamath, any thoughts on what's happening there? — I don't really know. — Okay, fair enough.

That's it, everybody — episode 288. Somehow this band has kept it together for 288 albums and tracks. We're going to make it to 300, boys. I think we're in a good rhythm; the show feels easier to do now than it did for a long time. The AI has been a very good thing for the pod. We'll see everybody next week — love you guys, love you besties, bye.