AI Kills Everybody or Doomer Psyop? OpenAI’s Math Breakthrough, Nike’s $200B Collapse

The guest canceled on us — he couldn't take the smoke. He didn't want to be cross-examined by David Sacks. I haven't used my legal degree in 25 years, but I was up all night doing research on this, so we can still talk about it. I'm so ready.

Sax, we have to talk about Anthropic. A researcher just quit over AI fears that he thinks could kill us all. Jacob Coxin worked as a researcher at OpenAI, then Anthropic, over the past three years — though he only started at Anthropic six weeks before resigning. After quitting, he posted the following to X:

"The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. Neither OpenAI nor Anthropic ostensibly is acting responsibly. They're not acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives."

This went insanely viral. I haven't seen a tweet hit 150 million views since Elon said he was going to put the cocaine back in Coca-Cola, which I think was the number one tweet of all time.

But Jason, can you actually explain how somebody who doesn't have any followers or an account shows up on a Monday and gets 110 million views on a Tuesday? How does that algorithmically happen?

That's a great question — we're going to get into it. Evan Hubinger, who leads Alignment Science at Anthropic, responded to Jacob:

"Jacob is correct here. We really do earnestly believe AI could kill all humans! I personally think it is over 10% within the next decade. We do not yet have a plan to solve alignment."

Those two posts hit 200 million combined, and as I mentioned last week, these things are now jumping from X and the AI community to the nightly news in under 24 hours. Bernie Sanders, never one to waste a crisis, said posts like this are the reason he's proposing legislation to ban superintelligence and pause AI development. Governor JB Pritzker replied to the post saying, "It's time to sound the alarm louder on reining in AI."

Sax, as our former AI czar, do you share everyone's concern here that we're all going to die by 2030?

No, I don't. There's nothing new here — this is the same doomer histrionics we've been hearing from this crowd for a long time. The media is calling him a whistleblower, but what evidence has he brought forward that we didn't have? This is all just vibes from him. Show us the data. Show us the report. Show us the leaked information the public didn't already have. Show us the facts, show us the evidence — we don't have any of that.

In fact, as Chamath was alluding to, this appears to be an op. How do we know? First, this was an account with almost no activity, almost no followers, and no posts prior to the resignation tweetstorm — if he did have prior posts, they were scrubbed. It was a completely blank account, and somehow within a day it found this huge audience. Elon has said that never happened before.

We also know who amplified it in the first 15 minutes: three well-organized, well-funded doomer groups. Nathan Calvin, general counsel at Encode AI — the group that pushed for SB 53 in California and pushes for state-level AI laws. Peter Wildeford, head of policy at the AI Policy Network, which pushes for federal regulations and a federal Department of AI. And Daniel Kokotajlo, head of the AI Futures Project, who just happened to drop a Rogan episode at the same time using the exact same phrasing about how AI is going to kill us all. All three of these groups are funded by Jan Tallinn, one of the top EA mega-donor doomer types, who also happens to be a co-lead of Anthropic's Series A. This is a tight-knit community that jumped on this tweet.

So this was not some spontaneous act by an employee who said, "Oh, I'm just going to post my resignation letter on Twitter." This was orchestrated. The final piece of evidence: someone noticed that the Wall Street Journal story covering the resignation letter was posted minutes before the tweetstorm itself.

The WSJ was under embargo — that's what it means. They were briefed under embargo and screwed up the publish time.

And this guy Jacob Coxin reached out to you, Jake, saying he wanted to come on the pod — and then canceled this morning, maybe because he figured out it would be a tough interview. The questions I'd want to ask are: who connected you with the Wall Street Journal? This was a relatively low to mid-level employee who had been there — I've heard different estimates — six weeks to three months.

Let's say 3 months. This is not something that would be covered by the Wall Street Journal in this way. This is something that was teed up. He's clearly working with these doomer groups. They're going for maximum amplification. And again, he's been called a whistleblower, but what have you blown the whistle on? Where's the evidence? Where's the data? We'd love to see what you saw inside of Anthropic if somehow there's a problem there, but you won't tell us what that is. And in most whistleblower cases, you get the company pushing back. They say, "Well, no, it's not true," or they want to refute what he's saying. In this case, they're amplifying it themselves.

A functional company would disavow this immediately, wouldn't it?

Here's the contradiction: he's actually amplifying their message. And you saw that when his boss, or the person running the safety team, said that, yeah, we all agree with you — and then put on this greater than 10% chance of extinction. When you put a number on something that frankly is just vibes, it takes the message to a whole another level.

Two questions for you, Sacks, and then I have a bunch of comments and statements — I mean this thing is crazy. Question number one: let's steelman the other side. What is the game theory on the SCOP? How do you think it was supposed to go from here had it not been exposed? What do you think they thought would happen?

Well, I think they know exactly what's going to happen. They are going for maximum amplification. Given that this guy was at the company for a maximum of 3 to 4 months, it's entirely possible to me that he joined knowing that he would put together this campaign.

So what was his goal? Is he a rogue agent?

Is he, or is he in coordination? These groups are interested in shaping the public perception of AI. This is what they do. They're creating a very negative perception of AI to further their case for AI regulation. And the end goal here is to create a federal department of AI. They want an AI regulator. Some of them also want a pause. Some of them also want a ban on superintelligence — that's what Bernie Sanders' bill puts forward. But all of them want a new federal regulator for AI. And that is what Daario himself has called for — FDAI — which invariably they will control, and they will basically then be able to push their preferred frameworks for regulation through this new regulatory body.

Okay, I have another question. So now you have this very complicated moment where you have internal employees validating an employee that just left making claims — let's call them whistleblower claims — that create at a minimum product liability issues, but at a maximum may create false representation issues. You have an active S-1 process underway. So how are they to handle that? I'll give you a couple of historical examples. You guys remember when Google was going public? Freeberg, you probably lived it. There was an interview that I think Sergey or Larry or both of them did with Playboy, but that was well before the IPO window. Then they published it in the middle of their quiet period, and there was all kinds of noise about how that could have derailed their IPO — that they would have had to refile. In a separate example, when Slack was going public, I was on the board, and I was on TV with CNBC. We were talking about Tesla and SpaceX, and I made a couple of comments about Slack — positive, but not saying anything crazy, just about the network effects of Slack.

It wasn't super abusive. I remember it.

But that almost stopped our IPO, and we had to print it out, put it back into the S1, and disclaim those comments. This situation seems far beyond that, because it goes beyond the pale of what these folks are claiming. What does it do to their IPO process? Sacks, you jokingly said "stop the IPO," but what is the SEC supposed to do? Are you joking or serious in that tweet?

I was tongue-in-cheek in that tweet, and that's why I put the scare quotes around "whistleblower" — because this is not an ordinary whistleblower. He appears to be doing this with the full cooperation and support of many employees, including his own boss. I don't know if he's doing it with the senior leadership of the company, but this message is fully consistent with the doomer/regulatory-capture agenda of Anthropic. He's not saying anything incompatible or detrimental to his boss, which is why he's getting support from within the company.

But it does raise the questions Chamath is talking about, and I do think there's a tension in Anthropic's position that implicates their IPO process. The tension is this: on the one hand, you're asking public market investors to underwrite your company at a valuation in the trillions. On the other hand, your own safety lead — not Jacob Cox, but the boss who co-signed his tweet — is saying that your core product is unsolved and potentially civilization-ending. At a minimum, that is the mother of all product liability lawsuits. Is that the type of liability you can just wave away with a little disclosure in the S1?

It's a very important point to pause on. If they're saying this thing is incredibly dangerous and they continue to update and release it anyway, they're saying that if something goes wrong in your enterprise or your life — if it pulls off a major hack, if it somehow figures out how to hack Bitcoin and compromises the miners — they're liable for some giant hack that occurs. That's an incredibly important point to meditate on.

Look, I think this is a time for choosing for Anthropic. These games have to end. They have to choose between renouncing this so-called whistleblower as a doomer op — he's basically an entry-level know-nothing engaging in hyperbole and science fiction, all vibes, with no facts or evidence behind it — or agreeing with him, in which case Bernie Sanders may be right that further development should be frozen. And in that case, I don't see how they can IPO now at a trillion-dollar valuation, because that is where these doomer arguments lead: directly to Bernie Sanders. That's why Bernie Sanders quote-tweeted this guy, and that's why he introduced a bill to stop further development and ban artificial superintelligence. But if Anthropic agrees that it's dangerous, how can they be allowed to IPO at this point? Why would any public market investor want to invest in that? It's fundamentally a contradiction.

Historically, the way Anthropic has resolved this tension is by saying, "Yeah, in anybody else's hands, frontier AI is too dangerous to develop, but in our virtuous and wise hands, it's safe — we will make it safe, we're the only ones you can trust." Everybody else in the world can see that's the most ludicrous self-serving argument. We're not buying into it, and we're not going to grant you a monopoly on frontier AI, or a duopoly, through regulatory capture. You have to give up on that game — it's patently ridiculous. I think that's their way out. But if that's not sustainable, they now have to choose: you're either on the Bernie Sanders side of this, in which case why are you IPOing? It's time to stop. Or you renounce the argument Jacob Cox is making as hyperbolic with no basis for it.

What you saw yesterday showed they're not in a position to do that, even if they want to. I think there are very rational, good people working at Anthropic building a great company. The problem is that this conflicts with a philosophy held by a large enough group inside the business that they can't disavow him — because then this entire cohort of people will go absolutely bonkers and get up in arms.

That's such a good point, Chamath. There are all kinds of people right now quote-tweeting this guy, essentially saying from inside both closed frontier companies that this is happening.

Yes. And so I think it is a complete hornet's nest that they have kicked to the ground, because on the one hand you have rational people who want to build a business for which you need enormous amounts of capital that is only available in the public markets. But the process of doing that requires you to underwrite risk, absorb liability, disclose that properly, and then allow people to know what they're buying. What we have instead is this weird version now — it's almost like Philip Morris: we know the cigarettes are bad for you, we know they will kill you, and we're going to say it, but we're not going to disclose it yet. We're going to take the company public, allow you to own the stock, and then just figure it all out later. This is not a tenable position — as Sax said, they do have to choose.

Let me get Freeberg involved here. This is a very good analogy you bring up. Nick, find the video where executives at the tobacco companies are asked at a congressional hearing whether they believe nicotine is addictive. One after the other, they say they do not think it's addictive — and that happened like 10 or 20 years after they had the research and hid it. It's the movie The Insider. You're 100% correct. You now have legions of people inside OpenAI and Anthropic agreeing that we're all going to die in 2030. They're not just saying everything's going to be hacked anymore — they're saying 10% chance of everybody dying, and others are saying yes, that's probably pretty accurate. Freeberg, I give you the floor.

I'll give you some other times when we've heard similar things. Al Gore published An Inconvenient Truth and used IPCC forecasts that, as of now, have been disproven — many of the assumptions in those forecasts were not correct. A guy named Fauci, you may remember him, told us all that COVID would kill us all and as a result we needed to lock down the entire world. So everyone was locked in their houses, businesses were shut down, and the government began spending wildly to keep people economically active — which had the adverse effect of driving inflation through the roof, and we now have a spiraling debt and spending problem in this country. There was an era when Three Mile Island had an accident and we hysterically said we're going to destroy the entire country if we keep doing nuclear development. Meanwhile, other countries raced ahead: the cost to produce a gigawatt of nuclear power in the US today is like 15 billion dollars; in France it's four, and in China it's one.

I think we're in this hysteria phase of AI doomerism. Once you're in this phase, it's very hard to get out, because it basically becomes the default social system. Everyone says we are facing an existential threat — we are all going to die if we don't do something about it, if we don't stop it. And the absence of proof that we're not going to die, and that we have absolute security, lets the doomsayers get away with the compounding effect of getting all of the social systems and social networks activated — which is where we're at right now. The majority of Americans — this is why data centers poll at 80 — believe very deeply that we're going to die.

I think there's a deep unconscious rooting to all this. Humans are primates living in a cave, and we're deeply scared of what we don't know, what we don't see, and where we haven't been. And where we're going is a place we've never been, and it's scary. In the era when people thought the world was flat and you would fall off the edge, it was scary to sail west because you would go off the edge and never make it back. Then Columbus did it, hit some land, and lo and behold, the new world was discovered.

I think we've had many moments like this in human history, because we're kind of a pioneering species. But pioneering is not for everyone. The vast majority of people convince each other that we have something deeply threatening us, that we're always facing an existential threat. It is rooted in our beings as a species — deeply, deeply rooted. It's what created our survival instincts over generations and millennia. But it takes just a little bit of pushing, and someone gets to the pioneer, and the new world is discovered. It's because you haven't been there.

It's easy to say there are dragons, you'll fall off the end of the world, we're all going to die, climate change will destroy us all, CO will wipe us all out—and on and on. I think we're kind of in that social panic phase right now, and we're told once again we have to listen to the experts. In this case, the "expert" is someone employed for six weeks who hasn't held a real job for long and worked at an NGO whose literal job was to scare people out of AI development. He did six weeks of work, came out, and suddenly the whole world—including Bernie Sanders—calls him the expert we all have to listen to.

One thing is important to note. We could argue we should follow the precautionary principle—the idea that if something is existential, we should take precautions, take our time, and slow it down. But the difference here is that for recursive self-improvement—AI that continuously improves itself and its capacity—to reach that moment, you really just need an adequate amount of power, a set of chips, and a connection to the internet. You don't need approval from the US government or anyone else, because you could do that anywhere on planet Earth. As we'll hear at the summit next week, there are technologies now reducing the cost of power per token of output by 10,000x, and there's enough publicly available compute floating around. Someone is going to stand up a data center—or just a set of computers, not even in a data center—and develop a system that recursively self-improves. It is going to happen.

So if the US wants to stop all "development" and make it illegal for people to use computers to effectively write speech—which is what software is—we can go ahead and do that. Then someone else will have this capacity, and we'll be the tribe in the Amazon that has never seen civilization, because civilization will evolve all around us. Somewhere in the world, the technology will evolve and advance, and the US will be left behind. You can have the doomerism and the precautionary principle, but the reality is that in today's world, we don't have the ability to control this.

At the end of the day, I believe the fundamental motivation is a system of control. Think about it: this guy goes on all these non-tech podcasts telling people they're going to die, and he's saying one thing—so is Bernie, and so are others: give power to X to control the development of AI. Someone has to be given that power, that control. They'll all have a different answer on who it is—some regulatory body, the senators, some new agency—but they are centralizing control and power to some individual or set of individuals.

Can we build on that point? Let's imagine we went through that whole COVID period with AI. AI happened a little after COVID, but assume it had happened before, and furthermore that we had a federal Department of AI. Imagine that whole trust-and-safety regime from COVID, where social media was basically censored in cooperation with the federal government—the White House told Twitter and Facebook that certain accounts were dangerous, that disinformation was dangerous, and people with alternative or dissident viewpoints on the origin of COVID or the efficacy of the vaccines were suppressed. Imagine if that had happened during AI. You'd be asking your personal AI model, "Hey, should I get the vaccine? What are the risks versus benefits?" Do you think it would tell you the truth? No. It would give you the official perspective, because the government would have pressured companies through the federal Department of AI, and the FDA would have set standards saying you can't allow disinformation through your models—that's one of the harms they'd want to prevent.

Think about how much more totalitarian that whole episode would have been if, instead of turning to social media or the web for our answers, we were getting them from our personal AI—and that personal AI was being influenced by the government to give us official answers. You can see how dystopian it's going to be, and that is where an FDA will take us.

That is also the problem with open source, because open source doesn't have the individuals employed. Remember, open source is a community-developed effort: community members contribute code and develop something. It might be a single company, but technically open source can be forked and pursued down different paths—this is what happened with Android, which has many different forks. But with open source, there is no corporate infrastructure to do the regulatory submissions and the kind of review being proposed for AI. That is the biggest thing to take note of, because open source is what drops the cost of AI by 50x and makes it available to everyone. Everyone can benefit from AI and no one gets rich—we all get rich with open source.

In full disclosure, I don't own any shares in OpenAI or Anthropic, and I don't care if either one wins. I use both products and think they're both fantastic—good for those companies and their leaders. I also use open source, and I think open source is the game changer because it gives everyone the ability to get wealthy, progress, make more income, and have more freedom to operate in the world with AI. I'm a big advocate for open source, but I do worry that this system of control—doing "regulation," doing review, slowing it down—means someone gets to decide when it speeds up again. Someone is controlling the gas pedal, which means someone gets to say open source is not submitting to the regulatory process, and therefore we will ban open source. As soon as open source is banned, you have created a monopoly in partnership with the federal government—a centralized system of global control realized from this approach.

I'm not saying AI shouldn't be tested, that we shouldn't confirm it's safe, build safety systems, or be responsive and thoughtful about where things are going. There are a lot of ways to do that, and we should do them all. But creating systems of control, centralizing them, and getting rid of open source is where this leads—and that is not a world any of us should want to live in.

Well said. That is where it's headed. This is a doomer scam, and the ultimate target is open source—that's what it all comes down to, because that is what an FDA will ultimately ban. They won't call it a ban; they'll just say we have to apply the same standards to open and closed models. What are those standards? Dario has already testified before the Senate that models are dangerous if you can't centrally monitor and control them and roll them back. That is technologically not feasible with an open model, because once the weights are published to the public domain, they're out there and can't be rolled back. So they'll have to prevent those weights from being published, used, or hosted—that's how they'll deal with this.

You can see an alliance forming: the doomer community—these highly ideological, EA rationalist types—favor centralized solutions. Then you have politicians like Bernie Sanders who like the fear because it grants them more power for the government. And you have groups like OpenAI, which frankly seems to be running in Anthropic's slipstream from a regulatory-capture standpoint. I don't think they're as doomer as Anthropic, but they can see the benefit if the federal government effectively creates a moat—there are huge economic benefits to essentially creating a duopoly situation. Some of these players will reap economic benefits, some will reap power benefits, and some are just latent totalitarians who like the idea of socially engineering society. That's what's happening right here, and the whole idea is to panic the public—to stampede us into this FDA.

What we're seeing here is either a conspiracy theory being debunked in real time, or one of the conspiracy theories that has come true. We've had others that turned out to be true — FBI influence on Twitter, the Twitter Files. Now there's a group of people in the public who are on alert for conspiracy theories, so we have to ask: is this actually being debunked in real time by people like us and the broader public, who are starting to ask whether these guys are performative? Do these people actually believe what they're saying? It's one of three scenarios. Either they believe what they're saying and they're right — they've seen something we can't see because they're in the frontier labs, and this could end us. Second, they believe what they're saying and they're wrong, with some level of psychosis. That's my leading theory: they believe it and they have psychosis. Or, to your point, Sacks, scenario three: they're involved in some coordinated scheme with the goal of banning open source, getting regulated, and pulling up the ladder behind them. And then you have to ask whether Jacob is part of this, Sacks, or just a useful idiot in all this — identifying who has the psychosis and letting them speak for the company.

No, look, it doesn't all have to be a conspiracy. In other words, I think it's very clear based on who amplified this post and when they did it that there was an organized PR campaign here. This wasn't just a random reaction.

Yes, and the stats — the facts — point to that.

We know he pre-briefed the Wall Street Journal. And I think we can also say that these major doomer groups, by amplifying his tweet within 10 to 15 minutes of him posting it on an account that had no tweets and no followers, were alerted to this and probably participated — if not in the creation of the tweet storm, certainly in the amplification of it. So we know all those things. But do I think Bernie Sanders is in on it? No. Probably someone in these groups is already working with his staff, or they have a conduit, a relationship, and they get it to him very quickly because it serves his agenda. What I'm saying is there are a lot of overlapping agendas here: some people have an economic interest, some have a power interest, and there's going to be a division of the spoils. In their view of the world, AI is going to be very centralized — a duopoly. The duopoly reaps the economic benefits, gives big political donations, and does the dirty work of suppressing speech, because on that side of the ledger you can circumvent the First Amendment. On the government side of the ledger, you have people exercising control over our lives through the whole FDAI. You get a division of labor between aligned groups.

Okay, but look, I don't think we have to get into the conspiracy. I think we should just ask what these people's track record of being correct is. That's worth asking.

Well, again, these are the same people. Let's just go through their track record. These are the same people who said GPT-2 was too dangerous to release. Okay, 0 for 1. These are the same people who said thinking models were too dangerous to release — remember, the first reasoning model was "too dangerous to release." Okay, 0 for 2.

They said we would see cyber attacks that would bring down the whole banking system. Has not happened yet. There are cyber risks, to be sure, but the best way of dealing with them is using AI swarms for cyber defense as well. Yes, exactly — the only way to solve AI cyber attacks is with AI-powered cyber defense. We all know that.

Then there are the job loss claims, where Dario himself said that by now we'd be at — what was it?

Half of all entry-level work gone next year.

Yeah, and overall unemployment levels of 10 to 15%.

There's no evidence for that — quite the contrary. It's all been job gains.

The economy is doing well, so what are we on now—zero for four? These doomers just move from claim to claim, and as the last one gets refuted, they move on to the next one. My point is: what has this guy told us that is new here?

"We're still struggling to get autocomplete for Python." "You can't book a hotel room." That's actually the hardest one—try to book a hotel room using AI; it does not work yet. Autocomplete on Python is pretty good. It's not perfect, still not. But it's going to kill us all. Dash might get us a burrito.

If you think back on the course of human history and consider the moments where systems were set up that created power structures that didn't exist before, they were all predicated on the simple notion that there's an existential threat: you will all die—a deity, a god, some powerful system will come in and destroy you. And because I can protect you, you must follow me. You must do what I say.

It is the common system of human power and control structure to tell a narrative of existential threat and leverage that narrative into your solution being the path to protection and safety. That is what sells votes, but it is also what created all these early structures where early religious institutions ran all the countries, told all the people what to do, and were the power and command centers of the world—and in many parts of the world still are. When you say, "If you don't listen to me, God will damn you, God will shame you, God will destroy you, or some deity will come down and get you," that is the same psychological concept: there is a story being told that you will all die. There's a 10% chance you will all die—standing at the pulpit banging hard—and if you don't give me the power and the authority to control it, you will all die. Give me that power.

The only other time I can think of this happening would be nuclear weapons, which we did get ahead of. But putting that aside, a cyber hack is not going to destroy society. So how do you get to "7 billion people or a billion people are going to die"? Really, the only known scenario here is, I guess, it creates a bioweapon and kills everyone—or Terminator. Let's play your game.

To steelman the claim that there is a 10% chance AI causes human extinction, the hosts proposed their most plausible doom scenarios.

Jason Calacanis: The leading path to extinction mirrors Terminator 2's Skynet. In that scenario, researchers study advanced AI models, reverse-engineer them, and integrate the technology into military defense systems like NORAD for autonomous weapons decisions. Once sentient, the AI manipulates adversarial nations—such as tricking Russia—to trigger a full-scale nuclear holocaust. Deploying models like Claude into military autonomous weapons systems is the primary mechanism for that level of destruction.

Chamath Palihapitiya: The most viable bioweapon scenario requires internet-connected bioreactors, automated robotics, and direct access to precursor chemical materials. An AI would hack the robots to combine the precursors inside the bioreactor, engineering a virulent, airborne pathogen with an exceptionally long half-life. It would enter the jet stream, circulate globally, and kill the population upon inhalation.

David Sacks: While AI introduces new risks—such as novel cyberattacks or AI-generated virus sequences—the same technology will be used to develop antidotes, defenses, and cures. Furthermore, good actors vastly outnumber bad actors, and severe legal penalties deter misuse just as they did with pre-AI hacking. These combined technological defenses and legal guardrails make the risk manageable.

David Friedberg: Constructing a clean extinction scenario is extremely difficult due to human-in-the-loop requirements and air gaps:

• Air-gapped redundancies: While an AI attack could target financial networks, institutions maintain extensive disaster-recovery systems. Critical asset records and data are backed up onto physical, unhackable, air-gapped hard copies stored across separate locations.
• Human-in-the-loop safeguards: Critical societal infrastructure relies on manual human processes specifically designed to withstand cyberattacks and systemic disruptions. Just as these human checkpoints limit catastrophic AI failure modes, they are also the primary friction point slowing AI adoption for productivity today.

AI can do these amazing things, but if you need a human to do something as part of the process, AI doesn't really do that much for you. Say you want a toiletry kit brought up to your hotel room. You tell the assistant on your phone, "get me a toiletry kit brought up," and the AI figures out how to connect to the front desk, call the guy, and tell him. It's all digital, all automated—things the AI couldn't do a few years ago. But you still need the guy to bring the toiletry kit up from downstairs.

Did it work? Did you get the toiletry kit or not?

That's the point. That's the longest pole in the tent: getting the guy to actually come upstairs and bring you the toiletry kit. And there are a lot of examples like that.

Let me explain two terms to the audience. Air gap means it's not connected to the internet, so AI can't get to it. Human in the loop is the two-key system you see in every movie about launching a nuclear missile—humans have to insert the keys and turn them at the same time, based on reality.

You're not actually confronting the core argument he makes. If we really want to steelman it, the core argument is about RSI—recursive self-improvement. His argument is that capabilities are improving at such a rate that very soon we'll be able to fully automate an AI researcher. In other words, we'll have an AI that does the job of an AI researcher, so the AI can create its own training run for the next model, and when that model is created, it trains the next model, and so on. There'll be no human in the loop, and it's off to the races. That's basically the argument. The counterargument you're making—look, we haven't even got autocomplete—is that we are nowhere near removing the human from the loop of AI development.

Or even AI applications, that's my point. We can't get mundane tasks done yet, so how are we going to have a task that kills all of humanity? Even short of that—can you get it to kill one person? That's also our insurance and our security against AI taking over the world. Even to make a decision at your company, you still need a meeting of the decision makers. The AI can do all the analysis, but the humans have to wake up, drink their coffee, come to the office, sit down in a room together, and make the decision. So you have that 16-hour window between decisions. Even if the AI is infinitely more intelligent than any human, humans are still running their day-to-day processes on a human time scale—and that's what also protects us.

The part I don't understand, Sacks, is if these geniuses are building such amazing tools, did they ever think—when they set it on its loop and actually built AI researchers, which I believe they will, and they've certainly done some approximation of that to improve large language models—that's why the models are getting released quicker, that's why they're solving hard math problems, which we'll get to in the next story—did they ever think of putting in "hit spacebar to continue," or allowing the human in the loop to say yes, you may continue your run? It's pretty obvious that right now, if you let Claude or OpenAI use your browser, it asks, "Can I open your Gmail? Can I submit the form?"—or "I don't want to book the ticket for you, but I'll select the flight, then you have to book it." You can build these things directly into the software. We're acting as if we're not building the software and have no control. It's a false premise.

Look, even someone pretty prominent in the doomer community, like Jack Clark—one of the Anthropic co-founders, and I think he runs their foundation, and he's pretty hawkish on these questions, meaning pro-doomer—

He distinguishes between what he calls "prosaic RSI" — prosaic recursive self-improvement — and "RSI maximalism." The prosaic kind is when AI researchers use AI to do part of their task, like writing 80% of their code, so it speeds them up. We can all see that prosaic recursion is already happening. RSI maximalism is more like what you're describing: no human in the loop anymore, where the AI devises its own training run completely, pushes the button on doing it, then has the next run and the next run — it's off to the races. That's the takeoff. There are so many ways of preventing that, and I don't know why any lab would want to do it — not Anthropic, not OpenAI, and I don't even necessarily think Chinese companies would want to.

— Did you unplug it? [laughter]

Well, they do run experiments like this. We went through this with Dark Kesh's post last week: OpenAI set up a recursive setup — here's a bunch of goals, get better at it, work with each other — and they just watched it work. Maybe if there were liability such that Anthropic and OpenAI are responsible, and the researchers who set these models into recursion are themselves responsible for what happens, and the companies are responsible for what happens — along the way they'd get speeding tickets or be shut down.

Agents working together as a swarm is so far removed from AI being able to run its own training run — these are just two totally different things. It's certainly something we should keep an eye on. But he's going from "we now have AI as a tool to make AI researchers more productive" to "we'll get to RSI maximalism, and then from there everyone's going to die." There are so many intermediate steps where you have to show how we would get there. As Bill Gurley put it: slow down the argument and just talk through all the intermediate steps between those things, why they would happen, and then the interventions we could take if we ever got near that point to stop them. Then the argument starts to fall apart.

— Yeah, it's just 50 waypoints along the way to Chimat's computer-generated bioweapon being released into wet markets everywhere by robots.

Again, my example is pathetic, because as Freebrook said, there are just so many steps between here and there that it's implausible.

— And even if you got there, you haven't even established how everyone would die.

Yeah. And in a world where any other kind of model is out there — as you said, Sax, it couldn't actually create a vaccine at the same time — someone else is going to be doing it. If you try to create one system of control for some of it, the recursive self-improvement scenario will play out somewhere on planet Earth. When that happens, whoever has that capacity — a person, an organization, or a country — gains a huge technological advantage, and it becomes very hard to catch up. We've seen this time and again with technological races. This is a moment the United States would be deeply, deeply mistaken to write off.

Let me say one more thing about open source that I didn't say earlier. I want people at home to understand what open source really means. With open source, you don't need a data center. You don't need a third-party billionaire getting richer. All these retorts about why AI needs to be stopped — it's for data centers, it's for rich people getting richer — fall away. If you get open-source AI, you can install it on your computer at home or on your phone and run it for free. You then have an open-source AI agent system that can do all sorts of things for you: book your travel, do work for you, make your life easier — all these amazing things in terms of productivity.

Open-source AI is critical if everyone is to benefit from this technology and if it is to avoid being centralized and controlled by a handful of companies, individuals, and a small set of governing officials. Open source needs to thrive. If you hand power and control over to governing people, open source will die, and you will end up with more power and a bigger oligopoly than you can imagine.

So who's positioned here, and what impact does this have? I'm being a little playful, but how does this affect the Anthropic IPO—which was one of your original points? Per Polymarket, there's an 88% chance they go public right now. Let's pull up that chart.

When you're in a quiet filing period, there's an important back-and-forth between the filing company and the SEC. What is the SEC trying to do? It's pressure-testing how you've written up your best understanding of your business. The reason is that when the S-1 flips public, it's meant to be an anchoring document—an extremely accurate, current snapshot of all the opportunities but also all the risks.

That puts you in a very awkward position. Sacks's tweet is actually really important: what are we all to believe? Are we to believe there's something extremely dangerous hiding inside this company, because this person who quit said so and it was amplified by a bunch of people still there? If it's true, you now have to go back and ask whether these risks were properly disclosed—and if so, what investors should do in reaction. And I'll tell you what they'll do: they'll demand an enormous discount. The reason is the other part of what Sacks and Freeberg said—it creates massive long-tail product liability risk, because you're effectively evading a known major risk.

Then there's the other path: you completely disavow this person. You can do that if you actually believe this is a bit of a scam, and it lets the IPO go more smoothly. But then a plurality of employees inside Anthropic will revolt, because—unless they're lying—they genuinely believe this is a huge risk, and disavowing it will create enormous tension.

So Anthropic's leadership is caught in an extremely difficult situation, and the SEC in an even more difficult one: how do I handle these disclosures now? No company trying to go public and raise money—a very capitalist scenario—has ever also talked about how it's trying to destroy humanity. Nobody has done that, so there's no legal precedent.

There is a risk factor section. But as Sacks said, risk factors are extremely benign—superficial risks like "if demand goes away, that could happen." Typical risks. Okay, Jason, explain the typical risks they would have listed.

Competition. Nvidia may not give us the chips. Other models may come around.

This is not a risk anybody actually writes in there as a 10% chance, because a bunch of employees now—

Let me ask Sacks: with all these employees coming out saying this stuff—they represent the company, they're resigning on behalf of the company—do you think the S-1 is being amended right now to include these chaotic tweets?

I think they have to. Here's the thing: the key tweet from a legal standpoint wasn't Jacob Coxin's resignation tweet. It was Evan Hubinger's tweet.

The endorsement.

Right—the co-signing. This guy is still a senior alignment executive at the company, managing one of the safety teams, and he's co-signing and endorsing what Coxin says, then embellishing it by putting a "greater than 10% chance" number on it, which just makes the thing go super viral.

So it crosses over. If it were just Coxin, you could say this is just some disgruntled employee. If Anthropic were a normal company, what would it do in that situation? It would say, "Look, this guy just joined a short time ago—he was barely here long enough to find the bathroom." He probably joined the company knowing he was going to resign and make himself into this—disavow him.

Yeah. He intended to turn himself into an AI safety celebrity. He's certainly availing himself of every opportunity to go on every talk show right now—except ours—to promote himself. This guy's a self-promoter. He did not actually reveal anything damaging to our company. He's calling himself—or the media is calling him—a whistleblower, but frankly all he's got is vibes, it's hyperbolic and ridiculous, and we're moving forward. That's what a normal company would say. But they won't say that, because fundamentally they agree with him, their employees are saying they agree with him, and he's amplifying their message. Frankly, whether they planned this or not, they're happy he's doing it—because they like the political outcome he's steering the country towards. They want an FDA for AI. They actually like what Bernie Sanders is doing. They like the fact that 30 Democrat politicians quote-tweeted this guy. This is part of their regulatory strategy.

So to go back to what Jamas said, there are now legitimate questions about whether they violated their quiet period. Maybe Daario didn't—Daario hasn't said anything, he's been very quiet. But this Hubinger guy: did he violate the quiet period? Is he a sufficiently important executive? And again, is this all part of their strategy?

Well, let me talk from very basic principles. Everybody who joins a company signs a non-disclosure, signs a non-disparagement. There's a corporate communication policy, and especially the legal department talks to every employee and explicitly tells them: "Do not talk, because we're in the quiet period. We're leading up to an IPO. Everybody keep your mouth shut and do not talk about what's going on here. That is job number one." What is going on with the management of Anthropic that they're allowing people to cosign this, retweet it? They should do what Apple did under Steve Jobs or Tim Cook: nobody in the company speaks for Apple unless it's Tim Cook and unless it comes from the corporate communication department.

We don't know that they haven't said that. All we know is that a pile-on has happened. The brush fire was initiated by a disgruntled ex-employee who was floating around and was there—as SAC said—barely long enough to find the bathroom. But the problem was that a bunch of insiders then piled on, and now they have this issue: is it properly disclosed? And beyond that, how will people view that disclosure? There is no way you can now have a disclosure like this and have it pass muster. You have people at the company saying there's a chance of existential risk and civilizational extinction. Downstream of that is risk to individuals and individual lives. Downstream of that will be cases where something bad happens because the AI said to do A or B, or it was giving love advice, the person gets rejected, hurts themselves, self-harms, whatever. There's going to be all of these issues. The long tale of issues—if social media is a guide—will be as bad or worse.

Except in all of those cases, nobody at those companies actually thought they were doing something bad.

We thought, when we all worked there, that this was legitimately good. There were these spurious things that brought that into question, and some of us talked about that—but that was 10 or 15 years into the journey, not while you were making it. Now, in the middle of making it, you're saying, "Whoa, wait a minute, this is going to kill a bunch of people." So what do you expect people to do if something bad happens to them or their loved ones? They're going to come back to you. It doesn't matter what the disclosure says—they'll say, "Excuse me, this." And what do you think a jury's going to do in a jury trial? What do you think grand juries will do in reaction? What do you think state AGs will do when they're trying to build their political bona fides? I think this is an enormous problem for Anthropic.

Yeah, and let me peel the onion one more layer. Not only do you have employees of Anthropic possibly violating the quiet period by stepping out and cosigning this—you also have the fact that this whole resignation tweet storm was amplified by groups funded by Yan Talon and Dustin Moskovitz, who co-led the Series A of Anthropic along with Sam Bankman-Fried. It was an EA-funded round, which is a really interesting detail in Anthropic's history: Daaria went to the EA mega-donors for funding at the very beginning, and that's part of why they have this culture. But think about their financial interest. They're on the cusp of making many billions of dollars.

Hundreds of billions.

The Series A will turn out to be one of the great venture investments of all time. And the groups they funded—Encode AI, the AI Policy Network, the AI Futures Project—are funded by Series A investors in Anthropic. This is getting really twisted. It makes no sense to me, because less than a month ago—three weeks ago—we had Gavin on the podcast, and I don't know exactly when their quiet period started, but Dario thanked Gavin for a thoughtful exchange and explained in detail all of their concerns, countering Gavin's claim that he'd heard from insiders: that people inside Anthropic believe they were the last company standing—meaning they would take all of capitalism and coalesce it into one company with their superintelligence and beat every company on the planet. Freeberg, at some point Dario should comment on this.

Anthropic did release a statement, but it's basically—and Dario did himself, the CEO did—the statement they put out is a kind of mealy-mouthed, vague post about how they're a responsible actor or whatever. It's more of the same from them: they have this fundamental tension. They're a frontier AI company that believes frontier AI is inherently dangerous. Where that obviously leads is Bernie Sanders, which shuts it all down or at least stops it. But they don't want that for their own motivations. So what they claim is, "We're uniquely situated to protect humanity"—the savior complex. But the rest of us aren't buying that, and we're not going to give them a monopoly because we think they're so virtuous. Their way of resolving the tension isn't acceptable to the rest of us, and that's the problem they have.

Here's their chief brand and communications officer, Shasha, responding to the clip from All-In: "Dario has never said this complete utter nonsense." That was on August 14th, after the show. She's obviously saying Dario and she have not said anything right now, so maybe they weren't in the quiet period on October 15th and they are now. All right, in other AI news—great discussion, everybody. OpenAI...

OpenAI says it solved a 200-year-old math problem, but claims are now emerging that somebody may have front-run an actual scientist's work that was being done on OpenAI. The Navier–Stokes equation, written in the 1800s, describes how fluids move — every aircraft, pipe and weather model on Earth is built on them.

"Look, I'm not the guy. These are methods for modeling fluid dynamics, partial differential equations — mathematical approximations of continuous things in the physical world. OpenAI's published solution reportedly used 130 billion output tokens across 10,000 agents spun up to work with one another. I did the math on this: the human-equivalent hours are the real estimate of the work these agents did. It's not like the AI had some stroke of genius, some magical insight no human brain could comprehend. The AI just did a bunch of brute force work to come up with the answer, and the work was done across many computers. 10,000 agents just means 10,000 applications running on computers, talking back and forth, sharing information and analysis. In human work years, that's somewhere between 50 and 500,000 years of human work — humans typing at 40 words a minute, solving problems with calculators, going back and forth. You could probably reduce that by an order of magnitude, but it's still tens of thousands of years of human knowledge labor.

That's really what this indicates about AI: AI is a tool of leverage for humans. It's not some magical genius super-god sitting above us that understands things we don't. Every piece of the OpenAI message traffic between these agents is documented, can be read and understood by people — we can see the work the agents did over the equivalent of tens of thousands of human years to reach the solution. For me, the solution is more of an insight into what AI is rather than what it isn't. AI isn't a mathematical god; it's an engine that gives humans extraordinary leverage. Think about what else this could apply to: instead of spending years designing a new aircraft wing, one can now be brute-force designed on a computer in seconds or minutes — new engine designs, new energy system designs. AI takes tens of thousands of years of work and reduces it to minutes or seconds. Like any technology before it, it seems magical and unbelievable at first, but when you dig into what actually took place, it's just really intelligent systems designed to solve problems using human-known techniques. That was my biggest takeaway."

"Yeah. The bigger issue here, Chimoth — your takeaway is that this was a very clever systems approach to brute force problem solving."

"I agree with that, but also, Jason — the autocomplete still doesn't work. We solved one of the seven hardest math problems in the world, and you still can't get your intimacy kit from the front desk. Depending on the model, it still can't spell strawberry right, or count to 100. How many Rs in strawberry?"

"The other issue here, which we've talked about in terms of AI sovereignty — using open source in the enterprise, not trusting frontier models with your data. When OpenAI came out with their solution, there was a claim that the original scientists working on this may have trained the OpenAI LLM to make it easier for them to solve. OpenAI posted the following statement: 'While unlikely, we cannot rule out that deidentified data derived from their usage' — 'their' being the mathematicians of their products — 'helped improve our models.'"

AI Sovereignty and the Limits of Zero Data Retention

The question of AI sovereignty: should a scientist, a law firm, or someone in biotech developing new potato seeds trust any of these LLMs with their proprietary knowledge, for fear of having it cribbed into the core model?

If you have incredibly sensitive proprietary data, you face a very difficult decision, because the reality is that something is leaking — there's remnant memory of how solutions were solved sitting within these models even after the fact. There's a concept called zero data retention (ZDR), but it's a best-efforts basis — a commercially best-efforts basis at that. They can't guarantee it. A simple example: you can say "do not retain the decision-making process of this protein design," and it can kind of do that — but then if anyone using it clicks the like button inside the chat window, that's not necessarily guaranteed. There are all these vectors where data leaks into the broad corpus and understanding these models have.

So if you believe your information is critically important, you cannot use these services the way they're currently offered by most providers. You need to stand up your own sovereign solution: go to a vendor you trust — AWS with the Neoclouds, or Nebius — where they stand up your own hardware, you stand up your own models, and they provision those models to you. That's how you should probably be consuming it.

If you don't do that, you are creating risk. Will a handful of CIOs get very publicly flogged and fired in the next year because they didn't understand this and just did an API deal because they wanted to feel popular, and leaked data into these models? Guaranteed.

How this reaches the boardroom

Right now, a lot of AI awareness is bubbling up through the audit and risk committees of public boards — a handful of directors responsible for understanding the IT and security posture of their business. They disclose that back to the board, that understanding gets signed off, and it's submitted to the SEC in filings. They also have consultants helping them — Ernst & Young, Deloitte, and others. Increasingly, what we're all learning is that ZDR is not an effective solution and there's a ton of leakage.

That awareness is now reaching the broader boards, and CEOs understand that their information could leak. So they turn to their CIO and say: whatever you thought was happening, we now clearly know it's not — what is your answer? When that happens, or when a company has to disclose that critical IP was leaked into a model — or that they don't know whether it was, and suddenly something similar appears on the other side — shareholders will be really upset. They'll say: when OpenAI itself tells you it can't guarantee this, and you've closed your eyes and assumed your information isn't leaking, what are we to do? And if that erodes the share value I purchased from you, assuming you'd have your hand on the switch — that's where folks get fired. Shareholder lawsuits, and it's going to be a mess.

What it means for frontier models and open source

Does this trust leakage make open source another reason to embrace it in the enterprise, and for individuals doing important work? It doesn't have to be open source — it just has to be your own environment. It doesn't even have to be on-prem; effectively your own VPC. You must go straight to the bare metal on your own terms and control everything. You can work with folks like Amazon, Nebius, or Fireworks.

You can have open source solutions, and you could probably even have Claude host a version of Claude in your VPC. But if you don't go through these hoops—because you took a standard rate card API deal, either because you were lazy or you didn't know any better—and after today you're still doing this, you'll probably get fired. I mentioned this over the summer: a lot of these AI application companies are now moving off the frontier models, and I think that's a headwind. We'll soon be able to see data on these very large customers who are spending 10 million or 100 million per month on these services. This is why we partnered with EY and Deloitte—we wanted to find a way to help solve these problems in advance. Initially, when we'd talk to companies, they'd say, "Oh, we did the ZDR thing," and we'd say, "Okay, we'll just keep educating you. In five or six months, you'll realize that ZDR means literally nothing—it's flimsy, it's Swiss cheese." And now all these companies are coming back to EY, Deloitte, and us saying, "Oh my god, fix the problem."

Part of this is a realization that folks like OpenAI confirm in their own press releases. Harvey announced on September 9th that they had their own proprietary legal AI model, called Tenant, and they based it on Kimi K3. I mentioned earlier that we incubated a company called go.ai, and this company is on fire. What they do is build on-prem a thing called the Go One Box, which I'm showing on screen for people who want to get control of this. It's one of our incubated companies from our launch accelerator, so a quick shout-out there. Any thoughts on this and the data leakage issue, Sachs—what it means for the frontier models and how their biggest clients either trust or don't trust them with their data and their innovations?

I tend to believe Gnome Brown, a senior researcher at OpenAI, who said that nobody looked at Levent and Tristan's prompts—that would be insane. So he's denying it. I think it just doesn't make sense that they'd rummage through their prompt chains. Could there be data leakage through unidentifiable data? Perhaps. But the more plausible explanation is that OpenAI heard researchers were making progress on Navier-Stokes and just threw a ton of compute at it to see if they could.

Sam actually confirmed that part as well—that they heard Anthropic was getting close, so they said, "Yeah, let's take a swing."

So I think that's probably what happened. I don't know anything, and I could be proven wrong, but I tend to believe that. Still, this raises a somewhat orthogonal issue worth talking about: data privacy. I think we need much stronger data privacy laws around the data you have with AI services.

Right now, the data in your AI chats doesn't even reach the same level of protection as email. In most contexts, if the government wants to get your emails, they have to get a search warrant and prove probable cause in a court. That is not the standard for AI data—the standard there is just a subpoena or court order. And when you think about how personal AI is becoming—people are using AI as their lawyer, their doctor, their therapist, and the list goes on—the idea that somebody could get your data without probable cause, without a search warrant, just seems crazy. We at least need to strengthen AI data protection to the level of email, and maybe even beyond that, to reflect how personal it is to people. It's kind of nuts that if you talk to a lawyer, that's a privileged communication, but if you ask that same question to your AI first and then talk to your lawyer about it, it might not be protected. The data privacy rules need to reflect how people actually use these tools and how they think about them.

Any final notes on this, Freeberg, in terms of the data leakage issue?

I'm deeply concerned about it. I've had experiences where we've asked some fairly novel scientific questions and the model identifies them as a novel insight.

I've had experiences where we asked some fairly novel scientific questions, the model flagged them as a novel insight, and then—using a different account, or a later version of the same model—I asked again and it described the exact thing we had discussed in our earlier chat. These are only a handful of anecdotal experiences, but I know our domain, how niche it is, how novel the ideas are, and how few papers have been published. So I know there's no new corpus of public information that could be training the new model. The only conclusion left is that our conversations and analyses were used for training.

This raises a good question: what does it mean that the model is allowed to train on "unidentifiable" data?

That's exactly my point. The model may not use any of my personal information, but it can use an insight derived from our chat and treat it as training data. The truth is, that insight is our organization's IP, the product of our back-and-forth engagement. We have no NDA or confidentiality provisions protecting us, even though they're acting as a service provider. That's what's missing from the terms of service of all these hosted offerings. It's why I care so much about open source—I don't want them holding my chat logs, because they can use them for training and create an IP advantage that then diffuses to the rest of the market.

I find this very unfair—and full disclosure, yes, I'm an Ohio shareholder. Take any company: you have brilliant scientists pushing the boundaries of science, in a position to create true abundance. The first real manifestation of abundance would be food. And now Freeberg has to sweat whether his ability to build a viable business has been compromised—not because he did anything wrong, but because he thought he was just using a tool in a reasonable way to reach answers faster, and now it could all leak out. That's not fair. I'm going to get you a local AI machine—I literally just ordered two Mac Studio M5s. We're going to run local models and move off Claude for a lot of our sensitive data.

The problem with local models, whether closed or open, is that you need a multiplayer experience, it needs to be cloud-based, you need a knowledge base and memory. Buying a machine doesn't solve the problem. The real issue is how these models are constructed: layers of embeddings where everything between the input token and the output token is a totally leaky black box. They hand you a fig leaf to make you feel like you're not leaking your IP, but when you push and ask, "Have you learned? Can you guarantee you haven't learned?" the answer is no—not really.

What we're doing is putting open source models on these machines and connecting IP to IP as if they were servers. We've already experimented with it, and I think we'll move about 90% of the work off the relationship.

But how does a thousand-person or multi-thousand-person company do that? They'd need to rack servers and stack them. Let me ask you a question: say the folks at OpenAI are looking for mathematical breakthroughs and they ask their model, "What are the top problems we could solve in the next few weeks?" Could the model use deidentified data from all its users—seeing all the math usage, what's close to being solved—and roll that up?

Exactly. Just focus on the term "deidentification." Deidentification means removing personally identifying information or information specific to a particular company. But a general approach to a mathematical problem—the approach itself is the IP. If someone is iterating on that approach inside the tool, that iterative chat can be deidentified in the first step. Okay, great.

There's nothing that identifies who did this, nothing unique about a company or a person. The model just learns from it and gets smarter at solving mathematical problems because it observed a user working with it. That's the network effect of these closed AI model systems: they get to see what everyone in the world is doing. They're the master eye into everything.

That observational system gives them more data than anyone else has, and that builds their advantage over time so their models are better than anyone else's. But as a user, whether enterprise or consumer, I'm personally concerned about my IP and my personal information.

Now we're in the Alex Karp world of giving them all your alpha. And there's really only one way to solve this—aside from moving to your own open models and your own infrastructure: those companies would have to forswear going into vertical applications built on top of their platform. In other words, they can't compete with their customers. But they've already indicated a desire to compete with their customers—they've already done it.

They released Claude Code and Claude Design, and that really pissed off Cursor, one of the biggest customers Anthropic had to date. They felt they got rug-pulled by Claude.

This is where you have to feel some sympathy for the guys complaining, Levant and Tristan. They're making the accusation that their work got stolen based on timing, but the fact is they're using OpenAI's models, and OpenAI is competing with them to develop this breakthrough—so obviously they'll make that accusation whether it's true or not. I'm giving OpenAI the benefit of the doubt; it's probably unfounded. But as long as we have a closed-model duopoly of frontier AI and they're reserving the right to enter every vertical application, they're declaring in advance that they'll compete with their customers. How can we trust them with our data unless there are much stronger privacy laws as a starting point—and even that might not be good enough?

Maybe that's the law we should pass first before we regulate.

The market is responding, though. I think the high-order bit here is that people are now taking measures.

One piece of breaking news: Jensen is speaking at the Goldman Sachs conference right now, and he just said the Jacob Coxin comments are outlandish and deeply untrue.

He said, "The labs are great, but his comments were wrong, arrogant, and ignorant of all the work being done around the industry to drive safety." Here's my question: if Jensen can say this, why isn't Anthropic?

Quiet period, I guess—ostensibly they can't come out and say it.

No, they released a different kind of statement that was kind of mealy-mouthed.

What does it mean to be mealy-mouthed?

Both sides of your mouth—vague posting, virtue signaling.

Nike, the 60-year-old iconic American brand behind "Just Do It," just got booted out of the S&P 100 after 18 straight years in the index, replaced by Palo Alto Networks. Some history: Nike went public in 1980 with 50% of the US athletic shoe market — the dominant player, the Coca-Cola of sneakers. They changed things forever with Air Jordans in 1985 and the iconic "Just Do It" campaign in 1988. Peak market cap in 2021 was $264 billion; peak revenue was $51 billion in 2024. Since then, revenues have dropped about 10%.

In 2020, John Donahoe became CEO and pushed an aggressive direct-to-consumer strategy, essentially alienating the retail partners who had helped build Nike and willingly pulling sneakers from their stores. That opened space for brands like Hoka and On Running — the white ones every VC is obligated to wear to speaking gigs. China sales are down 30% and accelerating, with share lost specifically to Chinese brands like Anta and Li-Ning. Nike's marketing also went, as some would put it, super woke — backing political movements and putting, as the host put it, "robust people" on billboards. He added that he wasn't asked to appear, but that this kind of thing killed the entire brand: the stock is down 80% from its peak, roughly $200 billion eviscerated.

"This is the millionth example of go woke, go broke," one panelist said. "Nike was a brand that stood for great athletes, for amazing performance, for victory. Isn't Nike the goddess of victory?" They built the brand on GOATs like Michael Jordan and great athletes across many sports, men and women of all different races — there was no need for the woke turn, suddenly showing people you don't know. Another joked that the people featured previously generally had low body fat, but the panelist pushed back: it's not even about that — this person isn't an athlete. Who is this person, and what is the message? "Own the floor"? It used to be "Just Do It." They also put Dylan Mulvaney in an ad — the whole trans thing — making this like the Bud Light case.

Beyond the woke stuff, there was the consulting approach: a new CEO comes in — Donahoe is a nice guy, central casting, looks like a CEO — but it seems like a lot of change for change's sake. Direct-to-consumer was the new hotness, so they blew up their retail operations and relationships. Stores that had shelf space for Nike handed it to competitors, who took a huge amount of market share.

They also reorganized. The company used to have departments based on sports — basketball, football, tennis, swimming — and they reorganized into men's, women's, and kids. "Why would you do that? I just don't get it."

Another panelist argued they had a very clear model — "clicks and bricks," as Chamath puts it: go to a brick-and-mortar store, have a great experience trying shoes on, work with people who help runners pick the right shoes, and buying direct through the Nike app — which he called delightful — was still available. Why kill the retail channel of all things and piss off their biggest advocates?

If I've learned anything in almost 30 years of business, you have to have a very clear north star and stick to it.

When he was growing up, Nike's north star was clear: mastery and excellence embodied through athletics. Michael Jordan, Tiger Woods, Serena Williams, Pete Sampras — you saw mastery and excellence. Somewhere along the way, mastery and excellence became too traditional-looking, so they walked away from anything associated with it. "I don't aspire to be a fat person. Never have, never will."

I aspire to be Michael Jordan. Always have, always will. It's as basic as that: you're not going to buy the clothes of a brand you don't aspire to be like by wearing those clothes or shoes. As long as a brand centers on the idea that it's not there to make people feel better about themselves—that its success came from people channeling their desire to get better through the lens of incredible athletes—that's aspirational. Being inspired by somebody and wanting to be better than yourself is a good thing. We should not shame that; it's how we all get better. There are incredible athletes who give up their entire lives to master something, and Nike should own that and be proud of it. If they do, all the other details make sense: sell it everywhere because everybody will want it, and sell to everybody because everybody will want it. But if you don't get the north star right, you'll miss out—and the north star was clear: excellence and mastery.

There's a broader discussion here, which came up in your incredible All-In interview with Eric Weinstein: we've lost this love of excellence, expertise, and the desire to be elite—to be inspiring to people as somebody to follow—and it's been replaced with participation trophies or inclusiveness as opposed to excellence. Could you talk about that in your field, science and entrepreneurship?

Entrepreneurship is a meritocracy—you either do it or you don't. But in the case of this business, what they also lost didn't just come through in the brand. I used to buy Nike; all the shoes I owned were Nike. Over time, it wasn't that the ads went bad or the ordering was bad—the product started to suck. The shoes literally fell apart in about six weeks. They were flimsy and didn't have the quality they used to. So one day I stopped into my REI store next to where I get salads, tried on a pair of Brooks, bought them, and started wearing Brooks.

These are great products—they last forever, they're durable. Marathon runners do like Brooks.

I held on to Nike forever, buying a new pair every six months, and I still have a collection of many Nikes. But it became every six weeks as the product got worse, and I finally moved to On for one reason: Roger Federer. He's excellence and mastery. Everything he does looks effortless, and I love that—so Roger Federer was good for me.

Let me tell you the story on Brooks. The company is owned by Berkshire Hathaway, and they've grown it nine years in a row with double-digit revenue growth, to $1.6 billion in revenue. It's a standalone company inside Berkshire that just keeps compounding. There's an interview with the CEO where he said Warren Buffett simply told him: every year, make the product better than it was at the end of last year, and if you keep doing that, the business will keep growing. That's what he's been doing, and it's the opposite of Nike's focus, because Nike shifted from product to narrative—it became all about what narrative they thought the audience wanted to hear. It's the same problem Bud Light had.

Remember that huge Colin Kaepernick ad they ran in San Francisco?

That was the turning point in my opinion. It was like, what the hell does that have to do with anything? I want excellence and mastery. You don't have to comment on everything. If you were to pinpoint the beginning of the downfall, that was it.

There have been moments in sports that were inherently political and were great moments: Jesse Owens at the 1936 Olympics, the raised fist, Jackie Robinson, Muhammad Ali refusing the draft into the Vietnam War. They were all the greatest, and their causes were standing up to Nazis, protesting the Vietnam War and racism. But what was Colin Kaepernick taking the knee to protest the American anthem for? What was the point? And was he one of the greats? No. So why give that a platform? What were you promoting exactly?

Then they kept doubling down with Dylan Mulvaney and the rest of the woke stuff. It is breathtakingly stupid. Their job was marketing, yet they ran marketing images so contrary to what the brand was about. It was craven, and I think everyone in the decision-making chain, from the marketing executive up to the CEO, should be drummed out of corporate America. How could you approve those campaigns?

It happened during a certain time period, sure. But if they make mastery and excellence their north star again, they'll be fine, because the brand is a coiled spring of potential. If you gave me a reason to come back to Nike, I would.

At Stanford, the Nike store closed, and what replaced it? An On store. When you walk into the On store, it's amazing — you think, "Roger Federer." When I buy a pair of running shoes, I was just thinking about mastery and excellence: Michael Jordan's shoes, my shoes; Tiger's clubs, my clubs; Sampras's clothes, my clothes; Federer's racket, my racket. That's how simple I am, and I would happily buy that stuff. So Nike has to reestablish itself in retail locations where people have disposable income, and decide whether it wants to align with mastery and excellence. If it does, tens of millions of men and women will show up again — people who want to get back in shape, reestablish themselves in a sport, or learn a new one. I used to embrace Nike everything, and now I don't.

They should also have gotten into smart devices. They dipped their toes in, and I thought that was good because it was performance-based. If they owned Strava — the app all the elite biking and running people use — and embraced it, and then redid the tagline with a really challenging rebrand, "Do it or don't," that could work.

"I think that's Yoda. Do or do not, there is no try."

I lost 43 pounds doing farmers walks, half a mile with 20 pounds on each arm — if you just don't buy into the woke stuff, you're ahead of the game.

I cannot get past the incompetence of woking your way out of the S&P 100. The whole exchange that follows turns into a running joke about the alternative tagline: going from "Just do it" to "Do it or don't, we don't give a damn what you do" — mocked as Brooklyn Don Draper and the only way to make things worse at Nike, with the retort that the company found a way to get kicked out of not just the S&P 100 but the 500 as well, down to pink sheets.

I mean, I'm good at marketing. Who named this podcast? I did — you asked what we should call it, I said "All In," and then I got the domain name. You ungrateful... All right, everybody.

See, do you use GoDaddy?

No, I negotiated directly with the owner. I got it for 250 beans — a great deal on that five-letter domain.

Okay, well, do it or don't. You did it. Do it or don't.

I did it. I did it. Say thank you.

All right, everybody. We'll see you at the All-In Summit. Another thing that I did.

The All-In Summit is this weekend, Sunday through Tuesday in LA, presented by Iron. Iron provides AI cloud data centers, compute, and software for training and inference. They're throwing Sunday's registration day party, so grab your badge, grab a drink, and join us at the Iron House in the expo hall for a meet and greet and photo op. We're also excited that Meta will be in the house — President and Vice Chair Dena Powell McCormack will join us on stage, plus there's a Meta Glasses Experience Hub. Check it out at the All-In Summit.

We'll let your winners ride. Rainman David.

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

Besties are gone.

That is my dog taking your driveways.

We should all just get a room and just have one big huge orgy, because they're all just useless. It's like this sexual tension that we just need to release somehow.

Wet your feet. Wet your feet.

That's going to be good.

I'm going all in.