人工智能冷战、Signalgate、CoreWeave IPO、关税终局、萨尔瓦多驱逐行动

When I wasn’t able to use the R-word, I would use “degenerate” for 2 people: Jason and Friedberg’s dog, Marshall Friedberg. Oh, come on. I can’t stand Marshall Friedberg, that little bastardino. He jumps on the table and eats the nuts. The worst. I hate Marshall Friedberg. I hate him. I can’t stand this bastard guy. Look how he sits. He sits like a degenerate. Look at this dweeb.

Nick, show them my dogs. Beautiful, beautiful dogs. Valentina and Zanini. She’s the breeder of breeders. Look at these 2 beautifully elegant dogs. Look how well-behaved they are. You don’t see them jumping on the table to eat the main course, Friedberg.

No, no. Marshall almost ruined our Christmas dinner.

This is why I’m holding a real grudge against Marshall Friedberg. He ate the nuts.

No, and Alison caught him off guard. She was like, “Marshall Friedberg, get off the table.”

I just love that he’s got a full name: Marshall Eugene Friedberg.

Let your winners ride, Rain Man David. Instead, we open-source it to the fans, and they’ve just gone crazy with it.

Our guy Gavin Baker is here. You know him from Atreides Management, which does private and public investing, with $4 billion under management. A lot to talk about with you, GB1. Good to have you.

CoreWeave’s IPO is happening soon. By the time you get this, CoreWeave will have gone public. You were at NVIDIA. You were the 1 analyst, Gavin, that Jensen pulled up. You’re loved by Jensen. What’s it like to be loved by Jensen? Tell us everything.

First, there was Alfred F. from [unclear], and then they had a nice sell-side guy whose name is escaping me at the moment. But it is technically true that I was the only public-equity investor on the buy side that Jensen asked.

I’ve known Jensen for 25 years. He’s kind of the same guy he ever was, just maybe slightly calmer.

Can I ask you about all this stuff where people talk about the balance sheet questions that some folks have about NVIDIA, some of the accounts-receivable issues and things like that? Is there any legitimacy to those issues? What’s the real story from someone who peers into that P&L and understands it?

Let’s take them in reverse order. I really don’t think that if NVIDIA had not put any money into CoreWeave or other neoclouds, it would have impacted their revenues at all. They would have sold more to Meta, more to Amazon, and more to Microsoft.

The reason they did it is that, even 3 years ago, there was a very stable 3-player oligopoly among the cloud-computing players: Amazon, Google, and Microsoft. If you’re NVIDIA, that’s not great. You don’t want to have just 3 big customers.

Those big clouds, particularly in 2023 when there was such a rush to get GPUs, each wanted to do its own custom version of an NVIDIA server. By giving CoreWeave a standard reference design, NVIDIA got GPUs into the market a lot faster. When you use NVIDIA’s reference design, it’s generally smoother and easier to stand them up.

The big 3 hyperscalers’ incremental share of revenue went down, and that was nothing but good for NVIDIA because it now has a more fragmented base of buyers who have less power over it.

What does that mean when accounts receivable jumps from $1.5 billion or so to $5.5 billion year over year? Is that concerning, or is that just the business?

It’s never good, but if you think about NVIDIA going through the biggest product transition in history, in terms of Hopper to Blackwell, this is a business doing tens of billions of dollars at scale. There’s never been a product transition like this in the history of semiconductors.

The only precedent for this on planet Earth is the iPhone. There’s something called the Osborne effect: when Apple comes out with a new iPhone, for the 3 months before it comes out, nobody buys the old iPhone. I think the only reason NVIDIA was able to grow through this product transition is because of reasoning models like DeepSeek-R1, which are so compute-hungry.

Accounts receivable is now around $23 billion. It’s never good when accounts receivable goes up, but this is probably the most understandable time for it to happen in a product-refresh cycle.

A Hopper server is a rack that’s about 7 feet high, weighs roughly 1,000 pounds, consumes 60 kilowatts—which is 60 American ovens—and is air-cooled. Blackwell weighs 3,000 pounds. The rack is 8 feet tall, maybe 5 feet deep or 4 feet deep, and consumes 120 kilowatts, so twice as much power. It has roughly the same footprint, 3 times the weight, and it’s liquid-cooled.

This is like an iPhone upgrade cycle where, to get the new iPhone, you don’t just have to change your connectors from Lightning to USB-C. You have to put in a new generator, a new boiler, and a whole-house humidification system to make it work. It’s a monumental product transition, and I think that’s a lot of why receivables went up. They’re recognizing Blackwell revenue while it’s still getting into customers’ hands.

If that trend continues past the July quarter, I would say that’s a reason for concern. But this is a very understandable time for it to happen, from my perspective.

So, basically, accounts receivable builds as you go through this complicated product transition. Then, at some point, it peaks, you work through all of it, it moves off the balance sheet, and it gets recognized. That accounts-receivable balance should then shrink.

To clarify, accounts receivable is the money owed to you by customers. The reason they’re not collecting it is because they haven’t delivered it?

The accounting gets very complicated with these systems, but yes, they may have delivered it and recognized the revenue, while the customer is saying, “We’re not going to pay you until we get them plugged in and working.” Who knows exactly what it is? It’s a new thing.

Total accounts receivable has simply gone up more than sales, and that’s a function of the complexity of this product transition.

The other part of what you’re saying is very similar to what Intel did to scale its chip dominance in the 1980s and 1990s with the Intel Inside program. In hindsight, Intel probably didn’t need to do it, but it helped support an ecosystem and diversify its buyer base so that fewer people had power over it.

A lot of this so-called round-tripping, from your perspective, is diversifying the ownership pool of NVIDIA?

Absolutely, and it’s sending NVIDIA GPUs to people who are adopting NVIDIA’s reference architecture, which they believe is the best architecture for those GPUs.

Each hyperscaler has its own slight tweaks to that server. The hyperscaler does that so it looks more like its other servers, but maybe it doesn’t run quite as well as the standard NVIDIA reference architecture, with all of their chips—not just the NVIDIA GPUs.

The Intel Inside analogy is interesting. I hadn’t thought of that. Thank you.

He is definitely doing keynotes like Intel did and trying to brand the actual names of the units. Intel Inside meant people would go buy a computer and ask, “Does it have a Pentium? Which Pentium does it have?” Now Apple is doing the same thing with the M4 series. People are actually asking what’s inside.

There is something to the idea that some portion of the consumer base wants to know what’s inside. It doesn’t cost much to brand.

Since we’re here, let’s talk a little bit about the CoreWeave IPO. CoreWeave is going public on Friday. We’re taping on Thursday, so when you listen to this episode, you’ll have the actual data on what happened with the IPO.

They’re going to raise $1.5 billion. They’re a neocloud. Neocloud means not just cloud computing, but a new cloud—hence “neo.” The company has really been growing quickly. It was reported that they wanted to raise $2 billion, and I guess they stepped it down to $1.5 billion at a $23 billion valuation.

They had $2 billion in revenue last year, so they’re up 8 times. They’ve got a lot of debt—that’s the other issue people note with CoreWeave—almost $8 billion in debt. One customer, Microsoft, accounts for over 60% of their revenue, which is always a bit of a red flag. NVIDIA accounts for 15% of their revenue, and NVIDIA owns 6%. Six percent seems small enough not to make a difference.

What do you make of CoreWeave and its timing? What does it say about the markets? This is the biggest IPO we’ve had in a while. The last notable ones were probably Reddit last year and a couple of other companies. Stripe still hasn’t gone public.

Is this a major milestone, or are they going public because they have to, because of all this debt? This is also against the backdrop of whether we overbuilt and whether these language models are going to require as much hardware as we thought last year.

We’re definitely not overbuilt yet. Since DeepSeek-R1 came out, China has been buying every GPU it can. One of the largest Chinese server manufacturers warned last night that there’s about to be a shortage of GPUs in China. The price of DRAM goes up every day, and OpenAI yesterday said it was gating its new image-generation service because it doesn’t have enough GPUs.

We’re not overbuilt today, point 1. Blackwell is going to be a very successful product cycle.

There is a little bit of a prisoner’s dilemma where you’re going to spend on Blackwell almost no matter what, because if you don’t, you’re afraid you’ll cede a big advantage to a competitor. If you’re Meta and you don’t spend while Google spends, and Google’s AI is a lot better than yours, you’re worried you may not catch up.

I would say that’s already the case.

That’s already the case. But this is a brand-new architecture, which is a lot better. As we went through, it’s not only really different; it’s much better.

All these GPUs are hard to get working at the beginning. Hoppers are finally tuned, and everybody knows how to work with them. With Blackwell, it’s almost like you’re skipping ahead. Imagine each one of these GPUs is a Formula 1 car. You’ve skipped 10 years into the future and now you have to learn how to drive that new Formula 1 car. It takes a while before you can drive it as well as the old one.

Any other thoughts on the CoreWeave IPO, which will have happened by the time people listen to this episode?

The sentiment among investors on X is pretty negative on CoreWeave. I think there’s maybe a little bit of negativity in the market. They had to reduce the range and lower the price, and everyone is very convinced this is a commodity business: loads of debt, loads of capital expenditures, one really concentrated customer, and a company that’s completely undifferentiated.

I would make one point. In America, you can pick any retail category, and if you can run 1,000 stores in 50 different states, with different preferences, and have those stores well-lit, clean, stocked by friendly employees with the right stuff in stock, you’ll create a business worth well over $10 billion. That sounds easy, but very few companies have been able to do it.

What may be underappreciated about CoreWeave is that it’s really hard to run these big training clusters. Everybody thinks it’s easy, but synchronizing tens of thousands of GPUs, with GPUs melting or cables being unplugged, means you can lose a bunch of training data when something goes wrong.

I just think it may not be the commodity that everyone thinks it is. It may turn out to be much harder to do than people think.

Wouldn’t the analogy be that people thought AWS would never work at the beginning? They asked, “Who’s going to trust Amazon with their compute?” It was difficult, and developers didn’t understand exactly how to abstract their services into the cloud.

The world was convinced AWS was a terrible business for the first few years. When Microsoft announced that it was going to transition to the cloud, it was very controversial because you go from a capex-light model to a capex-heavy model.

Wall Street consensus has been wrong before. I do think CoreWeave runs these big GPU clusters as well as anyone, and there aren’t that many people on Earth who can run them well.

It’s relatively easy to look at the net present value of their contracts and the asset value. Everyone who is confidently putting out all the negatives here should consider that there are some offsetting positives. It’s good for people to be aware of and consider them.

I think it’s a great business, and I think they deserve to be successful. It’s great to see a competitor to the existing 3 or 4 clouds that are out there. It’s also great to see a company go public. M&A with Wiz, and now CoreWeave going public—the roadblock on M&A and IPOs is over.

You have to give the administration a lot of credit for perhaps removing the roadblocks to M&A and IPOs, or at least people perceive that. I don’t know if perception becomes reality, but M&A does seem to be a bit on fire.

People are talking above and beyond Wiz. CoreWeave bought Weights & Biases, which I think is a really interesting acquisition for them. That further differentiates them and maybe decommoditizes them a little bit.

There have been quite a few acquisitions by big companies of smaller companies in the last month. Since none of the companies that have been bought were public, it may be flying a little under the radar, but it’s happening in an accelerating way. This would be very good for the market, animal spirits, and everything we care about.

Do you think there are any issues with the export-control list? A bunch of companies were added 2 days ago. Some were in quantum computing, but a lot of it was in the NVIDIA ecosystem. The U.S. government is really trying to make sure these next-generation GPUs don’t go directly to China, but also don’t go to countries that could redirect them into China. Did you get a chance to look at that, or did your team look at it?

Yes. I think this will be part of the upcoming negotiation between America and China. It is hard, though. If America cannot keep illegal drugs out of our own country, these GPUs are arguably dramatically more valuable per unit. A Blackwell is about this size, and it’s extremely valuable.

America is trying to keep drugs out, while China is trying to bring the GPUs in. It’s significantly harder to prevent than smuggling drugs. Everyone does their best, but the United States is always going to be playing a game of cat and mouse until there’s some sort of grand bargain.

Right now, they’re allowed to sell certain kinds of GPUs into China, and I’m sure they’ll continue to be allowed to sell some GPUs there.

Should we be doing that? If you were advising the president, would you say, “Just let NVIDIA sell them. It’s a fool’s errand to try to stop them because they’re all going to Singapore or Vietnam and routing their way around”?

I think we’re putting enough friction into the system that it does theoretically give America an advantage, at the cost of creating tremendous incentives for China to develop its own semiconductor ecosystem.

Necessity is the mother of innovation, and these export controls are creating an immense incentive for China to be really algorithmically innovative. You saw that with DeepSeek, where there were some real algorithmic innovations.

Said another way, if we squeeze too tightly on letting them buy NVIDIA chips, they might make a better NVIDIA. They might become resilient.

I think they’re trying now.

What are the chances they succeed in creating something competitive with or better than NVIDIA? What would that say about the AI race over the next 5 years?

I think the chances are 0. It’s really, really hard. Over 10 years, who knows? If you’re the Chinese Communist Party, 10 years isn’t that long. If you’re America, 10 years is an eternity.

They’re thinking in centuries, and we’re thinking in decades.

That’s probably correct.

Let’s pivot over to AI agents. We have a company that seems to be having a moment called Manus. It’s currently in private beta, but they’re creating agents. We’ve been talking about this agentic revolution—little jobs going out and doing things for you.

It seems to be working, and it’s pretty impressive. What do you think about this company specifically, if anything? Are we getting to a ChatGPT-2.5 moment with agents? If we do, what will that look like in terms of employment and how companies are run?

If agents materialize as a reality, Manus may be a little bit of a ChatGPT moment for that. OpenAI and Anthropic developed something called the Model Context Protocol, which OpenAI just adopted. I think it will become a standard.

It makes it very easy for an LLM to interact with a service. Stripe can integrate with MCP, and then any LLM that uses MCP can interact with Stripe. This solves a big problem for agents by making them much easier to use and much more standardized.

If agents become a reality, the ROI on AI and Blackwell will be very high. It will not be good for human employment, but the rate-limiting factor will be compute. In a world where we all have agents doing things for us all day long, it will be a long time before we have enough compute in the ground for that to become a really widespread reality.

I think that’s why OpenAI was talking about pricing its first agents at $10,000 to $20,000 a month. The ROI-on-AI question, to me, really does come down to agents in the short term.

Friedberg, your thoughts on these agents and the impact they could have on running businesses, employment, and efficiency? We’re starting to have some standards. MCP stands for Model Context Protocol. How might this impact architecting a business, which you’re doing right now with Ohalo? How might this change how people work on a day-to-day basis? Let’s give some examples for the audience.

I’ll take a slightly different view than Gavin. I’ve said this in the past, but maybe I’ll contextualize it a little bit.

I don’t think the big unlock with these agentic systems is necessarily replacing the easy human tasks, which everyone thinks are the obvious application. I think what we’re missing is the ability to unlock really complex tasks that aren’t manageable today.

I can’t come up with a grand project-planning exercise with the team I have today. I would have to go hire 30 people. Let me give you an example: I want to go build a plant-breeding facility under the ocean. How do I do that?

Today, I have to hire people who understand oceans, engineering, underwater engineering, materials science, physics, and construction project planning. There’s a whole group of complicated people I would need to put together to execute on that opportunity.

Now, a smaller group—maybe 3 or 4 people, or even 2—can use these agents to help build the project plan and spec out every step of the process, define how much everything will cost, define what workloads are needed, and determine who is going to do what. Suddenly, that difficult project becomes a reality.

Here’s another good example: the California high-speed rail program, which has ballooned to a $100 billion project. What if we had agents organizing and running that program? Maybe the cost and reality of that program would make it possible. Maybe it would become a program we launch, complete, and use to build the railroad, whereas all the people, complications, confusion, and whatever grift has been going on have led to a $100 billion budget with no outcome.

I think complex projects can be tackled more easily. That unlocks a tremendous amount of building and opportunity for new employment. Small teams of 3 or 4 people could do the work of 300 or 400 deeply sophisticated, highly technical people with a much smaller, less technical group.

I think there will be a tremendous amount of opportunity in biotech, engineering, manufacturing, urban design, transportation, and many other areas, where small groups will be spun up that can use these systems to do really complex projects. It’s less about replacing call centers—sure, maybe—but what will we be able to do with these things that I can’t even think about doing today?

Chamath, let’s level this up and talk about the bigger picture. Our giant rival, China, is catching up and in some cases exceeding us. If they get to AI first and then are able to execute on it, given their top-down controls, the size of their population, and their advantages in manufacturing—which we’re seeing now with Xiaomi making cars, BYD doing self-driving and cars, and Chinese chips competing with NVIDIA—almost everything we’re building here in the U.S. could be copied at a very fast pace and then exceeded.

What’s at stake here, big picture? Then we can look at what our philosophy should be geopolitically.

Let’s do an economics lesson. Why does anybody invest? They invest because they want to generate a rate of return that’s better than all their alternatives. The riskier it gets, the more of a return you want. That’s the basic idea.

If you invest a dollar in bonds, you can get 5% or 6%. If you invest a dollar in an early-stage startup, you want to get 50% or 60% returns. That’s obvious.

The return you generate is partly driven by how profitable the company is. If you have a 300-person group that has to do something, it’s a lot harder for it to generate profit and a rate of return than it is if you have 2 or 3 people.

The Chinese strategic imperative is no different from small startups versus an incumbent: be as economically disruptive as possible.

The big risk isn’t necessarily Chinese, although many of these alternatives will come from China. Most incumbents are American companies. If you look at SaaS software, that’s a perfect example. There are trillions of dollars in the software industrial complex, as I call it. It’s a $3.5 trillion or $4 trillion industry, growing by 10% a year. Every year, it adds $300 billion of so-called enterprise value.

Is that value really being created? I think a lot of customers would say no. If you’re a startup or China, what is the most disruptive thing you could do? You could take a 3-person team, disrupt a 30,000-person team, and blow a hole right through the side of that trillion-dollar economy.

That’s the big risk. Whether the risk comes from here or from China, what you’re seeing is that if these agents can scale, the actual operating expense—the actual load of making something—will go down by an order of magnitude, maybe even 2 orders of magnitude. That is incredibly disruptive because existing incumbents cannot compete with that cost scale.

To bring this around the horn, China is playing a different game. They’re subsidizing, for example, their cars and their entire car industry. The margins on cars here in the United States, Germany, and Japan are their opportunity. They don’t actually need to make a profit. They can break even.

There are now $10,000, $20,000, $30,000, and $40,000 cars being made there. This leads us to ask whether Americans will buy them, whether Trump will block them with tariffs, and what our tariff policy should be.

We have the lowest unemployment of our lifetime, but wages are obviously not where we want them to be. We know the tariff on autos is 25%, but do we know what the broader policy is? It seems to be in flux. Maybe we start there. We’ll see what happens on April 2, because Trump keeps tweaking or adjusting it in public. What should the tariff policy be if, as Chamath points out, they can put a giant hole through our software industry, car industry, and pick another industry? They may be able to do it because they don’t need to make a profit; they just need to create jobs and break even.

Traditional economics would say tariffs are not a good idea and that we should have free trade. Everybody understands the principle of comparative advantage.

I do think the Trump administration is convinced that approach has not worked well for America over the last 20 years. Maybe it worked well for knowledge workers, but it hasn’t worked well for ordinary Americans. They’re very committed to changing that and trying to bring back good, high-quality manufacturing jobs to America.

They see tariffs as an effective way to do that. Are they right? A lot of what we’re seeing in factories is automation, so perhaps tariffs could have brought back high-paying middle-class jobs over the past 20 years, but we can’t go back in time and change that policy.

Isn’t where the puck is going that we’re going to have Optimus robots, or some robot company, doing all this work?

Tariffs are a level-setting mechanism that fixes a historical imbalance. The reality is that we have had meaningfully lower tariffs on products coming in than the reciprocal tariffs on our products going into those other countries.

The 1 thing I’ll say about Donald Trump is that, whether or not you agree with the tariffs, he has been incredibly consistent. I was on YouTube yesterday and stumbled into an interview he did with Larry King in 1987. He ran a full-page ad in The New York Times talking about this exact issue, then went on Larry King and walked through the entire trade imbalance 40 years ago.

He’s been on this for a while.

He has. What they want is to create the economic incentives to reshore as much industry as possible into the United States.

The delicate balancing act is that, after 20 years of globalism and a perverted version of free trade—it’s not what it is today; it’s been perverted—it’s incredibly difficult to do that without other problems emerging, whether inflation, retaliatory tariffs, or consumption taxes.

The structural reason tariffs make sense is not necessarily to overly penalize one country over another. It’s to say, “If you charge 5%, we charge 5%. If you charge 10%, we charge 10%.”

One thing I’m always conscious of: at one of my wife Becky’s college or high-school reunions, there was a guy who had been in the Navy SEALs and had deployed to 80 countries. I asked him what he had learned, and he said the one thing he learned was that everybody in America is always trying to make America better; having been to 80 different places, our only goal should be not to screw it up. The first thing I’d say to the administration is that every time it says the word “tariff,” it needs to say “deregulation” two or three times. The best way to maximize the odds of this policy succeeding, despite the headwinds from automation, is to make it easier to do business in America. I think it will be a long time before we have hundreds of millions of robots; even between China and Tesla, it will take a long time to make vast numbers of humanoid robots.

There’s a third thing I would add. We need to figure out the difference between manufacturing and intellectual property. We want to make sure we trap the real value back in America as well.

I’ve told you guys this story before, but when I was helping run Facebook, I was a signatory when we were setting up Facebook abroad. We exported all of our IP to Ireland. Now, who do they sue? Zuckerberg, Meta, and me. We’ve been in a 10-year—maybe a 15-year—lawsuit with the IRS because they’re trying to come back and say, “Facebook, you owe all of this money.” Why? Because we exported all of our critical IP to Ireland and trapped it there.

That should not have been able to happen. We can fix it separately. Many companies are abroad but live inside the American market. They would import their IP into the United States if there were a mechanism to do so.

Not only can you have manufacturing; you can also have the critical knowledge—not just of that manufacturing, but of the product and supply chain. If we get these incentives and nuances right, it could be a renaissance for the United States.

Let’s talk about other solutions. Friedberg, perhaps raising the minimum wage or increasing corporate taxes would help us build back the middle class. That’s the criticism some people have.

When Trump did the Tax Cuts and Jobs Act, the corporate tax rate was 35%, and we brought it down to 21%. Repatriation of cash from overseas, as Chamath pointed out, is only around 15%. Everybody can afford that.

Wouldn’t another way to solve the problem of the bottom half or bottom third not making enough be to lower their tax rate, which is already pretty low?

The proposal I’ve heard from this administration, and that we heard from Howard Lutnick when we met with him, is that they’re going to cut all taxes for people making less than $150,000 a year, bringing the tax rate to 0%.

Do you believe that? Do you think it’s realistic?

I do. The key question is the timeline.

This is a first- and second-derivative problem. When you take money out of the economy, you want to make sure you’re also putting money back into the economy in another way to keep the economy stable or growing.

If you’re taking money out of the economy through tariffs, you’re putting money into the government’s coffers and reducing people’s spending because everything costs more. Consumers will spend less. You want to make sure the money you’re repatriating goes back to consumers so they have more money to spend.

That’s why you want to cut taxes at the same time you increase tariff rates. If you increase tariffs and things cost more, that’s not good. But if you cut my taxes and I have more money than before, I can still buy the things I was buying even though they’re a little more expensive. The economy should be able to remain stable or grow.

If you’re cutting government spending, the government can afford to take in less tax revenue. These things are all related. To Gavin’s point, you have to cut regulation at the same time so the incremental dollars flow into automation and onshore production. Then the cost of things comes down.

It’s not a simple equation. I would call it a grand economic experiment.

The analogy for me feels like 4 freight trains passing each other at very high speed while you’re trying to pass all the ingredients to make a sandwich between the windows. You’re trying to get them all over so you can make this delicious sandwich and get to this golden age.

That’s the configuration they need to assess. They have to thread the needle, and it’s really hard to do. You used the phrase “experiment,” Gavin, and you said, “Don’t screw up this amazing country.” This sounds like a very high-risk, high-dexterity maneuver. It feels like a fighter plane trying to land on a very small aircraft carrier.

Can they get tariffs working without inflation spiking or employment being affected, while also eliminating people’s taxes? Does this sound too complex and too much too fast? What’s your honest thought as you watch this as a capital allocator? Are they going too hard and too fast, or is it worth pursuing this experiment?

Scott Bessent is probably the single most respected macro investor in the world—or was the single most respected macro investor. He really understands the markets, how they flow into policy, and how policy flows back. He’s an extremely smart and talented guy.

They have a coherent theory. I thought Howard and Scott did a great job in the last 2 episodes explaining it.

The 1 thing I would say, having watched the first Trump administration, is that Trump is adaptable. He adapts to circumstances. The midterms are 15 months away, which is around the corner.

They’re trying to do a lot as quickly as they can, but if after 6 or 9 months it’s not working, I think they will adapt. There’s such an emphasis on reciprocity. Vietnam came out and lowered its tariffs on American goods, and that’s good.

The potential exit is, “Maybe it’s not going quite the way we hoped. We’re going to declare a series of grand bargains with these countries and declare victory.” Tariffs on American goods may be slightly lower, tariffs here may be slightly higher, it may feel more fair, and then they can move on to firmer ground: deregulation, balancing the budget, cutting government spending, and cutting government taxes commensurate with spending cuts.

I think they’re trying the risky stuff first, and if it doesn’t work, they will adapt quickly.

That was my perception, especially after hearing Bessent on the All-In podcast. Great job getting those long-form conversations.

I think this is where the administration is strong. When they do a long-form discussion and you hear their thoughts, it doesn’t sound as crazy as when, with all due respect to the president, the president says, “I’m going to go easy on them,” or “I’m going to go hard on them.” It becomes like a reality TV show: “Who’s going to get fired this week?”

You actually have 2 incredibly thoughtful, lifelong Democrats around Trump. I had this crazy observation, Gavin. I don’t know if you noticed it too. Trump lost his reelection campaign and filled his cabinet with Republicans. This time, he’s filled his cabinet—and everybody around him—with Democrats.

He’s arrived at a unique concept for solving the United States’ problems: get the working man and woman a higher salary and lower their taxes. This could be an incredible grand bargain that Bessent is the architect of, with Lutnick as the communicator and executor.

Do you think they have a chance of pulling this off? If you had to handicap it, what would you say?

I hope they do. As an American, I hope they do. They definitely have a chance.

They have a solid theory of the case, but they need to execute well. DOGE’s efforts are important. If you can cut $1 trillion of waste, fraud, and abuse out of government spending, that’s good.

We should all agree as Americans that, for whatever level of government spending we want, we want it to be as efficient as possible. If you’re a Democrat, that means more services. If you’re a Republican, maybe it means a smaller government.

They have a solid theory of the case, and I think there’s a reasonable chance it will succeed. But if it doesn’t, I think they will adapt quickly. The 1 thing they can’t afford is a recession. If they’re focused on the deficit, a recession blows out the deficit. They will ultimately be sensitive to market-based forces: interest rates, the stock market, and foreign-currency markets.

John Arnold had a post this morning on X saying the deficit this year is going to be $1.9 trillion. His best guess for 2026 was $1.9 trillion, and he was asking his followers what they thought.

If you listen to Howard Lutnick, he said, “I’m going to raise $1 trillion, and Elon’s going to cut $1 trillion.” If that comes to pass, we have a balanced budget.

If you ask the sharp money on Wall Street, the estimates range from $170 billion to $300 billion on the revenue side. Let’s take the midpoint, around $200 billion to $250 billion—say, $250 billion.

If the sharp money is right, tariffs and all of that will raise only a quarter of a trillion dollars. That means there will be even more pressure on Elon. He’ll have to find not $1 trillion, but $1.75 trillion, assuming that the $1.9 trillion deficit exists next year.

That’s a $2 trillion bogey, and they have to fill it somehow.

My intuition is that waste, fraud, and abuse are probably meaningfully more significant than we think. When we were there, we heard about 1 consulting firm where more than 95% of its multibillion-dollar revenue came from time-and-materials contracts with the U.S. government.

We heard throughout the week—I reposted it on X—that $65 billion is paid out to consulting organizations beyond just that 1 company through these time-and-materials, cost-plus contracts. If you start adding all of these up, they could be substantial.

We just heard that the $2 billion grift to Stacey Abrams was canceled, and the Justice Department is beginning an investigation into how those dollars were given to her.

It may be the case that, if there’s an upper bound on tariffs, there will be more than $1 trillion for Elon to fill in. I think it’s also possible they’re looking at military spending and perhaps cutting it by 8% a year for a couple of years. That’s where the big dollars are sitting.

Obviously, nobody wants to touch Social Security.

I do have to point this out for my besties: You guys did a great job with that interview. I couldn’t make it because I had ski week, and these last-minute interviews happened very rapidly. There’s no conspiracy. I want to say “F-U to the MSM” for trying to spin it as though Lutnick was trying to take away people’s Social Security.

They literally took a 1-hour-plus interview and clipped it to make it seem like he was saying the opposite of what he said. He was saying that the fraudsters would be the first to respond—not my mom. My mom would assume there were good intentions. She would eventually call and say, “Can I…?” The fraudster would be thinking, “I’ve got to get here.”

I think the same thing happened with the response to your interviews. These were great, substantive, long interviews with Democrats, one of them a gay Democrat—the highest-ranking gay Democrat ever to serve in an administration.

The left and the media should be praising these interviews and studying them. They’re doing the exact opposite. They’re trying to spin them into something they weren’t. It’s a gift the All-In podcast gave the American people by showing them long-form interviews with the administration explaining these details. Don’t spin it.

I’m sorry. I’m pissed off about it.

I don’t think you should be pissed off. I think it’s wonderful that these people lie. I think it’s incredible that they just put it out there and further erode what’s happening. If Bobby puts a nail in the coffin of pharmaceutical ads on TV, bye-bye Anderson Cooper, bye-bye all these guys. I think it’s great.

I’m so pissed off. I’m sorry. When I see people lie like that, they should be lying. They should be lying. It’s great because they will not exist.

Howard Lutnick literally said that if we can find a lot of waste, fraud, and abuse, we don’t have to raise the retirement age. Everyone is talking about raising the retirement age to 70. If we administer these programs more efficiently, we can keep it at 65, and that’s good for everyone.

It’s also interesting to me that there’s so much doubt about this when the inspectors general under both Biden and Obama said there were hundreds of billions of dollars of waste, fraud, and abuse.

Someone in my immediate family had their Social Security identity stolen. Someone is using their Social Security number to collect unemployment benefits in Alabama. This has been going on for 2 years, and they’ve been unable to stop it.

You don’t want to rely too much on anecdote, but I think this is really happening.

The third thing I would say is that I do think nobody has ever raised my taxes more than Trump did. I think they’re genuinely focused on normal working-class Americans and making their lives better. That is their North Star.

You can disagree with the theory of the case and the way they’re communicating it, but I think their intentions are good.

Absolutely. He even said in the interview, if you really listen to it, that he was disgusted by the idea of raising the Social Security age.

I think we should raise it, by the way. Any rational person would say maybe it should go up a year or 2, and maybe we should have a little austerity on the margins. But of course, we can’t talk about that.

The thing this administration could do better is communication. Those 2 long-form interviews were a great step in the right direction. They should do more of them. DOGE should do a better job explaining the waste and fraud.

They’re doing a good job—I’d give them a B-plus—but there’s more work to be done. There’s waste, overspending, fraud, and abuse. These are different categories.

To get to fraud, you have to build legal cases. All of that will have to be referred to people, and we have a justice system. It will take years to figure out whether somebody actually stole something. But waste and overspending can be found pretty quickly.

It’s a process, people. I trust the process because we’re overspending massively.

I think a quarterly or semiannual DOGE report, with everything audited and triple-checked, would be helpful. These are the 3 categories, and this is what we found.

They also need investigative journalists. I’m happy to help. I’ll do a DOGE pod. I’m not part of the administration. Everybody knows I was a Never Trumper, I’m independent, and there are plenty of disclaimers.

I would go in there and do the DOGE pod with Elon or whoever is in the group, going point by point through these things to explain them to the American public.

Why do you keep saying you’re independent and then throwing yourself a label that says you’re not independent?

I’m telling people I was a Never Trumper. If Trump is going to surround himself with moderate Democrats, are you truly independent? I like to think so. I try to call balls and strikes here on the program.

I’m not going to bend the knee to the administration. I’m going to be critical of Trump Coin. I’m going to be critical of these deportations, which we’re going to get to. I will say critical things.

Did you vote for Trump?

I don’t want to bring my vote into it. I did not vote for Kamala. I couldn’t vote for her. I don’t want to say it, but let’s keep going. I didn’t want to vote for somebody who was selected. I wanted to vote for somebody who had a primary. The Democrats disrespected democracy by not having a primary.

Signalgate is the biggest story in America right now.

I think 70% of what they’re doing is right and 30% is wrong. I’m going to get to the 30%, and human rights is a big part of it.

I think you just admitted you voted for Trump.

Anyway, Signalgate. On Monday, The Atlantic published a story titled “The Trump Administration Accidentally Texted Me Its War Plans.” Somehow, Atlantic editor-in-chief Jeffrey Goldberg was added to a high-profile Signal group with the secretary of state, vice president, Tulsi Gabbard, Scott Bessent, the CIA director, secretary of defense, and others.

How is this possible? They added a journalist—the editor-in-chief of The Atlantic—to their war-planning Signal group. They shouldn’t be using Signal to do this, obviously.

Inside this group, they discussed plans to strike Houthi targets across Yemen. Nobody noticed or checked that the editor-in-chief of The Atlantic was in the group chat. They included the exact targets, actions, and timing. It was apparently cut and pasted from another system by Hegseth, our secretary of defense and former Fox commentator. That’s not a dig; it’s just a fact.

People are losing their minds over this. It’s taking up all the oxygen in the room, in the media, and in group chats.

What’s your general take on what happened? How stupid and ridiculous is it? Does it matter? Chamath, start us off.

What people have stopped talking about entirely, and what is worth touching on, is what was actually going on that necessitated this group chat being created in such haste.

Somewhere after the war in Gaza, Iran continued to fund the Houthi rebels in Yemen. They started to invade and flood into the Red Sea and block the Suez Canal at different points, making it very perilous for freight traffic.

Ships were either forced to go all the way around the Cape of Good Hope or risk attack through the Suez Canal. Traffic through the Red Sea and the Suez Canal fell off a cliff.

Why is that important? Because prices started to skyrocket. If you were trying to ship things into the United States, prices rose 30% or 40%. Going to Europe, prices rose 300% or 400%.

As Gavin said earlier, we’re in a very delicate moment where we’re fixing all these parts of the economy. The thing we can’t have is productivity going down. We can’t have a recession. If input prices go up or trade slows down, you bring the likelihood of a recession to the forefront.

The second thing included in the group chat is that nobody was in a position to fight off the Houthis except the United States. The British and French naval capacity wasn’t strong enough or adaptable enough to fight these people off.

The United States had no choice but to attack them and reset the situation. Traffic has now recovered. Volumes are back up by the third week of March, to where they were a year ago. The price of shipping is contained, inflation risk is contained, and the productivity hit is contained.

That’s the context for why they did the attack.

I find it crazy that we have 95 articles on Signal and less than 1 article on the actual situation. I wanted the smart people listening to our podcast to understand it.

The second thing is that we should send the bill to Europe because this is really a European problem. We have the whole Pacific Ocean and can accept things through ports in Texas.

The point is, we were on a short clock. People had to act quickly.

Here’s my big problem with Signal: I think it was a screw-up. I think it was an important set of decisions that had to be made, but using Signal was a mistake for 1 narrow reason.

The Signal desktop app is total and complete garbage. When we build apps, we’re told they need to be on every endpoint: Android, iPad, phone, and desktop.

The problem is that you might have a computer at home, a computer in your den, and a computer at the office. Someone comes and helps you install this stuff, and suddenly, instead of having 1 attack vector, you have 7. That’s insanity.

It was a stupid mistake, but I think it was a mistake. In 2 Scaramuccis, we’ll forget about it.

Should somebody resign or be fired?

No. These are mistakes. The administration should establish clear, normal procedures for how this works.

John Ratcliffe was very clear that there are procedures they inherited. Some security IT person writes them and they get approved. Rewrite those procedures to be more normal.

This issue will become one of whether government officials can use Signal and similar apps for communication. This controversy has been around for a while and is being tried in the courts right now. There are 4 active cases.

The Federal Records Act was originally passed in 1950. An amendment related to electronic communications was passed in 2014. The Federal Records Act says all government officials and employees have to create, maintain, and preserve adequate and proper documentation of activities, transactions, decisions, policies, and other business.

That’s the law. Adequate and proper documentation is defined as records that clearly document the government’s decisions and actions and can be preserved and retrieved for accountability, transparency, and historical purposes.

Then we have FOIA laws, which allow citizens and journalists to make requests under the Freedom of Information Act. This is necessary for a functioning democracy.

People who work for us should not be able to end-run their communications, and that’s the law.

FOIA provides access to government records because our taxpayer dollars pay for the government to do its work, and therefore we have access.

The question is whether the particulars of the communication taking place among these individuals—in this case, people updating one another or talking about things—are subject to preservation under the Federal Records Act and the 2014 amendment.

There have been a bunch of lawsuits. One was the Competitive Enterprise Institute against the Office of Science and Technology Policy in 2016. The case challenged the use of email accounts and communications that weren’t properly archived or preserved. The D.C. Circuit Court of Appeals ruled that officials who use private email for official business must copy or forward those communications to government accounts to comply.

Citizens for Responsibility and Ethics in Washington sued the Trump administration in 2017. That’s still ongoing in multiple forms in the courts. There’s a lot of procedural complexity, but the courts have generally said these officials are obligated to preserve official communications and that failure to do so violates the law.

There’s also Judicial Watch versus the Department of Homeland Security and National Security Archive versus Mike Pompeo and the State Department, from 2019, which is also still in the courts.

From these cases and this particular incident, there will be much greater clarity about what communications need to be preserved and what communications officials are allowed to have off the record.

What do you think, Friedberg? We understand the law. What do you personally think? What does David Friedberg, the Sultan of Science, think?

I think it’s counterproductive to have every piece of communication subject to FOIA and kept in the public record.

When you’re making a business decision—say, an HR decision—the management team might be discussing whether to give someone a raise or terminate someone. You don’t want to have a record of that which employees can pull up and see, because it would have a chilling effect on the authenticity, productivity, and quality of the conversation.

I think the decision and the action should be recorded and subject to FOIA. But all the background material, noise, and dialogue that lead to that decision should not be subject to FOIA.

You need to justify the actions and decisions you make as a government official, but you don’t need a personal record of everything that led up to that action or decision. I think that would be deeply unproductive.

This is a serious issue. They shouldn’t be doing it for obvious reasons. It’s very easy to hack personal phones. They do have a way to do this, which is called a SCIF, and there are other places for secure communication.

They work for the American people, so there should be a record of this. They’re breaking the law by doing this. The administration should not be breaking the law.

Hegseth clearly took a report of the attack plan, cut and pasted it from some server, and pasted it into the chat. Your clipboard is a known attack vector. Anybody with basic security training knows that your clipboard can be hacked very easily.

They need to take ownership of it. This administration and this group do a poor job of taking ownership. Instead of taking ownership, they attack the journalist.

The Trump playbook is to attack the journalist. They take no ownership and attack. The whole anti-Hillary movement was based on the idea that she did this exact same thing with her email. Republicans had a long list of examples.

It’s complete hypocrisy, and it’s distasteful to attack the journalist you added. This is incompetence. Own it. It’s a mistake. Say you’re going to fix it and move on.

The phones of every person involved should be examined. They’re not doing a good job communicating what they’ll do to make sure someone hasn’t gotten onto everybody’s personal phone.

Why are they doing this on personal phones? As Chamath pointed out, these are attack vectors. They should use phones issued to them for these communications.

I think you’re wrong about the fact that they don’t have an adequate claim against this journalist. It’s fair to ask what ethical standard the journalist should have had. If you’re added to something and know it’s inappropriate, what do you do?

You don’t know how he was added. He could have been injected into a laptop. Nobody knows yet. That’s why I think people should let the investigation proceed. You don’t need scapegoats and heads.

But it is important to ask why this guy was lurking around. What ethical standard should he have followed? He had none.

That’s ridiculous. You’re so wrong about this. He tried to figure out what was going on, and then, when he did figure it out, he contacted them and removed himself. He didn’t stay on indefinitely.

If he had done that, it would have been wrong. But he may have been added to the group while he himself was being attacked.

The other reason you’re wrong is that they went on the attack immediately and denied there was anything wrong. They did not take ownership. That’s why we’re having this debate right now.

Marco Rubio did say someone made a big mistake. There were no consequences, but they learned from it and would do better. He’s the secretary of state.

The other people attacked a journalist. That’s what I find distasteful.

Donald Trump said it was a mistake too. I don’t think these people are denying that it was a mistake. Fix it and move on.

The bigger issue is the security protocols these people rely on. I’ll go out on a limb: As expert as they are in policy, I don’t think they’re experts in technology.

There is probably somebody responsible for how these people communicate, how software is installed on their multiple devices, and how new devices are issued. I’m guessing that person isn’t one of the principals.

I think the problem is that they’re using personal devices. Why is anybody in government using a personal device to share plans?

They said it was their personal device. It’s been established that they did this on personal devices.

They have SCIFs for this.

That’s the problem. Why don’t they use them?

The bigger issue is that they had to operationalize quickly. They weren’t in a position to be together, and the importance of the issue is getting lost.

I thought the contents were interesting. There’s been a lot of talk about the Trump administration cozying up to Russia and trying to blow up the post–World War II alliance with Europe.

There were 2 things in the chat. First, they said, “We are the only people on our side of the ledger who can deal with the Houthis.” By “our side of the ledger,” they meant the traditional European alliances. I thought that was interesting. Despite all the rhetoric, they’re conscious of who the home team is.

Second, they made it clear to everyone in the chat—J.D. Vance raised it—that 3% of America’s trade goes through the Suez Canal, while 40% of Europe’s trade does. We’re really doing this to help Europe. It helps us a little, but it helps Europe a vast amount.

The United States is helping Europe with something Europe cannot do for itself. That made me think the rhetoric about the U.S.-European schism is overblown. We’re still helping them.

That’s important to note.

Let’s move on to our final 2 stories. The Trump administration deported 238 alleged gang members to a maximum-security prison called CECOT in El Salvador. This was done under the Alien Enemies Act of 1798, which allows the president to detain and deport individuals without due process from an enemy nation during an invasion or predatory incursion.

It has been used before, but only during actual wars. Franklin Roosevelt used it to deport thousands of Germans, Italians, and Japanese during World War II.

The Trump administration claims the gang members were sent to the United States intentionally, calling them a hybrid criminal state. Not surprisingly, most of them—but not all—had criminal records in the United States. This was confirmed by ICE.

Some number of them appear to have been misidentified as gang members or terrorists. Coming from a law-enforcement family and looking at the statistics, out of 238 people, you’re not going to bat 1,000. Maybe you get a couple wrong.

Former professional soccer player Jerce Reyes Barrios fled Venezuela after being tortured by the government for protesting. He was a refugee and was mistaken for a gang member, according to his family, because he has an arm tattoo supporting his favorite club, Real Madrid. There was also a social-media post that ICE officials believed showed him making gang signs, but according to his attorney, it was actually the sign language for “I love you.”

A second example is shoe salesman and social-media influencer Norberto Rodríguez. He fled Venezuela for Colombia out of desperation because he and his family were starving. He’s a father of 3. He was suspected because of a tattoo featuring playing cards and dice. His family says he got it to cover a scar on his forearm.

The third example is stylist and makeup artist Andrés Hernández, represented by the Immigrant Defenders Law Center. He says he has no criminal history and sought asylum after fleeing Venezuela, where he was persecuted for being gay. His arrest was based on tattoos that his attorneys claim are not gang-affiliated.

This caused a U.S. district judge to pause the deportations. The administration apparently continued them and has even gloated by producing videos threatening people with the prison in El Salvador.

Chamath, what are your thoughts on this deportation process?

As long as there’s a mechanism for mistakes to be rectified, it seems like the overwhelming majority of these people should not have been in the United States and were part of illegal criminal organizations.

There should be a mechanism to correct mistakes.

If innocent people were sent to a prison like that, it’s a terrible mistake. That happens, and it shouldn’t happen. Hopefully it gets rectified.

My bigger thought is that this administration has an ambitious agenda. It wants to fundamentally change America in ways it thinks will be better for blue-collar, normal working Americans.

With things like Signalgate, if it made a mistake and sent innocent people to this prison, there are only so many mistakes it can afford before losing the mandate necessary to accomplish its goals.

We’re only 66 days into the administration—about 4.5% of the second term—but execution matters. To accomplish its goals, the administration needs to execute at a high level and communicate clearly and effectively.

If innocent people were sent to this prison, that’s terrible. Signalgate may be a big or small mistake depending on where you sit. I thought Marco Rubio’s response was good, but execution is important if they want to accomplish their goals.

I do not agree with sending people to a prison or detention center without due process. I felt similarly about Guantanamo Bay. I don’t think that matches the rights and values the United States should stand for.

If you believe people are here illegally and part of a criminal enterprise, and you don’t want to try them in courts here, then ship them back to a port of departure somewhere else—perhaps Venezuela—and let their courts try them. But I don’t think they should be sent to a prison without due process.

I’ll talk about the El Salvador motivation. The prison they established for gang members had a radical impact on living standards in the country.

In 2015, the homicide rate per capita in El Salvador was 103 people per 100,000. They introduced an aggressive policy of rounding people up without due process and putting them in this prison. Through that action, they reduced the homicide rate to below 2 per 100,000—1.9 last year.

It had a radically positive effect on society, but at the cost of a value we hold dear in the United States: due process. All it takes is 1 innocent life being locked up.

You’re saying there was no due process in El Salvador?

There was no due process. There are plenty of stories, if you watch the documentaries about this prison, of people who were later released after being found completely innocent. They were randomly rounded up or caught in a sweep.

I saw 1 fascinating interview with someone who spent more than a year in the prison and was completely innocent. He said it was worth it because the government had been able to clean up the streets.

He had to say that to get out. Who knows?

It was a profoundly difficult place to live in El Salvador. We’ve all heard these stories.

I do think this administration is taking a cue from it—that there’s a way to very quickly reach a positive outcome with respect to immigration and crime. But Gavin’s point is right: We have to maintain some boundaries, particularly as they relate to human rights, that we value in this country.

As long as there’s a court trial, a judge, and some sort of evidentiary hearing to determine that someone is likely to be this criminal, that makes sense. It doesn’t need to be a jury of peers.

But if there’s going to be forced movement of people, I don’t think they should necessarily be moved into a prison. They should be moved to a port of departure and tried in another country if they committed a crime, or tried here and put in prison.

I agree with what you just said. Due process is important. Human rights are a core American value.

The outcome in El Salvador was good: homicides are down 99.9%, and many lives were saved. There needs to be a balance.

Let me ask you both a pointed question. If you could implement an El Salvador-style incarceration system tomorrow and it would have the exact same effect on American crime, with a 4% or 5% error rate, would you do it? Would the benefit of a safer society outweigh the error rate?

Most people would say yes because they would think it wouldn’t affect them. The 2% to 5% of people whose family members are innocent and get locked up would feel differently.

What do you think?

I wouldn’t compromise that value.

I wouldn’t compromise it either. It’s incredibly dangerous.

We’ve been riding on a bit of a high. There are a couple of things this administration could do better: communicate when it makes a mistake, and recognize that human rights are a core value.

I believe there’s a sadistic nature to this administration that is an Achilles’ heel. When you see them producing videos from these prisons saying, “This is going to happen to you,” and the prison is known for torture and profoundly abhorrent conditions, that sadistic nature needs to be dropped.

It will derail them. People in this country care deeply about human rights and due process.

There’s also a constitutional issue because Trump and this administration are disobeying the court when it says to slow down. There is zero cost to Americans to send these people to a holding station and investigate each case to make sure we don’t send a gay hairdresser to El Salvador to be tortured in prison, if that is in fact what happened.

It is against all American principles not to have due process and to risk sending innocent people to prison. Republicans should say that this will derail the administration. These are the kinds of things that build up over time.

People in America who are rooting for this administration to do good things, including me, will stop supporting it when it does things like this. I represent a large swath of people, and this loses them votes. I predict that’s why they’ll lose the midterms if they continue doing this.

This isn’t virtue signaling. I cared about human rights long before the All-In podcast existed. My first job was at Amnesty International. This is a giant mistake by the administration.

This whole debate is a question of ends versus means. We all agree that a 99.9% reduction in the homicide rate in El Salvador is a good outcome. It’s a good end.

This is 1 of the great philosophical questions. There has to be a balance between the 2. Human rights are a core American value, and I agree with much of what David said.

But if you go back to the homicide rate, the Trump campaign spoke to the mothers and families of people who were brutally murdered or raped and murdered by people who were illegal immigrants, convicted criminals in their home countries, and known criminals who somehow still weren’t apprehended and then committed terrible crimes similar to those they had committed before.

You have a lot of compassion for the victims too. All the administration has to say—and I think more due process is good—is, “If we made a mistake, it’s terrible, and we’re going to fix it immediately.”

That gives you so much credibility.

Agreed. But isn’t due process happening? How else would you know that 3 out of the 238 were misidentified? The fact that you know that means the process is working.

The question is what about the other 235 people? Their families spoke to the press. The information didn’t come from the administration.

You could hold all 238 people in humane prisons and process them. Their attorneys could review their cases. You could make the cases public and say, “Here are the people we’re deporting and why.”

So your specific issue is where they’re being sent?

Yes. They’re being sent to one of the most sadistic prisons in the world.

Chamath, what’s your answer to the 5% trade-off question? If you sweep up 5% innocent people, is it worth that trade-off for lower crime?

I think that probably happens in America’s prisons today. It’s probably more than 5%.

The question is what an acceptable rate is. It’s a little like the trolley problem. There are many variations. This is probably above our pay grade, but it isn’t above the president’s. That’s part of what he was elected to do: make that call.

He didn’t hide behind rhetoric. He was very clear about his intentions. He brought the Laken Riley Act, and he was explicit about what he intended to do.

It still feels like due process requires being more careful with this, just as with Signalgate.

We’re 4.5% into the administration. I’m not saying that mistakes are okay if innocent people were sent to this prison. That’s terrible, and I hope it gets rectified quickly.

If there were an acknowledgment saying, “If we made a mistake, we’re going to fix it,” that would go a long way. Instead, we’re seeing promotional videos about how brutal this prison is.

An important point about the administration’s strategy is that this may not be the beginning of a continuous process. They may be sending a signal they expect will cause people to immigrate out of the country if they’re committing crimes or at risk of being caught for gang associations while in the United States.

The strategy may be to chase people out. There’s a calculation that if they do this to the first 240 people, even if 4 innocent people get caught up, there’s a net benefit because it reduces the number of gang members coming to the country. Some people may choose to go home.

They clearly believe this will save innocent lives in America. That’s their belief system.

Here’s Secretary Kristi Noem from the prison at CECOT:

Kristi Noem

I’m visiting this facility today, and first of all, I want to thank El Salvador and its president for their partnership with the United States of America to bring our terrorists here, incarcerate them, and impose consequences for the violence they have perpetrated in our communities.

I also want everybody to know: If you come to our country illegally, this is 1 of the consequences you could face. First of all, do not come to our country illegally. You will be removed and prosecuted. But know that this facility is 1 of the tools in our toolkit that we will use if you commit crimes against the American people.