Trump–Musk决裂余波 + 一位DOGE程序员发声 + ChefGPT

Well, Casey, I don't know if you know this, but today is my birthday.

I did know that. Happy birthday, Kevin.

Thank you.

And what better way to celebrate than by making a podcast?

It's true. Today, I am 38, or as they call it in Silicon Valley, a senior citizen.

And what is it like to be entering your sunset years? Do you have enough put away for retirement yet?

I have actually been rebalancing my 401(k), which is a very 38-year-old-coded activity.

It really is. I was trying to consider what to get you for your birthday, and I thought, "Why not ask the chatbots if they have any good ideas?" So I just said, "Hey, my friend is obsessed with AI. Do you have any gifts that might be funny?" And it gave me a lot of things that were not funny. But one suggestion it did give me was that I would hand you a book.

Hmm.

On the cover of the book, it would say, "What to do after AGI?" And then you would open it, and it was just a blank notebook. I thought, "That's actually pretty good."

That's pretty good.

Yeah.

So did you do that for me?

No, it was very expensive. I thought, "Hmm, no thanks."

Okay. Well, it's the thought that counts.

Well, the nice thing about a podcast is you can just say it. It's as if it were real.

Yeah.

You know what I mean?

So here, pretend you're handing it to me.

Here, enjoy.

Oh, thank you.

Yes.

I appreciate it.

Crinkle, crinkling the paper as he unwraps. Crinkle, crinkle.

What a good friend you are.

Yeah, so enjoy that.

All right. I'm Kevin Roose, a tech columnist at The New York Times.

I'm Casey Newton from Platformer.

And this is Hard Fork.

This week on the show, Elon and Trump's feud goes nuclear. What does it mean for them and for America? Then, entrepreneur Sahil Lavingia joins us to discuss his time at Doge and why it came to an abrupt end. And finally, The Times' Pete Wells is here to explain what some of America's best chefs are doing with AI. I hope it involves cooking us dinner.

1. Trump And Musk Break Up

Well, Casey, the big tech news this week is that there is trouble in Washington.

That's right, Kevin. We're having a broligarchy blowup of the highest order.

For many months now, there has been speculation about if and when President Trump and Elon Musk would come into conflict, would see their bromance fade, and that is happening today in a way that is pretty crazy.

Yeah, and I have to say, Kevin, I was somewhat cynical when all of this started about how real it was, but by approximately midday on Thursday, all of it seemed extremely real.

Yes, it's very real. We are recording this on Thursday, June 5, around 5 p.m. Eastern Time. By the time you hear this, things may have changed. It's all happening very quickly. But Casey, let me set the scene for you.

Yeah.

I'm in a meeting. I'm doing an interview. I'm not looking at my phone or Twitter or any of my incoming Slack messages or emails.

Mm-hmm.

I get out, and I have about 200 people yelling at me that Elon Musk and Donald Trump are fighting. This was something that we had been keeping our eyes out for. Obviously, these are two very combustible personalities who like to fight, especially on the internet. But Casey, what are they fighting about, and how did this start?

Well, it all started with the One Big Beautiful Bill, Kevin. That is the giant budget package that President Trump has been trying to move through Congress. It passed the House and is now being debated in the Senate, and it emerged over the past several days that Elon Musk did not like the bill. In fact, he did not think it was beautiful at all, and earlier in the week he called it, quote, "a disgusting abomination."

Yes, so he's not a big fan of this bill, but that by itself does not seem like grounds for a legendary falling-out with one of your former top aides and political allies. So what else is going on here, other than a dispute over this budget bill?

Well, let's mention a couple of specifics about the bill, though, Kevin, because it is sort of the best evidence we have for what might be driving Elon to be so angry. One thing is that this bill would eliminate an electric-vehicle tax credit, which is, of course, beloved by Tesla owners. Elon is, of course, the CEO of Tesla. So this could potentially be a big financial hit to Tesla at a time when its stock has been falling precipitously.

In addition, Axios reported this week that Elon was seeking a deal with the Federal Aviation Administration to get it to use Starlink for its communications infrastructure. Starlink is, of course, the network of satellites owned by SpaceX, another Elon company. So there we have 2 examples of what would have been fairly big deals for Elon in 2 of his different businesses that he pushed very hard for and did not get.

And that leads me to, I think, the emotional or psychological answer to your question, which is: Why is Elon so mad? To put it simply, Kevin, this is not what he thought he was paying for when he tried to buy the 2024 election.

Right. He was out on X this afternoon saying that basically Trump was only president because he was involved in the election. He donated all that money. He did all that campaigning. He is claiming credit for not only Trump's presidency, but also the Republican congressional majorities that that party currently enjoys. I saw him claiming credit for that stuff, and I thought, "Ah, man, Trump is not going to like that." And as it turned out, Trump did not like that.

No, he did not. He posted on his own social network, Truth Social, a number of things. He said that Elon was, quote, "wearing thin," and so he had asked him to leave the administration, which I think was pretty well known. Later, though, he escalated things by saying, quote, "The easiest way to save money in our budget, billions and billions of dollars, is to terminate Elon's governmental subsidies and contracts. I was always surprised that Biden didn't do it."

Right. And then Musk fired back, talking about how Trump is mentioned in the Epstein files, and that's why they have not been released. He's also, more recently, talked about needing a new political party. Most recently, as of this taping, he shared a video in which Donald Trump and Jeffrey Epstein are talking in 1992.

Well, and don't forget, Kevin, he also quote-tweeted a right-wing influencer who had asked in the wake of all this, "Should Trump be impeached?" Musk said, "Yes." He was essentially agreeing with a post that said, "Trump should be impeached and replaced by JD Vance." That was the post where I was like, "Okay, these are not two guys playing drama for the media. They're actually really mad at each other."

Yes. This is a big fight, and it has got a lot of people talking. But while we wait for the dust to settle here, I don't think it is too early to start thinking about what this means, not only for the Trump administration, but for Elon Musk and for some of the projects that Elon Musk had undertaken while he was in Washington. So, Casey, what do you think this feud and this breakup between Donald Trump and Elon Musk means for Elon Musk? Let's start there.

2. Musk Faces A Reckoning

Well, I think it has the potential to do a lot of damage to Musk. Remember, one of the reasons why he made his way out of the administration was because his businesses are increasingly in trouble, and that begins with Tesla. Tesla has a number of challenges that we don't have the time to get into today, but there's a real sense that Musk has not been focused on that business and needs to return to it.

Instead of doing that, though, he just amped up his feud with Trump. Today, Kevin, Tesla's stock declined 14%, so I would not be surprised if it continues to decline in the days ahead as people process what this means.

But let me throw out an even spicier possibility for why this could be trouble for Musk. I want to say that Elizabeth Lopatto noted this in The Verge, but she reminded me that The Washington Post had reported that Musk began his career in the United States by working here illegally. Elizabeth is wondering: Could this actually complicate his naturalization? Would Trump actually go after Musk and try to strip him of his citizenship? Let's just say, if that happened, it wouldn't be the first person who thought he was a citizen of the United States who found himself deported during this Trump administration.

Right, and that seems a little crazy when you just say it like that, but allies of the president are actually calling for this now.

In fact, just recently, Times reporter, my colleague Tyler Pager, says that Steve Bannon, the former Trump adviser, was calling for President Trump to not only cancel all of Musk's contracts but to initiate a formal investigation of his immigration status and maybe deport him. So I don't know that this is likely to happen. This seems like maybe a bit of wishful thinking.

But I do think it is going to be a very painful breakup for Elon Musk, not only because of his exposure through Tesla to some of these policy moves and the billions of dollars that he would stand to lose if Tesla's stock slide continues, but also because of things like government contracts for SpaceX, right? This is a man who has billions of dollars in exposure to government contracts. If some or all of those contracts are canceled by President Trump, he does stand to lose out that way. So—

Yeah.

Not a good day for Elon Musk and his portfolio.

Well, and let me throw in one more legal risk, Kevin, because Democrats on the House Oversight Committee have been trying to subpoena Musk. And recently, Representative Nancy Mace, who is a Republican from South Carolina, suspended the hearing so that Republicans could come and vote against the subpoena. You can now imagine a world where Trump goes to the Republicans and says, “Hey, you know what? Go ahead and let that subpoena take place,” right?

So there are a lot of legal risks here. And I would just say that, often in cases where authoritarians take power, it's the billionaires in the country who think, “I know how I can make myself … I'm gonna cozy up to this person.” And then they find that when something goes wrong, they're actually in more danger than anyone else.

Yeah. Casey, what do you think this breakup means for President Trump?

3. Trump Loses His Heat Shield

Well, think about how Musk functioned during the first few months of this Trump administration. He was Trump's heat shield. He went out, did a bunch of really unpopular things, and largely took the blame for it. He took the blame for DOGE dismantling wide swaths of the federal government, right?

Now Musk is gone, and so there isn't that same kind of heat shield out there. Trump is gonna have to own his decisions more, and I think his presidency is going to look different because of that. While this feud is undeniably entertaining and actually hugely consequential, I worry that something that's going to be lost is how useful these two men were to each other during the period they were useful to each other. Because they both got a lot out of this arrangement.

Absolutely. It was a very lucrative arrangement for both of them, in some sense. And I wonder if that doesn't end up mending the feud here if these guys get all of their rage out on each other and then decide, “Well, if I'm Elon, I don't really have anywhere else to go,” right? “I've burned my bridges with Democrats and on the left, and the Republican Party now belongs to President Trump.”

And so I think if you're him, you may decide to stay out of politics altogether. But if you wanna stay involved, you're kind of politically homeless if you're Elon Musk. And so maybe you do end up wanting to come back to Trump at some point. But it's too soon to say, I think.

And in the meantime, he's promising to create a third political party that will represent 80% of all Americans, and let's just say I'm looking forward to seeing how that one plays out.

4. The Tech Right Splits

So Elon Musk no longer has a formal role in the Trump administration, but plenty of his allies and people that he brought in are still there. I'm curious what you think this breakup will mean for people like, for example, David Sacks, who is the president's adviser on AI policy and crypto and a longtime ally of Elon Musk's. What do we think happens to the tech right if Elon Musk, who was responsible for some of that realignment behind Donald Trump, is no longer in the president's camp?

Well, you know, Kevin, have you ever been in a situation where two of your friends broke up with each other and then you had to choose which one you were still gonna be friends with?

Yes, the friend divorce.

Yes. Well, so there's about to be a major friend divorce in Washington, D.C., and we're gonna have to see who sides with whom.

The way that Trump tends to operate is by showing loyalty to anyone who is loyal to him. And so if, in the days ahead, the David Sackses of the world go to him and say, “Elon is wrong, you are right, Make America Great Again,” I can imagine those people continuing to have a role in the administration.

If, on the other hand, they feel loyalty to Elon for whatever reason—and Elon is the world's richest man and could give people a lot of reasons to be loyal to him—then you might see more of a split. So this is one where we truly don't know the answer to it, but I think it's one of the most interesting questions to watch in the week ahead: How do the members of the tech right whom Musk had brought to Washington respond to what has happened today?

5. DOGE Faces An Insider Threat

Well, and we should also talk about what happens to another group of people that Elon Musk and his allies brought to Washington, which is DOGE. DOGE still exists. They are still out there doing cost-cutting measures in various federal agencies. If Elon Musk goes, does DOGE go with him?

Yes. Whither Big Balls might be another way to put it, Kevin. Big Balls, of course—

As Big Balls goes—

Yeah.

So goes the nation.

Big Balls, of course, is one of the most famous employees of DOGE, and you can look up his accomplishments online.

So here's why this is an interesting question. I was seeing some chatter about this online over the past couple of hours. Because many of these folks who are part of DOGE are considered to be Elon people, Elon loyalists, I'm sure there is now a contingent within the Trump administration saying that DOGE is now what they would call in Silicon Valley an insider threat. This is a group of people that you now have to be worried about doing some sort of sabotage, right? Showing loyalty to Elon somehow.

And so I would not be surprised if this was actually a conversation being had right now at the White House: Which of these folks is loyal to us, and which are loyal to Elon? So we may actually see a big DOGE purge coming. A DOGE purge, Kevin.

It is just so wild to me to watch this breakup happening more or less in real time. I remember when Elon Musk threw his weight behind the Trump campaign, people were saying, “Oh, this will be short-lived. There's no way these two can coexist. Neither one of them is good at sharing the spotlight. This is gonna be a combustible situation.”

And I just thought there would be enough mutual benefit in it for both of them to keep this thing going for at least longer than it appears to have gone. But I don't know. Casey, are you surprised by what you've seen over the past couple of hours?

Not really. I wrote a year-end predictions post. One of those predictions was that the broligarchy would blow up. I am far from alone in making that prediction. I think maybe even most observers assumed that something like this would happen, just because we have enough data to know about how Donald Trump and Elon Musk tend to view other powerful people in their orbit. Typically, when they don't get what they want, they get super mad and turn on each other.

And so when it finally happened, initially I was skeptical just because they were kind of playing footsie with each other a bit. I thought, you know, Elon has a lot of good reason to make people like me believe that he is mad at Trump and is creating this distance between himself and Washington. Maybe that makes it a little easier for him to sell Teslas to Democrats again.

But when it went fully nuclear, there was actually a sense of relief because in a year where so many news developments have made me feel like I am crazy and no longer live on the same planet that I was born on, this was the moment where I was like, “Okay, some laws of gravity do still apply. Sometimes you can just trust what you see in front of your eyes.” And this was that moment for me.

Yeah, it does make me really feel for the Tesla owners out there who are trying to figure out which bumper sticker to put on their Tesla to indicate that they're not on board with Elon Musk's political agenda here. I'm not sure that any of them have changed their bumper stickers in the last few hours, but it does seem like Elon Musk has both alienated Democrats, who no longer wanna buy Teslas, and now, as of a couple of hours ago, he has probably alienated the MAGA wing of the Republican Party. And you can assume that they will not be buying his Teslas and Cybertrucks anymore either.

Well, Kevin, I'm sure you have a lot more questions for me, but I really should get back to reading social media posts and sharing them in my group chats. And if there was a social media post that seemed really important that we didn't talk about, it's probably because it was posted after 5:11 p.m. Eastern Time on Thursday.

When we come back, his DOGE days are over. Sahil Lavingia is here to tell us what he did to the federal bureaucracy.

6. Inside The DOGE Experiment

Well, Casey, speaking of Elon Musk and DOGE and the Trump administration, we have a very exciting interview to share on the show today.

Yes, an interview that we should say we conducted before all of the dramatics on social media on Thursday between Musk and Trump.

Yes. So last week, before any of this happened, when Donald Trump and Elon Musk were still friends, a former DOGE employee, Sahil Lavingia, published a fascinating blog post about the 55 days he spent working for DOGE before he was fired. He was stationed inside the Department of Veterans Affairs as a software engineer. He's a veteran startup worker. He worked at Pinterest, and then he started a company called Gumroad, which is an e-commerce platform.

I think this is a really important conversation to have because despite the fact that Elon Musk is now persona non grata in the Trump administration, DOGE is still there. Its employees are still stationed throughout the federal government. I wanted to take the chance to have a rare conversation with someone who has actually worked for DOGE inside the federal government and who has seen what it's like up close.

Let's bring him in.

Sahil Lavingia, welcome to Hard Fork.

Thanks for having me.

So, Sahil, you have been around Silicon Valley for a while. You had a pretty good thing going. You were one of the first employees at Pinterest, and then you went on to found your own company, Gumroad. What made you interested in stepping away and moving to Washington, D.C., to work for the federal government?

Yeah, I've always wanted to work for the federal government. I think it's such a cool opportunity. As a software engineer and product designer, which is my background, I think it's hard to think of another opportunity for such a large impact.

When I was even in high school, I always felt like I would eventually want to work for the federal government. When I left San Francisco, I remember feeling like the only things that private businesses were responsible for were the little jewel boxes on Mission Street.

Everything else was powered by public infrastructure. Eventually, if you're a competent software engineer and you want to really maximize your impact on the world—the Steve Jobs sort of dent in the universe—you'd have to go work for the government eventually.

I can imagine that the past few administrations might have been happy to accept your help. Why was this the moment that you wanted to step in?

Yeah, I've actually applied to work at USDS before, around 2015, when we failed to raise our Series B at Gumroad, laid off most of the team, and I was trying to figure out what I wanted to do next. I applied to USDS, but I think my background at the time was still very startup-y.

Mm-hmm.

It wasn't exactly the Venn diagram overlap of skill sets with what you may need in the federal government, the largest company on Earth. So I applied, and I never heard back.

When this DOGE thing started to happen, I felt like this met the criteria of shipping code for the federal government, and I wanted to work there.

Yeah. Now, how did the hiring process go? Did they do any kind of ideological or political filtering? Did they ask you, “What do you think about foreign aid or USAID or anything like that?”

I was asked, “Did you vote for Kamala Harris?” for example.

Hmm.

Basically, if you voted for Kamala Harris, you shouldn't waste your time here. You're not going to get through whatever the process ends up looking like.

This was before the inauguration, so it had to flow through the Trump transition team. I said, “Look, I actually didn't vote for Kamala. I didn't vote for Trump either, so if they need that, it's not going to really work out.”

But anyway, they asked these sorts of questions. My guess is this came from Trump and that side of the administration. I don't think DOGE cared too much.

So they approve your application, and then what happens? Do you just get assigned a role?

Yeah. It's interesting. I feel like I never even got an approval.

Hmm.

It wasn't like, “Congratulations,” or, “Here's your offer letter, here's your salary, here's your boss, here's your department.”

Basically, what happened was I passed, I guess, to some degree. Then eventually they said, “Hey, we have a role for you at the VA as a software engineer. You should go talk to this DOGE team lead at the VA, and if that seems like something you'd want to do, let us know.”

That's kind of what happened. I talked to them, and eventually they said, “If you're interested, you have to get fingerprinted here in D.C.”

It's this chicken-and-egg thing where you're not going to get any information until you move to D.C., and once you move to D.C., you're here. So I never actually—I mean, even today, I still don't know what my salary was. I assume it's zero, but every once in a while I get a check in the mail for some de minimis payment, which I assume is to make sure I'm not violating—

Wait, you didn't talk about your salary before you agreed to take this job?

I'm a phenomenal negotiator.

Yeah, that's strange.

I didn't know. I'll tell you the facts I knew.

Yeah.

So I knew that I was supposed to show up at 7:45 a.m. at the VA—the VACO, the Veterans Affairs Central Office—in D.C. on March 17, and actually, that's all I knew.

So, yeah, I didn't know who my boss was. I think I learned on the third day or something. I didn't know my role. It doesn't say your role on your badge.

I basically got my PIV card and a government laptop, which they call GFE, or government-furnished equipment. Then the DOGE people there walked me through what I was supposed to be doing and what they were working on, and I was sort of there to offer my technical expertise.

Though the laptop I got, for example, couldn't run Python. I couldn't run Git, and I couldn't download anything. So I had to work through this exercise of being able to cook—

Right.

Without any equipment.

I want to ask a little bit about the orientation. You mentioned that some of the folks at DOGE explained to you what they were working on. You're there presumably to serve the veterans of the United States. Did you have discussions with DOGE about, “This is who we serve, this is what they need, this is how we want to make things better for them,” or was it more focused on cutting costs and that sort of thing?

The DOGE conversations were around cutting costs. I think there was a Venn diagram overlap of, at the end of the day, I'm an employee—at least that's what my badge said. I'm an employee for VA. You go through VA-specific onboarding. There's no DOGE-specific onboarding. You get a VA laptop. I have the I CARE memorandum on my backpack now.

So I was always like, this has to meet the filter of what the VA wants. Nothing's going to happen if the VA doesn't want it, and so it has to meet 2 criteria for something to be interesting: You have to save money and make the veteran experience better, or at least not make it worse.

But it wasn't an explicit conversation of, “Okay, how do we actually ship software to make the veteran experience better?” I had a lot of those conversations because I was interested in doing that, so I would meet with the Office of the CTO, which was actually pretty well developed at the VA.

Mm.

Because USDS OGs had kind of built a coalition there. They use GitHub, they have Slack, that sort of stuff.

That’s really interesting because I think the larger narrative that people have heard from DOGE and from Elon Musk and people who are affiliated with him is that once you start looking under the hood inside these government agencies, you find all kinds of waste, fraud and abuse, and skeletons in the closet. Basically, what you’re saying is that you were surprised by how well the VA worked. Am I hearing that right?

Yeah, I was surprised by how effective the government is. At the end of the day, though, you’re defining effectiveness by what you’re asked to do. I do think the government is asked to do a lot of weird stuff. They tend to be quite effective at doing those things, but I think Congress has a role to play in this.

You can think of Congress as the worst PM you’ve ever met, right? You’re asked to build an app with all these weird specifications. You’re doing a job, but you’re playing Twister as an organization to do it. But surprisingly, there were many meetings I was in where I came in a little bit antagonistic and adversarial: Why is this contract $40 million to process claims or something? And I would leave thinking, “Oh, I get it.”

Actually, it’s pretty impressive.

You wrote in that post about this DOGE all-hands meeting where Elon Musk was asking the DOGE employees for ideas about how to improve the public perception of DOGE, which I imagine at this point was getting all kinds of criticism for gutting various agencies, laying people off, cutting foreign aid and things like that. Did it seem to you at any point during this experience that Elon Musk and other DOGE leaders were interested in solving actual problems? Or were they just concerned about the appearance of productivity and efficiency?

I think maybe I’m naively optimistic, but I think there was truly a desire to solve problems, because that’s how you save money at the end of the day: You solve problems, you kill two birds with one stone. That’s what software can do.

My guess is that, for example, at the meeting I suggested, “We should livestream this meeting.” This would help, right? If people saw how the sausage was made and how we weren’t scheming on how to kill a bunch of African children, I think it would help the brand image. And he embraced the idea. He was like, “That sounds awesome. We should do it next week. Let’s do it next week. We’ll start doing it next week.”

Unfortunately, there was no other meeting that I was a part of, so I never got to follow up on that “next week” thing. I didn’t get the invite. Did it happen? I didn’t hear about it on Twitter. What’s going on?

My guess is that the most optimistic, good-faith take would be that they were truly interested in these kinds of ideas, but the Trump administration wasn’t. I think people realized pretty quickly that this was not going to be a walk in the park, and that they should just focus on what they could do, which turned out to be cutting contracts. At the end of the day, a bunch of software engineers aren’t really necessary for that effort.

Well, so tell us a bit about the cutting-contracts piece. You have this situation where folks like yourself, who are public-spirited and want to see if they can make a contribution, parachute in, get assigned a role at an agency they have no prior familiarity with, and are then just asked to hack millions of dollars out of the budget. How did that work?

Yeah. One of Trump’s EOs says that the DOGE team gets to sit in on a lot of contracting meetings, so we helped them review as many contracts as they could. Beyond that, it was just lending my technical expertise. It was basically sitting in on a meeting where someone would say, “We’re paying $4 million for this chatbot on VA.gov. How?” And I’d say, “Well, that would take me a week to build and integrate, and could you explain why that costs so much more money?”

A lot of it was just that there was somebody in the room who had coded before. I think that was a lot of it: You’re paying a lot of money for things without the domain expertise to understand how much they should cost. So I was, hopefully, somewhat, the credible “expert,” quote-unquote, who would come in and talk to Booz or Deloitte and just be like, “Hey, do you guys use Cursor? What are the efficiency gains? Could you pass them on to us? Could you shave 25% off the bill?” That sort of thing.

At the end of the day, we didn’t have any true power. I think a lot of people think DOGE had a lot of power to fire.

Mm-hmm.

And at least from my perspective at VA, we did not. The EOs don’t give us that power. Unless VA is willing to give us the keys, we can’t drive the car. But we could sit in on meetings, be annoying, and ask questions.

I wanted to ask you about these meetings because I think there’s this popular image that was going around during the early days of DOGE that DOGE was just this group of kids in hoodies who would barge into these meetings with very experienced civil servants, demand to know what they were working on, and just run in and rip things up. It offended a lot of the career employees of the agencies where they were stationed.

Was that what you witnessed? Were you the archetypal guy in a hoodie coming in and telling these 30-year veterans of the VA how to do their jobs?

I was not that person. I did not wear a hoodie. I wore a shacket, a little bit of an upgrade.

But yeah, I’m sure to some degree that was happening, frankly. I’m not saying these stories are generally made up. I think they come from truth, and there was generally this vibe of, “They’re not going to help us,” or, “They’re going to give us the wrong information and cause us to go down this path,” which I think is probably common in a transition period.

I think every transition has a bunch of political appointees who come in, and you have a new boss who doesn’t really know what they’re doing. Everyone’s had that manager who doesn’t really understand why someone is taking so long or why they’re doing something. But I don’t think it was as much this idea of an Office Space interview process.

Even the RIF, right?

RIF is a reduction in force, also known as layoffs.

Exactly: the reduction in force, the Workforce Optimization Initiative, where Trump wants to cut the federal workforce. That entire EO is basically about making a recommendation to the White House about what to do. It’s not about actually firing anyone. Specifically, for that EO, it gives DOGE zero power to fire anyone.

So this idea of these Office Space conversations where you’re asking, “What are you doing here?”—there’s no point in that. However, I will say I was on the older side of the engineers. Of the people who were identified as software engineers, I think there were maybe 2 or 3 people who were above the age of 30, and most people were relatively young and probably hadn’t worked a ton in private companies either.

It’s just different, too. I think there was this culture clash, this lost-in-translation thing.

Are there any anecdotes or stories from your time at DOGE that would help illustrate the culture clash?

I would swear.

Mm-hmm.

I remember someone telling me something like, “That will get you fired.” It was just interesting—the rules and regulations that exist that sort of define conduct. There are just a lot of those things.

Even the DRP massively did not work. I think everyone was—

What’s the DRP?

The DRP, the Deferred Resignation Program—the fork-in-the-road emails.

I see.

Right? From someone who works in private industry, that’s an incredibly good offer: You get 8 months of severance. That’s a lot. That’s pretty good, you know? But the number of people who took it was drastically low, and I think there’s just this cultural disconnect about why you’re here. Well, you’re here to work for the government for life.

An extra 8 months doesn't compute. You're on a completely different career path. It's not like you're just going to exit the government and then go work at X. There are completely different ways of thinking about how you want to live your entire life.

It took time for me, at least, to understand that these people love their jobs, and I think that's a lot of the cultural disconnect: inefficiency versus efficiency. Everyone sounds the same, looks the same, dresses the same, but the way they make decisions manifests very, very differently.

Well, what was making you swear so much, Sahil?

I swear a lot. I just swear a lot. I think I have a kind of San Francisco pirate mentality or something like that. But it wasn't in anger. It wasn't at anybody, but just, why the fuck is this so weird? It's this kind of surprised emotion that I can't really help but express in that way.

Mm-hmm.

I can't believe you just swore on our podcast, Sahil.

I know. Please cut that out.

You're fired. Get out of here.

I'll give you one example: a little bit of a culture of sarcasm.

Yeah.

Mm-hmm.

Sarcasm just doesn't exist in the government workplace for some of these reasons. I'd come in and they'd be like, "What are you here for?" And I'm like, "I'm here to riff everybody." And it's dead.

Wait, you thought that was going to go over well?

You gotta read the room, dude.

I know. That's part of the culture. I'm not very good at reading the room. I'm good at writing code, not reading rooms, right? It's a kind of skill. It tends to be a different—

Yeah.

—set of skills.

I can see how that one didn't go over well. "Hey, I'm new here. I'm gonna let you all off." Ah.

So maybe the story was about me.

Yeah, maybe workshop that one next time.

Yeah, let's—

Work on that. I'm curious. I think a lot of people, Sahil, will be listening to this and thinking, "This guy sounds really naive." He hears about this DOGE thing. I think a lot of people, especially on the left, think that DOGE is sort of just a pretext for an ideological purge. If they look at what the agency has been doing inside these various federal bureaucracies, trying to essentially purge any hint of leftism or wokeism in what these agencies are doing, they will say, "Sahil should have known better than to expect that he was going to go in and actually be valued for his expertise in coding. He did not know—or should have known—that he was signing up for something very different than that." What is your response to that?

Of course The New York Times would say that.

Get his ass.

No, I mean, look, at the end of the day, when I joined Pinterest, people were like, "You know that startup's not going to work, right?" And it's just like, statistically, no, but maybe yes. If the maybe has an outsized impact, then the math ends up working.

That's what I felt. I felt like, look, I think all startup people are naive. I think that's part of what makes us startup people: we're not willing to just listen to the traditional, "This is why we don't do that." It's just like, no, we're not going to do that thing, and then you learn the hard way, you know? But also, every once in a while, you get an iPhone or you get AI or something like that.

Mm-hmm.

But yeah, if people are like, "Hey, you failed," I'd be like, "Yeah, that's what happens." Lots of things are a waste of time, and I have time, so I'm going to try.

Let's just close the loop, Sahil, on why you were let go from the government. You suggested in your blog post that your leading theory for why you were RIFed, as you put it, is because you talked to a journalist at Fast Company. Is that still your understanding of why you were let go?

Yeah, I never heard anything from them. All I heard was, "Did you run this by anyone at DOGE?" I said, "No, I did not run it by anybody." Then, a day later, I just got an email from GitHub that was like, "You're no longer a member of the VA on GitHub." And I was like, "Oh, shit. What happened?" It turned out I was RIFed.

But yeah, that's all I've heard. I've never heard from anyone affiliated with DOGE since then about anything I've done. They just kind of ghosted me.

That seems very sudden as a way to get laid off or to find out that you have been laid off. Did that process, that ghosting, as you described it, change the way you felt about your time with DOGE?

Yeah, it honestly made me feel like I wasn't going crazy. I felt like I was misreading the room, and I thought, clearly, we have a good goal. Why do we keep making these kinds of mistakes? Why are we so unempathetic?

And when I just got ghosted—just disappeared, my access just stopped working, no response—it honestly made me feel good in a way, that I wasn't the problem here. It's just this culture of this weird entity. It's this emergent behavior of a lack of empathy that exists in these systems, I think, and everyone gets to blame somebody else and kind of opt out of their own…

But yeah, I think it gave me the sense that my wife was right. This was not exactly the best use of my time. You know, the FOMO was real, right, at the end of the day.

It's a powerful reminder that I think a lot of other Silicon Valley CEOs could learn, which is: listen to your wife. She's onto something.

I think it's helpful. I think it is. Now, when she tells me something, I'm like, "My default is yes."

Well, if DOGE has done one good thing, it's made your marriage more harmonious. So congratulations.

Thank you.

I'm curious how you think about the theory of DOGE. I think one thing that often occurs to me when I hear people in tech talking about the government and bureaucracy is the idea that efficiency is the ultimate good, right? If you have a system that is slow-moving and has friction involved in it, and it's using outdated technology, it's an unquestionable good to go in and smooth those bumps, to make things faster, more efficient, cheaper.

There's also, I would say, something about government that is potentially inefficient by design, right? Part of the reason that things in government take longer than they do in the private sector is because governments are not businesses, right? They have to account for real constituents. They have to weigh various priorities against each other, sometimes conflicting priorities. They have to do the deliberative process before they embark on something that could potentially change people's lives.

I'm curious if you still think this Silicon Valley attitude is needed in Washington, if that is something that the federal government should have more of, or if you're coming to the view that maybe these are just 2 different agendas.

I do think there's a synthesis here, which is: I think if you define the goals accurately, then efficiency is better. If you're able to meet those objectives but spend less money doing it, or make it faster, or require people to fill out fewer forms, then it is better. Efficiency is better.

I think the lost-in-translation thing is that Silicon Valley doesn't serve every person, but if the government has a program, it's going to have to serve everybody, right? So I think efficiency is still good. I think it's just really important to understand what the constraints are.

I personally think the better word would be modernization or simplification. Efficiency now has such a loaded connotation to it: just "cut," "get rid of," things like that. But I do think simplifying things—

You know, I think what tends to happen is the government just stacks things on top of things that exist. It doesn't often refactor code and combine systems. There are a lot of agencies, each agency spun up to do a new thing, and I do think it would be good to simplify every once in a while and sort of say, "Hey, these 3 things that came up in 3 different eras, we could now merge."

I'm sure many people are sympathetic to the idea that the government is not as efficient as it could be, and they would love to see various aspects of it modernized so that the government was just easier to use.

My instinct for how to do that would be to do it slowly and carefully by bringing in a lot of expertise and then giving those people the time to work. The DOGE approach was to bring in a bunch of people who don't have the relative expertise, ask them to get a ton done in a hurry, and apply this kind of ready, fire, aim approach to the federal government.

So I think there are a lot of Americans right now who are looking at DOGE and seeing it as a kind of catastrophe, where we've disabled vast swaths of the federal government to no apparent benefit. And I wonder what you say, as somebody who worked on it, to people who just see DOGE as a catastrophe.

Yeah, it's unfortunate. I don't regret my time there because I didn't work on those specific projects, and I think I had a positive impact. It's like, do you throw the baby out with the bathwater? Hopefully, I was more the baby.

But, yeah, it's unfortunate. As another DOGE engineer said, mistakes were made. I don't know if the people who made the mistakes felt they were mistakes, but from the perspective of people who really wanted to come in and just work hard and make changes, I think it was less than ideal.

But at the end of the day, this is a Trump administration. That's who Americans voted for. Some Americans are very happy with what's happened and would like more of it to happen. We live in a democracy, and I'm just not an opt-out type of person.

I'm only on Earth for a short amount of time. I can't wait for some person or some party that I really agree with, and I'm only going to work there if I really align with everything they do. I have a lot of faith, frankly, that there are a lot of checks and balances, that either these systems will come back if they were blocked.

There are a lot of things that prevent this ready, aim, fire approach. That gun never actually goes off. It looks like it goes off. Everyone reports that it went off, but then a judge rolls it back.

So I think it is this naive optimism, and unfortunately, it's like, I am who I am. I'm not going to change. I really believe in using software to make the world better.

Well, I think that's a good place to end. Sahil, thank you so much for coming, and welcome back to civilian life.

Thank you. It was great.

Thanks, Sahil. After our interview, we reached out to the White House for comment on Sahil Lavingia's experience at DOGE, in particular, about a couple of things that he said about his vetting process. The White House said that you are allowed to ask people political questions when they're being vetted for political appointments, like roles at DOGE, and that DOGE employees understood their general responsibilities to cut waste.

We also asked for more details about Sahil's termination. The White House referred those questions to the VA, but the VA did not get back to us.

When we come back, let him cook. With AI? Pete Wells says America's top chefs are starting to do it. Cook with AI, I mean.

7. Chefs Find Their AI Muse

Well, Casey, I know you are a big foodie.

Mm-hmm.

You're always telling me what restaurants to go to. We went out to dinner this week, and you had gotten the clam chowder. I don't know where I'm going with this.

That's a classic foodie thing to do, to get the chowder. Actually, the most foodie thing about me, Kevin, is that many weeks after the show, we go downstairs and eat a sandwich from the Subway Corporation, which is known for its exquisite gourmet delicacies.

No, but you are a foodie. I know this about you, and so this week there was an article in The New York Times that made me think, "Well, I gotta talk about that with Casey."

Yes, and I love this story, too. It comes from Pete Wells, the great former food critic for The New York Times, who is now a food writer at the paper, and he had a great piece about how some of America's most famous chefs are starting to use AI in their work.

So because this is going to be a segment about AI, let's do our disclosures. I work for The New York Times Company, which is suing OpenAI and Microsoft.

And my boyfriend works for Anthropic.

I love this genre of story that we've started to see about how people in a specific occupation or field are using AI. My intuition is that you could do a version of this story for almost every profession, like how people are actually using this stuff, because people are using this stuff, and they're not always talking about how they're using it. I think it is a valuable service and a valuable journalistic mission to go into various industries and say, "Hey, what are you doing with this stuff?"

Exactly. A lot of our listeners are going out to restaurants, and there is now the possibility that you're eating dishes that have been crafted, or at least assisted in their development, by AI. I thought it might be interesting to hear about how that works.

I wondered why I was seeing glue pizza on so many menus. Let's bring in Pete Wells.

Let's bring him in. Pete Wells, welcome to Hard Fork.

Well, hello.

So how did you get interested in how chefs are using artificial intelligence?

There was this curious silence about it. Nobody was talking about it. I wrote emails to a couple of chefs, and right away one of them wrote back and said, "I use it nonstop. It is my favorite kitchen tool. But strangely, none of the chefs I'm in touch with ever mention it to me."

So he had the impression he was the only one, and of course I knew he couldn't be the only one. I started asking more people and got some pretty interesting responses back.

And tell us about this chef that you contacted who said that AI has become his favorite kitchen tool.

Oh, sure. This is Grant Achatz, whose famous restaurant is Alinea in Chicago. He also has Next, and in the fall he opened a place called Fire. He has been using it for all kinds of things.

When he was opening Fire, the idea was that everything would be cooked in some way on, over, or in flames—coals or embers. So he asked ChatGPT, "Tell me some unusual cooking fuels from around the world, where they're used, and what they're used for." It started giving him back things like avocado pits and corn cobs, some of which he knew about, but a lot of which he didn't.

Of course, it also comes back with cow dung, which is a cooking fuel, but not one that you can charge a lot of money for.

Yeah, I did that once in my house. Let's just say that smell's never coming out of the upholstery.

Yeah.

One of the reasons that this is so interesting, Pete, is that Grant Achatz is one of the most famous chefs in America. Alinea is one of its best restaurants. Have you ever eaten at Alinea, Kevin?

No. Have you?

I have, and it was one of the great meals of my life. I imagine you've eaten there once or twice.

Yeah. I've been writing about Grant since before he opened Alinea. I've been paying attention to what he does for a really long time, and he is probably one of the chefs most focused on creativity and inventiveness of anybody in the world.

Yeah. I think the last dish I had at Alinea was an edible balloon that they had made. Every dish at Alinea contains some sort of surprise, right? This is not like a steakhouse.

Everything there is designed to be original, which is what made it so interesting to me that he is one of the people at the vanguard of using AI in his cooking. I could imagine Grant Achatz saying, “No, no, no, the most precious thing to me is the ideas behind what is in my food. I’m going to make sure that’s always coming from my own brain.” What do you think it is that has made him and some other chefs say, “Oh, no, no, no, this can actually be an excellent creative partner”?

Well, he has always relied on collaboration. When I first met him, before he opened Alinea, he had this crew, this wrecking crew of bright, young, ambitious chefs. Every night after service was finished, around 11:00 or 11:30, when the kitchen was cleaned down, they would sit around a table and talk about ideas: “What do you got? What do you got for—what if we did this? What if we tried that?”

That’s always been a huge part of his process, so he’s already very accustomed to the idea that ideas come from within, but they can also come from outside. They can come from traveling, from some amazing meal that you had in Spain. They can come from a childhood memory of your grandmother. But they could also come from somebody else’s childhood memory. They could also come from your sous chef’s last trip.

It’s not a huge step, at least for him, to start asking the same kinds of questions he would ask his sous chefs and type them into ChatGPT. When your sous chef suggests something, you think about it. It’s not going to go straight on the menu anyway. You’re all going to critique it, discuss it, play around with it, elaborate on it, and refine it before it gets anywhere near the dining room.

I think it made a kind of sense to him: Let’s get this machine that can suggest crazy stuff sometimes, but will also suggest things that might turn out to be good with a tweak here and a change there. Let’s just bring it into the process.

Now, Pete, you mentioned that you’d found, in addition to Grant Achatz, a bunch of other chefs who were using AI tools in the kitchen. What were some of the ways they were using them, and who are these chefs?

There’s Ned Baldwin, who has a restaurant in Manhattan called Houseman. He sent me a whole bunch of transcripts of his conversations about sausage making and different things you can do to affect the texture of sausage—to make it firmer, or, if it’s too firm, to make it a little softer. If you want that springy, elastic feeling that you get in Asian fish balls, which is its own thing, how do you get that?

From there, it was, “Okay, I’ve got my basic recipe, but now I’d like to change the seasonings. What would I do if I were making this in Malaysia? What are the spices I would bring into it?”

In our culture, chefs are often presented as these figures of towering authority who know how to do anything. If you just toss them a few ingredients, they can make something. I think what your story gets at is that these chefs actually have all sorts of questions about how to do things, and if they want to expand their skill set, they don’t always know what the next thing to do is.

But chatbots are proving to be pretty good at that. As you point out, those chatbots are extremely patient. You can ask them a question at any time of the day, and it sounds like the chefs have been pretty satisfied with the answers they’re getting.

More than 1 chef used the phrase “out of the box” to describe what they go to it for. They know what they know, right? They know that zucchini and basil go together, right? There are certain combinations that are just go-tos for chefs over and over. What they can get out of ChatGPT is something they’d never considered before.

For a chef, that’s really valuable because they’re in a high-pressure, high-volume environment. Every night, they’ve got to get the food out to the tables, and it’s a grind. There’s not a lot of time to just sit and think.

Mm.

It’s just not part of the job, and when you’re under pressure, of course you go to what you know: I know this is going to work. I’ve done it before. I’ll just do it again. But if you do that too often, you start to seem like you’re not thinking anymore.

Whatever made you fresh and exciting to the world in the beginning starts to go away if you just rely on your proven formulas.

I think that’s so interesting because one of the knocks on these AI tools is that their outputs can sometimes be generic, because they’re essentially taking the statistical average of whatever data they were trained on. This has always been my hesitation about using AI for cooking.

I’m a home cook. I’m not a very good one, but I do use ChatGPT sometimes to inspire myself, and I’ve found that it often recommends the most generic thing possible. It’s like, “Oh, I have these 5 ingredients. What should I make?” And it’s not all that surprising.

Maybe I’m just not good enough at prompting the models. Maybe the chefs who are spending serious time with these tools have figured out how to get more creative, out-of-the-box ideas out of them.

I think I’ve started playing around with it just since I started reporting this story, and what I’ve found is that the first response you get might be, “Yeah, thanks. I could have done that myself.” But when you go back to it and say, “No, not that. Can you do better?” you start to get interesting responses.

Before we got on, I was thinking about dinner, and I said, “Look, I’ve got a couple of fennel bulbs. I’ve got an onion. I’ve got some garlic, olive oil, a couple of other things, and a can of sardines. What can I do for dinner?” And it said, “Well, pasta with sardines and fennel.”

I said, “I’m not so into the carbs. What can we do that’s lower?” Then it gave me basically the same dish without the pasta. I said, “Hmm, what if we just do a raw fennel?” So I get a fennel salad. I said, “Oh, that looks interesting. Is there a way to take this in a Turkish direction?”

Then it started to get really interesting. It was like, “Yes, you could add some braised zucchini. You could add some fried eggplant.” The more I went back to it, the more it gave me.

It never gets annoyed. It doesn’t say, “Hey, that recipe I gave you was a really good recipe. What’s your problem, dude?” It will just keep batting the ball back at you, which is interesting. The more specific you can be, the more surprising it will be.

Now, Pete, there has been a lot of resistance in creative fields to using AI. A lot of people in animation, art, or music just think, “This is taboo. We shouldn’t be allowing these things into our creative process.” Do you find anything similar among chefs and people who work in restaurants? Is there a cultural resistance to AI?

There’s a huge cultural resistance. Although I think in some of those other fields, like animation in particular, it has a lot to do with the fact that this thing can do their jobs now. It can do their jobs already.

It can’t really do what a chef does, not at the high-end creative level that the restaurants I wrote about are dealing in. It’s not there yet. But there’s still a huge resistance just to the idea—to the romantic idea that all of this stuff comes from the heart of the chef, who’s this heroic figure standing on a mountaintop, and lightning strikes him, and that’s dinner.

That bothers people. People are bothered by what they see as the laziness of it, or just all of the famous kind of crap that AI generates. We don’t want that crap in our food. It’s actually really deep.

I ran into a couple of food writers last night, and one of them walked up to me and said, “I hate that AI piece.”

Hmm.

I think it was not the way I wrote it but the facts in it: “I hate what the reality is that you reported.”

Well, here’s what I would say to that person: What’s lazier, using AI to try to come up with a new dish or putting a flourless chocolate cake on your dessert menu?

Right. Right.

Right.

If what you want is constant invention, reinvention, constant moving forward, those ideas can come from anywhere.

8. The Steak Temperature Test

I have 1 question that’s not actually about AI, but it’s on my mind, and I figured this would be a great time to ask the great Pete Wells about it.

Please.

Okay, Pete, so here’s the story. Last night, I went to [bleep], a wonderful institution. I went there with 3 folks, and we did, of course, order the famous chicken. But there were 4 of us, so we knew we were going to need a little something else.

And so they had a rib eye on the menu, and it cost $90. We said, “Well, that’s a lot of money for a rib eye. But you know what? There’s 4 of us. It’s a special occasion. Let’s go nuts.” We ordered the rib eye, and we ordered it medium-rare because that’s the proper temperature that you should order any piece of meat at, in my opinion. So the rib eye—

No, you’re right. This is factual.

Okay, thank you.

You heard it here.

You heard it here first. So the rib eye arrives, and we get into it, and it’s rare. It is not medium-rare. It is rare. And so, you hate to be those people, and I never feel worse in my life than when I have to ask somebody in the service industry for help in this way.

Right.

But we keep looking at each other. We’re like, “We did pay $90 for this, and it feels like it should come at the temperature we ordered.” So we say, “It’s kind of rare.” “Okay. We’ll put it back on the grill.” So here are my questions for you, Pete. Number 1, there was some controversy at my table because one person was surprised that they just grilled the piece of meat again, because by this point we’d hacked off half of it, and they took away that piece of meat rather than bringing us a new one. So question 1 is, is that the common way that this is addressed? And then number 2, is any sort of extra warranted in this? Is this a situation where the restaurant should throw in a free dessert or do something nice, or is that just a, “Hey, these things happen in the kitchen sometimes. We’re going to grill your damn steak a little warmer. Get over yourselves”? Your thoughts.

Well, yeah, I didn’t understand the first part of the question.

Yeah.

They took some of the steak back but not the entire steak?

Well, yeah, because we’d hacked off a few pieces. We had a few bites.

So he’s asking whether they should have given him and his friends a whole new rib eye.

That’s right.

Ooh. I mean, that’s a lot of money for the restaurant, but it’s a lot of money for you.

Yeah.

It’s typical, especially with something like that, to just take the uneaten part and send it back to the kitchen.

Yeah.

I don’t believe I’ve ever seen a whole new steak come out of the kitchen. I’ve been to a lot of steakhouses.

That’s what I thought, for what it’s worth.

Yeah.

Yeah, I’m on your side.

I’ve been to a lot of steakhouses.

Okay. Yeah.

And then, as far as doing something extra, I mean, no, they’re not obligated to—

Yeah.

They are, I think, kind of obligated to get it right.

Yes.

Right?

Yeah.

This is what we ordered, and it’s not what we ordered. And especially with something like that, did it ruin your night to have—

No.

A couple bites of rare steak, right? So I wouldn’t call that a traumatic experience. When it gets into trauma, like—

Wow, Casey. Pete Wells just said, “You and your friends, get over it.” That’s great. Listen, this is exactly what I wanted. I wanted my expectations to be managed by a professional, and now they’ve been managed.

Right. Where I think you start to need to, if you’re a certain kind of restaurant and you really want to go over and above, is to send something out to the table or do something extra or knock something off the check. That’s where they’ve kind of suffered a little bit.

Yeah.

They waited 30 minutes, right?

Right.

They waited 30 minutes for their steak, or somebody spilled wine on them.

Okay. So in your view, something pretty major has to happen to get into the free-dessert territory.

Yeah, I think you have to become sort of unhappy, right?

Yeah, yeah.

And it doesn’t sound like you got there.

I would say I didn’t feel wronged, but I will say that, not that we needed some grand apology or anything, I was sort of expecting somebody would be like, “Oh, we’re sorry about that. Let us take care of that.” Just a real quick thing. That didn’t really arrive, and so that was just curious to me. I don’t know.

Yeah, that’s definitely a service issue, because you would hope your server would say, “Oh, I’m so sorry it’s not the way you’d like it.”

Yeah.

Even if they don’t admit wrongdoing.

Exactly.

They don’t have to say, “Oh my God. I can’t believe that. Chef, I don’t know what’s going on back there.” But just to say, “I’m sorry it’s not what you were hoping for. Let me take care of that.”

Instead, I felt like we were getting the looks like, “Oh, here we go with these people.” It was like, “Oh, it’s one of these kinds of tables. They’ve got opinions about the meat temperature.”

Yeah.

Yeah.

Now, in my experience, that sort of very fine-tuned sense of hospitality—

Mm-hmm.

Of like, “I’ll apologize even if I don’t think the restaurant’s in the wrong. I’ll apologize because it’ll make them feel better,” that is sort of what separates 95, 96, 97 percent of the restaurants from the really, really good, careful ones.

There you go.

So that’s a rare experience, and you’re lucky if you get it.

Yeah. No, the rare experience I had was that the steak was too rare.

Yeah, you were hoping for more of a medium-rare experience.

Yeah, exactly. Let’s just say it wasn’t well-done.

Pete—

Thank you, Pete.

Thank you so much for coming. Great to have you.

Appreciate it.

Thank you. This has been so much fun.

Yeah.

I’m glad you thought to call me up.

Yeah.

Anytime.

Well, we’d love to have you back.