Can We Trust the Jobs Report?

0:01Stripe is built for whatever comes next, because as your business scales, so does the complexity of fraud. New customers, new markets, new AI-driven attack patterns. Stripe Radar protects your business as it scales with the scale of the Stripe network. The more payments Stripe sees, the smarter its AI gets at detecting those patterns as they emerge. Stripe's network is so vast, it has seen$1.9 trillion in payment volume across millions of global businesses. In fact, when a card is used, there's a 92 % chance the Stripe network has seen it before. All of this means fraudulent patterns across payments, accounts, and customer abuse can be detected, flagged, or blocked easier.

0:44Based on the rules you set up, of course. Blocking bad actors, welcoming good customers, keeping more of your revenue, your revenue. Learn more about fighting fraud with Stripe at stripe.com slash radar.

1:01This episode is sponsored by Smart Travel, a new podcast from NerdWallet. You know that one friend who always finds the best travel deals, picks the right cards, and somehow ends up in first class for the price of coach? Smart Travel is like that friend, but in podcast form. They cover all things travel finance, like the quick math for knowing when to pay with points or just swipe, or which travel cards actually pay you back, and which just make your wallet heavier for no reason. Travel smarter and spend less with help from NerdWallet. Follow Smart Travel wherever you get your podcasts.

1:45If you want to know how passionately Ben Castleman feels about the jobs report, all you need to know is this. I was between jobs for like a month when I left one job and took another. and I covered a jobs report on Twitter from a bus because I couldn't bear to miss one. Ben is the chief economics correspondent for The New York Times. He is also a jobs report nerd. And I say this with total admiration. In a previous life, I also covered the jobs report intensely, so I get it. The reason I called Ben now is that there is a political shadow over the jobs report. A shadow in the shape of one Donald J.

2:32Trump. I've always had a problem with these numbers. Last year, Trump fired the commissioner of the Bureau of Labor Statistics. That's the agency that produces the jobs report. You know what I did? I fired her. And you know what? I did the right thing. He said the numbers were rigged to make him look bad. I believe the numbers were phony just like they were before the election. First of all, that's not true. Second, the pressure on a new BLS commissioner is so intense that the Wall Street Journal called the job radioactive. So I think it is wild that we're using words like radioactive because the Bureau of Labor Statistics is, and I say this with all the love in the world, just the most boring, nerdy agency imaginable.

3:23They count stuff. They count jobs and prices and, you know, the inner workings of the economy. When Brett Matsumoto, who's President Trump's pick to run BLS, had his confirmation hearing on June 10th, he said, it is important for the public to be confident that decisions at the BLS are being driven by science rather than politics. And all I could think was like, buddy, that has been the BLS all along. It's the guy who nominated you who sort of got us into this crazy pants situation. And it makes me wonder, do you think in this environment can Matsumoto convince Americans that the jobs numbers are real?

4:09So, whether he can do it, I think, is an open question. Brett Matsumoto is this, we were talking about nerdy, right? He is a data nerd in the fine tradition of the BLS. I watched his confirmation hearing. He's not a big personality. He's not going to be somebody who's, like, standing in front of cameras and, you know, deftly explaining how they go about their work. Maybe that's okay, right? Maybe we want the kind of nerdy guy who just does the work and we don't hear from a lot. But at the end of the day, the BLS has to figure out how to do two really, really hard things. It has to figure out how to improve its actual data collection at a time when its old methods are showing signs of erosion.

4:58And it's got to do it in an environment where there's tremendous skepticism of everything that the government does. certainly all the numbers that come out of the government. And so they've got to convince us that whatever changes they make are being made in good faith and in an effort to get the real numbers and not in an effort to skew them for some political purpose. And that's a really tall order. Today on the show, the next round of jobs numbers drops on August 7th. Ben says, yes, they will be real and you should trust them. But whether the BLS remains untainted forever, that is a much harder question to answer.

5:38I'm Lizzie O 'Leary, and you're listening to What Next TBD, a show about technology, power, and how the future will be determined. Stick around.

5:58Stripe is built for whatever comes next, because as your business scales, so does the complexity of fraud. New customers, new markets, new AI-driven attack patterns. Stripe Radar protects your business as it scales with the scale of the Stripe network. The more payments Stripe sees, the smarter its AI gets at detecting those patterns as they emerge. Stripe's network is so vast, it has seen$1.9 trillion in payment volume across millions of global businesses. In fact, when a card is used, there's a 92 % chance the Stripe network has seen it before. All of this means fraudulent patterns across payments, accounts, and customer abuse can be detected, flagged, or blocked easier.

6:41Based on the rules you set up, of course. Blocking bad actors, welcoming good customers, keeping more of your revenue, your revenue. Learn more about fighting fraud with Stripe at stripe.com slash radar.

6:59This episode is sponsored by Climate Exchange's 11th Annual Electric Vehicle Raffle. Climate Exchange is a nonprofit that helps states protect communities from pollution and rising energy costs. This year, Climate Exchange is raffling off three cars, including an EV worth up to$120 ,000. It's all part of Climate Exchange's commitment to put more EVs on the road. And one of them could be yours. Tickets for the raffle start at just$100, and Climate Exchange pays the taxes and fees. So check it out. Visit www.carbonraffle.org slash what-next to learn more and win your dream EV.

7:47Ben Casselman: This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One.

8:22For people who just hear, the monthly jobs report is blah, blah, blah, blah, how would you explain what the jobs report actually is? The first thing I think that it's important to understand is just what it is we're trying to measure here. So, you know, you hear every month, employers added 150 ,000 jobs last month, non-farm jobs, we sometimes hear, right? Well, so what are we trying to measure there? There are sort of, you know, 150 million jobs in the country, right? And in one month, it'll go from 150 million to 150 million, 150 ,000, right? We're measuring these sort of very tiny little moves relative to the size of the overall economy.

9:12And we're measuring how many jobs there were in May with some margin of error and the number of jobs we're measuring that we have in June with some margin of error. And then we're trying to compare those two numbers. Right. So what we're doing is really hard. We do it through with two surveys. We survey employers and we ask them, how many people do you have on payroll? and we call up households and we say, are you working? Are you looking for work? And we pull all those numbers together and we come up with this report that everybody focuses on one or two headline numbers, but there are tables upon tables upon tables that break this down by industry, that break it down by age and race and education and they give us wages and hours and we pull all that together and we pour over it.

10:04we and economists and policymakers and the Fed, and we try to say, what does this tell us about the state of the labor market right now? Or at least, what does it tell us about the state of the labor market in the midpoint of last month? One of the things that I think is important, maybe for lay people to understand, and you got at this with describing the two different survey methods, right? Employers versus households. is kind of the different stories that those different surveys tell, because one tells us about trends in industries, and the other tells us kind of about people. Why do you think having two different methods of measuring the workforce is important?

10:46So there are a couple of reasons, right? One is that we want to ask the people who will give us the most accurate answers the pieces of information they're going to know the best, right? So employers know how many people they have on payroll. They know how many hours they paid them for. They know how much they paid them. They know what industry they are in. And of course, you ask one Walmart about how many employees they have, that's a lot more efficient than asking thousands upon thousands upon thousands of Walmart employees, right? So we ask them that, but that only tells us about the people who are working, right?

11:28Walmart can't tell you how many people aren't working at Walmart. They can't tell you how many people, you know, would like to work at Walmart. They can't tell you anything about those people. So if we want to know who's out there looking for work and what they look like, so we end up needing to have both pieces of information that together can give us sort of a fuller picture of what's going on in the labor market. I think an illustrative example might be, I mean, I'm going to pull from the time that I did this regularly during the financial crisis and resulting recession. You had sort of one number from the payroll survey, but then you had all these people who were, as the nerds would say, marginally attached to the labor force, which means that they're in like a job because they can't find anything else.

12:17Or they're not counted because they stopped looking at, you know, there's all these kind of numbers and sub numbers. where you get a really granular picture of what is going on that you might not get if you just kind of said, all right, Walmart, give us all your payrolls. Well, so I think this is a great illustration of the challenge here that, you know, we talk a lot about measurement challenges, right? But this is like a conceptual challenge of what are we trying to measure. So ultimately, right, what we really want to know for the most part is like, does everybody who needs a job have one? So some of that, right, we know pretty clearly, right, I have a job, you have a job, and we know other people who are actively looking, who've gotten laid off and are hunting for a job and putting in job applications, right?

13:12So we have employed people, we have unemployed people, all very neat and tidy. And we know some other people, right, who are retired and are just happily retired and they're done with work. And clearly, we shouldn't think of them as unemployed, right? They're done. We also know toddlers, right? We don't want to count them. They're not unemployed. They're very busily employed doing something else. Destructive. Destructive. Destroying things. But then there's all this gray area. And this is where things get tricky, right? You can work a couple of part-time jobs. You can work as a gig worker or a contractor.

13:52You're mostly retired, but you do some, you know, work on the side. So how do we count? Are you employed? Then we have all this complexity around, well, I'm not looking for work right now, but if the right thing came along, or I've given up looking for work for the time being, but, you know, I'm going to go back to school, and then I'm going to, you know, go out and look again in the future. Do we count those people as unemployed? Do we have rules around all of this in our official statistics? but it's conceptually complex. And these are not things that we can just go to ADP, the big payroll processor, and say, hey, you've got a big count of how many people work for all the employers that you do.

14:34Could you also tell us how many of them, how many people are out there looking for work but can't find a job right now, right? They don't know that answer. The only way we can get that information is by going out and talking to people. I think a casual news consumer can understand right? The unemployment rate, we get that. We understand what that means. When you look at how important the jobs numbers are to economic measurement, I want to unpack that a bit, because I do think this one report ripples across Wall Street, across the Federal Reserve, across big employers in a way that other pieces of economic data don't necessarily.

15:17Why is this so important? I think it's important for a couple of reasons. One is that this report is really timely. We get the jobs report typically on the first Friday of the month. And so it tells us about what happened just a couple weeks earlier. And so if we're trying to gauge in real time what's going on in the economy, we have to wait another week or two before we get the inflation data. We have to wait, you know, potentially months more before we get GDP, the kind of big sum-it-all-up number in the economy. But we get the jobs report pretty quickly, and so it's kind of an early window in.

15:56It's also just so important to everything, right? I mean, on some level, like, what matters more in the economy than whether people have jobs and how much they're getting paid in those jobs? Yeah. That kind of drives, you know, you want to worry about consumer spending. Well, consumer spending is heavily dependent on whether people have jobs and whether they're making any money doing them. You want to worry about inflation. Well, that's going to be heavily influenced by, again, by people's spending, which is driven by their pay. Like everything that happens in the economy is driven by and driving what happens in the job market.

16:35And so it just sort of filters through the entire economy. Let's talk about politicization. I am sure that your inbox is full of ranty emails about how none of this data can be trusted. But I used to get those in the Bush administration. So, like, the idea that the numbers were somehow rigged is not a new issue. But I think we should talk about why that theory exists and then how having the head of the federal government say that is different. Has there ever been any evidence of politicians putting a thumb on the scale to make the numbers appear one way or the other? In this country, in modern times, no.

17:28There have been plenty of international examples of politicians in various ways messing particularly with inflation numbers. Those seem to often be the chosen numbers to futz with. One could imagine why that is. We've seen plenty of examples in this country of politicians choosing their numbers very carefully, right, and interpreting them in all sorts of different ways and sometimes putting out their own analyses that, you know, tell their story, right? But the actual sort of core numbers that come out of the Bureau of Labor Statistics and the other statistical agencies, at least in kind of recent decades, have really been not political in any way.

18:15And I say that with a lot of confidence. So that brings me to the current president. And about a year ago, he fired the head of the BLS, Erica McInturfer. Why? Well, so we got a jobs report came out that was weaker than expected and that included some big negative revisions to prior months. But they said basically the job growth has been weaker over the last couple of months than we expected, than we previously believed. And we were all sitting here writing about that and covering that. And then in the middle of the afternoon, all of a sudden there's this truth, I guess, from the president. announcing that he is firing Erica McIntyre for.

18:59In my opinion, today's jobs numbers were rigged, all caps, in order to make the Republicans and me look bad. Yes, he accused her of rigging the numbers to make them look bad. I mean, we should be super clear here. That is just not true. I mean, I think we can just say outright that that is not true. The revisions that came out last summer were large, but they were not out of line with things that we had seen historically. And there was certainly no evidence anywhere that this was political bias. But yeah, the president was mad that we got these bad numbers that, you know, look bad. And so he fired her.

19:42So let's talk about revisions, because as we've said, the jobs report is sort of a snapshot in time from these two different surveys. But then a month or two months later, we get, oh, whoops, actually, there were an extra 150 ,000 jobs created. Or whoops, our bad, actually, we lost 280 ,000 jobs. Why does that happen? I mean, so I think these revisions have been a really big problem over the last couple years. And I think sometimes people downplay them. They say sort of, oh, this is part of the process. and that it is part of the process, but it's a really big issue because it undermines faith.

20:23And also, if what we're trying to do is understand the economy, then if it turns out that our numbers are wrong, that's a big problem. Right. So there are a couple of reasons. You know, for one, these are surveys. Surveys, as we all know from political polling and everything else, have error. These are much larger surveys and they're conducted differently than your sort of traditional, you know, New York Times Siena poll. but they are, they're still surveys, right? So they have error. When it comes to the jobs number specifically, sometimes employers don't respond or they respond late. And so the responses that come in for that first estimate are based on kind of whatever they have.

21:03And sometimes the employers that don't respond don't respond for reasons that make them not look like all the ones that did respond, right? So for example, if it's a really tough economy right now and businesses are not responding. Well, are they not responding maybe because they're actually literally have gone out of business and we don't know it yet? Maybe they're not responding because they're in the process of a big layoff and like responding to the BLS survey is not top of mind or they're like desperately trying to keep the doors open and again, so they're not responding. And so maybe all the businesses that respond are the ones that are doing pretty well and the ones that are struggling are the ones that don't respond.

21:46And so the BLS kind of has to assume that everyone who didn't respond is basically the same as the ones who did. And then as the numbers start to come in, they're like, oh, things were much worse than we thought. It can go the other way, right? It can be, if it's a boom time, maybe, oh my God, we're hiring like crazy. We don't have a chance to respond, right? And so we actually see this pattern where during good economic times, revisions tend to be upward. and during bad economic times, revisions tend to be downward. But you kind of only know that in retrospect. You don't know that in the moment.

22:20There's one other thing I should note, which is all of these measures are based to some degree on models that sort of assume that the world is continuing to look somewhat similar to the way it has over the last few years. Well, over the last few years, right, we've had COVID. We've had a huge wave and then reversal of the wave of immigration. and all sorts of other big economic changes, now AI, all the rest, that may be changing some of the traditional patterns in the economy. And the BLS models will catch that eventually, but it takes some time. And so for a bunch of different reasons, we may be having some larger revisions now than we have in the past.

23:02After the break, are there better ways to do this?

23:15Ben Casselman: This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade-in value. Advanced, intuitive, and deployed. That's how they sack. That's technology at Capital One. Hey, we're Nick and Jack, hosts of The Best One Yet. Behind every episode, there's research, email, spreadsheets, and our shared Google Doc, the headline hammer.

23:52With Google Workspace with Gemini, Google's AI is built right into the premium versions of tools we already use, like Gmail, Docs, and Sheets. No new tools or extra logins, just personalized AI for the way we work. That means less time on the work, around the work, and more time finding the stories worth talking about. Get AI tools for work. Upgrade to Google Workspace with Gemini at go.gle slash workspace with Gemini. That's go.gle slash workspace with Gemini.

24:21I think someone listening to this could fairly say, wait a minute, this is the biggest economy in the world. Is this really the best we can do? And I wonder what you would say to that. I would say that the numbers on the whole are actually a lot better than I think, maybe even this conversation sometimes makes them sound. We've had these big revisions, but even with these big revisions, the basic story that we told people about the economy last year, before the revisions, was basically accurate, right? We were talking about slowing job growth, but a low unemployment rate and cooling wage growth.

25:03We picked up on all of that accurately, right? If you're trying to measure a really big economy, you've got to focus on the larger trends of what's happening and not, you know, we warn people all the time, I think, right? Don't over-interpret one month of data. Don't get focused on, oh, the teenage unemployment rate, you know, jumped, you know, three-tenths of a point last month, right? Like, just set that aside. Focus on the bigger picture. That said, this is an economy that is changing very rapidly. We barely mentioned AI, right? I know. Which I love that we're having a whole conversation about the labor market that's not about AI, so I'm not going to make it about AI.

25:41But we know that there are these big changes coming in. We don't know what they're going to look like. So we would really like to do a better job of understanding the economy in real time. And we know that our existing methods, one, survey response rates have been falling. Well, that's what I wanted to bring up with you. The San Francisco Fed released some research about survey rates falling. So the response rates to these surveys, if you go back to even the early 2000s, the response rate to the survey of households was well over 90 percent. Wow. It has fallen down to the 60s over the course of, I mean, it kind of, it edged down very, very gradually.

26:27But yeah, I mean, you know, there are a lot of reasons why people are not necessarily, like, eager to turn over a bunch of information to anybody, including to the government. You're also just getting asked to do a million surveys these days, right? I mean, we're all constantly getting text messages asking us to fill out a survey, and so I think there's some fatigue there. But so these survey response rates have fallen, so that makes the data less good on some level, right? Businesses have also gotten less willing to respond to these surveys. The BLS is not necessarily fully kept up with sort of changes in the economy, right?

26:59We don't measure gig work very well. We struggle often to measure immigration flows. We kind of assume that those continue at basically constant rate, and obviously we know over the last few years that's not at all true. There are a bunch of really interesting proposals out there for dealing with this. Yeah, tell me. So, I mean, there are some really interesting proposals, but none of them are a sort of quick, easy solution, is, I think, the sort of headline to look at this. Which, for an enormous economy, seems entirely fair. Sure. Right? This is tough. This stuff is hard. So, I mean, people have probably heard about the numbers from ADP every month, right?

27:41Sometimes those get reported. So this is a big payroll processor, right? You might get your check from ADP or from another big payroll processor. They obviously know how many people work here. surely we should be able to get useful numbers from them. Well, ADP is not going to be able to replace what we get from BLS because they aren't representative of the entire economy, right? A small employer probably isn't going to use ADP. A new startup probably isn't going to use ADP. But maybe there is a way to incorporate their numbers into the official numbers and then supplement that with surveys and dedicate our survey resources to talking to the kinds of businesses that don't use ADP, right?

28:27Why bother reaching out to the Walmarts of the world if we can focus our survey attention on the smaller employers? So there are ideas, similar ideas around collecting inflation data, retail spending data, kind of all of these other areas. But doing all of that is going to require research to make sure that the numbers are consistent over time. It's going to require funding to do that work to probably buy the data from these companies. It may require legislation to require companies to turn over data because what you don't want is to base your entire monthly jobs report on data from a private company who then one morning gets a new CEO and they're like, yeah, I'm kind of done turning over this data.

29:17We don't want to share with you anymore. We don't want to do it anymore. That happened basically last year. ADP has shared data with the Federal Reserve for years. And they, for a period of time, stopped doing it. And that left the Fed not in the dark, but a little more in the dark, right? So we don't want to be in a situation where we're just sort of dependent on some private company to, you know, beneficently hand over data. Well, that's kind of what I was going to ask you. It's like, does this have to be the government because being at the whims of a private company or even, you know, research institution seems risky?

29:57Most economists that I talk to are firmly convinced that we need a, we need official statistics from a federal agency where government statisticians are pulling together the numbers and putting them out. Where the underlying data comes from is a more complicated question, right? That doesn't have to come from surveys that are conducted by government survey takers, right? That could come from private sources. We already use private data in some of our information. If you look at the consumer price index, the car price data there comes from, I think it's Kelley Blue Book. Some of the flight data may come from some of the airline providers.

30:44So there is data that is incorporated into this. But you really do need somebody in the government to be pulling it together, to be making sure that it is comprehensive, and to be making sure that consistent methodologies are being used so that we can all trust this. Because otherwise you end up in this world where I've got my numbers and you've got your numbers and the president has his numbers and nobody can even agree on what the basic numbers are, let alone then, of course, how to interpret all of them. There's another question that I wanted to ask you. I remember moderating a panel of economists right before the 2024 election.

31:23And overall, the economic numbers were looking pretty solid then. But the vibes were off, economically speaking. And I think there is often a disconnect between how people feel about the economy and what their lived experience is and what the numbers show. And I wonder if you think there is room to improve the way we collect jobs data, other data, or if it's a question of communicating that better, or if there will always be some gap between lived experience and official numbers. Well, so we have a 320 million person economy, right? So there's always going to be a range of experiences. I think that's inevitable.

32:14You can have the boomiest boom ever and you'll still have people who are struggling. But I do think that our numbers, and this is where collecting more data is going to be helpful, our numbers do not always make it easy to understand how the world looks for different groups of people. we want to be able to disaggregate these numbers and to be able to say, what does this look like for lower-income people, for middle-income people? What does this look like in cities versus rural areas? What does this look like in, you know, for people with college degrees and people without college degrees or people in different industries?

32:54And, look, we've had a lot of time over the last several years where the economy looks really, really different depending on where you sat. I would love to be able to see, you know, data on consumer spending divided by renters versus homeowners. Right, you think about a divide in the economy over the past, you know, 10 years. You know, what is larger than the one between homeowners sitting with their, you know, cushy fixed-rate mortgages and, you know, anybody who's been paying rent? We can't do that in our data, right, on a regular basis. I would love to, you know, be able to divide on a regular basis between people in AI-exposed occupations and not, right, and understand how things look different.

33:42That's going to require collecting a lot more data. It's going to require being able—so that we can then parse that data in all these different ways. and I think we also do need to do a better job, those of us who are in the business of communicating this data, to communicate what exactly it means, right? When consumer spending is strong, is that because a huge amount is being spent by wealthy people and that's enough to carry the entire economy, but actually everybody else is struggling? There's been a big debate recently, by the way, and I don't think the answer to that is clear, But, like, we should be able to pull these things apart and understand how things look different depending on where you sit.

34:27And we shouldn't be in the business of saying, oh, no, you're just wrong. Like, you're, you know, actually, the economy's great. What are you thinking? I think a lot of times we need to be doing a better job of listening to why it is that people feel like they're struggling. So we started this conversation talking about this idea of BLS commissioner being a radioactive, a politically radioactive job right now. What would it take to have it not be? I'm just thinking about, OK, new BLS commissioner, how could we get back to these are just the numbers? This is what they are. This is done with a very careful, professional process.

35:11because any sliver of doubt, at least to me, means a tiny shadow, a shadow of some size on America's economic standing in the world. I think, you know, this is always the challenge, right, is that once it's difficult to build a reputation, it's easy to destroy it, right? And so I don't think there's any sort of simple solution to restoring faith here. I do think there are a couple of things. I mean, one, I hope that we are on a path to getting the revision situation a bit more under control. And so it would help if we sort of had a period where we didn't get big, big revisions that kind of cause everybody to have questions about the data.

35:56We're going to need a lot of transparency from these agencies. These agencies are, for the most part, very transparent, right? They publish their methodologies. they are very good about taking questions and explaining how things are. I'm sure you've had the experience of being on the phone with folks at BLS and saying, can you explain how this number comes about? And they're great. But I think the more of this that gets published publicly, that gets explained publicly, that when methods are changed, that they are broadcast ahead of time, you know, we're going to be making this change in this way, on this date, and we're going to publish what it would have looked like under the old methodology as well as the new methodology so that you can compare, right?

36:39That all gives economists an opportunity to dig in and understand and in some cases even recalculate and say like, okay, well, I would have preferred these assumptions and that's how it would have looked different. And that over time, I think, builds confidence, right? That whether you agree with every decision or not, you at least believe that they're not being done in a skewed way. But it is going to take time for that. It's not like one day everybody wakes up and they say, OK, I guess everything's fine now.

37:15Ben Castleman, thank you so much for your time. Thank you for having me. Ben Castleman is chief economics correspondent for The New York Times. And that is our show for today. What Next? TBD is produced by Rob Gunther, Evan Campbell, Madeline Thames-Ducharm, and Patrick Fort. Paige Osborne is the senior supervising producer of What Next and What Next TBD. Mia Lobel is the executive producer of podcasts here at Slate. And Ben Richmond is the senior director of podcast operations. I am Lizzie O 'Leary. Thank you so much for listening. Talk to you soon.

38:05Ben Casselman: This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge, and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade-in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One.

39:02GLE slash workspace with Gemini. That's G-O-O dot G-L-E slash workspace with Gemini.