‘A.I.-Washing’ Layoffs? + Why L.L.M.s Can’t Write Well + Tokenmaxxing

We gave Times employees a preview of Crossplay from New York Times Games, and here's what they had to say.

I can finally play with other people.

Kevin RooseHost0:08

I'm pretty competitive.

Kevin RooseHost0:09

It's fun to beat friends and coworkers.

I have a J for 10 points.

Casey NewtonHost0:14

I'm guessing Tenga is not a word.

Casey NewtonHost0:16

Let's see.

Casey NewtonHost0:17

Tenga is a word.

Casey NewtonHost0:18

Oh.

As an English as a second language speaker, I like to learn new words.

Crossplay, the first two-player word game from New York Times Games.

Download it for free today.

Kevin RooseHost0:31

I just read the most heartwarming news this morning that I wanted to share with you, Kevin.

Casey NewtonHost0:35

What's that?

Kevin RooseHost0:36

The UK government has withdrawn a proposal to let AI companies train on copyrighted works after a backlash from artists like Dua Lipa.

Kevin RooseHost0:45

Did you see this?

Casey NewtonHost0:45

No.

Kevin RooseHost0:46

Dua Lipa said, "Don't start now with this AI."

Casey NewtonHost0:51

[laughs]

Kevin RooseHost0:52

My sugar boo, she litigating, Kevin.

Casey NewtonHost0:54

[laughs]

Kevin RooseHost0:54

She's like, she's making some new rules, and she's saying, "We're not gonna train on my copyrighted works."

Casey NewtonHost1:01

Wow.

Kevin RooseHost1:01

And that's why she is a queen.

Kevin RooseHost1:04

And so Dua Lipa, if you're listening, we salute you.

Casey NewtonHost1:07

Yeah.

Casey NewtonHost1:07

Dua Lipa, you're a Dua keeper.

Kevin RooseHost1:10

Yeah.

Kevin RooseHost1:10

[laughs] Period.

Casey NewtonHost1:11

[laughs]

Kevin RooseHost1:11

Dua Lipa said artist rights.

Casey NewtonHost1:14

Wow.

Casey NewtonHost1:15

[upbeat music]

Kevin RooseHost1:20

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

Casey NewtonHost1:22

I'm Casey Newton from Platformer.

Kevin RooseHost1:23

And this is Hard Fork.

Casey NewtonHost1:24

This week, a big wave of tech layoffs is raising the question, has AI job loss truly begun? Then, writer Jasmine Sun is here to help us answer the question, why are chatbots bad at writing? And finally, it's token maxing time.

Casey NewtonHost1:38

Why tech companies are building leaderboards to measure who is spending the most on AI.

Casey NewtonHost1:44

[upbeat music]

Kevin RooseHost1:54

Well, Casey, for years now, we've been monitoring for signs of an AI job apocalypse.

Casey NewtonHost1:59

Yeah, we've been monitoring the situation.

Kevin RooseHost2:01

It's true.

Kevin RooseHost2:02

And over the past few weeks, I think we've gotten some early indications that something is happening in the labor market, especially for tech workers.

Casey NewtonHost2:09

Yeah, we have certainly heard CEOs of companies announcing layoffs invoking AI as a reason that it is happening, and so that has, uh, gotten our attention.

Kevin RooseHost2:17

Yeah, so just a couple examples from the last few weeks.

Kevin RooseHost2:21

Last week, Atlassian announced a 10% reduction in its staff, about 1,600 jobs, that they said were going to help them fund further investment in AI and enterprise sales.

Kevin RooseHost2:33

That came on the heels of a big round of layoffs at Block, the financial tech company formerly known as Square, which said that it was cutting its staff by about 40%, or about 4,000 jobs, uh, saying that they were shifting the way that they were working to use smaller and flatter teams.

Kevin RooseHost2:50

And then the big one that folks are expecting, uh, maybe as soon as this week, is that Meta is reportedly poised to lay off 20% or more of the entire company.

Kevin RooseHost3:01

This was reported by Reuters last Friday, who said that their sources had told them that Meta was preparing to cut as many as 16,000 jobs, the largest layoffs at that company since late '22 or early 2023, when they laid off 20,000 people.

Kevin RooseHost3:19

So as of this recording, that hasn't happened yet that we know of, but I know that people at Meta are very on edge and are awaiting the further news about their jobs.

Casey NewtonHost3:29

Meta, after this story came out, told Reuters that it was, quote, "speculative reporting".

Kevin RooseHost3:34

Which if you're not familiar with the language deployed by Meta Communication staffers, means this is happening, but we don't wanna tell you it's happening yet.

Casey NewtonHost3:42

[laughs] Correct.

Kevin RooseHost3:43

So Casey, I wanna hear what you make of these layoffs, but first we should do our disclosures.

Kevin RooseHost3:47

I work for The New York Times, which is suing OpenAI, Microsoft, and Perplexity.

Casey NewtonHost3:51

And my fiance works at Anthropic.

Kevin RooseHost3:53

So okay, Casey, what do you make of the fact that all these companies are referencing AI in some way as a reason for their layoffs?

Casey NewtonHost4:01

Well, I think it's a little different at each company, Kevin, and I think we can make a decent case for and against the idea that AI is really driving the show at each of them, so maybe we should get into that.

Casey NewtonHost4:14

But at the highest level, I would say companies do continue to tell us now that AI is a significant factor in the reduction of these workforces, and sooner or later, I do think we're going to have to believe them.

Kevin RooseHost4:27

Yeah, I think this is the early warning sign for a lot of people, especially in the tech industry, who are, I think it's fair to say, going to be some of the first people to see their jobs change or disappear because of these new AI tools.

Kevin RooseHost4:42

But let's get into some of the specifics here.

Casey NewtonHost4:44

Yeah.

Kevin RooseHost4:44

So Casey, let's start with Atlassian, the first company I mentioned.

Kevin RooseHost4:48

Their CEO, Mike Cannon-Brookes, said in a company blog post, uh, that the bar for what great looks like for software companies on growth, on profitability, on speed, on value creation, has gone up.

Kevin RooseHost5:00

He said, "We are choosing to adapt thoughtfully, decisively, and quickly to drive durable, profitable growth." He claimed that AI was not replacing people, but he said it would be disingenuous to pretend that AI doesn't change the mix of skills we need or the number of roles required in certain areas.

Casey NewtonHost5:19

Yeah, so I take him at his word.

Casey NewtonHost5:21

It seems like he himself is trying to walk a middle path there, right, and sort of not denying that AI is a factor here, but also not saying, like, this is the only reason this is happening.

Casey NewtonHost5:32

I think some other context that is worth having is that Atlassian is one of the companies that could be part of what we've been calling the SaaSpocalypse around here, right? This is a company that makes tools for businesses.

Casey NewtonHost5:43

A lot of its products are essentially structured workflows, and there are those who believe that sooner or later, you're just going to be able to code your own pretty cheaply.

Casey NewtonHost5:52

Now, maybe you will still choose to buy a product from a company like Atlassian, but maybe you're not gonna be willing to pay nearly as much as you would before.

Casey NewtonHost6:00

And so the company's stock price has just been battered over the past year, and I think that has left them, one, uh, hurting for cash a little bit, but two, and probably more importantly, looking for a different story that they can tell the stock market about what they're doing.

Casey NewtonHost6:13

And so today that story is, "We're gonna get rid of some of these workers, and we're gonna figure out how to make our remaining workers more productive."

Kevin RooseHost6:19

Hmm.

Kevin RooseHost6:19

So there's this term that's been floating around called AI washing-

Casey NewtonHost6:23

Mm-hmm

Kevin RooseHost6:23

... um, which is basically when a company wants to lay a bunch of people off, or maybe they don't feel like they need as many people, and-

Casey NewtonHost6:29

I, I thought it was when a software engineer finally took a shower.

Kevin RooseHost6:32

[laughs] And basically the thesis is, like, these aren't really layoffs about AI.

Kevin RooseHost6:38

This is just sort of a convenient excuse that these companies are using.

Casey NewtonHost6:41

Yeah.

Kevin RooseHost6:41

Do, do you think Atlassian qualifies as AI washing?

Casey NewtonHost6:44

Um, I would like to get a little bit more detail on exactly who they are laying off here, whi- which is a detail that we do have about some of these other companies that helps us answer that question.

Casey NewtonHost6:53

So I don't know exactly how it is happening inside of Atlassian, but I think that their CEO was relatively straightforward a- as these things go in saying, like, "It's a little bit about AI.

Casey NewtonHost7:04

It's not entirely about AI, but, like, yes, keep your eye on AI." So to me, that just reads as honest, and so I'm gonna give them a pass.

Kevin RooseHost7:10

Okay.

Kevin RooseHost7:11

Let's talk about Block.

Kevin RooseHost7:12

Uh, Jack Dorsey, the CEO of Block-Gave an explanation about their layoffs.

Kevin RooseHost7:18

He said, quote, "We're not making this decision because we're in trouble.

Kevin RooseHost7:21

Our business is strong, but something has changed.

Kevin RooseHost7:24

I had two options: cut gradually over months or years as this shift plays out, or be honest about where we are and act on it now.

Kevin RooseHost7:31

I chose the latter." Casey, your take.

Casey NewtonHost7:34

So something to know about me and Jack Dorsey is I have a bit of a bias against him as a former Twitter user who misses that website dearly.

Casey NewtonHost7:43

At this point in 2026, I would not hire Jack Dorsey to run a lemonade stand.

Kevin RooseHost7:46

[laughs]

Casey NewtonHost7:47

Okay? But if you wanna talk about Block specifically, this is a company that tripled its headcount from about 3,800 people in 2019 in what seems like just kind of classic, like, inattention to what was happening in the business during pandemic era boom times, right? And I wonder if you saw this detail, because it truly took me out, Kevin.

Casey NewtonHost8:07

Five months before they announced the layoffs, Block spent $68 million to fly 8,000 people to an in-person event with Jay-Z.

Kevin RooseHost8:15

Come on.

Casey NewtonHost8:16

Yeah.

Casey NewtonHost8:17

[laughs] So that's the kind of famous attention to detail that has turned Jack Dorsey-

Kevin RooseHost8:21

[laughs]

Casey NewtonHost8:21

... into one of the greatest visionaries in tech.

Casey NewtonHost8:24

So look, is this about AI? Again, you know, what does Block really do? They have, uh, those little iPads at the coffee shop.

Kevin RooseHost8:32

Yeah.

Casey NewtonHost8:32

And then they have Cash App, okay?

Kevin RooseHost8:33

Mm-hmm.

Casey NewtonHost8:33

How many people do you really need to run those products? Probably fewer than 10,000.

Kevin RooseHost8:38

Mm.

Casey NewtonHost8:39

Is that about AI? Uh, I don't know.

Casey NewtonHost8:42

Maybe if you squint.

Casey NewtonHost8:43

But again, this is a company whose stock price was cratering.

Casey NewtonHost8:46

They needed a different story to tell the market.

Casey NewtonHost8:49

And I do think you can make a case that AI will make the remaining workers more productive.

Casey NewtonHost8:53

So again, this is another one where it's like you could use AI to justify what's happening, but you also could just say, "This company has been mismanaged for a while now."

Kevin RooseHost9:01

Yeah, you could use AI washing or Jay-Z washing-

Casey NewtonHost9:04

[laughs]

Kevin RooseHost9:04

... which just seems to be what they, what they are doing here.

Casey NewtonHost9:07

Mm-hmm.

Casey NewtonHost9:07

Yes.

Kevin RooseHost9:07

So this did seem to have an effect on their stock price.

Kevin RooseHost9:11

In fact, the day after Jack Dorsey announced the layoffs, Block's stock shot up 17%.

Kevin RooseHost9:17

It's gone down a little bit since then, but they're still up from where they were before these layoffs.

Kevin RooseHost9:22

And I think we should just say, like, this is also a part of the equation here, right? These are companies, largely public ones, that have investors', uh, attention.

Kevin RooseHost9:32

And right now, there's sort of this narrative power around AI where if you seem like a company that is investing heavily in the AI tools and the AI way of working, your investors say, "Oh, that company is really forward-looking.

Kevin RooseHost9:47

They must have a plan for how to navigate this transition." And so I think there's sort of, they're seeing the power in telling the story that all this is related to AI.

Casey NewtonHost9:58

Yeah, which by the way reminds me of, like, the peak of crypto mania when, like, some public traded companies would just add, like, a crypto term to their name-

Kevin RooseHost10:05

Yes.

Kevin RooseHost10:05

[laughs]

Casey NewtonHost10:05

... and their stock price would shoot up by, like, 40,000%.

Kevin RooseHost10:08

Yes.

Casey NewtonHost10:09

It turns out that the public markets actually can just be tricked that easily.

Kevin RooseHost10:12

Yes.

Casey NewtonHost10:13

That would give me some relief if I was a CEO, just knowing that I could fool people like that.

Casey NewtonHost10:16

But anyways...

Kevin RooseHost10:17

So let's talk about the third large tech company that is reportedly conducting layoffs, uh, Meta.

Kevin RooseHost10:23

We don't know exactly who or what teams are being affected by these layoffs, but, uh, this is a significant part of their workforce, and they seem to be saying in their communications with the public what all of these other companies are saying, which is, "We are going all in on the new way of working, and we are going to have to make some cuts to make that work."

Casey NewtonHost10:44

Yeah, on a recent earnings call, Mark Zuckerberg said that, quote, "Projects that used to require big teams now can be accomplished by a single, very talented person." And, uh, and we should also say that this cut is coming alongside this massive AI infrastructure investment, right? They're gonna spend $135 billion on capital expenditures this year.

Casey NewtonHost11:03

And even for a company of Meta's size, like, that is real money, right? So I know that they're trying to be careful, again, trying to not spook the stock markets too much.

Casey NewtonHost11:13

This is obviously the biggest bet in the company's history, and I think that making sub- substantial cuts are gonna signal to the market, like, "Hey, don't worry.

Casey NewtonHost11:20

We're not, like, completely losing our minds here.

Casey NewtonHost11:22

Like, we're going to keep some of these expenses under control."

Kevin RooseHost11:25

Yeah, I think that's a really important point, because what we're seeing here at some of these companies is that they are not actually sort of cutting costs in the aggregate by using these tools.

Kevin RooseHost11:36

They are just shifting the cost from human labor to AI.

Casey NewtonHost11:40

Right.

Kevin RooseHost11:41

They are plowing this money that they are going to save by laying off these thousands of people into the building of data centers and other AI infrastructure.

Kevin RooseHost11:50

And basically, the bet they're making is these new AI workers are going to be faster, more efficient, maybe cheaper in the long run, maybe not, um, but they are going to be able to do the work that used to require many thousands of people.

Kevin RooseHost12:03

And that is a profound shift in the way that companies are talking about their workers.

Kevin RooseHost12:10

I recently talked to a venture capitalist who said that a lot of the AI startups that he sees, the most AI native companies, are spending more on AI tools than they are on payroll.

Casey NewtonHost12:20

[laughs]

Kevin RooseHost12:21

Um, and that may be an outlier, but I think that is sort of where these companies believe that we are headed, where the majority of your expenses will not go to paying the salaries of human workers.

Kevin RooseHost12:31

It will go toward buying the AI tools and the tokens that your company runs on.

Casey NewtonHost12:36

Yes, I think that's absolutely the, the bet that they're making.

Casey NewtonHost12:38

I also just think it is worth noting that this is still purely mostly speculative, right? Like, in the case of Meta specifically, this is a company that has arguably been struggling when it comes to AI.

Casey NewtonHost12:50

They had to abandon their last model, Behemoth, because it wasn't very good.

Casey NewtonHost12:55

The Times reported last week that it's delaying the release of its latest model, Avocado, because it hasn't been hitting its performance targets.

Casey NewtonHost13:04

It's apparently barely outperformed Gemini 2.5. What is this, last March?

Kevin RooseHost13:09

[laughs] Yeah, that model is really the pits.

Casey NewtonHost13:12

So-

Kevin RooseHost13:12

That's an Avocado joke.

Casey NewtonHost13:13

[laughs] That's very good.

Casey NewtonHost13:15

Thank you.

Casey NewtonHost13:16

So again, this is not as simple as saying they're able to cut 20% of their workforce because they've just made these massive gains.

Casey NewtonHost13:24

I'm sure there are individuals there who have made massive gains.

Casey NewtonHost13:27

But as a company, it still seems like it is somewhat mired in dysfunction.

Casey NewtonHost13:30

They just did yet another partial reorg of their AI teams.

Casey NewtonHost13:35

Uh, and, and that just always sort of makes me, uh, raise my eyebrows.

Kevin RooseHost13:37

Yeah.

Kevin RooseHost13:38

I, I will say, like, one thing that's been surprising to me about this recent round of layoffs is that the companies that are making them are not the ones on the frontier, right? It is not the OpenAIs, the Anthropics, the Googles.

Kevin RooseHost13:53

Those companies are not laying off people en masse because of these AI tools which they are building and presumably have even better models than the ones they're releasing to the public.

Kevin RooseHost14:04

So you have to think that part of this is just companies that are sort of lagging behind their competition saying, "Well, maybe if we just use a bunch of AI, it'll help us catch up."

Casey NewtonHost14:13

Yes, but also, like, OpenAI and Anthropic are much smaller companies than some of the ones that we've been talking about today, at least in number of workers, right? Like, I think it, it is interesting to think that Atlassian is, like, bigger than OpenAI in terms of the number of people who work there when you look at the, you know, relative, like, value of what, what they're generating.

Kevin RooseHost14:33

DocuSign has 7,000 employees.

Casey NewtonHost14:35

There's no funnier sentence that is true-

Kevin RooseHost14:37

[laughs]

Casey NewtonHost14:37

...

Casey NewtonHost14:37

in all of tech journalism.

Casey NewtonHost14:39

As somebody who, who has a paid subscription for DocuSign that I, that I truly resent paying for-Get to work over there, people

Kevin RooseHost14:47

Or get not to work.

Casey NewtonHost14:48

Get not to work.

Casey NewtonHost14:48

H- here's another question that I would ask, Kevin.

Casey NewtonHost14:51

Okay, so we're seeing a bunch of layoffs.

Casey NewtonHost14:52

Like, are these AI related or not? Does it actually matter if the effect on workers is the same, right? Like, you know, if you're the worker, like, whether it's about AI or not, you're still out of a job.

Kevin RooseHost15:03

Yeah, and it's not clear to me what workers can or should be doing to sort of protect themselves against these layoffs.

Kevin RooseHost15:11

One, uh, person I talked to said, you know, they're, um, they, they work at one of these big tech companies and they're like, "Well, there's just a lot of jostling and fear and anxiety right now." People don't know if they should be, like, using the AI tools a ton because then it shows that they're, like, getting with the, the program or whether that just means that they're proving that their work can be automated.

Kevin RooseHost15:31

Like, I think there's a lot of fear and suspicion and mistrust inside these companies right now.

Kevin RooseHost15:37

And, and for good reason.

Kevin RooseHost15:38

Their executives are planning to lay them off.

Casey NewtonHost15:40

Yes, and by the way, I think at, at least some of these companies, that is maybe not an explicit reason for these layoffs, but some of the executives there would see that as a, a positive byproduct, right? Because, you know, if you're like Mark Zuckerberg, you lived through the 2020 era, you had these restive employees that, like, wanted a lot of things from you and they wanted to have a lot of control over what the company could and could not do and how it did it.

Casey NewtonHost16:05

And, you know, I just know that executives over there really resented that sort of thing.

Casey NewtonHost16:09

And once, uh, Meta entered this new era of massive layoffs, employees over there did get really scared for all of the reasons that you would assume.

Casey NewtonHost16:16

They were like, "Oh, God, like, you know, maybe I actually am gonna lose my job." And all of a sudden they got a lot more quiet and you started to see a lot fewer protests over there.

Casey NewtonHost16:23

So I'm not gonna say that, like, these occasional mass layoffs are a way of, like, keeping the workforce in line, but I have noticed that it seems to be having that effect.

Kevin RooseHost16:32

Totally.

Kevin RooseHost16:32

And it makes me wonder whether something that I predicted was going to happen, you know, a year or two ago that did not happen, which is the sort of sudden and mass unionization of workers at these companies may actually start to happen in the next year or two.

Kevin RooseHost16:47

I think one major difference between what's happening now at these tech companies and what has been happening for decades at manufacturing companies, car companies, you know, factory workers, is that those workers were by and large unionized.

Kevin RooseHost17:00

And so when the employers said, "Hey, we're gonna lay a bunch of you off," they were able to negotiate.

Kevin RooseHost17:06

They were able to say, "Hey, maybe instead of laying us all off, maybe you could find other jobs for us.

Kevin RooseHost17:11

If our jobs are being automated, maybe we should be allowed to sort of retrain to do something else." And that was largely successful.

Kevin RooseHost17:18

There were still layoffs, of course, but not the number that we're seeing today at these tech companies.

Kevin RooseHost17:22

So do you think there's any possibility of that or is that just sort of a union fever dream?

Casey NewtonHost17:27

Here's what I will say.

Casey NewtonHost17:28

I cannot think of anything that would make Mark Zuckerberg more mad than a union of software engineers at Meta.

Kevin RooseHost17:35

[laughs]

Casey NewtonHost17:35

And I think the software engineers at Meta should use that information how they will.

Kevin RooseHost17:38

You think that would make him more mad than getting booed at a UFC fight?

Casey NewtonHost17:41

[laughs] Abso- absolute-

Kevin RooseHost17:44

[laughs]

Casey NewtonHost17:44

I think that probably just made him really sad.

Kevin RooseHost17:47

[laughs] Well, there you have it.

Kevin RooseHost17:49

If you wanna make Mark Zuckerberg mad, Meta employees, sign your union card.

Casey NewtonHost17:56

When we come back, why aren't chatbots as good at writing as I am?

Kevin RooseHost18:00

We'll ask Jasmine Sung.

Casey NewtonHost18:01

[laughs]

Kevin RooseHost18:06

[upbeat music]

Casey NewtonHost18:11

This is A.G. Sulzberger.

Casey NewtonHost18:12

I'm the publisher of The New York Times.

Casey NewtonHost18:14

I oversee our news operations and our business, but I'm also a former reporter who has watched with a lot of alarm as our profession has shrunk and shrunk in recent years.

Casey NewtonHost18:25

Normally in these ads, we talk about the importance of subscribing to The Times.

Casey NewtonHost18:30

I'm here today with a different message.

Casey NewtonHost18:33

I'm encouraging you to support any news organization that's dedicated to original reporting.

Casey NewtonHost18:39

If that's your local newspaper, terrific.

Casey NewtonHost18:41

Local newspapers in particular need your support.

Casey NewtonHost18:44

If that's another national newspaper, that's great, too.

Casey NewtonHost18:47

And if it's The New York Times, we'll use that money to send reporters out to find the facts and context that you'll never get from AI.

Casey NewtonHost18:56

That's it.

Casey NewtonHost18:57

Not asking you to click on any link, just subscribe to a real news organization with real journalists doing firsthand fact-based reporting.

Casey NewtonHost19:06

And if you already do, thank you.

Kevin RooseHost19:12

Well, Casey, over the last couple of years we've talked on this show about how AI models are getting better at so many things.

Kevin RooseHost19:20

They are getting better at coding, at competition math, at solving novel physics problems.

Casey NewtonHost19:26

Mass domestic surveillance-

Kevin RooseHost19:27

[laughs]

Casey NewtonHost19:27

... autonomous weapons.

Kevin RooseHost19:29

Yes.

Kevin RooseHost19:29

[laughs] And I think the story of the last few years in AI has been one of sort of rapid, steady progress, but these systems are still sort of jagged and they have flaws and weaknesses, and one place where they arguably haven't improved that much is in writing.

Casey NewtonHost19:48

Now, that's our domain.

Kevin RooseHost19:49

Yes.

Kevin RooseHost19:50

At least that is the argument that Jasmine Sung made in The Atlantic this week.

Kevin RooseHost19:54

She is a freelance journalist.

Kevin RooseHost19:56

Her piece was called The Human Skill That Eludes AI, and it's her attempt to understand why despite so much progress in all these different areas the models of today don't seem to be writing anything particularly good or compelling.

Casey NewtonHost20:12

Yeah.

Casey NewtonHost20:12

And while I think the question of are LLMs good at writing is highly subjective and dependent on the use case, I do think Jasmine makes a really interesting technical case for why these models write the way they do.

Kevin RooseHost20:28

Yes.

Kevin RooseHost20:29

And we should say before we bring her in, Jasmine is a friend of mine.

Kevin RooseHost20:32

She has also been my researcher on the upcoming book that I'm working on, and I just think she's, like, one of the best people writing about AI today.

Kevin RooseHost20:40

She writes on her Substack, which is called Jasmine News.

Kevin RooseHost20:44

It's J-A-S-M-I dot news, and you can read much more of her writing there.

Casey NewtonHost20:50

All right.

Casey NewtonHost20:51

I'll allow it, but I do wanna balance it out.

Casey NewtonHost20:53

By next week, bring on one of your enemies.

Kevin RooseHost20:54

[laughs] Okay, let's bring her in.Jasmine Sung, welcome to Hard Fork.

Thanks for having me.

I'm excited.

Casey NewtonHost21:05

Hi, Jasmine.

Kevin RooseHost21:06

So you wrote this great piece in The Atlantic this week about the human skill that eludes AI, and I wanna start by challenging the subtitle of your piece.

Mm-hmm.

Kevin RooseHost21:16

Why can't language models write well?

[laughs]

Kevin RooseHost21:19

Can't language models write well?

So I do say in the piece that most writing, period, is very bad, and so I think that language models are definitely better at writing and language than most humans are.

But the question that I was really curious about is why can't they write at a sort of literary, creative fiction level? Because the thing is, if you listen to these AI leaders talk about their aspirations, they say, "We're gonna cure cancer.

We're gonna solve physics.

We're gonna build a superhuman coder." They are not shy about, "Oh, our AI models are gonna be better than 75% of human coders." They're saying, "No, we will literally build a self-replicating factory tomorrow." And then Tyler Cowen asked Sam Altman in an interview from last October, "When do you think GPT will be able to write a Neruda poem?" And Sam Altman says, "Maybe in the future, ChatGPT will be able to write, quote, 'A real poet's okay poem.'" So that was the thing that fascinated me is even these guys who are more bullish than anybody else about the capabilities of their technology, they're very reserved about how much literary writing their models can do.

Kevin RooseHost22:24

Mm.

And so that was the gap that I was really interested in.

Kevin RooseHost22:26

Hmm.

Casey NewtonHost22:27

Hmm.

Kevin RooseHost22:27

And you start your piece with this interesting provocation, which is that in some ways-

[laughs]

Kevin RooseHost22:32

...

Kevin RooseHost22:32

GPT-2 was the peak of AI-

[laughs]

Kevin RooseHost22:35

...

Kevin RooseHost22:35

when it comes to creative writing.

Kevin RooseHost22:37

So explain that.

Part of what got me interested in this piece was, um, I was actually doing research for your book, uh, and I was going through all of these previous generations of models and reading the outputs.

And the thing that really shocked me is that, like, in a way, the writing style of GPT-2 and GPT-3 I found so much more compelling than ChatGPT today.

It doesn't have any of the annoying tics.

It doesn't have the em dashes, the tripartite lists, the it's not this but that.

The tone was much more variable.

Like, it would actually surprise you.

It would be funny.

It would be poetic.

And that shocked me to sort of like go back a few generations and realize that maybe, you know, they were also lying all the time and all sorts of other things, but from a writing style perspective, I kind of preferred it, and I wanted to investigate that.

Kevin RooseHost23:18

They were weird.

Casey NewtonHost23:19

That shocks me.

Casey NewtonHost23:20

What...

Casey NewtonHost23:20

To me, talking to GPT-2 was like talking to somebody who had just fallen down the stairs.

[laughs]

Casey NewtonHost23:24

You know what I mean?

Kevin RooseHost23:25

[laughs]

Casey NewtonHost23:25

When it was like, "Do...

Casey NewtonHost23:26

I think...

Casey NewtonHost23:26

Do we need to get you to the hospital?"

Kevin RooseHost23:27

[laughs]

Casey NewtonHost23:28

"Do you smell toast?"

[laughs]

Kevin RooseHost23:29

Yeah, you...

Kevin RooseHost23:30

There's- There are these amazing prompts from this, like, early OpenAI prompt library where they would say, like, you know, "I just won $175,000 in Las Vegas.

Kevin RooseHost23:38

What do I need to know about taxes?" And GPT-2 would say, like, start just writing some short story about, like, an orphanage and, like [laughs]

[laughs]

Kevin RooseHost23:46

It was, like, complete-

But again, they were surprising.

Kevin RooseHost23:48

Yes.

They were, like, nutty.

They were, they were, they were weird.

They would absolutely be a terrible corporate assistant, horrible, like, coding intern.

It can't do any of the things that modern LLMs can do that I'm very grateful for, but, like, from a pure writing style perspective, they were very good.

So GPT-3 in particular, like, they, um, there's...

I found this, like, set of samples that some guy did where it was like, oh, write in the style of Paul Graham, write in the style of, uh, Richard Dawkins, whatever, and it could style match much better than modern LLMs can.

And particularly because so much of sort of literary writing comes from voice and style, that was one of the things I was really interested in was, like, what did we lose that the LLMs can no longer emulate Paul Graham's style or whoever's style? Because I would put in the same exact prompt that this guy gave GPT-3, put it into ChatGPT 5.4 Thinking or whatever, and it would be God awful.

Kevin RooseHost24:36

Hmm.

And I was like, that's really weird.

Kevin RooseHost24:37

So tell us about what you learned about what happened after the GPT-2 and 3 era that changed the way that these models respond to us.

Yeah, I mean, I think the, the answer is post-training, basically.

So they started adding a post-training layer, which is basically saying we have these, like, crazy unpredictable, like, nutjob, concussed models-

Kevin RooseHost24:55

[laughs]

... and, um, they need to learn how to behave because a model that can't behave is a very bad corporate assistant.

And so the AI researchers give them example dialogues and scripts to learn from.

They give them words that they can and can't say.

Uh, they do RLHF, which is a process by which human graders will rate, like, which response is the most helpful sounding or something like this.

And so now these post-trained models have been trapped in a way or trained or guided towards a very particular character or persona that is a very helpful assistant but might be very bad at writing in creative and surprising ways.

Casey NewtonHost25:32

Hmm.

Kevin RooseHost25:33

I mean, the way that you described it was that there is a, a, a phase within the post-training phase where these AI models are evaluated by humans.

Mm-hmm.

Kevin RooseHost25:43

And that's part of what they call RLHF for reinforcement learning from human feedback, and w- what struck me in your reporting is that you actually talked to some people who have done this kind of feedback, uh, to the models who say that they're, they're just being asked to grade things that, in ways that don't make sense.

Yeah.

Kevin RooseHost26:01

Right? Tell us about that.

Yeah, I mean, this is super interesting 'cause, like, th- these job listings you'll see on, like, places like Mercor or xAI, Elon's company, will list them directly.

It'll be like, "Creative writing expert, $45 an hour, must be a New York Times bestseller, and have, like, a Starred Kirkus Review" or something like this.

[laughs]

Casey NewtonHost26:16

Have you ever gotten a Starred Kirkus Review, Roose?

Kevin RooseHost26:18

I think so.

Casey NewtonHost26:19

Okay, good job.

Kevin RooseHost26:19

Not sure.

Casey NewtonHost26:20

All right.

You might qualify to help Elon-

Casey NewtonHost26:22

There you go

... to help Annie from Grok write a little bit better.

Casey NewtonHost26:25

Yeah.

Casey NewtonHost26:25

We're gonna get on that job listing.

Casey NewtonHost26:27

But okay, you were saying.

[laughs] Um, yeah, so anyway, so these companies, because they realize that these AI researchers are really good at knowing, like, what good coding is, but they don't actually know what good writing is, so they're like, "Why don't we hire some humans to find out?" And so they'll commission, like, MFAs and published authors and sometimes just, like, random guys with a blog or whatever.

And one of the people I talked to who was a contractor for Scale AI as a writing evaluator, and, um, he was doing this for one of the bigger labs, he said that the rubric just didn't make any sense.

He would be told things like, uh, you have to grade them based on the number of exclamation marks that there are.

And so if something has three exclamation marks, that's too many, and so you have to ding that one.

Casey NewtonHost27:06

Yeah, and I have to say, generally not bad writing advice.

[laughs]

Kevin RooseHost27:09

Yeah.

Casey NewtonHost27:09

I mean, I guess it depends on the length of the text, but three feels like a lot for many scenarios.

I mean, this is what they tell women in business communications-

Kevin RooseHost27:16

Mm

... is, like, take all those exclamation marks, replace them with periods.

[laughs] Like, we are just gonna remove all of the i's and points.

Casey NewtonHost27:21

We teach women to shrink themselves.

Exactly.

Casey NewtonHost27:23

Yeah.

Um, but yeah, so he was, he was sort of, like, being asked to grade these things, or another one was he got a bunch of fan fictions and he was supposed to grade them on their factuality.Uh, since that was one of the criteria.

Um, I do imagine that one could, you know, devise better rubrics than this particular evaluator was given.

But I think it does show at least that, uh, some of these like very big companies that are very well-resourced simply do not know how to think about what good writing is.

Casey NewtonHost27:47

Briefly, like I, like I, I wanna underline that, 'cause to me that seems like the whole story.

Casey NewtonHost27:51

We are taking the entire internet, and we are grading it on factuality.

[laughs]

Casey NewtonHost27:55

And like so the, the LLM that you're gonna get out of that is just probably not gonna be all that creative.

Kevin RooseHost27:59

Well, and I, I wonder how much of it is related to this sort of verifiable reward-

Mm-hmm

Kevin RooseHost28:04

... system that a lot of these companies are using where you, you have a system generate a bunch of code, and then you have another evaluator model check the code to see whether it's good or not.

Kevin RooseHost28:14

And that works in domains like programming where the code either runs or it doesn't, but creative writing doesn't work that way.

Kevin RooseHost28:21

You can't have an evaluator tell you, um, you know, with any sort of consistency whether something is good or not.

Kevin RooseHost28:27

And so it may just come down to preference.

Kevin RooseHost28:29

And so I guess I'm curious like, do you see this as a technical problem that the labs are frustrated trying to solve, or is this just demand-related? Is this just what people want chatbots to sound like, and in every test where they pit different models against one another, the one that sounds like a bland corporate assistant wins, and so they go with that?

I think both are true.

Um, it's, it's like the majority of writing that we are asking the models to do is, "Write this email for me," right? And like they excel at that.

They are truly great corporate email writers.

They are much better at the whole like passive-aggressive thing than I am.

At the same time, I do think, like you said, there is a technical challenge that has to do largely with verifiability, is like there are people who have spent decades of their lives attempting to articulate what makes Shakespeare Shakespeare, or what makes a Neruda poem a Neruda poem, and they will still not know in any kind of certain way.

They will still get into debates with their fellow academics and literary critics, um, about which writer is better than the other.

And because these things are subjective, because they are ineffable, because they are hard to put in a rubric, and that is the nature of art.

Casey NewtonHost29:33

And to that point, you know, y- you started, uh, this segment by talking about Sam Altman saying like, "Hey, you know, we just basically can't write a great poem yet." Sam Altman, a year ago, said the company had trained a good creative writing model and posted a short story-

[laughs]

Casey NewtonHost29:48

... on X.

Casey NewtonHost29:49

Many people found it compelling.

Casey NewtonHost29:52

Is Sam Altman just not being consistently candid with us, Jasmine?

Ooh.

[laughs] Uh, wouldn't be the first time.

But that short story, if you remember, I'm sure you guys recall, had some great lines, like talking about the seams of mirrors or Thursday, the...

What was it?

Kevin RooseHost30:06

It was like the liminal almost Friday or something.

Kevin RooseHost30:08

[laughs]

Yeah, the liminal day that comes almost Friday.

Casey NewtonHost30:10

[laughs]

Kevin RooseHost30:10

Wait, I, I have to actually look this one up-

It's so good

Kevin RooseHost30:12

... 'cause it was so good.

Kevin RooseHost30:13

Um-

While you're looking it up, like, you know, when, the, the thing about AI writing is like it comes up with all of these fun metaphors, and they are like kind of surprising sometimes with the metaphors.

But also the language is not grounded in the life, and that was my other thing is aside from the verifiability, fundamentally when I think about the writers who I really love, when I think about whether it's journalists or poets or whatever, like they are writing from life, right? Like a journalist goes out and talks to people, and they like see stuff and observe like the color of the sky in a particular way.

Or, um, like a poet is thinking about personal experiences that they've had.

Their, their writing has stakes.

It comes from an emotional place.

And the fact that like LLMs, while being very talented, grammatically pristine, whatever, they don't have lives.

That means that all of the metaphors they choose, all of the words they choose, the examples they choose, they're just ungrounded, right? Like it's not coming from a point of view or a particular experience or a particular community that makes the writing believable.

I think part of what voice and style is, is that it is very specific to the life that a person has had, and LLMs cannot get there in the same way a human who hasn't really lived that life like cannot get there.

Casey NewtonHost31:15

I don't know.

Casey NewtonHost31:16

I feel like it's-

[laughs]

Casey NewtonHost31:17

... case dependent, you know?

Okay.

Casey NewtonHost31:18

I'm a big music fan.

Yeah.

Casey NewtonHost31:20

And, um, over the past few months, I have enjoyed putting questions about music, and in particular, the sounds of certain bands to an LLM, which sounds like a joke prompt because an LLM has never heard anything, right?

Yeah.

Casey NewtonHost31:36

And yet I find that, in general, the models can have good conversations with me about the sound of music.

Casey NewtonHost31:42

Now, it may be that they are just pattern matching based on a bunch of public writing on the internet by people who do have ears-

[laughs]

Casey NewtonHost31:48

... and have heard, right? Like, I'm very open to that.

Yeah.

Casey NewtonHost31:51

But I, I, I, again, I, I have just sort of been struck about the way that it is able to like, um, sort of write about sensory topics in an evocative way that to, at least to me, like surpasses what I would predict they would be able to do.

Kevin RooseHost32:05

Yeah.

Kevin RooseHost32:06

I, I wanna pose a couple objections that I think-

Okay

Kevin RooseHost32:08

... someone might make to-

Perfect

Kevin RooseHost32:09

... your article.

Kevin RooseHost32:10

One of them is, this is cope.

Kevin RooseHost32:12

This is Jasmine, a writer, a very talented writer, sort of finding the things that AI, in her view, is not good at yet and saying, "This is categorical proof that they will, it will be very hard for AI to do these things." This is the same reaction that software engineers had when models started getting really good at code.

Kevin RooseHost32:29

They would say, "Oh, well, it can't do these other 10 things that I do," and that basically just wait a few years, and the models will be better than all of us at everything, including writing.

Um, I would love for it to be cope because I try to automate myself away all the time.

I have no sort of deep attachment [laughs] having to like...

Like, I like writing, but like I've tried over and over and over for the past three years to automate my own job away and to get Claude to do my job for me.

It cannot do it.

This is very frustrating.

Kevin RooseHost32:54

Mm-hmm.

It's not out of a lack of trying, you know? Um, and again, I'm going back to the CEOs themselves and the things that they themselves are saying, right? Like, it's not just me, a writer, it's Sam Altman saying, "This thing will cure cancer and solve physics, but it will not write better than a real poet's okay poem." And so like I think like that suggests that there is something that is at least perceived as a little bit different.

I think it's very possible the models will get much better at writing over the next few years.

I don't think it's like a never thing.

I do think that, you know, like reporting is hard to replicate.

I think that like having life experiences that are real and verifiable is hard to replicate.

I think the style stuff can be improved, especially if you fine-tune the models.

But I think what's also interesting to me about this piece is that-It shows how the, um, the market incentives, the demand incentives of these companies do shape what we see, uh, their abilities are today.

Mm-hmm.

Kevin RooseHost33:46

The other objection I, I'm imagining people might have, uh, who are very AI pilled-

Mm

Kevin RooseHost33:51

... is that this is all in the eye of the beholder, right?

Mm-hmm.

Kevin RooseHost33:54

There have been several studies now that have shown that if you give people a blind taste test of AI writing versus human writing, they prefer the AI writing until you tell them that it's AI writing, and then the value in their eyes plummets.

Right.

Kevin RooseHost34:07

I did one of these in a New York Times quiz just recently.

Kevin RooseHost34:11

So is it possible that the models have already become superhuman at writing, but that the minute we learn that they are AI models generating text and not humans writing words with their fingers, ah, we lose all interest in it, just because of the source, not because of the quality of the writing?

I mean, I think it's definitely interesting and true that people don't want to like AI writing, and that is part of what bothers them when they s- see AI text that is obviously AI, even though, like you said, like, in these quizzes and tests, AI can outperform, uh, human writers in those narrow scenarios.

I mean, my quibble with a lot [laughs] of these quizzes and tests is that, like, as a writer, and you guys are writers too, how much of your job is actually text generation? I think AI is a superhuman text generator, right?

Kevin RooseHost34:59

Mm-hmm.

My job, I am generating text probably 25% of the hours in my day.

I spend a lot of time interviewing people.

I spend a lot of time coming up with ideas.

I spend a lot of time reading, and not just reading indiscriminately, but, like, reading very particular sources that feel like the right ones.

And so, like, you know, usually at the point that you are doing one of these tests, you're saying, like, "Generate, like, one paragraph very specifically about, like, why Trump won the 2016 election, 500 words or less." And, like, you've already given the prompt, which I think is a critical part of writing, is, like, what are you gonna write about? You've often, like, supplied some of the evidence and the guidance in the form of it, saying, like, 500 words or less.

And at that point, I do think that AI is probably a better text generator than almost all humans are.

But again, when I think about it, you know, AI is still very bad at coming up with ideas for articles.

It is still very bad at reporting.

The non-text generation parts of the role feel, um, further away from automation.

Again, like, I'm not a never say...

Like, like, I'm, I'm sort of, like, never say never.

Like, maybe it'll get there.

Um, I would be totally happy if, you know, Claude was able to give me good ideas for my next essays, but it's not there yet.

Kevin RooseHost36:03

Well, we're, we're already s- seeing the LLMs make huge progress in genre fiction, right?

Mm-hmm.

Kevin RooseHost36:07

So, like, recently on the show, we talked to the author of a story in The Times about how authors of romance novels are now able to generate dozens of novels a year using, uh, LLMs.

Kevin RooseHost36:20

In fact, much of the discussion that we had was around how you just have to prompt them differently and sort of relentlessly in order to get what you want.

Kevin RooseHost36:28

You know, your piece, Jasmine, it made me wonder, like, how much of getting a model to just write weird can be achieved by repeatedly telling it-

[laughs]

Kevin RooseHost36:36

...

Kevin RooseHost36:36

in different ways, "Hey, be a little weirder."

Some of it, but not all of it.

I mean, so I talked to, for example, James Yu, who is the co-founder of Sudowrite, which is one of the earliest creative fiction AI writing assistants.

I talked to some other folks who similarly were in the fiction writing LLM space.

And like you said, to an extent, a lot of writers are already using these a lot, are already leaning on LLMs to generate large amounts of text, and it can be very successful, and it can meet readers' needs and whatever.

Um, but, like, even these people who I was talking to, they were describing to me how freaking hard it is to undo all of the post-training that the labs have done.

Um, so they are, like, applying immense amounts of engineering effort that clearly when- in, in my conversations with them, clearly frustrates them, that it is so hard to get these models to stop being so chirpy, so sycophantic, so PG-13 and everything, uh, in, in order to get them to this sort of, like, base model state where they're able to be weird again.

Um, so I think it's certainly possible, but I think the labs have made it quite challenging just because of the way that these models are trained.

The other thing that I think is important is I tend to think that writing and a lot of creative work is actually, like, the perfect use case for these centaur models, right? Like, the idea that the human plus AI collaboration is where you can get the furthest.

And when I listen to the interviews that you guys did about the fiction authors, I was thinking, "This is a centaur model," right? Because without the human prompting and bullying the AI into getting weird and getting sensual and whatever, like, it was not gonna do that on its own.

And, like, I, myself, like, I do use LLMs as a research assistant.

Like, I wrote about that inside the Atlantic piece about the way that Claude has now sort of helped me, uh, edit my own work in a way that I found incredibly useful.

But I do feel like the collaborative element is m- is important for any domain where the personal perspective, lived experience, whatever, really matters.

Kevin RooseHost38:22

Talk about that a little bit.

Kevin RooseHost38:23

You mentioned your editing process.

Kevin RooseHost38:25

How are you using AI to help you edit your work, and are you finding it useful?

Yeah, so I feel like I really cracked this over the last, like, couple months, which I'm very excited about.

Because, again, I've tried to make these things, like, write and edit for me over and over and over, and they've never really been able to do it.

So the thing that I realized was, um, if I make Claude into an editor that is not just trying to grade and give feedback on my work against some genericized standard of what good writing is, but actually we- what we did against basically what my personal, Jasmine's personal aspirations for writing are, it can give feedback that I find much, much more helpful.

So what I did was basically I fed Claude my entire Substack archive of the writing that I've previously done, as well as some of my freelance work, um-

Kevin RooseHost39:07

And, uh, just to get real specific, is this inside, like, a Claude project, or how have you set this up? 'Cause I know s- like, our listeners are gonna wanna try this.

Yes.

Um, I did it in a project-

Kevin RooseHost39:14

Okay

... on Claude's advice.

I was like, "Do I need to Claude Code something?" Claude was like, "No, that's overkill."

Kevin RooseHost39:18

Okay.

"Um, so you don't need to code or anything." Um, so in a Claude project, I gave it my whole archive of writing.

I also personally write retro notes to myself after everything I publish.

So I have a notes app that's just, like, me writing what was good and bad about everything I've ever written.

Kevin RooseHost39:32

Hmm.

Um, just a few bullet points.

Kevin RooseHost39:34

This is why Jasmine's gonna be our boss.

Kevin RooseHost39:35

[laughs]

[laughs]

[laughs]

Kevin RooseHost39:37

For sure.

I mean, th- these are very low-quality bullet points.

But, um, I also gave it that because I wanted to learn my taste.

Kevin RooseHost39:42

Mm-hmm.

I wanted to learn what do I aspire to be, and where do I see myself falling short, and where- what am I proud of, right? And so-From those two things, plus a little bit more information about, like, "Here's my audience, this is my beat, this is my goals," um, we were able to co-develop a rubric of instead of, like, how many exclamation marks does it have, it would say things like, "Does this take advantage of your quote unquote, like, insider anthropologist position in Silicon Valley?" 'Cause that's one of the things that Claude and I think distinguish my voice.

Or it'll also notice like, "Oh, Jasmine, you tend to move between registers.

You'll switch between, you know, startup jargon and, like, internet slang and whatever." And, like, I think the fact that you can do the high-low or move from, like, policy to personal scene, this is something that is characteristic of your writing.

And so again, we're co-developing these qualitative criteria.

Um, and then I split it into phases of, like, ideation phase, a rubric, structure rubric, prose rubric, final fact-checking.

And so what I do now, I put this all on a Claude project.

I said, "Your job is to evaluate my drafts based on this criteria, but not to do the writing for me, and to make sure to prompt out of me, like, what I can do better." I dumped a draft into Claude.

Claude will run, like, phase two structure on it.

It'll say things like, "Your conclusion is just a summary, and this is really boring.

In fact, in your piece about this and that, you actually ended on a scene, and I thought that was much more powerful.

So why don't you try ending this one on a scene?" And Claude will say, rather than inventing a scene, it will say, "What were you thinking when the plane took off? What were you feeling inside? Can you think of a scenario where you had a conversation with, say, a kid safety advocate about AI that really resonated with you? Because right now it sounds like dry policy explainer." And, like, that feedback I actually found incredibly useful.

Kevin RooseHost41:17

Hmm.

Kevin RooseHost41:17

It is.

Like, I'm still applying my own judgment to say do I take it or not, but I'm like, you know, this is about me becoming the best version of myself as a writer.

It's about, like, me self-improving and Claude pushing me to do that, which I found much, much more helpful.

Kevin RooseHost41:28

Hmm.

Casey NewtonHost41:29

Wow.

Kevin RooseHost41:29

I wanna ask you both a question as-

[laughs]

Kevin RooseHost41:31

... fellow writers.

Kevin RooseHost41:33

Do you feel the impulse to make your writing weirder-

Mm-hmm

Kevin RooseHost41:38

... because of AI to sort of stand out from the sea of slop? Because I, I find myself feeling this tug of like, "Oh, that's a little weird aside that probably I should cut, but I think I'm gonna leave it in," 'cause, like, Claude would never do that, right?

Mm.

Kevin RooseHost41:53

It's like, it's like a marker that I am typing these words, and I feel like that's sort of my imprimatur that I'm leaving.

Casey NewtonHost41:59

My answer to you is yes, I absolutely feel that way, and I've, I've, like, gone back and tried to edit sentences to, like, make them feel a little bit more, like, weird or, like, in particular to make them sound colloquial in a way that I know, like, an LLM generally would not be.

Casey NewtonHost42:12

And, like, yes, it is for that reason.

Casey NewtonHost42:13

I think that writing right now, like, we're all, not all, many of us are on such high alert for the, uh, prospect that we might be reading slop-

[laughs]

Casey NewtonHost42:22

... that I think if you're a writer who does not want to be producing slop, like, y- you should be asking yourself that question.

I, I think it makes me a lot more comfortable writing the way I want to write in the first place.

Like, I think, like, maybe unlike both of you, I didn't sort of come up through newsrooms where I was, like, learning a very specific house style and all of these norms.

Like, I can do news writing now.

It's something I've learned now.

But, like, I'm actually much more, quote unquote, like, Internet and blogging native, w- which is a form that is voicey and irreverent and not as pristine and, like, will make inappropriate jokes, and, like, you know, it's just a looser form of writing.

And so I think what it's actually done is made me more comfortable doing the bloggy thing instead of sort of always trying to write in, um, a more professionalized journalistic tone.

Kevin RooseHost43:02

Hmm.

Kevin RooseHost43:03

So I think we should leave this with a question for you, Jasmine, which is, you know, y- your, your piece makes the case very convincingly that today's AIs are not very good at the kind of writing that I think we all value.

Kevin RooseHost43:14

Do you think they will get there, and what should the companies do to make their models better at writing?

I think that if we separate out text generation from reporting, which I'm not that bullish on the models doing, and we're, are just talking about, say, literary fiction or here's a bunch of interview transcripts, write a magazine feature or something, I think that if they applied as many resources towards that task as they do towards coding agents and things that actually make them money, I think that they could get there.

Um, will the companies ever find it financially advisable to spend all their resources on that instead of automating 23-year-old software engineers? Probably not.

I would be grateful for that world.

I don't need them to take my job or these folks' jobs.

But I think it's possible.

Casey NewtonHost43:57

Look, they're gonna get around to it eventually.

[laughs]

Casey NewtonHost43:59

Okay? You know, it's like, I mean, I, I, I hear what you're saying-

Kevin RooseHost44:01

Have you seen what writers make in this economy, Casey [laughs]?

Casey NewtonHost44:04

It, eventually, like, they-

Kevin RooseHost44:05

Those aren't gonna pay for a lot of data centers.

Casey NewtonHost44:07

No. The-

[laughs]

Casey NewtonHost44:08

...

Casey NewtonHost44:08

there is economic value in writing.

Casey NewtonHost44:09

Like, and eventually the AI companies will want that all to themselves.

Kevin RooseHost44:12

You know what would be a f- very funny outcome of this, you know, uh, taking your, your point about the sort of guardrails of the models, maybe the great, the next great American novel will be written by Grok.

Oh, God.

[laughs] That's... [laughs]

Kevin RooseHost44:25

And with that, Jasmine Sang, thank you for joining us.

Casey NewtonHost44:28

Thank you, Jasmine.

Thank you very much, Kevin and Casey.

[laughs]

Kevin RooseHost44:31

[laughs]

Casey NewtonHost44:35

When we come back, everyone's spending money on tokens, Kevin.

Kevin RooseHost44:39

Great.

Kevin RooseHost44:40

Keep going.

Casey NewtonHost44:41

You've gotta be token, token maxing, that is.

Kevin RooseHost44:44

Keep going.

Casey NewtonHost44:44

[laughs]

Kevin RooseHost44:45

Yes, and.

Casey NewtonHost44:46

When we come back, what are you token about? That's the question being asked by a leaderboard that's sweeping Silicon Valley.

Kevin RooseHost44:52

That's sweeping.

Kevin RooseHost44:53

It's sh- it's really sweeping.

Casey NewtonHost44:55

Ah, sweep.

Casey NewtonHost44:56

[upbeat music] Well, Kevin, you've recently returned from book leave and are once again writing in The New York Times.

Casey NewtonHost45:14

How does it feel to see your name in print again?

Kevin RooseHost45:16

Feels great.

Kevin RooseHost45:17

Hasn't happened yet, but when it does, it'll be great.

Casey NewtonHost45:19

Well, I got to take an early read at a story that, uh, you are publishing about the fact that tech companies have now created leaderboards to show which employees are using the most AI tokens in their work.

Kevin RooseHost45:34

Yes, it's a token frenzy out there, and, uh, the employees of these companies are competing among their colleagues sort of informally and, and sort of for fun, but they're taking it very seriously.

Kevin RooseHost45:47

They want to be the people at their company who are using the most AI tokens.

Casey NewtonHost45:51

So let me just ask a basic question for listeners who may not be familiar.

Casey NewtonHost45:55

What is a token, and why is that something you might start keeping track of?

Kevin RooseHost45:59

So a token is the basic atomic unit of AI labor.

Kevin RooseHost46:04

It's basically a, a fragment of a word, and it is how AI model providers measure their consumption.

Kevin RooseHost46:13

So if you type in a prompt, uh, you know, "Help me, uh, write this essay," an old model might have given you a couple hundred tokens in response.

Kevin RooseHost46:23

That would be a couple hundred words.

Kevin RooseHost46:26

And-What has been happening over the past year or so as these agentic coding tools have started taking off is that the models are just much more token hungry.

Kevin RooseHost46:36

You can use now ah hundreds of thousands or even millions of tokens in a single session and so that is what is propelling these leaderboards is the idea that the more sort of coding you're doing, the more agentic tools you're using, the more simultaneous processes you're running, the higher your token count will be.

Casey NewtonHost46:57

One measurement I found useful was that apparently it takes about 10,000 tokens to generate 7,500 words, if that sort of, you know, helps to ground you at all.

Casey NewtonHost47:07

But as, as you just said, and I want to hear more about this, the more advanced systems are using way more tokens than that.

Casey NewtonHost47:15

So tell me about some of the numbers that some of the sort of token all-stars are putting up on the, on the boards.

Kevin RooseHost47:21

[laughs] So I don't know all of the exact numbers, but I did learn that at OpenAI, where they do track this kind of leaderboard, the highest employee token count over a seven-day period recently ah was a guy who used 210 billion tokens [laughs].

Casey NewtonHost47:40

[laughs]

Kevin RooseHost47:42

And this is, for rough scale, about 33 Wikipedia's worth of text.

Casey NewtonHost47:48

Mm.

Kevin RooseHost47:48

And now all of that is not sort of typing and receiving a response.

Kevin RooseHost47:52

Some of that is what they call cached tokens, so it's not all sort of, you know, being extruded from the model for the first time.

Kevin RooseHost48:01

But these are the kinds of numbers that I think even a year ago would have sounded completely insane.

Casey NewtonHost48:06

Right.

Casey NewtonHost48:06

Now, was this guy working on a new mass domestic surveillance program-

Kevin RooseHost48:09

[laughs]

Casey NewtonHost48:09

... for the Department of Defense?

Kevin RooseHost48:11

I don't know, and OpenAI did not make him available for interviews.

Casey NewtonHost48:16

Yeah.

Kevin RooseHost48:16

But what I wanted to do in writing this column was to try to call up a bunch of people or talk to a bunch of people who are in this sort of billion token club, right? The, the sort of extreme power users, and just ask them like, "Hey, how are you guys using all those tokens, and isn't that very expensive, and how are you paying for it all?" And I learned a lot.

Casey NewtonHost48:34

Yeah.

Casey NewtonHost48:34

Well, okay.

Casey NewtonHost48:34

Well, so tell us first of all just how expensive it is.

Kevin RooseHost48:37

Very expensive.

Kevin RooseHost48:37

[laughs]

Casey NewtonHost48:37

Yeah.

Kevin RooseHost48:38

In fact, um, I, I, ah, heard that the top user of Claude Code, the top individual user of Claude Code, uh, as measured by Anthropic, uh, spent more than $150,000 on tokens last month.

Kevin RooseHost48:55

So extrapolate that.

Kevin RooseHost48:56

That is like a, an employee making more than a million dollars a year.

Casey NewtonHost49:01

Yeah.

Kevin RooseHost49:01

And they are burning that in a month, and I heard similar figures from some of these other extreme coders who are spending something on the order of thousands of dollars a day on tokens from these models.

Kevin RooseHost49:14

Now, we should also say the employees of these companies get their tokens for free, right?

Casey NewtonHost49:18

Right.

Kevin RooseHost49:18

So they're not shelling out, their companies are not shelling out.

Kevin RooseHost49:21

But at other companies, this is starting to become an issue because they are at, sort of outstripping their, their budgets for these things.

Casey NewtonHost49:28

So there are companies where there are engineers who legitimately are costing their employers maybe $150,000 a week because they're getting tokens from one of the big providers.

Kevin RooseHost49:37

Yeah.

Kevin RooseHost49:37

I talked to a software engineer in Sweden who said that he probably spends more than his salary on Claude.

Kevin RooseHost49:46

So this is essentially becoming like a, a very expensive job perk for some of these coders.

Casey NewtonHost49:52

So talk to me about why employers want to create leaderboards to promote this to employees, because I could see other companies saying, "If you spent $150,000 on tokens last month, you actually don't work at this company anymore-

Kevin RooseHost50:07

[laughs]

Casey NewtonHost50:07

... 'cause we're bankrupt."

Kevin RooseHost50:08

[laughs] Right.

Kevin RooseHost50:09

So this was a big question that I had is like why is this going on, and it seems to be some combination of sort of employee motivation and worker tracking, right? There are executives at these companies who think that the more tokens you use, the more productive you probably are, and as we discussed in a previous segment on this show, these companies are very eager to have their workers start embracing the AI tools.

Kevin RooseHost50:38

And so at a number of these companies, I talked to people who said, "Yeah, this is just basically them trying to see who is really all in on the new way of programming."

Casey NewtonHost50:48

And you've talked to a number of people who are ranking high on these leaderboards.

Casey NewtonHost50:52

I realize you probably haven't dug deep into their code, but what is your sense of how productive they actually are? Like, what is the relationship between token usage and taking my company to the next level?

Kevin RooseHost51:06

[laughs] I mean, it's very unclear, right? Some of these people may be just generating like worthless projects.

Kevin RooseHost51:11

I think the, the thing that worries a lot of the people I talk to about these leaderboards is that they just incentivize you to like run up your token count, right?

Casey NewtonHost51:18

Yes.

Kevin RooseHost51:19

'Cause then you look like the special, you know, 10X engineer or 100X engineer who's like outperforming all your colleagues.

Kevin RooseHost51:25

So I think there are a number of, uh, companies that, that see this leaderboard business as a little strange and maybe counterproductive, but I do think that there is a feeling among the most sort of heavy token users that they are being productive.

Casey NewtonHost51:39

Yeah.

Casey NewtonHost51:39

I have to say, when I read your column, I thought this just seems like it would create the worst incentives.

Kevin RooseHost51:45

Yes.

Casey NewtonHost51:45

Right? There's this idea of Goodhart's law, right? Like, when a measure becomes a target, it ceases to become a good measure.

Casey NewtonHost51:52

I can't think of a better way to ensure that tokens usage becomes a bad measure than creating a leaderboard for it.

Kevin RooseHost51:58

Totally.

Casey NewtonHost51:59

What are the people inside the company saying about that?

Kevin RooseHost52:01

Well, some of them are opposed to this whole leaderboard thing.

Kevin RooseHost52:05

I also talked with some folks who defended the leaderboards.

Kevin RooseHost52:08

They said, "Look, it's never been all that easy to track the productivity of programmers.

Kevin RooseHost52:13

Some people have had their productivity measured by like how many lines of code they generate or how many pull requests they made.

Kevin RooseHost52:19

These are sort of these imperfect proxies for like how hard are you working, how much are you doing.

Kevin RooseHost52:25

But the employees of these companies also see this, I think wisely, as a key to their own success.

Kevin RooseHost52:32

A number of these companies are now using AI token use and consumption as part of the performance review cycle.

Kevin RooseHost52:40

So you go in for your annual review.

Kevin RooseHost52:42

Your boss says, "Hey, it looks like you only used, you know, 70 million tokens, uh, last month.

Kevin RooseHost52:48

What's going on?" And so I think the engineers at these companies are getting wise to the fact that if they want to have a long, successful career, they better start using some tokens.

Casey NewtonHost52:57

Yeah, but I imagine that some of them are really nervous about that, though, right? Because like i- it seems clear to me that at least some of these companies want to incentivize token usage because the companies themselves suspect that the more we can get them using this stuff, the less long we will have to employ the humans.

Kevin RooseHost53:12

Maybe, although I think it's less about like the AI systems replacing the humans and more about like it is just a radically different way of working, right?

Casey NewtonHost53:21

Mm-hmm.

Kevin RooseHost53:22

These are people who most of them have had long careers in software engineering.They grew up writing code by hand, they maybe grew up using some sort of, like, AI assistant, like GitHub Copilot and what people at these companies are saying is that these agentic engineering systems are just really different.

Kevin RooseHost53:41

You have to approach them in a different way, you have to spend a lot of time with them to understand what they're good and not good at and to them, this is sort of a way of motivating their employees to say, "Hey, go out and try the new thing."

Casey NewtonHost53:52

Yeah.

Casey NewtonHost53:53

I d- I don't know.

Casey NewtonHost53:53

I've been thinking a lot about this question of, like, if I were an engineer at one of these companies and I had this incentive to get on the leaderboard, like, how would I approach it? And I do think that, like, the instinct to just, like, waste a bunch of tokens to, to, like, rise higher on the leaderboard...

Casey NewtonHost54:08

Like, ultimately, if you rise too high, people are gonna ask you what you did-

Kevin RooseHost54:11

[laughs]

Casey NewtonHost54:11

... with all the tokens.

Kevin RooseHost54:12

Right.

Kevin RooseHost54:12

[laughs]

Casey NewtonHost54:12

If you're number one at, like, 10 billion tokens and you only managed to, you know, like, a v- you know, vibe-code a calculator or something-

Kevin RooseHost54:19

[laughs]

Casey NewtonHost54:19

... people are probably gonna get mad at you.

Kevin RooseHost54:21

Yeah, and I actually did talk to one person who speculated that actually the people at the top of the leaderboards are all doing side projects.

Kevin RooseHost54:26

[laughs]

Casey NewtonHost54:27

[laughs]

Kevin RooseHost54:27

They're starting their, uh, their-

Casey NewtonHost54:29

They're starting a new company

Kevin RooseHost54:29

... their side hustles.

Kevin RooseHost54:30

[laughs]

Casey NewtonHost54:30

They start a new company with the, with the boss's money.

Casey NewtonHost54:33

And if you're doing that, I just wanna say I salute you.

Kevin RooseHost54:35

Yeah.

Casey NewtonHost54:35

Like, that is the right way to work in 2026.

Kevin RooseHost54:36

Yeah, maybe don't be the number one on the leaderboard-

Casey NewtonHost54:39

[laughs]

Kevin RooseHost54:39

...

Kevin RooseHost54:39

if you're doing that.

Kevin RooseHost54:39

Maybe try to stick around six or seven.