Why Tech C.E.O.s Are Blaming A.I. for Mass Layoffs

Summary

This episode of Hard Fork covers three major stories. First, Kevin Roose and Casey Newton examine a wave of tech layoffs at Atlassian (1,600 jobs, 10% of staff), Block, and Meta (reportedly up to 20% of its workforce). The hosts debate whether AI is genuinely driving these cuts or whether companies are “AI washing” — using AI as a convenient excuse for layoffs that have more to do with cost-cutting and market positioning. They note that Block’s stock jumped 7.5% after Jack Dorsey framed layoffs as AI-driven, drawing comparisons to the crypto era when companies would add blockchain to their name for a stock boost. The key tension: companies on the AI frontier like OpenAI and Anthropic are actually hiring aggressively, while companies like Atlassian and Block are using the AI narrative to justify cuts.

Second, the hosts welcome journalist Jasmine Sun to discuss her Atlantic piece arguing that AI chatbots remain fundamentally bad at creative writing. Sun makes the case that GPT-2 was paradoxically the peak of AI creativity — those early models were “weird” and “nutty” in ways that produced genuinely surprising text. Post-training through RLHF (reinforcement learning from human feedback) made models useful but stripped away their creative edge, turning them into what Sun calls “a very helpful assistant that might be very bad at writing in creative and surprising ways.” She reveals that the humans evaluating AI writing quality are often grading fan fiction on factuality, highlighting how the training process itself sabotages creative output.

Third, Kevin reports on “tokenmaxxing” — the emerging phenomenon of tech companies creating leaderboards to track which employees consume the most AI tokens. At OpenAI, the top user burned through 38 million tokens in a single week. The top individual Claude Code user spent over $150,000 on tokens in one month. A Swedish engineer told Kevin he probably spends more than his salary on Claude. The hosts worry this creates perverse Goodhart’s law incentives, where the metric becomes the target and employees waste tokens on side projects rather than productive work, while some companies are beginning to incorporate token usage into performance reviews.

Highlights

”AI washing is the new crypto washing”

“It turns out that the public markets actually can just be tricked that easily.” — Casey Newton, 8:52

yt-dlp --download-sections "*8:43-9:05" "https://www.youtube.com/watch?v=Prm_V51XbPg" --force-keyframes-at-cuts --merge-output-format mp4 -o "ai-washing-crypto-washing.mp4"

”GPT-2 was the peak of AI creative writing”

“But they were surprising! They were like nutty. They were… 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 well. But they were weird and creative and kind of delightful in a way that the modern models are not.” — Jasmine Sun, 22:21

yt-dlp --download-sections "*22:21-23:15" "https://www.youtube.com/watch?v=Prm_V51XbPg" --force-keyframes-at-cuts --merge-output-format mp4 -o "gpt2-peak-creative-writing.mp4"

”The top Claude Code user spent over $150,000 on tokens last month”

“Very expensive. In fact, I heard that the top user of Claude code, the top individual user of Claude code as measured by Anthropic, spent more than $150,000 on tokens last month.” — Kevin Roose, 46:57

yt-dlp --download-sections "*46:57-47:55" "https://www.youtube.com/watch?v=Prm_V51XbPg" --force-keyframes-at-cuts --merge-output-format mp4 -o "150k-claude-code-tokens.mp4"

”He probably spends more than his salary on Claude”

“I talked to a software engineer in Sweden who said that he probably spends more than his salary on Claude. So this is essentially becoming like a very expensive job perk for some of these coders.” — Kevin Roose, 48:00

yt-dlp --download-sections "*48:00-48:30" "https://www.youtube.com/watch?v=Prm_V51XbPg" --force-keyframes-at-cuts --merge-output-format mp4 -o "spends-more-than-salary-on-claude.mp4"

”Grading fan fiction on factuality”

“He got a bunch of fan fictions and he was supposed to grade them on their factuality, since that was one of the criteria.” — Jasmine Sun, 25:59

yt-dlp --download-sections "*25:58-26:35" "https://www.youtube.com/watch?v=Prm_V51XbPg" --force-keyframes-at-cuts --merge-output-format mp4 -o "grading-fanfiction-on-factuality.mp4"

”What’s my token budget?”

“Well what’s so interesting is now it’s becoming part of job conversations for engineering jobs. People are going into new job conversations and saying, well what’s my token budget?” — Casey Newton, 58:25

yt-dlp --download-sections "*58:25-59:02" "https://www.youtube.com/watch?v=Prm_V51XbPg" --force-keyframes-at-cuts --merge-output-format mp4 -o "whats-my-token-budget.mp4"

Key Points

• Atlassian cuts 1,600 jobs citing AI investment (0:48) - 10% staff reduction framed as funding further AI investment, but unclear what AI actually replaced
• Meta reportedly planning 20% workforce reduction (2:00) - Reuters report that Meta called “speculative” — which Kevin translates as “this is happening but we don’t want to tell you”
• AI washing defined (5:00) - Companies using AI as a convenient excuse for layoffs that may be driven by other factors like cost-cutting
• Block’s stock jumps 7.5% after AI-framed layoffs (7:46) - Markets rewarding companies that frame layoffs as AI-driven, reminiscent of crypto-era name changes
• Frontier AI companies are hiring, not cutting (13:31) - OpenAI, Anthropic, and xAI are growing while companies further from the frontier are laying off
• DocuSign has 7,000 employees (14:26) - A fact that Casey calls “no funnier sentence that is true in all of tech”
• AI could trigger tech worker unionization (16:25) - Kevin wonders if mass AI-related layoffs could finally push tech workers to organize
• Post-training killed AI creativity (23:21) - RLHF made models useful but created a “helpful assistant” persona that is bad at surprising, creative writing
• RLHF raters grade fan fiction on factuality (25:59) - The evaluation criteria used in training actively sabotage creative output
• AI writing tools frustrated by their own guardrails (36:00) - Founders of creative writing AI tools like Sudowrite struggle to make models less “chirpy” and “PG-13”
• Jasmine’s Claude editing workflow (37:59) - Fed Claude her best and worst writing samples plus critic quotes to train a personalized editor
• OpenAI’s top token user: 38 million tokens in one week (45:54) - Employee leaderboards tracking token consumption as a proxy for AI adoption
• Top Claude Code user: $150K/month (46:57) - Individual token spending exceeding annual salaries at some companies
• Token usage entering performance reviews (51:00) - Some companies incorporating AI token consumption into annual review criteria
• Token budget as job perk (58:25) - Engineers now asking “what’s my token budget?” in job interviews alongside salary and equity

Mentions

Companies

• Atlassian (0:48) - Cut 1,600 jobs (10% of staff), cited AI investment
• Block (5:53) - Jack Dorsey announced layoffs framed as AI-driven; stock rose 7.5%
• Meta (10:01) - Reportedly planning to cut up to 20% of workforce as AI costs mount
• OpenAI (2:14) - Being sued by NYT; tracks employee token leaderboards; top user hit 38M tokens/week
• Anthropic (2:22) - Casey’s fiancee works there; measures top Claude Code user at $150K/month
• xAI (24:39) - Lists creative writing expert job postings for model evaluation
• Sudowrite (35:25) - One of earliest creative fiction AI writing assistants; frustrated by model guardrails
• DocuSign (14:26) - Has 7,000 employees, cited as a sign of enterprise bloat

Products & Technologies

• Claude Code (46:57) - Top individual user spent $150K/month on tokens
• GPT-2 (21:01) - Argued to be peak of AI creative writing; “weird” and “nutty” but genuinely surprising
• RLHF (24:08) - Post-training process that made models safe but killed creative output

People

• Jack Dorsey (5:53) - Block CEO who framed layoffs as AI-driven; Casey is skeptical of his leadership
• Mark Zuckerberg (10:42) - Said “projects that used to require big teams now can be accomplished by a single very talented person”
• Jasmine Sun (18:26) - Journalist; wrote “The Human Skill That Eludes AI” for The Atlantic; Kevin’s book researcher
• Sam Altman (28:10) - Said a year ago that writing a great short story was top priority; models still haven’t cracked it
• Aaron Zamos (9:02) - Former Block communications head who wrote NYT essay arguing the layoffs aren’t truly about AI
• James Yu (35:25) - Co-founder of Sudowrite; frustrated that models are too “chirpy” and “sycophantic”

Surprising Quotes

“Which, if you’re not familiar with the language deployed by Meta communications staffers, means ‘this is happening, but we don’t want to tell you it’s happening yet’.” — Kevin Roose, 2:07

“I cannot think of anything that would make Mark Zuckerberg more mad than a union of software engineers at Meta and I think the software engineers at Meta should use that information as they see fit.” — Casey Newton, 17:22

“To me, talking to GPT-2 was like talking to somebody who had just fallen down the stairs. You know what I mean? It was like, do we need to get you to the hospital?” — Kevin Roose, 21:54

“I talked to a software engineer in Sweden who said that he probably spends more than his salary on Claude.” — Kevin Roose, 48:00

“Do what Marc Andreessen will not and introspect.” — Kevin Roose, 59:02

Transcript

Kevin Roose: 0:00 I’m Kevin Roose, a tech columnist at the New York Times.

Casey Newton: 0:02 I’m Casey Newton from Platformer.

Kevin Roose: 0:04 And this is Hard Fork. 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 creative writing? And finally, it’s tokenmaxxing time. Well, Casey, for years now we’ve been monitoring for signs of an AI job apocalypse.

Casey Newton: 0:33 Yeah, we’ve been monitoring the situation. It’s true.

Kevin Roose: 0:34 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 Newton: 0:41 Yeah, we have certainly heard CEOs of companies announcing layoffs invoking AI as a reason that it is happening, and so that has gotten our attention.

Kevin Roose: 0:48 Yeah, so just a couple examples from the last few weeks. 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.

Casey Newton: 2:00 Meta, after this story came out, told Reuters that it was, quote, ‘speculative reporting’.

Kevin Roose: 2:07 Which, if you’re not familiar with the language deployed by Meta communications staffers, means ‘this is happening, but we don’t want to tell you it’s happening yet’. So Casey, I want to hear what you make of these layoffs, but first we should do our disclosures. I work for the New York Times, which is suing OpenAI, Microsoft, and Perplexity.

Casey Newton: 2:22 And my fiancee works at Anthropic.

Kevin Roose: 2:25 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 Newton: 2:33 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 it. But I will say that we’re sort of hearing this from more and more companies and sooner or later I do think we’re going to have to believe them.

Kevin Roose: 3:03 Yeah, I think there is probably some complications here and we should get into them, but like I think this is the early warning sign for a lot of people, especially in the tech industry, who are, I think, starting to look at this and say, oh, this is real. This is not a drill. So Atlassian, Casey, they blamed AI. They said the CEO said it was a quote, “hard call, but the right one.” Should we take them at their word?

Casey Newton: 3:59 Yeah, so I take him at his word. 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 we’re doing this.

Kevin Roose: 5:00 Hmm. So there’s this term that’s been floating around called AI washing, 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 instead of just saying that they invoke AI as the reason.

Casey Newton: 5:11 I thought it was when a software engineer finally took a shower.

Kevin Roose: 5:16 And basically the thesis is like these aren’t really layoffs about AI. This is just sort of a convenient excuse that these companies are using. Do you think Atlassian qualifies as AI washing?

Casey Newton: 5:26 Um, I would like to get a little bit more detail on exactly who they are laying off here, which is a detail that we do have about some of these other companies that helps us answer that question. So I’m reserving judgment for now.

Kevin Roose: 5:53 Okay, let’s talk about Block. Jack Dorsey, the CEO of Block, gave an explanation about their layoffs. He said quote, “we’re not making this decision because we’re in trouble, our business is strong, but something has changed, I had two options, cut gradually over months or years as this shift plays out, or act now.”

Casey Newton: 6:15 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. At this point in 2024 I would not hire Jack Dorsey to run a Bath and Body Works.

Kevin Roose: 7:11 Come on!

Casey Newton: 7:12 Yeah. So that’s the kind of famous attention to detail that has turned Jack Dorsey into one of the greatest visionaries in tech. So look, is this about AI? Again, you know, what does Block really do? They have Cash App and they have the little Square reader.

Kevin Roose: 7:46 Yeah, you could use AI washing or Jay-Z washing which seems to be what they’re doing here. So this did seem to have an effect on their stock price. In fact, the day after Jack Dorsey announced these layoffs, Block’s stock went up 7.5%.

Casey Newton: 8:43 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 and their stock price would shoot up by like 40%.

Casey Newton: 8:52 It turns out that the public markets actually can just be tricked that easily. That was giving me some relief if I was a CEO. Just knowing that I could fool people like that. But anyways.

Kevin Roose: 9:02 Totally. And I don’t think everyone is falling for it. Aaron Zamos, the former head of communications for what was then Square, wrote a guest essay in the New York Times recently basically saying these layoffs are not about AI.

Casey Newton: 9:31 Yeah, I have to say, like, what do we even think the future of Block is? Again, they make like a little widget you can plug into your phone so that you can sell bracelets at the craft fair, and then it sort of does some business stuff too.

Kevin Roose: 9:58 Don’t email Casey, just Cash App him instead. So let’s talk about the third large tech company that is reportedly conducting layoffs. Meta. We don’t know exactly who or what teams are being affected by these layoffs, but this is a significant part of the workforce — potentially up to 20%.

Casey Newton: 10:42 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 we should also say that this is happening at the same time that Meta is spending, you know, tens of billions of dollars on building out its AI infrastructure.

Kevin Roose: 11:12 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. They are reallocating spending from human labor to AI compute. I recently talked to a venture capitalist who said that the most AI native companies are spending more on AI tools than they are on payroll. And that may be an outlier, but I think that is sort of where these companies are headed.

Casey Newton: 12:23 Yes, I think that’s absolutely the bet that they’re making. 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 had its year of efficiency. They’ve been through layoffs before. 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. I’m sure there are individuals there whose jobs have gotten easier because of AI.

Kevin Roose: 13:31 Yeah, 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 and the Anthropics and the xAIs of the world. Those companies are all hiring like crazy.

Casey Newton: 14:08 Yes, but also like OpenAI and Anthropic are much smaller companies than some of the ones that we’ve been talking about today.

Jasmine Sun: 14:26 DocuSign has 7,000 employees.

Casey Newton: 14:28 That there’s no funnier sentence that is true in all of tech than that one.

Kevin Roose: 14:31 Listen, as somebody who has a paid subscription for DocuSign that I truly resent paying for, get to work over there, people.

Casey Newton: 14:40 Here’s another question that I would ask Kevin. Okay, so we’re seeing a bunch of layoffs, like, are these AI related or not? Does it actually matter if the effect on workers is the same, right?

Kevin Roose: 14:55 Yeah, and it’s not clear to me what workers can or should be doing to sort of protect themselves against these layoffs. One person I talked to said they work at one of these big tech companies and they’re like, well, there’s just a lot of jostling now for who can show they’re the most AI-savvy.

Casey Newton: 15:29 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 positive byproduct, right? Like if you can scare the people who remain into adopting AI faster, that’s kind of a win for you.

Kevin Roose: 16:25 Totally. 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 in the tech industry. Like is that possible now that there are sort of more concrete examples of people losing their jobs?

Casey Newton: 17:22 Here’s what I will say. I cannot think of anything that would make Mark Zuckerberg more mad than a union of software engineers at Meta and I think the software engineers at Meta should use that information as they see fit.

Kevin Roose: 17:34 You think that would make him more mad than getting booed at a UFC fight?

Casey Newton: 17:40 Absolutely. I think that probably just made him really sad.

Kevin Roose: 17:45 Well, there you have it. If you want to make Mark Zuckerberg mad, then employees sign your union card. Casey, over the last couple of years, we’ve talked on this show about how AI models are getting better at all sorts of things. Competition math, at solving novel physics problems, mass domestic surveillance, autonomous weapons. And I think the story of the last few years in AI has been one of sort of rapid, steady progress in most areas. But one area where models have not gotten obviously better is creative writing.

Casey Newton: 18:25 Now that’s our domain.

Kevin Roose: 18:26 Yes. At least that is the argument that Jasmine Sun made in the Atlantic this week. She is a freelance journalist. Her piece was called ‘The Human Skill That Eludes AI.’ And it’s her attempt to understand why chatbots just can’t write well.

Casey Newton: 18:50 Yeah, 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 seem to plateau when it comes to truly creative output.

Kevin Roose: 18:53 Yes. And we should say, before we bring her in, Jasmine is a friend of mine. She has also been my researcher on the upcoming book that I’m working on.

Casey Newton: 19:06 Alright, I’ll allow it, but I do want to balance it out by next week bringing on one of your enemies. Okay, let’s bring her in.

Kevin Roose: 19:10 Jasmine, welcome to Hard Fork.

Jasmine Sun: 19:12 Thanks for having me! I’m excited.

Casey Newton: 19:16 So, you wrote this great piece in the Atlantic this week about the human skill that eludes AI, and I want to start by challenging the subtitle of your piece: Why can’t language models write well? Can’t they?

Jasmine Sun: 19:32 They can. 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 interested in is: why does it seem like even with all this massive spending on AI development, the improvements in mathematical reasoning and coding ability just haven’t transferred to creative writing?

Casey Newton: 21:01 Hmm. And you start your piece with this interesting provocation, which is that in some ways GPT-2 was the peak of AI when it comes to creative writing. So explain that.

Jasmine Sun: 21:13 Part of what got me interested in this piece was I was actually doing research for your book, and I was going through all these previous generations of models and reading the outputs. And the thing that shocked me was that these early models, the GPT-2 and GPT-3 era models, were genuinely weird and creative in a way that modern models are not.

Kevin Roose: 21:54 They were weird. That shocks me. To me, talking to GPT-2 was like talking to somebody who had just fallen down the stairs. You know what I mean? It was like, do we need to get you to the hospital?

Casey Newton: 22:05 Yeah, there are these amazing prompts from this early OpenAI prompt library where they would say like, ‘I just won $175,000 in Las Vegas. What do I need to know about taxes?’ And it would just go on the most wild tangent.

Jasmine Sun: 22:21 But they were surprising! They were like nutty. They were… 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 well. But they were weird and creative and kind of delightful in a way that the modern models are not.

Casey Newton: 23:11 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.

Jasmine Sun: 23:21 Yeah, I mean, the answer is post-training basically. So they started adding a post-training layer, which is basically saying we have these like crazy unpredictable nutjob concussed people and we are going to train them to be a character or persona that is a very helpful assistant, but might be very bad at writing in creative and surprising ways.

Casey Newton: 24:08 I mean, the way that you described it was that there is a phase within the post-training phase where these AI models are evaluated by humans, and that’s part of what they call RLHF, or reinforcement learning from human feedback.

Jasmine Sun: 24:39 Yeah, I mean, this is super interesting because like 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, four stars on Goodreads or a starred Kirkus review.

Casey Newton: 24:52 Have you ever gotten a starred Kirkus review, Roose?

Kevin Roose: 24:53 I think so. Not sure. All right.

Jasmine Sun: 25:01 So these companies, because they realize that these AI researchers, they’re 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 writers? But then the issue is that you give these writers these rubrics that are extremely reductive. 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.

Kevin Roose: 25:48 Yeah, and I have to say, generally not bad writing advice.

Jasmine Sun: 25:55 I mean, this is what they tell women in business communications. It’s like, take all those exclamation points, replace them with periods. He got a bunch of fan fictions and he was supposed to grade them on their factuality, since that was one of the criteria.

Kevin Roose: 26:21 And is that just briefly like I want to underline that because to me that seems like the whole story. We are taking the entire internet and we are grading it on factuality and like so the LLM that comes out is going to be maximally factual and minimally creative.

Casey Newton: 26:33 Well, and I wonder how much of it is related to this sort of verifiable reward system that a lot of these companies are using where you have a system generate a bunch of code and then you have another system verify it. And that works extremely well for code and math. But for writing, it just comes down to preference.

Jasmine Sun: 27:23 I think both are true. It’s like the majority of writing that we are asking the models to do is, ‘write this email for me,’ right? And they excel at that. They are truly great corporate email writers. But the creative side is just not where the economic demand is.

Kevin Roose: 28:10 And to that point, you know, you started this segment by talking about Sam Altman saying like, ‘hey, you know, we’ve just basically can’t write a great poem yet.’ Sam Altman a year ago said the company’s top priority was writing a great short story.

Jasmine Sun: 28:29 Ooh, wouldn’t be the first time. But that short story, if you remember, had some great lines, like talking about the seams of mirrors.

Casey Newton: 28:38 It was like the ‘liminal almost Friday’ or —

Jasmine Sun: 28:40 The liminal day that is almost Friday!

Jasmine Sun: 28:49 I mean, I think while you’re looking it up, like, the thing about AI writing is it comes up with all these fun metaphors and they are like kind of surprising sometimes with the metaphors. But there’s like a deeper level of creative surprise that’s about the structure, the pacing, like the emotional arc of a piece.

Kevin Roose: 29:52 I want to pose a couple objections that I think someone might make to your article. One of them is this is cope. This is Jasmine, a writer, a very talented writer, sort of finding the thing that AI can’t do yet and clinging to it.

Jasmine Sun: 31:17 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 to having to write. I like writing, but I have tried over and over and over to make these things write for me and they just can’t.

Kevin Roose: 32:27 The other objection I imagine people might have who are very AI-pilled is that this is all in the eye of the beholder, right? There have been several studies now that have shown that if you give people AI-generated text and human-generated text without labels, they prefer the AI text. But the moment you tell them that they are AI models generating text and not humans writing words with their fingers, we lose all interest in it just because of the source not because of the quality of the writing.

Jasmine Sun: 33:10 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 see AI text that is obviously AI.

Casey Newton: 34:47 Well we’re already seeing that the LLMs make huge progress in genre fiction, right? So like recently on the show we talked to the author of a story in the Times about how authors of romance novels are using AI to write way more books.

Jasmine Sun: 35:25 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. In my conversations with them it clearly frustrates them that it is so hard to get these models to stop being so chirpy, so sycophantic, so PG-13 and everything, in order to get them to this sort of raw, surprising place.

Kevin Roose: 37:08 Talk about that a little bit. You mentioned your editing process. How are you using AI to help you edit your work and are you finding it useful?

Jasmine Sun: 37:15 Yeah, so I feel like I really cracked this over the last couple months. Because again, I’ve tried to make these things write and edit for me over and over and it just didn’t work. What I did is I basically pasted some of my favorite things I’ve written and some of the worst things I’ve written and also pasted in some quotes from some of my favorite critics.

Kevin Roose: 37:56 And just to get real specific, is this inside like a Claude project or how have you set this up?

Jasmine Sun: 37:59 Yes. I did it in a project, but on Claude’s advice. I was like, do I need to Claude code something? And Claude was like, no, that’s overkill. I also gave it that because I wanted it to learn my taste. I wanted it to learn what do I aspire to be and where do I see myself falling short and what am I proud of.

Jasmine Sun: 39:00 It’s like, oh, Jasmine, you tend to move between registers. You’ll switch between startup jargon and internet slang and whatever. And the fact that you can do the high-low thing is actually your strongest suit. And I was like, that’s actually a really good insight about my writing.

Kevin Roose: 40:21 I want to ask you both a question as fellow writers. Do you feel the impulse to make your writing weirder because of AI? To sort of stand out from the sea of slop. Because 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 going to leave it in because like, that is what makes me human.

Casey Newton: 40:51 My answer to you is yes, I absolutely feel that way and I’ve gone back and tried to edit sentences to make them feel a little bit more weird or in particular to make them sound colloquial.

Jasmine Sun: 41:18 I think it makes me a lot more comfortable writing the way I want to write in the first place.

Casey Newton: 42:00 Today’s AIs are not very good at the kind of writing that I think we all value. Do you think they will get there and what should the companies do to make their models better at writing?

Jasmine Sun: 42:12 I think that if we separate out text generation from reporting, which I am not that bullish on the models doing, and we are just talking about literary fiction or here’s a bunch of interview transcripts, write me a piece — I think we’re years away from that being good. The labs would need to fundamentally rethink their training approach.

Casey Newton: 42:52 Look, they’re going to get around to it eventually. Have you seen what writers make in this economy, Jasmine?

Kevin Roose: 43:02 It’s not going to pay for a lot of data centers.

Kevin Roose: 43:13 You know what would be a very funny outcome of this, you know, taking your point about the sort of guardrails of the models. Maybe the next great American novel will be written by Grok.

Casey Newton: 43:22 And with that, Jasmine Sun, thank you for joining us.

Jasmine Sun: 43:25 Thank you very much, Kevin and Casey.

Casey Newton: 43:26 Well Kevin, you’ve recently returned from book leave and are once again writing in the New York Times. How does it feel to see your name in print again?

Kevin Roose: 43:40 Feels good. Hasn’t happened yet, but when it does, it’ll be great.

Casey Newton: 43:47 Well, I got to take an early read at a story that 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. What is going on?

Kevin Roose: 44:08 Yes, it’s a token frenzy out there and the employees of these companies are competing among their colleagues sort of informally and sort of for fun, but they’re taking it very seriously. They want to be the number one token consumer.

Casey Newton: 44:18 So let me just ask a basic question for listeners who may not be familiar. What is a token and why is that something you might start keeping track of?

Kevin Roose: 44:30 So a token is the basic atomic unit of AI labor. It’s basically a fragment of a word and it is how AI model providers measure their consumption. So if you type in a prompt, you know, help me write this email, that takes a certain number of tokens to process. And the more agentic tools you’re using, the more simultaneous processes you’re running, the more tokens you consume.

Casey Newton: 45:16 One measurement I found useful was that apparently it takes about 10,000 tokens to generate 7,500 words. But as you just said, these coding tools can burn through millions of tokens in a session.

Kevin Roose: 45:54 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 was a guy who used 38 million tokens.

Casey Newton: 46:26 Now was this guy working on a new mass domestic surveillance program for the Department of Defense?

Kevin Roose: 46:31 I don’t know and OpenAI did not make him available for interview, 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 top tier of token consumption.

Casey Newton: 46:53 Yeah, well, okay, so tell us first of all just how expensive it is.

Kevin Roose: 46:57 Very expensive. In fact, I heard that the top user of Claude Code, the top individual user of Claude Code as measured by Anthropic, spent more than $150,000 on tokens last month. So extrapolate that.

Casey Newton: 47:51 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 Roose: 48:00 I talked to a software engineer in Sweden who said that he probably spends more than his salary on Claude. So this is essentially becoming like a very expensive job perk for some of these coders.

Casey Newton: 48:13 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 here anymore.

Kevin Roose: 48:29 Right. 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. There are executives at these companies who believe that the more AI their workers are using, the more productive they’re being.

Casey Newton: 49:14 And you’ve talked to a number of people who are ranking high on these leaderboards. What is your sense of how productive they actually are?

Kevin Roose: 49:33 I mean it’s very unclear, right? Some of these people may be just generating worthless projects. The thing that worries a lot of the people I talked to about these leaderboards is that this is just Goodhart’s law in action.

Casey Newton: 50:24 Yeah, I have to say when I read your column, I thought this just seems like it would create the worst incentives. There’s this idea of Goodhart’s law, right? Like when a measure becomes a target, it ceases to be a good measure. What are the people inside the company saying about that?

Kevin Roose: 50:49 Well, some of them are opposed to this whole leaderboard thing. I also talked with some folks who defended the leaderboards. They said look, it’s never been all that easy to track the productivity of software engineers. Some companies are now using AI token use and consumption as part of the performance review cycle. So you go in for your annual review, your boss says, “Hey, it looks like you only used 70 million tokens last month. What’s going on? Are you not using AI enough?”

Jasmine Sun: 51:23 Yeah, but I imagine that some of them are really nervous about that though, right? Because like it seems clear to me that at least some of these companies want to incentivize token usage because they want to figure out which jobs can be automated.

Kevin Roose: 51:42 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? These are people who, most of them have had the same workflow for years.

Jasmine Sun: 52:24 Yeah, I don’t know. 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?

Kevin Roose: 52:51 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.

Jasmine Sun: 52:57 They’re starting their new company.

Casey Newton: 52:59 They’re starting a new company with the boss’s money.

Jasmine Sun: 53:01 And if you’re doing that, I just want to say I salute you. Like that is the right way to work in 2024.

Kevin Roose: 53:06 Maybe don’t be the number one on the leaderboard if you’re doing that. Maybe try to stick around six or seven.

Kevin Roose: 53:25 No, I think that’s a bad idea. For all the reasons that we just talked about, including Goodhart’s law, which is I think this is just going to lead to people just wasting tokens, doing side projects, and not actually being more productive.

Casey Newton: 54:00 This leaderboard just represents a new incarnation of something that the software industry has been trying to figure out for a long time, which is how can I figure out if my software engineers are productive.

Kevin Roose: 55:17 Yeah, I think it’s gonna be pretty soon, in part because the budgets are just getting very ridiculous.

Casey Newton: 55:34 You know, maybe the last question I have for you about this is just what implications do you think it has for the broader economy?

Kevin Roose: 55:58 I hope not. I think it’s really a bad move, not just for tracking actual productivity and output, but just for morale. I remember years ago when Gawker would have like a traffic leaderboard.

Jasmine Sun: 56:30 Well, I have to say, I worry that this idea of token maxing is going to spread into the broader economy. I was talking with somebody who works in marketing and she was telling me that her job used to be evaluated solely on creativity and then recently the performance review got a new AI section and everyone is being evaluated on how much AI they’re using.

Casey Newton: 57:39 Yeah, I think it’s going to be very case by case. I think there will be people who are token maxing, who are way more productive than their colleagues.

Kevin Roose: 58:13 Yeah, well I will say on the flip side, I’ve also heard of people in my social circle who have gotten in trouble for spending too much on Claude.

Casey Newton: 58:21 Wait, really?

Kevin Roose: 58:21 Yeah. And when I heard that I was like, oh like your company’s not going to make it, bro. Like you got to spend on this stuff.

Casey Newton: 58:25 Well what’s so interesting is now it’s becoming part of job conversations for engineering jobs. People are going into new job conversations and saying, well what’s my token budget? And for the employers that are paying attention, that’s actually a really good sign.

Kevin Roose: 58:47 Yeah, I mean those sound like real incentives and better than the ones at Meta. Do you remember when Meta was spinning up a superintelligence lab and they said you can sit really close to Mark Zuckerberg?

Casey Newton: 58:50 If I were them, I’d be like, I’ll take the tokens, thanks.

Jasmine Sun: 58:51 All right, well just to wrap this up, exactly how many tokens should a person use?

Casey Newton: 58:56 Uh, I think you have to look within yourself.

Kevin Roose: 58:59 That’s between you and your God. Do what Marc Andreessen will not and introspect.