My Tech Workers Have Unionized, 2026.10.05

Alex Hanna: Welcome, everyone, to Mystery AI Hype Theater 3000, where we seek catharsis in this age of AI hype. We find the worst of it and pop it with the sharpest needles we can find. Emily M. Bender: Along the way, we learn to always read the footnotes, and each time we think we've reached peak AI hype, the summit of Bullshit Mountain, we discover there's worse to come. I'm Emily M. Bender, professor of linguistics at the University of Washington. Alex Hanna: And I'm Alex Hanna, director of research for the Distributed AI Research Institute. This is episode 87, which we're recording on October 5th of 2026. And we're joined this week by JS Tan and Clarissa Redwine. They are the authors of a new book, Against Tech Oligarchy: Worker Resistance in the World's Most Powerful Industry. Emily M. Bender: JS is getting his PhD at MIT and formerly worked in tech, where he spent years organizing across the industry. He writes the newsletter Value Added, and his work can be found in The New York Times, The Guardian, and many other outlets. Alex Hanna: And Clarissa helped organize the Kickstarter union and has since dedicated herself to organizing the tech industry. She also co-founded and runs the annual tech organizing conference, Circuit Breakers, which I will be at later this month. Welcome to the show, both of you. So excited to have y'all. Clarissa Redwine: Thank you. We're excited to be here. Yeah. Longtime fans. JS Tan: Longtime fans. Very excited. Thank you. Emily M. Bender: We're gonna have some fun. Alex Hanna: Yeah. So this week, as you might expect, we're going to be focusing on tech workers and the labor that makes this industry run. So let's get into it. Emily M. Bender: All right, here comes our first artifact. This is from the Financial Times, July 30th of 2026, with the sticker Amazon.com, by Rafe Rosner-Udden. And the headline is, "Amazon finds cases of AI causing runaway spending on tech projects." Subhead, "E-commerce giant's staff said the new technology had caused budget overruns that sometimes took months to detect." Alex Hanna: That's hilarious. Emily M. Bender: Yeah. We're gonna save all this money by having you automate your work instead, right? Alex Hanna: Yes. So the first two paragraphs, "Amazon staff have identified cases of what they called, quote, 'catastrophically expensive,' end quote, cost overruns caused by mistakes in how the company has deployed AI and a lack of spending controls. Senior engineers told colleagues in a staff meeting on Tuesday that attempts to switch tasks from conventional programming to using AI models had led to, quote, 'unplanned spending,' said multiple people familiar with the matter. Engineers were working to create, quote, 'automated guardrails for future products,' amongst other measures to keep a lid on costs, the people said." All right. Thoughts on that, Clarissa and JS? Clarissa Redwine: This doesn't even count the wasted labor of engineers and other workers trying to fix the poor work, the poor output of AI that they're being forced to use. So the cost, I would assume, is actually much higher. JS Tan: Yeah, and we've heard of other cases, just yesterday speaking to a tech worker who came to one of our events, he was telling us that this kind of drive to automate engineering workflows and stuff like that has been happening throughout his company. But what they're finding is that they automate a bunch of stuff, they do a layoff, and then very quickly they realize that, "Oh shit, a lot of our actual domain knowledge has gone." And then they go back and try to hire some of the people. Emily M. Bender: All right, so let me get this right. The recipe here is, here's something that outputs code. We think our engineers output code. Let's just have the machine do it instead of the engineers. And, doesn't matter how much it costs to have the machine do it 'cause that's better somehow than having a person do it. So let's get rid of those people and, oh, oops, engineers do more than output code. Alex Hanna: Yeah. And that's the sort of thing, and I appreciate this, this thing y'all bring up. There's so much of this embedded knowledge that is contained with different types of engineers in this role. And one of the things y'all make as a point in your book, which I think is a really underappreciated point, is that so much of engineering work is doing automation. And so there's some ways where engineers can undercut some of the work they're doing. But to suggest that there's a one-to-one replacement really just misses all the points of all the different work that engineers do, and all the different work that people in the tech industry do that are really embedded type of knowledges. Clarissa Redwine: Yeah, absolutely. Emily M. Bender: I want a little more commentary on that third paragraph. So the idea of having automated guardrails to keep a lid on costs, if you were designing a way to avoid accidental overspend on tokenmaxxing that was more worker-centered, what would you propose? Clarissa Redwine: Oh my god. The projects that they're working on, not sure that those are necessary to begin with, so maybe cut some of those. JS Tan: Yeah, I think one of the big problems is that when Claude Code and these agentic systems started to blow up earlier this year, all these companies rushed to push this upon their workforce, and there are just so many cases that we've heard of where management started instituting these leaderboards to see who was spending the most tokens, and then later who was doing the most GitHub PRs- or, how do I say that non-technically?- GitHub merge requests into the code base. And that obviously messed up the incentives. And so I think people were just incentivized to just churn out a bunch of crap code, to hit those KPIs, and I think now going back on that and asking people to cut back on costs and keep the lid on costs is forcing them to now completely work in this other way. Alex Hanna: Absolutely. So let's see- we are on the fourth paragraph. So, "The episode highlights the difficulty that even the biggest tech companies are having integrating AI into their daily operations without overspending." And this quote is hilarious. So quote, "'It's difficult to figure out how much anything [AI related] costs,' one senior Amazon employee told the Financial Times." Sorry, I just think that's so funny. It's like, I don't know how much we're bleeding here. And I think the thing that's very funny about this is it's not just the idea that automation is going to do this work, it's that they're also just, they don't know how to do accounting for this work, too. It's just become so enmeshed into the organization. I'm wondering if y'all have thoughts on that. JS Tan: Yeah. One of the big problems I think is that, historically it's just been so hard for management to track, to monitor what software engineers or other tech workers do. And that's partly because, I think these companies rely oftentimes on their employees to do the work of innovating, to do that creative work. And if you're a manager and you're seeing the engineer who works for you staring at his or her terminal screen, you don't know if they are just zoning out for the day, or if they're coming up with a great new idea or a new algorithm. And so there's always been this really tricky task for management to figure out, how do I get these incentives in place to get folks to really be doing productive work? And I think a lot of that has actually changed over the past couple of years. Starting in 2022 when the interest rate went up, that was a moment in which the balance of power tilted very much so in the favor of employers. And that was a moment in which tech workers themselves lost a lot of labor market power. And so that, I think, was the first kind of period in which tech bosses started to wanna do things like increase monitoring or increase surveillance or put more and more people on PIPs. And I think that has only increased over the past year or two, especially now that these agentic coding systems are around. One of the things that we hear all the time speaking to people is that, at these companies, software engineers, they're finding that their bosses want to use AI systems as part of, as a replacement for more traditional forms of bureaucratic control. So being able to use AI to more minutely measure output or in some ways measure productivity. But I think even still, even with more capabilities in surveilling and controlling their workforce, there's still this really tough problem of, how do you set up the labor relations in a tech sector so that tech workers are actually productive, without completely controlling their time or completely controlling their day-to-day. Clarissa Redwine: Yeah, and crowding out the creativity that you rely on. And they're doing that in really ham-fisted ways. I'm sure y'all remember one of the first instances of this was Elon Musk asking engineers to print out code and judge based on how much code they produce. Yeah, it's a problem. Emily M. Bender: Yeah. There's something I wanna lift up from the chat here, and then I wanna share an old memory with you of my days in the tech workforce. HelleWalt says, "Now I get it. The AI writes code that nobody understands so that the AI cannot get fired." Which is pretty funny, but also speaks to the sort of institutional knowledge that is so important. But I worked for a brief stint at a startup that was doing automated customer service response based on natural language processing, but actually grammar-driven natural language processing. And my job was to expand the grammar for Japanese so that we could expand to the Japanese market. And as the startup was falling apart in 2001 into 2002, we got a new CEO installed by the board who really had no clue, and he went around to, meet everybody and see what we were doing. And he stopped by at my workstation where I had something open. I was gonna show him the Japanese grammar, and he was just completely enthralled with the front-end processor that allowed me to type in Japanese. And I'm like, "This isn't our product." But that's what he was curious about. I'm like, "You realize that's not the thing I'm making?" So funny. Alex Hanna: Yeah, I think it's also, like you write about in the book, 'cause y'all were organizing at Twitter, and then Elon Musk took over. And so Musk decided to, he had people print out the lines of code. But also what he did was like, let's fire a bunch of people and then see if the service stays up. And so this is very much trying to optimize using a very blunt instrument, rather than trying to do this type of scientific management that so many of these companies have been trying to do and control with AI and worker surveillance and whatnot. Clarissa Redwine: Yeah. The interesting thing about tech companies is you can do this bizarre raid of value from the company when you can lay off a significant number of workers and the platform keeps running for a bit. It may break in some areas, it may degrade, but it pretty much keeps running. And that's what we saw Elon take a bet on, this big thing that he bought, can he get rid of 80% of the workers in pretty much a short period of time and keep this money-making machine up and running? And it did, and that's something we talk about in the book is how that affects labor stoppage power of workers in tech. JS Tan: Yeah, and I think, obviously we're a bit biased having written this book, but I do think the plight and the struggle of tech workers is really worth paying attention to, especially for other professional workers who are interested in labor organizing. The day-to-day work of tech workers, for a long time, has been to automate for their employers the generation of profits. Whether you're a software engineer or even whether you're a designer, a data scientist, a product manager, all the labor that you're doing on a day-to-day basis is to create these autonomous platforms such that the workers can step away from their desk, and the company still be able to make money from that. And I think with the rise of tools like Claude Code and Codex, that kind of dynamic is actually gonna spread across the white-collar workforce. Today, let's say a lawyer... i'm gonna make this up a little bit 'cause I don't know exactly what a lawyer's work is, but their work is to, perhaps, review documents and provide insights on a day-to-day basis on whatever it is that they are reading. But moving forward, I could very much see a future in which increasingly they'll be pressured to create systems to automate that analysis, and be able to step away from their desks on a day, and have that continue without them. And so I think the question that we raise in our book is, given that this is the kind of day-to-day work, what is the leverage that tech workers, or more broadly professional workers, have in pushing back against their bosses, and actually stopping, attacking the profit centers of their employers? Emily M. Bender: Yeah, I think it's an interesting cautionary tale. I am actually very skeptical about the long-term sustainability of software projects built where the code isn't written by people. 'Cause you've got that, and when you move over to something like legal services, the work is not just the words, just like the work is not just the code, and it's gonna fall apart, I think, even faster. There's a quote in the chat, or a comment in the chat here that's really apropos from LozzaHouse, who says, "One of the things I find so bizarre is that CFOs, SLT, privacy lawyers, et cetera, are usually incredibly cautious and restrictive. But when it comes to implementing synthetic media generators and agentic systems, they lose all their boundaries and faculties." And it's like lawyers, for example, when the first lawyer got in trouble for using ChatGPT to extrude something that looked like a brief, I thought, okay, you could see how someone could fall for the marketing here. But when the second one got in trouble for that, I was like, "Don't y'all gossip?" Like, how is that not just career-ending to make that mistake? But so the point of, like, if management thinks that this can be automated, then how do we organize workers to push back against that? Clarissa Redwine: Yeah. Something we talk about all the time is there's also opportunity in this. The more slop that is injected into these platforms, the more maintenance, the more bugs that we'll find. And maybe this will create an opportunity for sites to go down, and for workers to refuse to fix them. And that's a new strategy for labor stoppage that might be in the next chapter of the tech worker movement Alex Hanna: Yeah. There's a few really hilarious incidents in this article that I'd love to read. So one of them, it says- so after that senior employer says it's difficult to estimate these things, they give a few different incidences. So one reads, "During a presentation this week, employees were told about one instance in which the company spent $1.8 million on matching author details with listings on the group's e-commerce website using Anthropic's Claude Sonnet AI model, despite the deployment failing, the people said." Emily M. Bender: Wait a second. This feels like a problem for a regular expression. Alex Hanna: Yeah, and I skimmed down in the comments in the article, and someone was like, "Wow, $1.8 million for five months of clerical work. I'm available for a summer internship." Yeah. When you're literally just matching things or doing it. And, "The spending represented an 860% cost overrun compared to the project's deadline. It took five months to detect, they added." Wow. That would, in any other case, be cause for a product manager to be fired and laughed out of the industry. And then "Amazon said, quote, ' As with any new technology, we're experimenting, learning, and improving how we use it, including how we drive cost efficiencies.'" Emily M. Bender: Okay, so why are heads not rolling with these ridiculous cost overruns? Do you have any insight into that? Clarissa Redwine: Yeah, I think all of these initiatives are being handed down from the top, and I think the executives at these companies are closing rank. So you know, if anyone's head is rolling, it's probably done quietly, in the lower parts of the company. JS Tan: Yeah, and I think it's also worth stepping back and seeing that this is actually happening in a financial environment that's actually quite different from what we had even 10 years ago. The ZIRP, the zero interest rate policy era, is over, and interest rates are now much higher than they were. And what tech CEOs learned, and what they found early in 2020, 2022, 2023, was that when they fire employees, they almost always get a bump in the stock, in the share price. And in fact, because of the way tech work operates, oftentimes they can fire employees without their platforms immediately breaking. And so I think now with AI obviously the big thing in the discourse, in the tech sector broadly, when a CEO comes out and says, "We're pushing through this automation, we're gonna cut costs on employees," that's gonna be very exciting to shareholders. And so I think that's that's the thing that they're forcing through, and we're seeing that across the industry. Clarissa Redwine: Yeah, I think they're, as the statement says, they're willing to take losses in order to continue showing that they are part of the AI race. They're still in it. They're willing to sink all these resources in order to get ahead. Emily M. Bender: This feels a bit like the "just a flesh wound" Monty Python sketch. Clarissa Redwine: Yeah, totally. Alex Hanna: Yeah. It's just a flesh wound! Clarissa Redwine: Flesh wound all the way down. Alex Hanna: Yeah. abstract_tesseract also has a good note in the chat where he says, "Plus, executives get credit for starting initiatives, not finishing them." Clarissa Redwine: Exactly. Alex Hanna: Yeah, very much. You get credited for launches. You don't get credited for landings. Emily M. Bender: Yeah, and BoxoMcFoxo says, "The most sunk cost fallacy in history." Clarissa Redwine: So good. I was just gonna say that the playbook is you launch a big initiative, and then you move to your next company. You get the heck out before it has outcomes. Alex Hanna: Exactly. You're only interested in exits. You're not really interested in... being more invested in launches and not landings, that's like a mantra for SpaceX, I guess. Emily M. Bender: Oh, and I found the just a flesh wound quote here. "The company added, 'Cherry-picking small isolated examples where teams are learning from one another and portraying them as business as usual doesn't reflect how teams across Amazon are using AI.'" Alex Hanna: Yeah, and it's all obscurantism. They're obscuring a lot. Emily M. Bender: Yeah. I think we should move to Meta. So this is in Reuters, by Katie Paul, August 26 of 2026, and the headline is, "Mark Zuckerberg had a bold plan to replace Meta staff with AI. Here's how it imploded." Yes, go for it. Clarissa Redwine: Bold is interesting. Alex Hanna: Okay, the subhead says, "To make its workforce, quote, 'AI native,' Meta explored slashing the size of many teams across the company by as much as 60% in two waves, internal documents show. Staff revolted, and hours before a big layoff in May, the CEO called off planning for further cuts." And the banner image is like a sickly teal background, and then there's Mark Zuckerberg's head- Emily M. Bender: Wait, teal, teal the color, not Peter Thiel? Alex Hanna: Yeah. Teal the color. And it's, actually, I don't know if y'all know that famous coffee cup that has the kind of squiggly from the '90s. It's got that teal. Millennial kids will know what I'm talking about. And then it's got this magenta, like, splotchy version of Mark Zuckerberg's face, and he's got on Ray-Bans. Emily M. Bender: The pervert glasses. Alex Hanna: Yes, pervert glasses. One side says enter, the other's exit, and there's this flashing DDR arrow, Dance Dance Revolution arrow. And then there are people, there's bodies with the sparkly AI logo emoji walking into the glasses, and then there's people with heads on holding, like, banker's boxes walking out. It's a very vivid image, and I just wanted to try to actually describe it with my mouth. Emily M. Bender: Wait, there's one more part of this, which is that around the pink image of his head, there's little bits of lighter teal coming out that you could almost read as galaxy brain happening. Alex Hanna: Oh yeah, and EliotL calls them jazz cups, and then kwamevtaylor said, "Vaporwaveee." Yeah, it's very vapor wave. It's got the vibe. JS Tan: I guess what's happening with the people is that the AI people are walking into Mark Zuckerberg's face, and they're leaving, perhaps laid off, right? Is that- Clarissa Redwine: Yeah. Oh my god, yeah. Alex Hanna: Yeah. Or like, the people living in Mark Zuckerberg's head rent-free have been evicted. Anyway, fascinating image. Emily M. Bender: All right, so this first part describes the basic story. Do you wanna read or summarize for us, Alex? Alex Hanna: Yeah. We can just start, the first three paragraphs to just give you an idea of what's happening. So, "In January, Meta CEO Mark Zuckerberg and his top lieutenants gathered for their annual leadership retreat at his Hawaii compound." Fucking colonialist. "There, they hatched a radical plan to reimagine work at the social media giant in the age of artificial intelligence. Codenamed Project OT-" so not overtime- "short for Organization Transformation, the plan envisioned a, quote, 'AI native' future for the owner of Facebook and Instagram. AI would take over much of the daily work performed by thousands of human employees. Virtual workers would be overseen inside Meta by smaller, quote, 'talent dense-'" Emily M. Bender: That bugged me. That really bugged me. Alex Hanna: Yeah. "-cadres of human staffers, according to one internal planning document reviewed by Reuters and three people familiar with the project. In scenario planning exercises, two of these people said executives explored slashing the size of many teams across Meta by as much as 60%. Some employees would be offered roles in new units, while others would be laid off as part of a culling that one human resources executive predicted would be as big or bigger than any of the company's cuts of around 25% three years ago, according to another internal document. But on the night of May 19th, just hours before the first layoff wave, Zuckerberg blinked. Meta laid off 10% of its employees the next day, but it called off planning for the November cuts, according to one internal document reviewed by Reuters." Emily M. Bender: Yes. And then can we talk for a second about "talent dense"? Alex Hanna: 100%, we should talk about talent dense. Emily M. Bender: So, is this supposed to be, if you have lots and lots of people in the company, some are more talented than others, and so the talent density is low, but if you manage to lay off the less talented people, you've raised the talent density? Is that what's going on here? Alex Hanna: I think that is what's happening. And my reading is, the talent density means... talent dense is such a ridiculous phrase that I wish I never had to read. But my idea is that is they are ascribing the talentedness with probably, a lot of more senior people that they're going to end up saddling with loads more work. And so then they are basically saying, "We're going to cut off any kind of entry and do this kind of cutting off of these less talented hacks? I don't know. Something of that nature. That's the implication here. Emily M. Bender: So in the chat, iratelump says, "Laying off all the janitors because they aren't talented enough." And ndrwtylr says, "Increasing talent density by shrinking desk size." Clarissa Redwine: So funny. Yeah, this is actually pretty ironic because when management announced this plan to ingest employee data to train AI that could do those employees' jobs, when employees revolted, Mark Zuckerberg in a big meeting said, "Hey, the reason that we need your data is because you're smarter than everyone else." So it was like, a way to sell their own exploitation back to them. And now he's saying, "Actually, quite a few of you are gonna go. I'm sorry." So yeah, so bad. Emily M. Bender: Ah! All right, this document gets even more ridiculous though. So here, "The 'AI Native Playbook.'" And I just want to pause for a second on AI native. So there's the whole colonialist idea of who's a native, which I think maybe they aren't trying to index here, but they're trying to make the company "AI native" in the sense that it's somehow a new company that's just starting right now in the AI era, which, like- you can't, Meta. You're a giant. You've been around a while. You don't get to be reborn. That doesn't work. But this first paragraph, just hilarious. "Ever since the release of OpenAI's ChatGPT in late 2022, Silicon Valley leaders have been imagining the revolutionary possibilities that generative AI might unleash for the future of work. They have been especially intrigued by the promise of, quote, 'agentic capabilities.' Agents, a nascent technology, are being designed to do more than just spit out answers to queries like typical AI chatbots. Instead, they are supposed to take autonomous actions such as shopping, booking travel, and creating new apps." So the paragraph, it starts with, we're gonna be able to do our work better with software that does shopping and booking travel. Alex Hanna: And I think this is- I'd love for y'all to speak to this too, Clarissa and JS, where so many of the examples, especially of AI agents, are things that are done by executive assistants. Basically, it's this work that's very feminized and seen to be relatively easy. And I'd love for you, in your experience too, on talking about how you've seen this in tech worker organizing and how the nature of that work also can be used to split a lot of organizing. I'm curious if y'all have seen that, too. Clarissa Redwine: Yeah, I remember when I first saw some of these early artificial intelligence products coming out, even back in 2016, where there were, like, email agents that would help set up your calendar. Pull something from your email and put a calendar event. Those were very feminized. The names of those products were always "Susie" or something like that. and yeah, I think it very much mirrors the kinds of divisions and systemic racism and sexism that are employed in our workplace to justify more exploitation for some groups rather than others. And that is something that is challenging to overcome in an organizing context in tech companies, because workers that are making 300, even like 900K, may want different things than their counterparts in maybe contract world, where there are data labelers that are making a lot less and really looking for bread-and-butter compensation. And overcoming those divisions has always been a challenge that workers have to overcome in order to build real leverage in the workplace. So yeah, it's something we see all the time. JS Tan: And I think right now, going back to this talent dense stuff, right? There is this kind of clear attempt, whether it's within a single company or within the industry more broadly, to stratify and polarize the existing tech workforce. You're either on the AI train and you're spinning up agents and you're using these harnesses, or you're not. Clarissa Redwine: Or you're a Luddite in the pejorative sense. JS Tan: Exactly. And you're being shut out, and I think that is what we're seeing. We are seeing not just one body of tech workers. We're seeing now two different groups of tech workers, those who are racing to stay in the game and who believe in things like the permanent underclass. And we're seeing another group. If the first group is mainly concentrated in SF and in the Bay Area, the tech workers across the rest of the country I think are now adopting a very different kind of politics, right? They're recognizing themselves as workers. They're embracing a politics of more affordability. And I think, increasingly, they're more open to the idea of class struggle and unionization. Clarissa Redwine: Yeah. But back to your original comment, Alex, it's very interesting for us to see, through the years of the tech worker movement, precarity has been used against workers to create divisions, to, say, "Hey, you better keep your head down or you're gonna be next in the layoffs." And that actually helped spur a lot of organizing in the tech industry. Because the more precarious groups, the groups that wanted to push back against military contracts or working with oil companies, knew that they were the most at risk for retaliation and had to join forces with other groups with more privilege and more structural power in the workplace. Yeah, so it actually spurred a lot. And what's interesting now is precarity is being weaponized against some of the most highly paid workers with the most individual leverage in the workplace. So that's a new shift that is causing even more solidarity across these shops. Alex Hanna: Yeah. No, that's a great point, too. And I think it's helpful to expose a bit of the cracks of what types of things are seen as okay to automate. And so when it was things that looked like more contract work or more feminized labor, the most highly paid folks were like, "That seems like it's easy." But now it's... of course, it wasn't easy, it never is. And now that their jobs are up for automation, or are seen to be at least on the chopping block, now they're like, "Oh, okay. This is something that we really need to band against." Emily M. Bender: Yeah. And I guess there's probably a little bit of a hurdle for some of these workers to see the expertise of their colleagues who are not in, or historically the more precarious positions, and to see that there is more power to be had by joining forces rather than sitting in their local optimum, relatively powerful position. What kinds of things do you see help people over that hump? Clarissa Redwine: Yeah. So I'll let JS jump in, but something that we try to stress in our book, and we're really trying to make a case for in our book, is the act of participating in class conflict where workers are actively trying to win material gains from their boss is very instructive for a lot of workers, because they have this clarifying moment where not only do they see the mask come off of their own bosses and their boss say, "Actually, no, we're not gonna protect the mission. We're not a family. Meritocracy doesn't exist, and I'm here to dominate you." That always crops up in these moments. But also, workers will often discover that their power is limited. So even these most, very highly paid workers understand that in order to actually win material gains, they have to join forces with their coworkers across the company, and sometimes their coworkers have a better understanding of how to wield their leverage than they do. For instance, Amazon warehouse workers know that if they get everyone to walk out, that is shutting down that warehouse. Whereas a lot of software engineers are like, "If I walk out for a day, I don't know how much would go down." So it's essential to engage in these moments of class conflict in order to learn where our leverage is and build solidarity. JS Tan: Yeah. Absolutely. I wanted to just zoom out and actually, I think part of this drive to- especially from people like Mark Zuckerberg and these other big tech companies- this drive to automate the work of their employees, I think a large part of it is about control, right? From the previous FT article, we see clearly it's not just about cutting costs. It's about control. And I think if we were to go back into the history of the tech worker movement, what a lot of bosses experienced between the years of 2018 to 2022 were employees organizing and pushing back against their decisions as bosses. And initially, these bosses, given the sort of political climate, were, they ceded to some of the employee demands, or they doubled down on trying to be these progressive CEOs. But ultimately, what they realized was that, forget the carrot. I'm gonna take out the stick, and I'm done with my employees telling me how to run my business and what contracts I can take. And I just imagine these tech bosses are drooling at the idea that they can have this agentic workforce that will never organize and push back against their decisions. Alex Hanna: Yeah, for sure. And there's a lot of terrible things in this article. Like first, one of the executives uses a basketball analogy, which I don't think actually makes sense. And which I take personally right now. It's WNBA playoffs. Let's go, Valks. But then there's also a section I would really love us to just chat about the employee resistance. This is a little further down in the article, in which people start to push back. So they talk a little bit about the tracking software. So it says, "Reuters also reported in April that Meta had mandated that tracking software be installed on US employees' devices to capture their keystrokes and mouse clicks to teach its AI agents to replicate how humans interact with computers. Believing they may be training their own AI replacements and irate about the lack of detail about the layoffs, many employees flooded Meta's in-house communications network called Workplace with angry screeds and gallows humor." And so then there's some petitions. There's a picture of a petition, like in a washroom, that says, "Want Meta to stop collecting employee data to feed their AI models? Sign the petition here." And then there's a QR code, and then there's some memes that folks are using. So I know y'all are longtime organizers, and I'm wondering if you could speak a little bit to that, and also different kind of strategies that workers have used, the resistance to these types of efforts and other efforts by companies to establish this kind of control. Clarissa Redwine: Yeah, one of the things that we see in tech worker organizing quite a bit is sometimes, agitation can come from the feeling of being exploited, of course. But also in tech especially, there's this craftsmanship element where tech workers really don't like to see things being built haphazardly or inefficiently. You know, that's part of the culture a little bit. And the first thing that happened at Meta whenever this got out, that management was trying to ingest all this data and train AI, was the security aspect of it. Workers were like, "Hey, has this gone through all of the security checks, all the processes?" And management made basically no attempt to unpack whether or not they had gone through these processes. They said, "Yeah, sure, we did it," but it didn't make sense with the timeline of how these processes take place. So first off, all of their employees were very concerned about their data being fed into this untested process. And that was one of the first cracks that really motivated workers to start posting in Workplace. And I think workers really sprung into action and created this petition that they put out online, and I think a lot of us were thinking, "Oh, this is gonna get maybe a couple hundred signatures at best over maybe a month." And I think, maybe overnight, or within a couple days, it had thousands. Yeah, that was a show. But I don't think that the petition itself could be the only explanation for why management not only paused- or said that they paused layoffs, we don't know if they're coming yet. I think the workers will have to be vigilant. But they did pause this AI program, where they were ingesting data. They no longer are feeding data into that. And part of that was what you were mentioning in this article where workers were openly asking, peppering management with questions, being critical, and also flyering in the office. And I think the real point of pressure that management was feeling was actually seeing some of their most quote-unquote, "valuable" groups of workers starting to organize. And yeah, especially the workers that were drafted to shape this AI model, which were high performers from across the company. When they started hearing them organize, I think that was a wake-up call. Alex Hanna: Yeah. Wow. Emily M. Bender: I don't think that I realized the extent to which Meta thought that they could create fake engineers doing this. But it would also make sense that engineers would find this viscerally offensive. Clarissa Redwine: Yeah, of course. And they did. And I think not just offensive that you think you can replace me, but also this idea that you didn't ask for consent to take my data. This is not okay. It's not part of the terms of my employment. Yeah, all of that was very insulting. And to top it off, after they announced they wanted to be like AI native, AI first, workers, in a big call with the CTO said, "Hey, how about we think about how this productivity can serve the company, serve company culture? How about we think about the four-day work week?" And their CTO didn't just say no to the four-day work week. He was like, "Hey, why don't you go ask your parents if that's a good career move to ask your boss for time off?" Just very insulting. Alex Hanna: What the fuck? Clarissa Redwine: Yep. And this was like, to an individual worker, to the entire company. JS Tan: Yeah, I think that's actually that sentence there, "Meta Chief Technology Officer Andrew Bosworth clashed with some employees." I think that's what that's about. Clarissa Redwine: Yep. Yeah, he was DM'ing people. It was very inappropriate. Alex Hanna: That's wild. JS Tan: It's just so crazy, because you hear the kind of rhetoric coming from the bosses and how AI is gonna bring about so much productivity that we'll be living in abundance, and at the first glimmer of actually bringing AI into a company, that's actually how the bosses respond. Clarissa Redwine: Yeah, like their first chance prompted by their own employees that are in open rebellion. They're like, "Hey, how about you give us this crumb?" And they're like, "Actually, no." It's just such bad signal. Alex Hanna: Actually, why don't you go ask your parents? 'Cause I guess their parents were maybe... Actually, that's such a confusing statement. Did they think that their parents were like, miners? Like living in pre-Taft-Hartley Act, were their parents just getting shot by Pinkertons? Or like, what is that reference to? Clarissa Redwine: Only when they ask for time off, Alex. Come on. That's the lesson. Alex Hanna: Only when they ask for... Yeah. Emily M. Bender: So, iratelump says, "Carrot costs too much and needs to be replaced too often. We've replaced the carrot with another stick." Clarissa Redwine: Chew on this. Alex Hanna: I wanna read one more thing before we- so this is a callback to an essay that we read two episodes ago in episode 85. So at the very end it says, "In his essay on the future of AI titled 'The Future Is for Everyone,' Zuckerberg predicts that there could be, quote, 'an abundance of jobs in the future.' But some companies might end up with fewer employees. Quote, 'Company sizes may shrink just as they did in the transition from industrial giants to tech companies,' he writes. 'But this doesn't mean fewer jobs overall. It implies a larger number of companies with fewer people each.'" And so it's just such a nice coda to this, 'cause he's effectively like, "Actually, industrial capitalism leads to more competition?" Question mark. Emily M. Bender: But we're gonna maintain a monopoly here, yeah. Alex Hanna: And don't dare you ask for more workers' rights or we'll fucking kill you. Emily M. Bender: And don't go crying to your mommy about it. Clarissa Redwine: Yeah. Part of the argument internally for workers has always been, including up to today from Meta, "Hey, if we don't do this, someone else will, and then we won't be on top anymore, and that'll be bad for you." That is always part of the argument. Alex Hanna: Yeah. And the point has always been to have monopoly. This is the Paul Graham, Y Combinator view. You just gotta do it. Emily M. Bender: So Alex, I haven't cast you as a demon in Fresh AI Hell for a while, so I think it's time for you to- And we may have done this one once before, but, a reprise anyway. You are a demon organizer in Fresh AI Hell, leading the demon workers on a strike. We can, if you're leading us in chants, we can chant with you if you want. Alex Hanna: Okay. I don't know if I could have a chant, but maybe I'm handing out flyers. I'm like, "Hey, we're rebelling against the AI training. I need you to, I need you here." And I'm handing out what is actually an infernal soul, and all it does is it jitters your mouse. Clarissa Redwine: Oh, excellent. Alex Hanna: "Here, put this next to your mouse. And if they ask you to type backwards like usual, like the name of Beelzebub, try typing forwards. It's gonna really throw them off." That's my organizing spiel to other AI demons. Clarissa Redwine: You got us. The demons are on board. We're all going, "Demon power." Alex Hanna: Demon power, yeah. Skin flayers local 666. Emily M. Bender: Okay. Alex, you can have this one 'cause I wanna land on the Seattle City Council one. Alex Hanna: Totally. So this is a quote tweet. The original quote is from 404 Media journalist Joseph Cox, and says, "New: ICE plans to spend millions of dollars on Boston Dynamics dog robots. ICE says this is to, quote, 'improve officer safety.' The robots can open doors, pick up objects, have 360 ca-" 360 cameras? That's so- I thought it was, originally read that as 360 degrees, but literally 360 cameras? That's obnoxious. "Infrared vision, et cetera. Boston Dynamics didn't respond to a question on its stance on ICE." And then Charles Logan, friend of the pod, he's been on the pod before, says, "School leaders still bringing in these robot dogs from companies like Boston Dynamics and Ghost Robotics for STEM lessons should (finally) reconsider." Emily M. Bender: Yeah. Yikes. Okay. Next we have some reporting from 404 Media, Jason Koebler this time, September 17th, 2026, and the headline is, quote, "Doom Loop: OpenAI and Microsoft Admits LLMs are Destroying the Web and Built on Theft." And this is some of the documents that have come out in the ongoing lawsuit between The New York Times and these two. And there's some pretty damning quotes where it's the executives on the Microsoft and OpenAI side saying things like, "An astonishing theft of unprecedented proportions," and, quote, "The largest theft of labor in human history." Alex Hanna: Which, chattel slavery is likely what that is, but that is a pretty damning statement that they're saying this shit at all. Emily M. Bender: Yeah. Oof. Clarissa Redwine: Oh, their stock price just went up. Emily M. Bender: Yeah. 'Cause they've been so successful at this biggest theft ever. All right, Alex? Alex Hanna: Lots of 404 Media today. So this is another 404 Media, also Jason Koebler from September 30th. "Someone, quote, 'torturing' LLMs in a robot prison has triggered the dumbest debate in AI yet." And so this is, basically what the guy did is he developed this, quote-unquote, "experiment" of a GitHub project based on some kind of experiment from a paper on pain. And so the- Emily M. Bender: But it wasn't a paper on pain, it was a paper on... The first paper was actually someone found pain vectors in the models and was, like, manipulating them. Alex Hanna: Oh, I thought it was originally, somebody was taking this from, actually, a psychology paper. Emily M. Bender: No, it was a model welfare paper. Alex Hanna: So, model welfare paper. And decided to basically torture the LLMs. And so then all the Effective Altruists and people who think there's AI consciousness were like, "This is so unethical. You need to take this GitHub project down." Just the most mind-numbing kind of thing you could imagine in this kind of debate, that just, this shit doesn't matter. Emily M. Bender: And talk about organized effort being directed at something completely pointless. Alex Hanna: Yeah, and LozzaHouse says, "Say I'm in pain. 'I'm in pain.' Oh my god." Emily M. Bender: Coliarmauve says, "It's a fractally dumb story." All right. So that's just silly, but this next thing is sad. So this is from GeekWire, which is a local publication here in Seattle. The journalist is Todd Bishop from September 21st of 2026, and the headline is, "Former Seattle City Council president launches AI-powered site to track and report on City Hall." And so, "Sara Nelson, former Seattle City Council president, is making her debut as a publisher and commentator with an eye towards boosting civic engagement with a little help from AI agents." So the site is basically "reporting" in quotes on what's going on inside Seattle City Council. So if the work were really being done, it'd be combing through Seattle City Council documents and making this information accessible to people. But she's getting drafts from an LLM first and then revising, editing, and approving them. And I am just so angry that somebody thinks that this is a contribution to civic discourse to give us papier-mâché made up out of City Council documents. And furthermore, she's very centrist, so it's like, "Everything else has spin. I'm gonna give you the unbiased view," which... Alex Hanna: That's really unfortunate. And it's really a terrible kind of thing where there's just this huge gap in local reporting, and you can imagine how stuff like the Google program to aid civic journalism could hold this up and say, "Look at what this person is doing!" when it is just not actually holding. And it's really actually frustrating too, because the quote from her says, "I'm trying to solve for the hassle of civic engagement." What the fuck do you think civic engagement is, my friend? Emily M. Bender: And "solve for" is such tech speak. Alex Hanna: Yeah. And then the rest of the paragraph, the rest of her quote says, she describes it as an effort to help create, quote, "More accountable, accessible governance." Ay, yi yi. Terrible. Emily M. Bender: All right. And then in the chaser spot we have something really fun. Alex, you wanna do the honors? Alex Hanna: So this is from the Seattle version of The Onion, which is called The Needling. And what is the subhead on this? Seattle's Only Real Fake News. Very good. And so the title of it is, "New AI Literacy Program Awards Students Free Pizza If They Don't Know How To Read Books Anymore." And it's got a picture of a sad personal pan pizza, and it has the Book It, like, a play on the Book It little sticker or pin. I remember the pin. I had the pin and it was holographic. Lots of '90s references today. And it has a picture of a book with a flame on it, and it says, "Ask AI." And the first paragraph says, "Thanks to a generous new grant from Microsoft and the Gates Foundation, a new AI literacy program is awarding Seattle students a free personal pan pizza after showing teachers they don't know how to read a single book anymore." Emily M. Bender: I just love the dogging on the Gates Foundation there. And how completely sad this pizza looks. It's like there's five pieces of pepperoni, but they're not evenly distributed. Alex Hanna: Oh, and abstract_tesseract actually perfected this. And it says, "Bookn't." Don't, but book not. It's like- Sorry, that's harder to say, but it looks correct. Emily M. Bender: Excellent visual humor. Alex Hanna: All right. Cool. That's it for this week. Clarissa Redwine and JS Tan are the authors of Against Tech Oligarchy: Worker Resistance in the World's Most Powerful Industry, which is available now from Haymarket Books. Thanks so much again for joining us. Clarissa Redwine: Thanks so much. It was so great to be on the show. JS Tan: Yeah, thank you. Super fun. Emily M. Bender: So nice to have you. Alex Hanna: Our theme song is by Toby Menon. Graphic design by Naomi Pleasure-Park. Production by Ozzy Llinas Goodman. And thanks as always to the Distributed AI Research Institute. If you like this show, you can support us in so many ways. Order The AI Con at thecon.ai or wherever you get your books, or request it at your local library. Emily M. Bender: But wait, there's more. Rate and review us on your podcast app, subscribe to the Mystery AI Hype Theater 3000 newsletter on Buttondown for more anti-hype analysis, or donate to DAIR at dair-institute.org. You can find our merch store there, too. That's dair-institute.org. You can find video versions of our podcast episodes on Peertube, and you can watch and comment on the show while it's happening live on our Twitch stream. That's twitch.tv/dair_institute. Again, that's dair_institute. I'm Emily M. Bender. Alex Hanna: And I'm Alex Hanna. Stay out of AI Hell, y'all. And sign your union card! Skinflayers Local 666, fighting for you.