On a Thursday evening earlier this month, an announcement went out to the 1,500 or so workers on the data annotation team at xAI. They were told to stop whatever they were working on and complete a series of diagnostic exams as a part of a sudden, team-wide reorganization to prioritize “specialist AI tutors” over “generalist” roles.
By Friday morning, according to Business Insider, the workers submitted their tests, which included categories like STEM, coding, finance, media, AI safety, Grok “personality and behavior,” and — this being an Elon Musk-led company — “shitposting and doomscrolling.” By Friday evening a third of them were fired.
The same evening termination notices went out, the official xAI account posted on X: “We will immediately surge our Specialist AI tutor team by 10x!”
The unusual move was not a standard cost-cutting layoff. Instead, it is an instance of what workforce data analyst Amanda Goodall calls a “sort-and-cull” model of workforce reduction, something she predicts will become a trend throughout the tech industry.
“Other companies are going to think, Wow, that’s really smart,” Goodall said. “It’s a lot less HR management, like they’re not having to deal with PIPs” — performance improvement plans — “or things like that. So I would not be surprised if we see that happening more often.”
The restructuring at xAI tells a larger story about the burden placed on tech workers in a moment where AI investment is booming, mass layoffs are commonplace, and more negotiating power is concentrated in corporate hands.
One part of that story is that “specialization” has become an industry watchword, particularly among software engineers whose field is increasingly being offshored, contracted out, and automated. Sometimes this automation is being done by the engineers themselves via a trendy new practice known as “vibe coding,” where human developers prompt AI to generate code and provide guidance for product development. If AI can write code, guide product development, create prototypes, and more, the human working with it had better be really, really skilled in something AI can’t do.
“Specialization is becoming the new buzzword,” Goodall said, citing how traditional ways of advancing through white-collar fields, like higher degrees, are losing value as AI shakes up the tech industry. “We’re seeing PhD holders laid off because they don’t fit where the company is going. Even top performers are vulnerable if their function is deemed replaceable.”
Goodall said many tech workers are beginning to think of “specialization [as] the new job security,” but predicting exactly which special skills will have staying power in today's employer's market is difficult. As businesses continue to figure out exactly what AI should be used for, corporate priorities are especially capricious. You could be the top shitposter/doomscroller one day and still find yourself locked out of Slack the next.
Kwaku Ofori-Atta, a machine learning engineer at the San Francisco-based media analytics startup Swayable, said entry-level software engineers will likely feel the need to market their special skills most strongly.
Eight years into his career, Ofori-Atta said that he has seen an “eroding tolerance” for waiting for new hires to learn the ropes, as well as a decline in training and mentorship for young developers.
“I think the best profile is being a former entrepreneur, where you’ve tried to build software end-to-end,” he said. “That’s more valuable than just being smart and green.”
Of course, building software end-to-end would require being a “specialist” in multiple areas; in other words, a generalist.
“Everyone says be specialized, but what companies mean is, be specialized in vastly different things, and then you end up a generalist,” Ofori-Atta said. “Generalist, specialist, those labels don’t actually mean a lot.”
In the current climate, it’s no surprise tech workers are hoping for new ways to protect themselves. Of the half-dozen or so software engineers I reached out to for this story, the majority declined to speak with me, some saying they had signed NDAs and others expressing fear of hurting their job prospects.
Earlier this year, Microsoft announced it would invest $80 billion in AI followed soon after by several rounds of layoffs, the most recent of which was earlier this month. Last month, McKinsey slashed roughly 5,000 roles from its workforce while activating thousands of AI agents. The same is happening at Salesforce, Oracle, Shopify, and Duolingo, the last of which boasted no full-time employee layoffs since it became an “AI first” company but has culled many of its contractors.
The current landscape for tech workers is grim, but there is nothing inherently illegal about those layoffs, nor about the xAI sort-and-cull event. Still, Shannon Liss-Riordan, a labor attorney known for bringing class-action lawsuits against Uber, Lyft, and Twitter after it was bought and hollowed out by Musk, said such moves can open companies up to discrimination cases.
“If whatever criteria are used [to decide who gets laid off] results in a disparate impact on protected categories of employees, such as women or employees of color, that could create a legal claim,” Liss-Riordan told Gazetteer SF.
Liss-Riordan has sued Musk multiple times for unfair labor practices, including stiffing canvassers who had worked for Musk’s America PAC before the 2024 election and allegedly withholding severance pay from laid-off Twitter employees in 2022.
Liss-Riordan said no lawsuits from former xAI staffers have come across her desk so far. If they do, she said her team will perform what is called a disparate impact analysis, which assesses whether ostensibly neutral policies, like testing workers on how good they are at, say, shitposting, created outsized negative impacts on workers in protected categories.
Liss-Riordan said she cannot speculate on any discriminatory nature of the diagnostic exams distributed before the xAI layoffs; she can only draw on what she’s seen in the past.
“Let’s just say that I don’t believe that Twitter ran any disparate impact studies before its mass layoffs when Elon bought the company,” she said.
Representatives from xAI did not immediately respond to Gazetteer SF’s inquiries into whether the company conducted a disparate impact analysis before distributing the diagnostic exams or whether laid-off xAI employees are indeed being paid through November 30 or the end of their contracts, as the termination notices stated.
Liss-Riordan said that the xAI layoffs have happened at a moment when labor power is on the decline, in large part due to President Donald Trump’s dismantling of the National Labor Relations Board, which began shortly after he took office in January. Organizers say this only makes the need for coordinated labor efforts more urgent.
"Organizing with coworkers is the most important thing that tech workers can do to improve their job security,” Stephen McMurtry, a senior software engineer at Google and the communications chair of the Alphabet Workers Union-Communication Workers of America, said in an email statement to Gazetteer SF.
The tech industry has never had a strong union presence within its workforce, and so far unionization efforts have proven easy to quash. Last week, more than 200 contractors working on Google’s AI products were fired without warning amid unionization efforts.
Hoping to protect their livelihoods, tech workers are left with few options. Specialization may ultimately be a phantom lever of labor power, but at the moment, it is one of the last ones left for workers to bet on.
“I call it the ‘corporate revolution,’” Goodall, the workforce data analyst, said. “You always think of revolutions as the people rising up, but it’s actually the opposite. Corporations are changing, and they’re changing so fast. We’re not ready for that, but with competition out there, there’s nothing you can say but, Okay, and do what the company says if you want the job.”
