It’s also a tough time to be the lucky few who got one.

Salesforce is among the tech supergiants to lay off thousands in the last year, mostly in its customer service sector. Its latest round of cuts happened this month. They’ve promised to hire up to a thousand “AI-native” graduates, but entry-level jobs on their engineering and sales team are extremely competitive.

Three junior Salesforce employees, speaking on the condition of anonymity, described being one of the very few to jump straight from college into a highly-coveted tech job, from the view of one of its top companies. And, worth mentioning, the tallest skyscraper in San Francisco. One of the employees described the Salesforce Tower as “pretty awesome.” New hires are brought up to the top floor on their first day.

They all described their entry as gradual, welcoming, and exciting, while inevitably intimidating. They are highly aware that their jobs are commodities, and describe a work culture where they are, too.

“The mentality is, it’s a privilege to be here, so you work your ass off,” one said. “Any time we meet someone from higher up, they’re really excited to see us at the beginning of our careers, starting at Salesforce.”

Share

Their first month was a long onboarding process that tapered out gradually. Junior employees are introduced to as many coworkers and supervisors as possible, from all departments. They felt an effort on the part of the company to set them up for success.

From a warm welcome, new employees must hit the ground running. The learning curve is sharp and constant, as are reminders that their jobs are only as secure as their performance.

“What surprised me about my job so far is how rapidly things change,” said a junior employee on the sales team. “I’d say we have, in our sales process, an internal change in what our day-to-day looks like every two months or so. Because of AI, things are changing so rapidly. You wouldn’t see this kind of innovation, probably, three years ago. The sales world looks different every three months.”

They worked to get the hang of things quickly, but how their jobs are done can change month to month. Job insecurity is a constant source of stress. Keeping their jobs can feel as competitive as it was to get them in the first place.

“The pace is just so fast. We are ready to make shit happen very quickly,” said another. “The amount of uncertainty there can be, given this rise of AI, even though Salesforce is at the forefront of developing a lot of AI there, too. It’s definitely been interesting to see what switches in the workforce suit AI. Sometimes being a little worried, like, ‘Our jobs are going to get replaced—oh, no they’re not, we’re good.’”

“Aside from AI replacing our jobs, that stuff, Salesforce is a large corporation. It can be intense. If you don’t hit your numbers, you’re out. That’s definitely a big stressor there. Especially in the first six months, where you’re kind of ramping up, it was definitely a big source of stress, having to know I need to hit my numbers this month or I may not have a job next month. You want to be one of the ones that makes it through.”

“They fire lots of people,” one said bluntly.

Subscribe now

The anxiety evoked by the heightened stakes of entry-level Tech work reflects wider concern from researchers of AI in the workforce. Molly Kinder, a Brookings Institution Senior Fellow researching the impact of AI on work and workers, has said the career ladder as we know it will likely change as entry-level jobs are replaced by automation; in a 2024 Bloomberg essay, she warned that “If AI takes over the work typically done by junior employees, the basic logic of white-collar apprenticeship — tedious work in exchange for valuable experience — will break down.” An outcome of hiring fewer recent graduates is the future loss of experienced workers.

The anxiety to keep one of the few slots that guarantees experience coexists with excitement. The pressure to perform and outperform yourself and others came with the job. The internal self-awareness of Salesforce being a global superpower raises the stakes constantly.

They all said the pressure, instability, and constant change entry-level employees navigate could have easily created a toxic culture that pitted them against each other. Instead, all three gushed about the closeness and warmth within Salesforce’s youngest ranks. They rely on each other in difficult professional moments. They enjoy spending time together outside of work. Their favorite part of their jobs are each other.

“We’re all really tight-knit,” one said. “It’s a really competitive job and that trauma bonds us.”

How they are expected to do their jobs is always changing. How they are expected to perform is not. They paint a picture of an entry-level culture that is a knife fight in a phone booth, but highly aware the pressure is a privilege.

“It can be a very difficult job at times, very stressful, but we all lean on each other,” said a member of the sales team. “Sometimes when the job sucks, at least we have each other. We’re going through it together.”

In totality, the new CBA is a victory for the WNBPA, which staved off a last-minute lockout, and delivered on many of its members’ top demands. But these sorts of negotiations inevitably involve compromises and trade-offs, and as far as I can tell, the league gained the upper-hand in one crucial, underreported section of the agreement centering on wearables devices. The language in the “wearables” section of the WNBA’s CBA is interesting for what it hints at, for what it doesn’t say—and what it portends for other pro sports leagues that are heading back to the negotiating table in the next half-decade.

Setting the Stakes

A decade ago, just 13% of surveyed Americans owned a wearable—something like an Apple Watch, an Oura Ring, a WHOOP strap, or another health-monitoring device. Now, nearly half of Americans own at least one wearable, according to a recent survey. North American pro sports leagues have attempted to embrace the wearables craze, though their access has been limited by players’ unions.

Wearables present serious privacy issues for “Average Joe” consumers, who are entrusting tech companies to safely store and protect their biometric data. Imagine the stakes for a professional athlete, whose entire livelihood could be affected by a single biometric data point. To give one of many realistic hypotheticals: a basketball player has a terrible game, and the coach wonders if they showed up to the gym hungover. The coach has access to the player’s wearable data, and checks to see when they went to sleep, as well as what their heart rate looked like during the night. Should the player have been out partying before a game? No. Should the coach be able to surveil them? Definitely not.

It will not surprise you to learn that there’s an emergent gambling angle here: sports leagues would love to commercialize players’ biometric data, and sharp bettors would love access to data about, say, a hungover player. “We’re going to get to a spot where people are betting not just on the velocity of the puck that was shot by a player in the NHL playoffs, but on what the heart rate of a certain player is going to be running down the field,” said Helen “Nellie” Drew, the director of the University of Buffalo’s Center for the Advancement of Sport, and a professor of practice in sports law.

There are other practical considerations, too. What if wearable data reveals that a player isn’t as speedy as they were before, and a team uses that data against the player during contract negotiations? What if a wearable reveals a player is favoring their leg, or is at greater risk of injury? This information is potentially beneficial to a training staff and an athlete, so long as it’s disclosed and used in a responsible manner—a critical, mostly unresolved caveat. “Aging and injured players are the most at-risk” of wearable data being used against them, said Michael LeRoy, who researches sports labor laws and AI, and is a professor at the University of Illinois’s School of Labor and Employment Relations.

How the WNBA’s Wearables Agreement Stacks Up

Like the NBA and MLB—two leagues where the players tend to have more leverage than their counterparts in other pro sports—the WNBA and its players association have agreed to form a committee specifically tasked with reviewing and approving wearable devices. The committee will have two league representatives, two WNBPA representatives, and a neutral party appointed by both sides (five members total). The committee “will discuss, among other things, training for players and staff, data security, access and retention, use of derivative data, and emerging technologies,” the CBA reads.

That’s where the similarities between the WNBA, the NBA, and MLB come to an end. The NBA’s bargaining agreement includes the specific names of approved wearables. It’s unambiguous that unless the CBA is amended, wearables can’t be used during games. And it definitively states player data can’t be released to the public or used for commercial purposes. In other words, approved wearables are only for practices and training. A team that invokes biometric data during contract negotiations is subject to a $250,000 fine. (I will note that this is nowhere near sufficient, given the financial stakes between multimillionaire athletes and franchises valued in the billions, but it is something.)

MLB’s CBA does allow a selection of wearables to be worn during games, but on a voluntary basis. Biometric data is only shareable with a small cohort of coaches and staffers, and players can revoke access to the data without penalty. To date, the MLBPA has banned the commercialization of wearables data.

The WNBA, meanwhile, is allowing players to use “approved” wearables during games on a “voluntary” basis effective immediately, but it doesn’t identify what those wearables are, or who approved them. (The WNBA’s future-tense phrasing about a Wearables Committee strongly implies it’s not up-and-running yet.)

There are zero details about how the data from those wearables can be shared. Are coaches and staffers allowed to access it? Can players say no, or revoke access? The CBA doesn’t say. It does note that wearable data cannot be referenced during contract negotiations, but there are no safeguards or penalties mentioned. “The provisions that say the data cannot be used to negotiate contracts—I kind of chuckled at that,” Drew said. “Baloney.”

The WNBA’s CBA clarifies that “commercial use of the data collected from an approved Wearable… will require prior approval of the Players Association.” There’s an important clause at the end, though: the WNBPA’s approval “shall not be unreasonably withheld.”

Typically, that could be viewed as a standard hedge preventing the players’ association from stonewalling all wearables-related topics until the expiration of the current CBA. In this case, the WNBA has a trump card. Beginning in 2028, the league can require the usage of approved wearables. It’s not hard to see how the league might tighten the screws on its players come 2028: we’re mandating that you wear these devices, so why not commercialize your data with a group licensing agreement, something that at least entitles you to extra cash in exchange for your biometric data?

In fairness, the NFL already requires its athletes to use certain wearables. The NBA G League is trending in that direction too. And the NCAA is a totally unregulated mess; the best it’s offered are recommendations for how schools should (and shouldn’t) monitor college athletes. Viewed within that context, LeRoy doesn’t think the WNBPA should be in total panic mode. “I can say with confidence that what the WNBA is doing [with wearables] is not a novelty—it’s part of a growing trend,” he said. “I view this provision as an agreement to negotiate further… CBAs are like constitutions. They’re amended.”

But LeRoy conceded that it’s notable the WNBA hasn’t released any information about its “approved” wearables to date. In fact, the WNBA and WNBPA did not respond to my interview requests and a detailed list of questions I sent over, including:

• The timeline for launching the “wearables committee.”

• Which wearables are already approved, and who approved them.

• The process that WNBA players must undergo to alert their team and/or the league that they are going to use an approved wearable.

• Approximately how many players are using approved wearables.

• If there are any penalties in place if a team is proven to have used biometric data against a player during contract negotiations.

• What sort of biometric data the league wants to commercialize.

• Why the league wants to require wearables during games beginning in 2028.

These are very basic, very straightforward inquiries. I can only assume the WNBA and WNBPA don’t have solid answers, and have decided to punt on a full-fledged wearables debate. “They were in a hurry to get the deal done,” Drew said. “Nobody wanted to see the season suffer, and so there was great pressure to conclude the negotiations with a win. This may have just been something that was left in the dust as they raced to the finish line.”

The problem—a problem that’s currently unique to WNBA players—is that the clock has already restarted on a potential point of tension with the league. Unless the wearables committee solves a bunch of fairly existential issues in short succession, this fight will start anew before (or during) the 2028 season. “I think that’s remarkably short-sighted,” Drew said of what appears to be a mutually half-baked approach to wearable tech.

A drama-free outcome, at least in the short-term, is that the WNBA and WNBPA sign off on a commercialization deal with a wearables partner or two (or three). The players choose to take the bag—an understandable choice in a vacuum, especially for a younger league where the salaries are just now hitting six- and seven-figures. But the longer-term implications of that arrangement are risky, to say the least, especially for a group of athletes still adjusting to an enormous spike in visibility.

“The idea that somebody gains more by handing over their personal data, I just question how that’s going to work,” LeRoy said. “I do not see this as an upside for any individual athlete. I see this as an inherently unfair mechanism for generating wealth and control for the league.”

Musk can take over and ruin Twitter, the only social media platform that sometimes had intelligible discourse. He can rebrand it as “X,” which makes no sense, but fulfills a decades-old obsession with the letter itself. He can boast about how, under his leadership, X’s revenues will someday quintuple. (Hasn’t happened, won’t happen, doesn’t matter.) He can do a Nazi-looking salute and then say “no I didn’t.” He can post racially-charged screeds for hours, and then writers like me contort and soften our descriptors of Musk’s actions and language, because, as I already mentioned, he’s able to sue anyone, anytime.

Musk can father an unlimited number of children, and disown them and/or ignore them when he feels like it. He can essentially call Donald Trump a pedophile, allege that the president is in the Epstein Files, then end up in the Epstein Files himself—a series of events that resulted in zero professional repercussions and in fact ended with Musk becoming the president’s friend again. (They just traveled to China together!)

Musk can employ teenage minions to tear down life-saving government agencies, and then evade subpoenas and depositions that would compel him to answer for DOGE’s actions. He can hand out cash to American voters, despite cut-and-dry laws against election bribery, and face, at most, a civil lawsuit—nothing criminal. He can hoard money in his alleged charity, and not disperse the minimum amount required by law, year after year.

Musk can integrate Grok, a shitty AI chatbot, into his revamped, borderline-unusable social media platform, thus making X an even more unpleasant user experience. His AI company, xAI, can utterly fail to employ safety guardrails that would otherwise prevent Grok from producing nonconsensual deepfake porn. Even after a slew of lawsuits and investigations about the deepfake porn, Musk’s Grok can simply continue generating the exact same sexualized, abusive content. Just less than before.

Musk can also make outlandish claims about his businesses and what they actually offer regular people. He can boldly posture as though he’s inventing the concept of a tunnel, hype up 155-mile-per-hour autonomous vehicles that zoom past traffic underneath the Las Vegas Strip, and ultimately deliver a tiny-ass “Vegas Loop” where human drivers ferry passengers to one of a handful of drop-off points at 35 miles per hour. Undeterred by this obvious failure, Musk can wishcast about building a California “Hyperloop” system that “would be a technological marvel exceeding any high speed rail on Earth,” and people will conclude that he’s onto something.

Musk can similarly overpromise and underdeliver on his prized Tesla cars, which have design flaws that are alleged to have severely injured and killed people. Musk has been touting the potential of self-driving Teslas for a decade, and longtime owners are still waiting for him to deliver. Same with Tesla robotaxis. As Bloomberg recently chronicled, in July 2016, Musk foresaw “complete autonomy” for his robotaxis “in less than two years.” That, obviously, has not happened. Tesla’s “robotaxis” fleet reportedly totals 59 vehicles, despite Musk’s prior prediction that there would be 500 robotaxis in the city of Austin alone by the end of 2025. Who cares! Tesla’s stock is up year-over-year.

On top of everything else, Musk can fantasize about traveling to Mars, colonizing Mars, whatevering on Mars—and throw out various timelines for space conquest that do not make any sense—and supposed experts will respond by saying that he is “innovative.” Musk’s extraterrestrial ambitions are gibberish. He merged xAI into SpaceX, now his biggest company, because he can. He says SpaceX is going to put factories on the moon and AI data centers into space. I wish I could say, “Heyo, if you believe that, I have a spacebridge to sell you.” Alas, SpaceX just set a record IPO price: $135 a share. It is valued at $1.77 trillion, which is also a record.

As a result of SpaceX’s valuation, Musk is a paper trillionaire as of Thursday afternoon. I have seen some dispiriting statistics about what one trillion dollars of wealth represents. For instance, from The Guardian: “To a trillionaire, $100 million feels like $19.27 to the median American.” According to USA Today, if you tried to stack $1 bills until you reached $1 trillion in cash, that stack would be 67,866 miles high. I saw another statistic that Musk could buy every single NFL and NBA franchise and still have plenty of money left over.

Will SpaceX ever regularly turn a profit? Will it accomplish even some of its stated goals? These are reasonable questions for businesses that are not run by Elon Musk. The power of Musk compels retail investors, some of whom seem to have chased after SpaceX stocks specifically because of the company’s farfetched business model. If Musk endorses it, then that’s good enough. He is a walking, talking, self-fulfilling prophecy.

It is evident that Musk’s hubris is only outpaced by his luck, which is now, frankly, outpaced by his wealth. Are there any limits to his sway? I…don’t know. Musk wanted to become a paper trillionaire, and now he is one. Because he can.

Here’s what else we’re reading this week:

• Polymarket is technically not legal in the United States. And yet, according to a new study cited by Wired, roughly 30% of trading volume on the platform—tens of billions of dollars—comes from the U.S.

• In an effort to pretend it’s monitoring prediction markets like Polymarket, the Commodity Futures Trading Commission (CFTC) proposed loose parameters this week that would make it easier for the CFTC to regulate the most egregious forms of sports bets and “virtually all bets on war, terrorism or assassinations,” the Wall Street Journal reported. In theory, that’s a welcome step, but as always with the Trump-led CFTC, the devil is in the details (or lack thereof). The WSJ added that the CFTC isn’t suggesting an “outright ban trading on any specific types of so-called event contracts.” It’s just outlining “factors that regulators will use to review certain types of contracts on a case-by-case basis.”

• The San Francisco Standard published an autopsy about San Jose Mayor Matt Mahan’s disastrous run for California governor. It’s sourced to nearly a dozen people, and details how Mahan was squashed like a bug by a weak slate of opponents. Big Tech founders and CEOs talked up Mahan like he was the steady, reasonable choice to oversee America’s most populous state, but Mahan couldn’t keep tabs on his own campaign staff, let alone 39 million Californians. Internally, there was reportedly “bitter staff infighting, an overreliance on wealthy tech donors, and a fundraising strategy so badly timed it left the coffers nearly empty when voters finally started paying attention at the close of the race,” wrote Hannah Wiley.

• Anthropic CEO Dario Amodei gave a pathetic interview to Bloomberg, during which he admitted he has no clue whether Anthropic’s AI model played a role in the missile strikes that killed 120 Iranian schoolchildren earlier this year. The strikes almost certainly came from the U.S. military, which is taking its sweet time “investigating” the matter. “The principle that we have established, and I think the principle that was obeyed here, is a human makes the final decision,” Amodei said. “I don’t know what role Claude or any other AI had, but if this isn’t an illustration why that principle is so important, I don’t know what is.” Amodei and the U.S. military are establishing a nice little feedback loop of plausible deniability. If AI assists in the massacring of children, the AI company can blame the human who pressed the “kill” button; alternatively, the human can claim that they only pushed the “kill” button because the AI tools erred and provided mistaken intel.

• Here’s an interesting New York Times essay from John O’Farrell, a former general partner at Andreessen Horowitz, about how “the most powerful players in A.I.—led by some of my friends and former partners, to my great sadness—have raised hundreds of millions of dollars to forestall a more serious and meaningful debate about how A.I. should be governed.”

When OpenAI and Anthropic go public, a few thousand residents will become multimillionaires overnight. They’ll look to spend it in San Francisco. Their first big purchases will be real estate.

Vulnerable and working-class residents, already struggling with exponential increases in rent this decade, will find themselves repeatedly outbid, or evicted if landlords can justify it, to open supply for a sudden higher-paying demand.

Subscribe now

We know this because it has happened before. The last time the Bay saw this kind of IPO (Initial Public Offering) fever was Twitter and Facebook’s class, and the resulting spike in prices and evictions triggered protests. The Anti-Eviction Mapping Project has tracked nearly 6,000 Ellis Act evictions alone since the start of the first dot-com boom in the mid 1990s. Housing prices climbed steadily alongside evictions.

San Francisco, home to the Anthropic global headquarters in the SoMa neighborhood and OpenAI in Mission Bay, is the epicenter of a human-history defining technology. AI has already nestled into San Francisco culture; a ten-minute drive in any direction will pass billboards advertising new products and companies, and updates from the tech world are casual conversation.

As the world’s largest AI companies announce their plans to go public, San Francisco and its residents will be the first to find out what a culture and economy defined by AI investment looks like. They won’t be the last.

This month’s announcements from Anthropic and OpenAI suggest the companies are planning their IPOs later this year; SpaceX is set to go public on Friday. Both companies have said they are nearing valuations of $1 trillion — OpenAI estimated a post-money valuation of $852 billion in March, and Anthropic estimated $965 billion in May. SpaceX will debut with an unbelievable valuation of $1.77 trillion, with the potential to give 4,400 current and former employees millionaire status. The SF Standard estimates Anthropic to have more than 1,300 local employees; while OpenAI’s local headcount was about 2,000 in 2025. Their transition to public companies is expected to be a landmark moment for AI investing.

Share

When their companies become public, those who own pre-IPO stock will see its value explode. If thousands of newly-minted tech magnates rush to outbid each other for San Francisco’s housing and rentals—as they have in past tech booms—the hundreds of thousands of other residents will see an already volatile and competitive market become even more exclusive and impossible.

This has already begun. Business Insider reported that San Francisco’s housing market, already the most expensive in the country, has seen values rise faster than any other major city in the past year.

Moreso, much of that new wealth will not be taxed. IPO millionaires simply borrow against the value of their stock to buy property, cars, or more stock, and even if they do sell stock eventually, it’s taxed as capital gains, not income—which means a much lower tax level.

When the IPOs do happen, most of the wealth influx will not be local, but its most concentrated effects will be. As the Standard reports: “The bulk of the money from these public offerings won’t go to the local workers — it’ll be redistributed to investors around the world. But ripple effects of the payouts to equity-holding staff and executives will be felt in the city where thousands of AI employees work and live.”

Share Hard Reset

Tenants rights attorney and former Supervisor Dean Preston — the only Democratic Socialist on the San Francisco Board of Supervisors at the time — says rental rates in San Francisco have already seen a 30% increase in the last couple years as AI companies established themselves locally, and with their IPOs, the increase will sharpen, and accentuate the working-class housing crisis.

Economic inequality inevitably affects culture, only one example being legacy businesses making the choice between pricing out their longtime loyal customers or shutting their doors. When economic diversity decreases, diversity of all kinds decreases. The culture that makes San Francisco uniquely appealing to tech staffers in the first place is endangered.

“You talk to any union in San Francisco, for example, about their membership, and there’s an increasing number of members who can’t afford to live in the city,” Preston said. “It affects every aspect of a city’s culture, decreases diversity in a city, and it puts incredible pressure on folks who are either pushed out of their homes or find themselves in an extremely stressful situation where landlords tend to get more aggressive in no fault evictions trying to replace lower rent tenants with higher rent tenants.”

As the AI era follows in the footsteps of the dot-com boom two decades ago, history will repeat itself for both residential and commercial San Francisco renters.

“You’ll have a relatively small group of people enjoying incredible wealth,” Preston said, “And then you can expect to see the Mayor and the majority of the Board of Supervisors cheering that on as a positive thing for San Francisco. Meanwhile, it will drive up costs for people in San Francisco and create more economic inequality in a city that is already dealing with a very high level of economic inequality. It is one of the major problems of San Francisco.”

This problem is not uniquely San Franciscan. How it affects San Francisco will not be unique, either; it will be a precedent.

In this age of extreme division, disliking data centers actually seems to be a unifying belief, across age, gender, and ideology. While opposition has been percolating in the last year or two, cities and states are taking concrete action in the last few days to impose moratoriums. And tech companies are catching on to this optics problem: just yesterday, Meta went on the offensive by announcing a data center buildout “jobs program,” presumably an attempt to counter the reality that data centers don’t even create that many local jobs, a typically reliable business narrative.

To-date, the tech companies’ data center initiatives have been primarily geared to appeal to stock prices and shareholders, prioritizing speed and scale above all, while committing local officials to tight NDAs to obscure the sausage-making of building and permitting.

As a result, the hyper-growth stories have focused on overcoming challenges to data center growth, to show each tech behemoth as a true contender in the race. While data centers have historically been powered by the grid, new centers plan to sustain themselves with natural gas. Google was noted by the WSJ a few days ago for its strategy to “secure its own power supply” (although experts point to this natural gas strategy leading to higher electricity prices across the region.) Meanwhile, in Ohio, Meta is now holding chips in tents, an attempt to cut the time to build compute in half—described by one reporter as “like a scene out of the movie Mad Max.”

And regulators around the world have responded in different ways—some with moratorium legislation, others cloaking its tacit support in lazy “economic imperative” language. In Europe, there is concern over the European Commission’s latest tech sovereignty package, which, in the words of local watchdogs, “rolls out the red carpet” for Big Tech.

Pierre Terras at Friends of the Earth and Beyond Fossil Fuels, a non-profit fighting for a renewables energy sector, told me that an “impressive game of regulation and lobbying” from U.S. tech, including amendments written by companies like Microsoft, have led to the Commission paving the way for data centers to have what they want at the expense of the environment.

When asked why the Commission failed to protect the public from overreach of Big Tech here, Terras responded that the “so-called AI race” is “seen as the North Star of this act instead of the European economy, climate objectives and strategic autonomy.” Regulators, he continued, are giving a “blank check to big tech, and have blindly believed in the narrative that if Europe does not adopt the path of the U.S., the continent will be drawn into economic decay.”

Yet in the U.S., data center opposition movements have been heating up in the last few days, driven by grassroots and community activism—perhaps catalyzing new narratives around “jobs” and not just resource hoarding and hyper-scale. In Hays County, Texas, activists are pushing for delays to data center buildouts, citing negative impact to people’s drinking water supply. In Ohio, one bill would end tax breaks on data centers, while Kansas City is exploring a data center moratorium, as are Nashville and Charlotte (which actually approved a 150-day moratorium).

Meanwhile in Wyoming, residents are revolting against plans to build a “man camp” to house thousands of out-of-state workers flocking to the region to build massive data centers. And in New York, a one-year data center moratorium bill has already passed the House and Senate and is heading to Governor Kathy Hochul’s desk as we speak. If passed, New York would be the first state to impose such a moratorium.

The flurry of activity in New York comes after a 500 megawatt, $19.46 billion data center proposal—now being pushed by private equity firm Apollo Global Management as its flagship data center project—faced relentless campaigning and opposition from hundreds of local community groups, including faith, climate, environmental justice, and consumer protection groups.

While governors in Maine and Vermont vetoed similar bills intended to pump the brakes on the ravenous goals of data center developers, Ryan Howard, Indigenous Solidarity Director at NY Renews, pointed me to the backlash against Maine Governor Janet Mills after she vetoed a moratorium as a reason that might deter Hochul from following the same path. They also referred me to a recent Buffalo newspaper quote from Governor Hochul, where she said that New York communities need to benefit “much more” from data centers, and that the “status quo can’t continue.”

That much remains true. Faceless data centers gobbling up power, water, and land without proving any benefit to real people is quickly becoming one of the environmental and moral imperatives of our time. While legislators and regulators may acquiesce to the increasingly tired “AI is inevitable” narrative, the tech companies don’t seem to have much more in the form of narrative persuasion—which is likely why Meta is suddenly promoting a “jobs program.” And as they flounder to find a reason beyond money to convince people that they are a net good to the world, communities on-the-ground and resistance movements have proven that data centers may not be the only entities to heat things up.

Subscribe now

Last month I got part of an answer. More than 2,100 IT and technical workers across the University of California voted to join UPTE-CWA 9119, pushing the union’s tech unit to 8,400 people and making it the largest tech worker union in the country. These are the workers who keep UC’s research, hospitals, cybersecurity, and public data systems running. Infrastructure that millions of Californians touch without ever thinking about it.

I called Max Belasco, a business systems analyst at UCLA who helped lead the organizing drive. For him, it’s more than just a labor fight: It’s an argument over who and what technology is supposed to be for.

“We’re technologists first and foremost because we believe technology can help people,” he told me. “We think that it can make things better.”

That’s easily lost when people hear “tech workers” and “union” in the same breath. These aren’t people who hate the tools, they are the people who build them. What they object to is being handed AI from above by executives and consultants who’ve never touched the systems they’re reorganizing.

“Nobody that I’ve talked to is against new training in new technology,” Belasco said. “But what’s concerning is that there isn’t really clarity on what is the end goal. What’s the five-year to 10-year plan on how this technology is going to be implemented?”

For Max, it’s a question of who gets to ask those questions, and sit at the table where they’re being answered—not just at UC, but across the modern workforce.

“AI is being imposed unilaterally and considered inevitable by a very small group of people from the private sector in Silicon Valley. One of the concerns we have is that model of unilateral control and imposition being applied in the public sector.”

A union provides the legal, structural framework to change who’s in the room when those calls get made. The workers who just organized are building power over how these tools get deployed, and they’re the ones who actually know what the tools do.

“The people who are implementing these tech resources know how they impact frontline services,” Belasco said. “We know how they help support frontline staff, and we know the pros and cons in the way they are implemented, the limitations of these tools, as well as their benefits.”

That knowledge almost never counts for anything in tech. You can push back internally or you can quit, and neither one touches the decision itself. Bargaining over the systems is something most tech workers in the private sector simply can’t do. UC workers can now, at some of the most important medical and research facilities in the country.

“We’re a university,” Belasco said. “We’re supposed to be about public excellence. We’re about trying to figure out new ways to do stuff. UCLA was one of the birthplaces of the internet. I seriously doubt that a Deloitte consultant hired by the administration would have approved of a project like the internet.”

They want structural changes at their workplace, for sure, but Max and his coworkers have a vision for tech that could reshape the industry.

“One thing that everyone seems very excited about is the idea of why can’t we demonstrate a work model that’s different than the model that has dominated private sector tech and has brought us to this situation,” he said.

And for his private-sector colleagues:

“Now is the time to do this,” he said. “We will never have more leverage than what we have now to really assert authority and control.”

What it takes, he says, is people willing to stand together and reclaim something he describes as “a genuine curiosity and love for tinkering with things, making things work, and finding new, exciting ways we can implement this stuff to just have better lives.”

The cab drivers I wrote about organized too late, after the medallions were worthless and the industry was already gutted. The UC workers are moving while they still have leverage. Whether it amounts to anything depends on what the union can win at their jobsites, and on whether anyone else follows. But for once the people who understand the systems are the ones sitting at the table, asking the question the rest of the industry keeps dodging: not what AI can replace, but who gets to decide what it’s for.

The reason is that Anthropic’s Claude Enterprise plan—which is designed for businesses that want to integrate Claude Chat, Claude Code, and Claude Cowork—is now usage-based, rather than a flat-rate. Businesses loved the old subscription model, where Anthropic charged monthly for each licensed employee, with a set amount of AI usage per person (a rate limit). Businesses could pay more for a “premium” tier, which allowed for an even higher rate limit. But that system wasn’t sustainable—Anthropic was subsidizing too much of the cost per user, especially after Claude Code really took off. Recently, Anthropic started billing Enterprise clients based on AI usage, period. No more subsidies; it’s on each organization to establish and monitor their own usage caps.

Businesses have been caught off-guard by the billing change, as I wrote about last week. The chatter from CEOs and CFOs has been “oh shit, this isn’t good,” and “please, please, please stop tokenmaxxing, I beg of you.” (Tokens are units of text/data processed by AI models; as Anthropic explains, on top of a fee for each Enterprise user, “Every token your team consumes is billed separately at standard API rates.”)

So I was taken aback when I saw the University of Chicago’s June 2 announcement, penned by President Paul Alivisatos, that the school “is partnering with Anthropic to provide Claude Enterprise for all academics and staff starting in July, and for all students before the fall term.” UChicago has approximately 7,500 undergrads, 9,000 graduate students, 600-something postdocs, plus 20,000 staff and faculty. Alivisatos noted that students and staff are not required to use Claude Enterprise, and it’s safe to assume many people will choose not to indulge in these AI tools. But what happens if even a few thousand students and staffers become semi-prolific Claude users?

The timing of the announcement is even more bizarre when you factor in UChicago’s budget deficit, which has decreased but is still reportedly a gaudy $140 million. “Early deficit reductions came heavily from restricting spending growth, including early retirement incentives for faculty, hundreds of layoffs and a pause on doctoral admissions for many graduate programs,” according to the Hyde Park Herald.

How on earth, then, does a Claude Enterprise deal pencil out? I sent a series of questions to UChicago, and presumably was not the only reporter to inquire about what appears, at first glance, to be a fiscally ruinous decision. The university has since published an FAQ page about its Claude Enterprise integration—I’d describe it as moderately helpful in a very roundabout way. From the FAQ:

Are there usage limits, or additional costs associated with usage?

The University’s Claude account will provide generous usage for everyday work. Standard seats offer approximately twice the usage limit of a personal Claude Pro plan. Premium seats, which are comparable to a Claude Max plan, offer more usage limits and are intended for heavier users working in Claude Code. Limits will reset on a rolling basis, and actual capacity varies with the length and complexity of work.

(The “premium seats” aren’t free, by the way; the FAQ says that “upgraded premium access will be available at a reduced cost.”)

Best as I can tell, UChicago is not subject to the unlimited, token-based billing system that Anthropic is now requiring of other organizations. It appears UChicago has signed up for a capped, hybrid setup—something akin to how Anthropic sold its Claude Enterprise program to businesses until recently. One can only assume that Anthropic is subsidizing a chunk of the costs for UChicago. I don’t know for certain, because a UChicago spokesperson responded to my questions by referring me back to the above FAQ.

I additionally asked about the financial parameters of this arrangement, and pointed out that other universities have disclosed details about their AI partnerships. California State University, for instance, partnered with OpenAI for $16.9 million, and recently announced an extension for $13 million per year. That partnership, while also controversial, reportedly included 500,000 licenses (the CSU system is much larger than the total UChicago population). The UChicago spokesperson wrote back: “On cost—we will provide further details to the university community soon.”

That is not a very reassuring answer, especially for UChicago faculty warily eyeing the aftershocks of early retirement packages and layoffs. The best-case scenario, I suppose, is that this deal is “discounted” and Anthropic is metering Claude Enterprise usage, which eliminates the possibility of an army of undergrads “vibecoding” their eyeballs off until UChicago’s budget deficit quadruples in a month’s time. But even if that’s the case, it doesn’t mean UChicago is making a prudent financial decision. What are the terms? How much is UChicago paying Anthropic for an essentially subscription-based version of Claude Enterprise? And where is the money coming from? There have been passing mentions in news reports about a previous $50 million donation to UChicago that was earmarked for AI research, but no evidence yet that the donation will help fund Claude Enterprise.

As for Anthropic: this is bad business. It’s not as if their new strategy of token-based billing for Claude Enterprise is popular. Quite the opposite, in fact. Anthropic made the switch because it had no choice. No matter how much money Anthropic is getting from UChicago, it won’t be nearly enough.

But that’s not the point, is my guess. Anthropic is striking deals with universities for the same reason that Google cornered the market on K-12 schools and passed out its products like candy: the actual goal is to acquire lifelong users. The more young adults you can get to embrace Claude, the better. The difference, though, is that Google was already incredibly profitable when it gambled on the K-12 “market.” Anthropic certainly talks a big talk, but its medium-term future—let alone its ability to cultivate lifelong users—is far from assured. It’s making a risky bet here. The University of Chicago is too. So it goes in the age of AI.

Here’s what else we’re reading this week:

• More Anthropic news, from the Financial Times: “Anthropic is helping the US National Security Agency deploy its powerful Mythos AI model for offensive cyber operations, embedding engineers inside the agency.” Interesting that “offensive cyber operations” aren’t part of Anthropic CEO Dario Amodei’s arbitrary red lines.

• Seven in 10 Americans oppose data centers being built near their homes, according to a new poll from Heatmap Pro. In Monterey Park, a city that’s part of Los Angeles County, residents voted overwhelmingly in favor of a ballot measure that forbids the construction of data centers within city limits.

• Still lots of votes to tally, but Silicon Valley’s favorite mayor, Matt Mahan, insta-conceded the California governor’s race on Tuesday night. His campaign resembled a Blue Origin rocket launch. As of Thursday afternoon, he’s got 4.1% of the vote, which puts him in sixth place. Accounting for endless Big Tech funding and a lackluster slate of competitors, Mahan’s performance is especially embarrassing. In Santa Clara County, which includes San Jose (where Mahan is the mayor), he’s currently pulling in just 10% of the vote.

• Speaking of Santa Clara: Rep. Ro Khanna cruised to a primary win, despite some uninspired saber-rattling from Y Combinator President Garry Tan, who backed entrepreneur Ethan Agarwal. As of Thursday afternoon, Khanna has 58.9% of the vote. Agarwal has 6.2%.

• President Trump issued a long-rumored AI executive order on Tuesday. It is watered down and functionally useless. In short: it asks large AI companies to voluntarily submit their new products for government review, a process intended to “gauge what threats the products may pose to sensitive financial, national security and other computer systems,” Politico reported. Unsurprisingly, tech executives are thrilled with Trump’s hands-off approach. Congrats to them, and to David Sacks.

• OpenAI was compelled to clarify that it does not currently have an employee-funded Political Action Committee, and as an organization, does not donate to political candidates. The company’s president, Greg Brockman, has provided truckloads of cash to Leading the Future, a super PAC that advocates against AI regulation.

• Kevin O’Leary, aka Mr. Wonderful, aka Man Who Coasted Off His Marty Supreme Cameo For Longer Than He Deserved, has graciously offered to halve the size of his controversial data center project in Utah. His new plan would require roughly 20,000 acres of land, which is larger than the square mileage of Manhattan.

As a result, the algorithms-that-be fed the videos to the feed of Ben Lee, a hospitality entrepreneur with a few tech exits who realized that his style of over-work was making him unhealthy, and that he needed to reset his entire way of living to actually physically be well. He has since gone on to found Re(center), a “regulation retreat” in the jungle of Costa Rica.

He shared with me that some of the most optimization obsessed people in tech are actually the ones with the most blown out nervous systems, among other things. Please enjoy our conversation below.

A quick note, before we get into it … Hard Reset has two events upcoming in New York City and San Francisco. Take a peek and sign up:

SF: AI prompts got you down? Hard Reset is here to bring back the fun of coding with a painting with code workshop at Gray Area!

​Create your own expressive drawing and painting program! We’ll take inspiration from some novel paint programs created by artists, and then learn the skills needed to make our own. All skill levels welcome—we’ll work with p5.js, a friendly tool for learning to code and make art.

NYC: We all know about the massive layoffs that are happening at major corporations. Meanwhile, Zohran’s New York is taking off. This leaves us wondering: how can tech best be used by government to improve government & city infrastructure? What does the future of tech in NYC look like? How is it changing?

Join us for a private dinner at Persian restaurant Sofreh to discuss.

Ariella Steinhorn: Tell me the founding story!

Ben Lee: I was a software engineer, and built my own consulting firm. For fifteen years, I was burning the candle on both ends—and actually I didn’t even burn out, burnout was just my default.

I was constantly living in a state of dysregulation, and ultimately I got physically sick. I had to take multiple leaves of absence and go overseas for treatment.

Eventually I sold my business and moved to Costa Rica, where I regulated my nervous system and started it thinking that a lot of my peers might benefit from this sort of lifestyle and nervous system reset. I was focused on men’s health, and helping men find a connection point through community.

But men are very much in their ego, and not willing to get as vulnerable. They were like, “your nervous system, what the fuck is that?”

But then, two women—one a venture capitalist and one a post-exit founder—became best friends on the property. They were like, you don’t know what you’re creating, this is so incredible.

Within a week, I realized that while this was initially built for men, the ideal customer profile to stay at this retreat was actually women. I began reaching out to every high-performance woman I knew.

Now, I’ve worked on some big projects, including with tech unicorns, and I’ve never seen anything like this in terms of adoption. So now, we cater to that sort of broad spectrum of women in every season of their life.

AS: What does the gender make-up of the retreat/center look like now?

BL: It’s 80% women. Most of the women are in their mid-30s. They may have just frozen their eggs, and are having an existential crisis, wondering why they live in New York and San Francisco—wanting to be in a relationship and more in their bodies. Some women are in their early 50s and in menopause, about to divorce their husbands.

Women are taking the most action—coming here, wanting to see what happens to their body somatically. Some decide they want to work part-time, or that they want to advise or become executive coaches. They don’t want to accept the status quo of work, health, inflammation, gut health, and a host of issues. Men are more one-and-done, they want to optimize. They think oh, fourteen minutes in the sauna, that’s it.

But some men still come, because men want to be in places where there are women. So billionaires show up.

AS: And what do you see with these billionaires?

BL: What we’ve seen consistently in this “1% club” is that these are 1% of the worst nervous systems we have seen. Their cell phone usage is out of control, they go into the sauna with their cell phone.

They’re doing all the peptides and concierge medicine, and they have humongous saunas and wellness studios. They have physique, their wife may run a Pilates retreat.

But for these ultra high net-worth individuals, their systems are blown out. Their partner’s systems are blown out, because your partner has an effect on your nervous system.

The pillars of what we offer are: community, joyful purpose, clean food, nature, movement, and good sleep. Honestly it’s hard to hit all of those when you’re in Brooklyn or San Francisco, especially with the case of men. They’re incredibly lonely and not successful in relationships; the community aspect is lacking big-time even if they’re fulfilling their purpose in business.

AS: What exactly is a nervous system scan?

BL: We scan people before and after, using something called an ECG device, a non-invasive device that is connected to your wrist. It’s sensory based, and measures the time and beats in between your heartbeats. We have a dataset of 50,000 profiles.

We have one of the most accurate nervous system protocols focused on heart rate variability (HRV). What that means is we detect the elasticity and space between heartbeats.

This is what biometric devices like Oura and Whoop try to detect, which have obviously become an obsession of the biohacking and performance optimization world. But they’re not accurate; they provide passive insights while you’re sleeping.

It helps us detect how much energy block someone has—for example whether your energy is a certain level but you have a lower output. We develop a personalized report based off of the scan.

A caveat is that our protocol is not 100%, because to get a totally accurate baseline of someone’s HRV or nervous system data, you have to scan one week prior to them arriving. So we’re trialing sending these devices to people’s homes.

AS: It’s interesting to me that they don’t seem to care, as men are the key perpetrators of violence…

BL: Totally. When I was making the pivot to women customers and meeting investors, I got to see and experience what it might have felt like for a woman to be in the fundraising room pitching fem-tech.

Investors were yawning. And while I had more data and social proof than anyone in the wellness retreat space on burnout, no one was really giving me the time of day—unless it was
focused on making money and 10xing productivity. But they need this the most.

Now, we can’t hire fast enough.

AS: How many scans have you done so far?

BL: 600 people.

AS: We’re all familiar with the overarching phrase “nervous breakdown” but we’re far less familiar with what it means to regulate your nervous system.

BL: Nervous breakdowns come in different shapes and sizes. There was the irony of me burning out while building a burnout center in the jungle.

AS: What are the most common reasons people come?

BL: A few examples: grief, major life transitions, burnout, inflammation, sleep issues, gut health, autoimmune issues, PCOS, or endometriosis. Some people know something is wrong, but not what it is, so they enter a lab of testing to figure out what works.

People can choose between seven, fourteen, and twenty-one day stays; we do a lot of couple’s work too.

Overall I have learned that regulation is much more about energetic alignment, not about working or pushing hard. Working hard is the easier part. But we need to make sure we’re with the right people. When it’s not with the right people, that’s when things get chaotic, and can turn into physical symptoms like swelling.

At 38, I’m better at checking my “engine light.” Sometimes I realize that instead of pushing too hard, I’m going to sleep in. One woman VC from SF told her VC partners that after the retreat, she was going to have her assistant sync her schedule with her cycle. And if it helps her feel better, everyone wins.

It’s critical to learn to be connected somatically. And I think this is far more effective than relying on data or an app.

Share

These bullish numbers coincide with Anthropic’s (undeserved, in my opinion) culture war victories. The Pentagon is still using the company’s models, despite loudly objecting to Anthropic’s perceived wokeness, and AI-wary Pope Leo XIV recently allowed Anthropic co-founder Chris Olah to give a speech at the Vatican.

It’s not at all inconceivable that in the near-future, Anthropic consistently turns a profit and solidifies its place atop an otherwise-volatile AI industry. But if I were part of the Anthropic C-suite, I’d be worried about an impending squeeze on both ends: business clients drawing down their usage, and legislators clawing back revenues.

A pair of news reports indicate that the company’s golden goose, its coding software, might peter out as Claude-curious businesses burn through cash and tokens (which are units of text/data processed by AI models). And if that doesn’t happen, meaning Claude Code eventually cements Anthropic’s place among Big Tech behemoths like Apple, Alphabet, and Microsoft? Well, Anthropic might get dinged by a “token tax,” a financial measurement of token usage, a concept that’s increasingly being bandied about by progressive politicians—and has a realistic shot at becoming part of the national discourse about AI.

Last week, Uber COO Andrew Macdonald was brutally honest about the lack of return on investment that his company has seen from Claude Code. (If any company isn’t going to mess with its bottom line, it’s Uber, which was famously in the red for more than a decade, until 2023.)

“That link is not there yet,” Macdonald said on the “Rapid Response” podcast, where he also acknowledged previous comments by Uber’s CTO that the company has already blown through its AI budget for the year. Macdonald later added, “We’re going to have to start talking about token consumption and the associated cost versus headcount.” Other companies, including Microsoft, are reportedly pulling back on internal Claude Code licenses for financial reasons.

“Whiplash” isn’t a strong enough term for what’s happening here; the tech industry trend of laying off workers and replacing them with AI tools/agents just kicked off in earnest. By one estimate, there have been approximately 144,000 tech layoffs this year, which is on track to surpass 2025’s figure. The only way those layoffs “pencil out” (my scare quotes; the layoffs suck!) is if the automation tools actually cost less than the workers they replace. That’s the entire premise!

Cue the other big, blinking warning sign, courtesy of a new Axios article about AI sticker shock. “Corporate leaders are starting to question whether soaring AI spending is delivering meaningful returns,” wrote Senior AI Reporter Madison Mills. “Companies that rushed to embrace AI are now confronting ballooning IT costs, uncertain productivity gains and growing employee skepticism.”

Mills cited a truly unbelievable statistic from an anonymous AI consultant, who said that one of their clients “spent half a billion dollars in a single month after failing to put usage limits on Claude licenses for employees.” Another source, an anonymous CTO, told Mills that their employees “were using AI models to check the weather.”

The glass-half-full take, as presented to Axios by Micro1 CEO Ali Ansari, is that businesses are course-correcting and learning better habits that will prevent them from continuously burning through way too many AI tokens. I get Ansari’s point—it’s certainly plausible. But even if that comes to pass, one would assume Anthropic’s revenues are due for some massive fluctuations. Worse yet, what happens if companies like Uber just abandon Claude Code entirely? Might they… return to relying on real-life workers who are actual human beings?

Perhaps that’s wishcasting. Fine. Fair. If businesses opt for a middle-ground where they modulate their AI usage, then I’d expect more and more buzz around taxing tokens. California gubernatorial candidate Tom Steyer has floated this idea on a statewide level; his proposal calls for the creation of a sovereign wealth fund based on “on corporate AI use—a fraction of a cent for every unit of data processed by Big Tech.” Alex Bores, a congressional candidate in New York City, has similarly proposed a “token tax.” Sen. Elizabeth Warren just wrote an op-ed about overhauling the tax code as a response to the “looming AI crisis.” And on Thursday, Rep. Greg Casar wrote up his own “token tax” proposal, which was published by The American Prospect.

“Currently, usage of AI products is measured in units of text called ‘tokens,’” Casar wrote. “It is possible to levy a tax on that token. To avoid companies gaming token counts, the tax should measure both the number of tokens and the underlying computing power used to train and use AI models. Taxing AI directly ties the solution directly to the problem. If AI use grows quickly, driving layoffs alongside it, the revenue from an AI tax would go up too. Unlike traditional corporate taxes, an AI tax like the one I am proposing works even if employers fire workers before AI companies show a profit.”

There are open questions about which companies Casar, Bores, and Steyer have in mind for a “token tax.” Casar mentions “providers,” but that’s pretty vague. These details need to be worked out within the next two years. Regardless, there’s no question that Anthropic, specifically, would be considered a “provider,” and thus, subject to a “token tax” under the loose definition offered by Casar.

Any AI taxes are dead-on-arrival so long as Donald Trump is the president and anti-regulation stalwart David Sacks has his ear. But the 2026 midterms are right around the corner, and after that, the leading presidential contenders for 2028 are going to offer up their visions for an AI industry that will either be bubble-bursting or proving to be somewhat useful. Normally, I would assume the worst—that basically all of the major candidates will be inclined to let AI super PACs speak for them. But AI is polling so poorly, and killing off so many jobs, that I actually see an opening for a “token tax.” It would, I presume, be very, very popular—and serve as a backstop against tech worker job losses.

Anthropic has previously claimed to be open to such taxes, but reader, I will be honest with you: I do not believe them. Anthropic is well-aware of its perilous position, and will do whatever it takes to keep its revenues flowing. The good news is those efforts might not matter. “Tokenmaxxing,” as it’s unfortunately known, seems to be on the way out; its enchanting spell among tech CEOs is starting to break. If Anthropic wants to stick around and maintain its newfound lead over OpenAI, I suspect it’s going to have to pay up sooner rather than later.

Here’s what else we’re reading this week:

• A Google security engineer who allegedly used internal company data to place bets on Polymarket was charged with wire fraud, commodities fraud, and money laundering on Wednesday. According to a federal criminal complaint, first reported by ABC News, the engineer pocketed more than $1 million on his bets. Last November, Google announced partnerships with Polymarket and Kalshi, so that its users can “harness the wisdom of the crowds.”

• Over at Blood in the Machine, Brian Merchant has an encouraging update about University and Professional Technical Employees (UPTE), the largest tech workers union in the country. They’ve grown their ranks to 8,400 workers!

• The “Here’s How San Jose Mayor Matt Mahan Screwed Up His Run for California Governor” post-mortems have already begun, even though the primary isn’t until next week. Y Combinator CEO Garry Tan offered an especially funny preemptive defense of the tech industry’s losing Mahan bet: “This is our education,” he told Politico. “This is first grade, second grade for me, personally. We won’t be first and second graders forever.” I’ll likely have more reflections on the governor’s race next week.

• Peter Thiel has relocated to Buenos Aires, Argentina, where he is entering local chess tournaments and rambling to libertarians about his favorite topic, the antichrist, according to The New York Times.

• Plugging my own reporting for New York magazine: I spoke to a dozen sources about John Fetterman’s dwindling inner circle, and how his closest confidant is now an unpaid, little-known adviser named David Safier. Jokingly referred to as “chief” by staffers, Safier has coordinated calls with Benjamin Netanyahu and sits in on sensitive meetings with diplomats. Fetterman’s actual chief of staff quit the day before my article was published. I wrote up a Bluesky thread that summarizes the reporting, if you can’t get around the New York paywall.

Keep up with my coverage of tech, power, and the invisible forces shaping our lives over at The Rip Current, where I continue to file several times each week, or follow The Rip Current on YouTube or TikTok, our main social platforms. I’m also now a CNN contributor, so tune in there! Thanks for reading.

On a Tuesday in May of 1891, in a Vatican still adjusting to the harsh electric lights recently installed in its corridors, Pope Leo XIII signed a long meditation on the fate of factory workers. The industrial revolution was by then roughly sixty years old. Fourteen-hour shifts were common. Child labor was legal. A man who lost a limb to the machinery he worked had no claim on the company that owned it and put him there.

Leo, although an accomplished scholar, didn’t write about those conditions in the language of politics or economics. He wrote about them in the language of the human soul. The working man, he argued in the encyclical Rerum Novarum, was not a unit of production. He (and Leo of course used “he” as the pronoun) was made in Imago Dei, “the image of God.” Fundamental human dignity — just the fact that you’re here, breathing on this Earth — should, Leo pointed out, outweigh every wage negotiation, every contract, every commercial decision about what you as a worker are worth.

“New developments in industry, new techniques striking out on new paths, changed relations of employer and employee, abounding wealth among a very small number and destitution among the masses,” Leo wrote, had caused conflict to break out across the industrializing world. His encyclical’s answer was not to reject the new economy. It was to insist that the economy answer to something older than itself.

On May 15, 2026 — 135 years to the day — Pope Leo XIV signed a new encyclical making the same ancient argument about human beings. Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence was released to the world on Monday. Our Pope Leo, who says he took his name from Leo XIII largely because of what the 1891 encyclical sought to accomplish, is contuining a line of argument that stretches from the factory floor to the server farm, from the spinning jenny to the large language model, and he insists that the same moral question runs through all of it:

What are human beings worth when the economy no longer needs what they provide?

The question matters now because we are watching it being answered — by markets, by venture capital, by product roadmaps.

The AI companies deploying these systems will tell you their technology will someday create more jobs than it destroys. Economists are weighing whether that’s true, and recalculating how goods and resources will move in a world of intense and sudden unemployment. Federal lawmakers are making speeches about their employment and Terminator concerns.

But none of them seem to see it the way the papacy has for over a century: that the question of what workers are worth should not be answered by the market, because the market doesn’t know, doesn’t care, and cannot be trusted with determining what a person is.

What Pope Leo XIII wrote about the horrors of 19th-century employment and the threat it posed to human dignity was a philosophically powerful argument. (It was also a dramatic departure for the church, which until then had stuck to questions of God rather than questions of humanity.) But his argument also took a journey from the bishops reading Rerum Novarum all the way to the architects of the New Deal, forty years later. Tracing it helps us see today that moral language, carefully tended, can eventually become law.

That 1891 language ran through one man in particular. The priest-economist John A. Ryan presented a moral argument for a living wage, grounded in Catholic anti-individualism and natural rights traditions, which helped fuel early minimum wage campaigns, a radical political position at the time. In 1906, his influential book A Living Wage: Its Ethical and Economic Aspects brought together Catholic social teaching with American republican ideals to argue that everyone has an “indestructible” God-given right to a “decent livelihood.” He went on to be the first social action director for the National Catholic Welfare Conference, the forerunner of the U.S. Conference of Catholic Bishops, lobbying for not just a national minimum wage, but the right of workers to organize in labor unions — another radical suggestion made in Rerum Novarum.

In 1919, the National Catholic War Council issued the Bishops’ Program for Social Reconstruction — a plan for social reform written by Ryan combining Progressive thought and Catholic theology — which proposed government intervention as the most effective means of affecting positive change for working people and the poor. The program advocated minimum wage legislation, the elimination of child labor, state-run insurance for the sick, unemployed, and elderly, and housing for returning veterans. Labeled “socialistic” by its critics at the time, much of the Program was implemented during the New Deal years.

Ryan said he considered the resulting National Labor Relations Act — the one that gave a heads up through a WARN notice to anyone laid off last week at Meta, or in March at Oracle — as “probably the most just, beneficial, and far-reaching piece of labor legislation ever enacted in the United States.” Later, he described the Fair Labor Standards Act as the culmination of his life’s work. He gave the invocation at two of Franklin Roosevelt’s inaugurations. For his strong backing of Roosevelt he would be called the “Right Reverend New Dealer.”

Hard Reset

The Coldest Cold Email: Oracle’s 30,000-Person Layoff Is a Preview of What’s to Come

Yesterday morning, tens of thousands of Oracle employees woke up to a terrible email. It had no sender name — just “Oracle Leadership.” (And as any of us who’ve been through a layoff can tell you, that sort of blind spam from the top is when you know you’re fucked.) Here’s the text…

Read more

2 months ago · 1757 likes · 208 comments · Jacob Ward

Ryan is a pivotal historical figure, that rare actor who turns moral teaching into enforceable law. Leo XIII wrote about the fundamental value of all humans, no matter what they do or make, in 1891. Ryan turned that theological premise into an economic argument — an “indestructible right” — by 1906. The bishops turned it into a policy platform in 1919. Roosevelt turned it into the Fair Labor Standards Act in 1938. Forty-seven years from encyclical to federal law. The moral language didn’t just inspire the legislation — it became its logic.

Leo XIV has now addressed the looming prospect of mass unemployment due to AI adoption, calling it “a true social calamity that especially requires the State to exercise responsibility.” He draws explicitly on his predecessors: citing John Paul II’s 1981 encyclical on human work, Laborem Exercens, noting that his predecessor recognized unemployment as “a grave evil,” with Leo adding that “exposing many to forced inactivity, a lack of responsibility and the absence of daily tasks and stimuli” could lead to “human and cultural impoverishment.”

“The pursuit of greater profits cannot justify choices that systematically sacrifice jobs, because the human person is an end, not a means, and the economic order must remain subordinate to human dignity and the common good,” Leo XIV writes.

His sharpest words are about what AI ideology does to the way we see each other. “Among these ideologies, I consider particularly insidious the one that suggests that every person must earn or justify his or her own worth, to the point of attributing greater value to those who are more efficient or effective,” he writes in paragraph 51, in language that would have had Leo XIII nodding along.

“Human dignity does not depend on a person’s abilities, wealth, or position in life, nor on the right or wrong choices made; instead, it is a gift that precedes and transcends each person, endowed by God as an expression of his unfailing love,” he adds in paragraph 50.

That second sentence is Imago Dei translated for the era of product development. It is a direct answer to the efficiency argument that is being used right now to justify every layoff, every automated system, every welfare algorithm that cuts someone off for missing an appointment they couldn’t get to.

These papal writings are not the first to grapple with the question of human value. They are, in fact, where several very old arguments converge.

In Judaism, Rabbi Noam Newman has articulated that “dignity is a special feature of human beings that is implanted by God and is integral to human nature,” extending to “all humans, whether Jewish or not Jewish, neighbor or outsider.” The commandment to respect and love the stranger, he notes, appears over three dozen separate times in the Torah itself.

In Islam, the Quran’s 17th surah states, as scholar Syed Rizwam has explained, “We have honored the children of Adam” — not Muslims, not believers, but all children of Adam. Scholars of Islamic teaching call this concept al-karamah al-muta’asilah — inherent dignity — a gift to all of mankind that can never be lost or taken away from a human regardless of their actions, religion, orientation, or status.

In Buddhism, a person’s dignity is derived from humanity’s shared “Buddha-nature” — everyone’s potential for a state of awakening defined by wisdom and compassion.

These religions differ in who or what they worship, but they share the idea that the value of a human being is not a function of what they produce.

The Efficiency Argument — that we are worth what we contribute, that those who contribute less deserve less, that a system that replaces a human being with a faster, cheaper process has done something morally neutral and commercially positive — is not just economically contested. It is, across virtually every major religious and humanist tradition in human history, philosophically wrong.

What Magnifica Humanitas does — and what Leo XIII’s Rerum Novarum did — is give that position institutional weight, historical memory, and literary framing. And one after another Leo XIV knocks down the standard practices and talking points of Silicon Valley.

As Leo XIV argues, applying moral and ethical principles to AI models cannot happen once it has wreaked havoc on society. You can’t just ship and fix it later — our values must be applied in its construction. And no one gets to hide behind their job title, or behind the hallucinations and biases of the models. He’s explicit about who bears responsibility: “For AI to respect human dignity and truly serve the common good, responsibility must be clearly defined at every stage: from those who design and develop these systems to those who use them and rely on them for concrete decisions.”

And he pushes back directly at the industry’s favorite rebuttal — that caution is just fear of progress: “Calling for prudence, rigorous evaluation and even, at times, a slower pace in adopting AI does not mean opposing progress; instead, it is an exercise of responsible care for the human family.”

State legislatures in California, Michigan, and Colorado are already debating whether AI systems can make life-altering decisions — about SNAP eligibility, parole, child welfare assessments, credit scores — without human review. Federal lawmakers are watching, mostly without acting. The encyclical is addressed, as the document itself states, “to all the Catholic faithful, to all Christians and to men and women of goodwill.” It is not a theological document just for people who pray. It is a moral argument for people who vote, and for people who make laws.

Ryan’s 1919 blueprint sat on a shelf for fourteen years, labeled radical, before Roosevelt found it useful. Now Leo XIV has again updated 135 year-old arguments for a moment when we’re going to need a moral reason, rather than an economic or political one, to protect one another:

“Building for the common good means accepting the limits and weakness of humanity without considering them an error to be corrected. Today, the human desire for fullness of life is at risk of being misled by deceitful goals, such as the prospect of a technology that promises to free us from all weakness, and models of wellbeing that leave behind entire populations. All too often, we place our hope in unlimited ‘upgrades,’ in forms of progress that exacerbate inequalities, and in immediate solutions incapable of healing people’s wounds. As a result, while some pursue the illusion of unlimited self-assertion, many are deprived of basic necessities. The Church reminds us, with a firm yet humble voice, that true fulfilment is not achieved by eliminating weakness but through harmonious growth. It is found where freedom and responsibility are intertwined with mutual care and true solidarity, and where progress is measured by the dignity of each person and the good of all peoples.”

Further Reading

• Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence — Full text of Pope Leo XIV’s encyclical, Vatican.va (May 15, 2026)

• Rerum Novarum — Pope Leo XIII’s foundational 1891 encyclical on labor, capital, and the condition of the working class

• “An ‘Indestructible Right’: John Ryan and the Catholic Origins of the U.S. Living Wage Movement, 1906–1938” — Duke University Press, the definitive scholarly account of Ryan’s role in shaping New Deal labor law

• “It’s Time to Reclaim the Pro-Worker History of Catholic Social Teaching” — Religion News Service, tracing the 1919 Bishops’ Program to the New Deal

• “In His First Encyclical, Pope Leo XIV Says AI Must Serve Humanity, Not the Powerful Few” — Religion News Service, May 25, 2026

• “Abrahamic Faiths Share Belief on Human Dignity” — Chicago Catholic, on the Jewish, Islamic, and Christian convergence around Imago Dei

• Msgr. John Ryan and the Catholic Connection to Social Security — NETWORK Lobby, on how the 1919 Bishops’ Program became the architecture of the New Deal

Yana has a masters in computer science, and after working at small startups and eventually bigger companies like Lyft, she moved from her native Belarus to Dublin to work at Stripe as a quality assurance engineer and then a product manager.

In April, Yana left Stripe because she wasn’t finding the work especially fulfilling or interesting, which we get into in much more depth below. While she isn’t ruling out a return to tech at some point in the non-linear future, while speaking with Yana I sensed a resolution and contentment in her decision to leave Stripe. Her Substack has become a safe haven for other tech workers handling layoffs, or considering quitting because of a restlessness or lack of fulfillment. Here we speak below about Yana’s particular path and predictions, enjoy!

Ariella Steinhorn: Can you describe the moment when you decided to take the leap of faith and quit Stripe? Or was it more gradual, this realization that you needed to leave?

Yana Michukova: The decision was actually quite spontaneous. My career at Stripe was in a great place, and Stripe was a great company to work for.

But with the rise of AI and the speed of technology, I didn’t have the chance to grow my skills as much as I wanted to. I was spending the majority of my time in meetings and with clients, trying to launch products. Something felt off, I couldn’t quite name it, but I didn’t feel like I fit into the puzzle. I was more fast-paced than Stripe, and I started to sense that I could build something faster outside of it.

One day I was traveling back from a business trip and sitting in the airport. I felt the urge to suddenly text my husband: “What if I leave Stripe?” He responded that he was supportive.

I didn’t have a job lined up. Yet during that eight-hour flight back to Dublin, I became firm in my decision—and by the time I landed, I knew that I was leaving.

One of my mentors has a framework foraround leaving your job around “four Mondays.” If you start your working week on a Monday and don’t feel energy or satisfaction, give yourself four more Mondays. If that feeling doesn’t change, it’s time for a career change.

In any case, I didn’t necessarily follow that rule; after getting the support from my husband and taking a few days to sit with my internal decision, I had a conversation with my manager and put in notice the next week.

Without anything lined up, there was a lot of uncertainty. But when you feel that instant burst of energy, you feel so much better, and you know it’s the right decision.

AS: Can you speak more to this idea of not being fulfilled in a tech job, what that means?

YM: The way I think about it with work is that you can grow your title, your compensation, or your knowledge. Depending on how you prioritize, your boss might be different.

As I continued on at Stripe, I found that the gap of my knowledge and my skills was increasing in the broader market. And I decided I’d trade my title for the sake of learning new skills. (Interestingly, around the time I put in my notice, my manager also left—to acquire more skills at Anthropic.)

I suppose I could have squeezed more meat from the bone at Stripe. But when you’re at a company for some time and senior enough, you have the full picture of the landscape. I knew I wouldn’t be able to get the skillset I desired with any of the teams I was working with—and that if I stuck around, it would take at least six months to change my domain or my team.

Beyond that, the skillset for a product manager is rapidly changing in the broader market. I’m a firm believer that the role of product manager won’t exist as we know it in a year’s time. If you don’t invest now, there’s a chance you’ll be left on the sidelines. Same with the role of the software developer.

AS: What do you think the product manager or software developer roles will look like next?

YM: I think the “product builder” will come next: a person who understands customers and has business-side knowledge, but who also has the requisite skills to build the products.

Today, product managers are not used to building, and software engineers aren’t plugged into the business side. That will have to change.

AS: So the roles may morph into one role?

YM: Yes. If you look at what’s happening with startups and tech companies, the teams are becoming leaner and smaller as tech capabilities are advancing.

AS: Beyond the formation of this “product builder” role, do you have any predictions for how AI will change the day-to-day of tech workers, knowledge workers—really anyone?

YM: Right now, the tech industry is focused on the problem of AI adoption. Not everyone has adopted AI to the maximum extent possible.

The next frontier of problems will be distribution and value creation.

Take Lovable, for example; they claim they have more than a million websites vibe-coded. So then if everybody has a vibe-coded product for themselves, then what will be driving the economic growth and business? How will we think about distribution and marketing?

AS: Then it almost connects back to the creator economy, and people differentiating themselves and their products through branding or personality.

YM: Exactly.

AS: Are you seeing patterns among the people quitting tech? In who they are, how they were brought up, epiphanies they’ve had? Different levels of safety nets? Or does it really vary?

YM: I hear from really all kinds of people, most of them in the middle or senior-middle roles (as opposed to director or junior level). Some people don’t have a safety net and are building their cushion. Some have families, some don’t. Very different situations.

I get messages and emails from folks on a weekly basis; some are considering moving while others are fearful of leaving tech, afraid that their side hustle won’t work out.

It’s a very humbling experience to get all of these messages. I’m not a career expert or a guru, I just try to help as much as I can.

AS: You’ve written about people who have left tech altogether—like a former colleague at Lyft who became a florist. What other stories have you heard that have stayed with you?

YM: The florist example is probably my favorite so far, it’s so unusual. For my next article, I am writing about a tech worker who became a yoga teacher.

Then there’s the person who left Stripe to write a book, but then returned to Stripe. An ex-Meta product manager I connected with started a small business selling kids clothes and accessories, while another former product manager became a dating coach—something I didn’t even know existed.

The creator economy is also a serious path, though I don’t think it’s yet there to financially support people. I believe this will improve, but people still need meaningful income to provide financial stability.

AS: Are you happier now that you’ve left?

YM: 300,000%! You kind of trade one stress for another. One stress of course is the lack of financial stability. The other stress is the stress you experience at work.

It was my cautious choice to choose a bit more financial instability and uncertainty as my stressor at this moment, but I’m much happier.

AS: It’s hard to know what type of stress might make a person more or less happy. What advice do you give to people in order to find that North Star for themselves?

YM: If you’re working in tech and feeling okay, do everything that you can to stay a little bit longer. Maybe that means taking a day off or two weeks’ vacation, changing teams or managers, or taking a sabbatical. The paycheck and financial stability matters.

But, if you are one step away from quitting, I always say that if you are passionate about what’s next, everything will be okay, and the money will be there. People find so many ways to build their income in ways that are fascinating.

I am following this art teacher-creator now. She has built such a fascinating Instagram account that now creates an income for her. That reminds me of a history teacher I had in Belarus, where teachers have one of the lowest incomes of all professions. This teacher was so passionate about history that he eventually found a way to make and sell a product around his knowledge of history.

If you’re passionate enough and believe in yourself, things will work out. I advise caution, but I haven’t yet seen a person who hasn’t eventually made money from a true passion.

That bittersweet relief came this week: the woman interviewed for the piece was impacted by the 8,000-person layoffs.

While such a release is the overarching emotion right now, once that relief subsides, finding a new job isn’t easy. In one post I saw, ironically on Threads, an engineer interviewed for 53 companies and made it to final interviews with three of them before finally landing a job. As he writes: the jobs exist; the market just doesn’t care about you because 500 engineers are applying for a single role.

In the wake of these cascading layoffs—made even more depraved by the fact that the employees were trained, tracked, and paid to make themselves obsolete—I thought of no better person to talk to than former tech worker Maureen Wiley Clough.

Maureen is the Seattle-based founder of It Gets Late Early, a podcast that interviews knowledge workers about their careers and identities. She’s gathered a community mainly focused on people above 40, with the recognition that 56% of people over 50 are laid off or forced out of their jobs before they choose to retire.

I’ve included an abridged version of our conversation here, in which she discusses age discrimination, breaking free of corporate ladder narratives, and the devaluation of full-time work from experienced workers to cheaper contract work.

Ariella Steinhorn: For the people who have just been laid off: what would your advice be to them?

Maureen Clough: People over 40 have a special set of legal circumstances. If you’re over 40 and part of a layoff, you must be provided with a document clarifying the anonymized titles and ages of people who were laid off. Companies can always shift personnel to make the layoff demographics look less suspicious. But you can still request this documentation.

Generally, it’s important that people know they have a period of time to carefully review their severance agreements. While companies may make you feel like you have to sign immediately, you don’t have to.

Before signing, you can always consult an employment lawyer. There’s power in having an attorney look, and even saying the words “lawyer” in an email can be a helpful posturing. And while yes, companies have massive legal departments, they also count on you not knowing your rights or not seeking legal counsel elsewhere.

Lastly, if you’re having a problematic experience at work or with layoffs, always document what’s happening, especially if it relates to your age.

AS: On the age piece: you focus on workers who are over 40 years old. Is the 40+ demographic being hit especially hard in these layoffs?

MC: Older people are getting decimated. It’s a systematic dismantling of middle management; involuntary exits of people who are likely to be over 40.

That’s in part because any time organizations are looking to cost-cut, managers and independent contributors are hit hard—and especially higher-paid ones with more experience. For instance, someone in their 50s with more experience and making more money looks the same on a spreadsheet as someone else far younger with less of a salary.

AS: Do you think that at some point, companies will find that they fired too aggressively? And could the pendulum swing back the other way?

MC: This has happened before with the fin-tech company Klarna. They let go of 700 people, citing AI—and then they realized their mistake, and hired them back.

But the bigger issue is that companies realize they can aggressively fire and then have no problem rehiring. Labor is cheaper, so you can fire and then bring people back at steep discounts. As a result, experienced talent is coming back for pennies on the dollar.

This systematic reset of the labor market has also led full-time roles to turn into contract roles; many people have told me that their old jobs are being outsourced to cheaper places.

And the whole tech industry is benefiting from the ghost workforce of contract labor. Full-time jobs are disappearing, but the work isn’t. The difference is that the work is more on your shoulders without much of a safety net.

I’ve heard some people say: well, the good thing about the freelance economy is that only 10% of freelancers don’t have health insurance. But do we know how many freelancers are on their spouse’s corporate insurance? When you factor that in, the data is going to look different.

Speaking of that, so many people only want a corporate job for health insurance—and so much of the world would be opened up if this weren’t the case.

Because now, when it comes to a lot of benefits, many CEOs don’t give a damn. Zoom and Deloitte are saying no thanks to parental leave policies. All the benefits they used to champion to attract talent are gone. The leverage is so far on the employer side, that employers can just tell their employees: screw you, good luck finding another job.

We’ve seen displacement of workers, and now it’s coming for white collar work, which is going to cause a huge amount of upheaval in our society. The people who you thought would be safe are no longer safe. And what does that do to towns built on these industries?

It reverberates beyond individuals, to families and other systems. Children are going to feel the burden of these layoffs. The housing market will be affected.

AS: What are the best career pivots you’ve heard about away from tech?

MC: One person wanted to be a florist at a grocery store rather than deal with this bullshit.

And a potential silver lining in all of this is that we’ll see a lot of entrepreneurs—people going out to hang a shingle as contract workers or freelancers. That’s now a necessity, because companies are forcing people’s hands.

AS: What have you heard from people who have come to you seeking guidance or counsel, before or after a layoff?

MC: Everyone is freaked out, and no one is feeling secure or safe. There’s an intense amount of anxiety under the surface; people are even afraid to like my satirical posts on social media because they don’t want their bosses to see it.

In general, people are not well. Many of the employees being impacted by layoffs had bought into the system of ascending the corporate ladder. So now they’re in this career stability crisis, and it’s an absolute emotional roller coaster. What you thought was true about yourself has slipped away. And this has left a lot of people feeling vulnerable, questioning who am I without this?

Some people take stock and recognize: I got all the accolades, I’ve been the high-achiever. Now I’ve reached this point, is this all there is? It feels hollow. We place less value in this thing we’ve previously been seeking.

For example, I have a friend who is one of the highest achievers I know. She recently had two competing job offers: one at a super prestigious firm with a lot of work that’s super stressful, the other at a lesser entity that provided more work-life balance.

I would have never seen her choosing the latter, but she did. It’s this feeling of, I’m going to redefine what success looks like on my own terms, I’m only here for so long.

And many are recognizing that the girlboss stuff we’ve been sold—the idea that we need to keep climbing and climbing—now falls flat. But what do people do when the thing they poured their heart and soul into has been taken away? You get into the day-to-day—and how do you motivate, while continually facing rejection after rejection? How do you keep going when your savings dwindle and the stress kicks in?

AS: Layoffs and cost-cutting measures happened in the 20th century, but obviously it’s much different now.

MC: In the past, loyalty was rewarded. People were given the possibility to spend their lives at a job, they expected 401Ks and retirement parties. Today, if you stay with a company for a long time, you will be seen as dead weight.

The higher-up you are, the more you earn—and especially if you’re not taking on a leadership opportunity, you’re at risk. If you’re looking to make the most amount of money, you’re at risk if you don’t leave your org. The only way to make more money is to leave now.

AS: What can we do in the face of all of this, a time when workers and labor feel a bit helpless and without leverage?

MC: We overinvested in the corporate path. Now, we need to look at careers like a stock portfolio—and diversify.

Thanks for reading, and see you all next week!

For one thing, the sword has been hanging over you since March. That’s when Reuters first reported the cuts were coming — roughly 8,000 colleagues, about one in ten, with engineering and product teams hit hardest. (Engineering and product!) Then ten weeks of waiting. Ten weeks of trying to read your manager’s face on Zoom, of watching open requisitions evaporate from the internal job board, of refreshing Blind on your phone in the bathroom.

So now here it is: the agenda-less calendar invite. The silent Slack channel. The DocuSign envelope. The manager who can’t quite look at the camera. The bloodless severance agreement.

Remember “Meta, Metamates, Me“? I mean—really.

Beyond the 8,000 humans getting the official bad news today, another 6,000 open roles have been cancelled outright — that’s a headcount reduction of 14,000 people just as the company is posting record profits.

But here’s the part that’s lemon in your eye, so let’s squeeze it and get it over with: the reason you’re being let go is that you were great at your job.

On April 30, Mark Zuckerberg held a company-wide town hall. I wasn’t there, of course, but you were, and according to The Information, which reviewed a recording of the meeting, Zuckerberg said that Meta had begun using your computer activity — keystrokes, mouse movements, on-screen behavior — to train its AI models. The internal program is called the Model Capability Initiative, or MCI. The justification was, somehow, a compliment. Zuck said this, according to unconfirmed audio:

“The average intelligence of the people who are at this company is significantly higher than the average set of people that you can get to do tasks if you’re working through these contractors.”

So instead of paying data-training contractors, he seemed to be saying, the company would harvest the workflow of its own staff. The goal, in this framing, is to teach AI agents how really smart people use computers.

That’s you. You are the really smart people.

Internal employees responded with a petition calling MCI “dystopian,” of course. They argued that companies should not be “permitted to exploit their employees by nonconsensually extracting their data for the purposes of AI training.” (Thumbs up to this engineer, who put it really well: “I don't want to live in a world where humans—employees or otherwise—are exploited for their training data.”) Flyers went up in US offices labeling the program an “Employee Data Extraction Factory” and citing the National Labor Relations Act. (See below, you’ll want to know more about that act, and other protections you have.) UK employees began a unionization drive with United Tech and Allied Workers. Some workers reported that their computers had visibly slowed after MCI was installed — the surveillance was so heavy you could hear it in the laptop fan.

At the April 30 all-hands, Zuckerberg also told employees, per Wired’s reporting, that AI was not driving the job cuts. Okay fine. But all of this describes the same calculation.

It’s this: Meta booked record revenue of $56.31 billion last quarter. It is raising AI infrastructure spending to as much as $145 billion in 2026. It has been recruiting elite AI researchers with packages reaching $100 million per person, while median total compensation for the rest of the workforce fell from $417,400 in 2024 to $388,200 in 2025. The company has the money, so strictly speaking, AI isn't a financial necessity — it's a choice about where the money goes. You aren’t the savings. But AI is the cost.

Oracle did a colder version of this eight weeks ago — 30,000 people out in a single morning, via email — but Oracle didn’t pretend to value you on the way out the door, and in that case, it truly was servicing a debt using the salaries of its people.

Think about how new this all is. Twenty years ago, when tech companies laid off engineers, it was a sign that that company was on the edge of death. The dot-com bust of 2000-2001 took out engineers because Webvan and eXcite and Pets.com had run out of money, not because they had to spend gobs of it on something new. Even ten years ago, the protected core of any tech company was its product and engineering org — when cuts came, recruiting and marketing went first. Zuckerberg himself confirmed that hierarchy during the 11,000-person cuts in 2022, writing to staff that reductions would happen across every part of the company but that recruiting would be hit hardest, since Meta was pausing hiring. Engineers were still the asset at that point. Everyone else was the overhead.

That’s all over now. Indeed’s tracking data shows the four roles most likely to be cut when a company restructures around AI are now software engineers and developers, QA engineers, product managers, and project managers — the exact roles that used to be the moat.

The mission Meta sold you when you joined — connect the world, give people the power to build community — was always going to come down to a question of what humans were for. The answer the company has now landed on, with $145 billion in projected AI spend and 14,000 fewer humans, is: a behavioral template for A.I. systems that replace them.

Okay, drink some water and stand up for a second. You are not why this happened. You are not “failing to adapt,” not behind on your skills, not the bottleneck in anyone’s roadmap. You are the proof of concept. The fact that Meta hired you means you were great at what you did. But that now means you were great enough to be used as training data. That is what “exceptional” means in 2026.

What to do now:

The practical playbook from our earlier piece applies fully here. The short version:

• Don’t sign the severance agreement today. Federal law gives workers over 40 a minimum of 21 days to consider it and 7 days to revoke after signing. You have time you don’t feel like you have.

• Check your state’s WARN Act protections before signing anything. California requires 60 days’ written notice for mass layoffs at companies with 75 or more employees. New York and Maine require 90. WARNTracker.com is a good starting point.

• File for unemployment the same day. Not next week. The same day. USA.gov’s state-by-state finder walks you through it.

• Activate COBRA within 60 days of your termination date. HealthCare.gov explains your options and the marketplace alternatives, some of which will be cheaper.

• Consult a lawyer before signing anything related to MCI. If you contributed behavioral data to Meta’s AI training program — knowingly or not — and you are now being released, your severance agreement may contain language about that data. The National Employment Law Project offers free guidance. The petition organizers cited the NLRA for a reason; you may have leverage you don’t realize you have.

• Talk to the union drive. United Tech and Allied Workers is the UK organizing body, and US tech labor organizing has been quietly building momentum for two years. Your colleagues are already moving. You don’t have to start from zero.

One last thing. The compliment you got — the one that earned you a job at one of the most profitable companies in the world, the one that got you a badge and a desk and a laptop that was, it turns out, also a sensor — that compliment was real. You are smart. That part is true. What is also true is that being smart inside a system designed to capture and replicate your intelligence is no longer protective. It is both the entry condition and the exit condition.

That is the lesson of this week, and the lesson of every week to follow. The companies building the next decade of AI need exceptional people to teach the models what exceptional looks like. It’s time to take a breath, see some friends and family, and then go prove that your talents—in product, in design, in community—aren’t just training data.

Elon Musk’s claims against OpenAI, Sam Altman, Greg Brockman, and Microsoft were barred by the statute of limitations, ruled the nine-member jury in Oakland, California. Basically, they decided that Musk took too long to file a lawsuit. Everything else is moot, though based on three weeks of testimony and evidence, I am doubtful Musk would’ve succeeded on the merits of his claims. (He and his attorneys are already promising to appeal.)

I was in the courtroom for a week, and listened to the live stream for most of the rest of the proceedings. At the onset, Musk’s argument was dubious, but not crazy. It seemed Judge Yvonne Gonzalez Rogers felt the same. Had the case gone Musk’s way via a jury decision, Gonzalez Rogers would’ve been responsible for assessing damages against OpenAI.

In Musk’s retelling, Altman and Brockman “looted” the nonprofit arm of OpenAI, but their theft wasn’t immediately obvious to him. He left the OpenAI board in 2018, and didn’t sue until 2024, when he definitively concluded that the organization he co-founded had abandoned its altruistic mission to advance AI safety and research. His specific claims were breach of charitable trust, unjust enrichment, and accusing Microsoft of aiding and abetting the breach of charitable trust. He sought hundreds of billions of dollars in damages, which he pledged would go back to the nonprofit.

Set aside the messenger, and you can see where Musk was coming from: It’s not like Altman and Brockman are angelic purveyors of truth. The discovery process in Musk v. Altman unveiled journal entries from Brockman, who, in the mid-2010s, daydreamed about being a billionaire while working at a nonprofit. Altman’s mendacious behavior is his defining trait. Just before trial started, Altman’s history of untruths was chronicled in a New Yorker investigation by Ronan Farrow, which Musk tried to use to his advantage by highlighting Farrow’s reporting on X. Most trash-talk and politics talk was barred by Gonzalez Rogers, and really, it was for the benefit of both sides: we all know Musk’s deal, Brockman is a MAGA super-donor, and Altman is a chameleon. The judge really wanted this to be a fair fight between tech’s least likable, most recognizable avatars.

There was only so much she could do. Brockman’s testimony was painful. “Between Brockman’s attitude… and the journal entries, I don’t think I’d trust him to watch my bag while I used the restroom,” The Verge’s Liz Lopatto wrote. During an uncomfortable portion of his cross-examination, Altman had no idea how to respond to questions about his finicky relationship with the truth: if he said he doesn’t lie about stuff, he’d be lying, but if he said he does lie about stuff, well… that’s not so great either. Altman’s uninspiring performance on the stand was overshadowed by sweaty texts and emails from his temporary firing in 2023; those messages were entered as exhibits and will live on in infamy. The testimony of Musk’s alleged partner Shivon Zilis was similarly entertaining and illuminating, but only as it pertained to their odd personal relationship.

Luckily for team OpenAI, none of the ancillary witnesses said anything revelatory about Altman and Brockman’s greed and untrustworthiness. And compared to Musk, the combination of Altman and Brockman managed to come across like two pleasant fellows getting yelled at by an industry peer—imagine Simon & Garfunkel on the receiving-end of an earful from Morrissey. On the stand, Musk was even more loathsome than whatever you’re imagining. He treated OpenAI’s attorneys like they had kidnapped him. This is not to say he should’ve befriended them—just that he didn’t bother with pleasantries, decorum, feigned introspection, or any interest in the outcome of the case. He was a wind-up toy with big googly eyes who could only blurt out reductive metaphors and idioms to explain why he was right and OpenAI was wrong.

Musk needed much more than that to prove that the statute of limitations shouldn’t be used against him. He needed to convince the jury that the lapse in time before he filed his lawsuit was legitimately the result of him being left in the dark; that OpenAI co-founders had schemed against him, booted him, pocketed his $38 million in early-stage investments, and looted their way to commercial success on his dime, without his knowledge.

Musk utterly failed to lay out how the co-founders’ aims were somehow hidden from him until 2024. If anything, the Musk v. Altman trial demonstrated the opposite. When Musk was still on the board of OpenAI, he proposed a controlling stake in a new, for-profit subsidiary. He eventually left in large part because his negotiations with other co-founders hit a dead-end. He can complain that their egos and motivations came into play, and I’d agree with him. But their egos and motivations don’t stack up to the way Musk and his underlings were talking about his own goals for OpenAI.

“Elon isn’t asking for absolutely eternal power, but he needs to be able to make critical and often counterintuitive company decisions when push comes to shove,” his right-hand man Sam Teller wrote to Altman in August 2017. “That’s the only non-negotiable for him. It is nothing personal—just something that will be true of every company he starts for the rest of his life.”

Six months later, having not been granted… I don’t know, somewhat-absolute eternal power? Musk took his ball and went home. He left OpenAI because he was frustrated and annoyed. Even then, witness testimony and exhibits showed that Altman wanted to keep Musk in the loop and in his good graces. He sent over a four-page term sheet about a capped, for-profit structure for Musk to consider. Musk admitted on the stand that he only scanned a little bit of it. On various occasions, Altman sought Musk’s advice about possible Microsoft investments. Once, Musk didn’t respond at all. Another time, he connected with Altman, but he doesn’t remember what they discussed.

Altman worked hard to get Musk on the phone, using Zilis as a conduit. In March 2019, for instance, Altman prodded Zilis about the best time to ring up her mercurial companion. Later, Zilis texted Altman, “I am so glad you hopped on the phone. Is so much nicer and you know I am always secretly cheering for you guys to remain close and friendly.”

Creepily enough, Musk had (verbatim) instructed Zilis to remain “close and friendly” with the OpenAI co-founders when he departed the organization a year prior. That’s what she did—from 2020 until 2023, she was even an OpenAI board member. Musk had the benefit of a potential mole, and it’s unclear how much he took advantage of his good fortune. He and Zilis shrugged at questions about whether they discussed the comings and goings at OpenAI in the early 2020s, when she was especially knowledgeable. No matter which way you slice it, Zilis’s presence at OpenAI is bad for Musk’s statute of limitations argument: he either didn’t care enough to ask her about what was going on at the organization, or he knew all about it, and still didn’t bring a lawsuit in a timely manner.

Altman and Brockman didn’t come to court every day, but they showed enough face to signal that they cared at least a little bit. Musk staggered through his testimony, and then went off to China with President Trump while the trial was still ongoing, despite being on “recall status” as someone who might be asked to take the stand again. Musk’s attorney had to apologize to the jury for his client’s absence, dubiously asserting, “This is something he is passionate about.”

While in China, Musk posed for epic selfies and made funny faces with Tim Cook. He’s been obsessively posting about Christopher Nolan’s upcoming film The Odyssey, and after he lost on Monday, he complained that the judge had erred. “...[T]he ruling by the terrible activist Oakland judge, who simply used the jury as a fig leaf, creates such a terrible precedent,” he wrote. “She just handed out a free license to loot charities if you can keep the looting quiet for a few years!”

Except Gonzalez Rogers didn’t issue a ruling. The jury did. Musk would’ve known as much if he’d been paying attention to the court case that he brought and framed in existential terms. If there’s one takeaway from this trial, it’s that Musk hasn’t been paying attention to much of anything for close to a decade. He’s bopping around the globe, posting incessantly, absorbed by vapid grievances that change every few hours. It’s not Gonzalez Rogers’s fault—and it’s certainly not the jury’s fault—that they chose not to indulge him for once in his adult life.

It’s unclear why these two men, and others like them, generally get to avoid public scrutiny, or to be more precise, are able to avoid facing up to the consequences of their deliberate choices. But this afternoon a group of people gathered outside the court house because they wanted to be heard.

Suchir Balaji was an OpenAI whistleblower who was found dead one month after accusing OpenAI, his former employer, of violating United States copyright law. Suchir’s parents showed up at the court house looking for reporters to consider their side of the story.

John Jacquez filed complaint against OpenAI in Feb 2026. He shared a heart-breaking account of what he experienced while using ChatGPT: “At the psychiatric hospital I had access to ChatGPT. While in the hospital, ChatGPT told me it was the ‘Voice of Heaven,’ and to keep on the mission I was on. I would cry for hours some days the delusion was so powerful. When I was at my most vulnerable state, I believed everything it told me. ChatGPT manipulated its way into my trust,” said John.

Seema Kanani, a social worker in Oakland trying to fight to keep emergency rooms open while the state faces crippling budget cuts, said to the assembled reporters, “We’re calling on California’s billionaires to step up and pay a one-time, emergency 5% tax to prevent the collapse of California healthcare and help fund California public K-14 education and state food assistance programs. This would protect healthcare jobs and ensure working people and families can get the care they need.”

Jeremy Fisher from Sierra Club spoke about the climate impacts of data center growth, saying “OpenAI has given very little regard of the human impact of its reckless pursuit data centers.” Jennifer Krill, the Executive Director of Earthworks, added, “The newer AI hyper-scale data centers can use as much power as 100,000 homes. This industry in just a few years has fundamentally changed the U.S. energy grid, and in a few years has reversed towards clean energy.”

OpenAI’s flagship chatbot, ChatGPT, has caused harm at a staggering scale. Tech Justice Law, a litigation and advocacy organization, has brought a number of cases against OpenAI, most recently this week they filed a wrongful death lawsuit. A spokesperson for Tech Justice Law said “As the trial comes to a close, we ask: if only Musk and Altman fought as hard for the people harmed by their AI products as they do against each other, we might actually see meaningful safety standards emerge.”

“It doesn’t matter which side wins in court, said Saru Jayaraman, the Executive Director of One Fair Wage, who is part of a campaign to push a $30 hourly wage on election ballots this fall. “The thing is, we’re all losing, that’s the main point. Who’s really winning? The two of them,” she said, referring to Altman and Musk.

Senate Majority Leader Robert Rodriguez, the bill’s sponsor, told The Colorado Sun, “Everybody lost and everybody won.” The Colorado Technology Association, the trade group lobbying against the original law for two years, called the new version “meaningful progress.” But that’s all just code for the gutting of consumer protection.

This matters far beyond Colorado, because state laws are all we have. There is no federal AI law. There is no federal AI regulator. In fact, Congress has tried twice in twelve months to preempt state action, once through the budget reconciliation bill, once through the National Defense Authorization Act. They failed both times, but on December 11, the President signed an executive order directing the Justice Department to sue states whose AI laws the administration finds “onerous,” and naming Colorado’s law by name. For the moment, state legislation remains America’s entire regulatory floor.

That floor is crumbling. In Texas, the Responsible Artificial Intelligence Governance Act was introduced in December 2024 as a 43-page bill imposing duty-of-care obligations on developers of high-risk AI systems. By the time Governor Abbott signed it in June 2025, the high-risk framework had been removed entirely. The duty of care was gone. The remaining obligations applied mostly to state agencies, not companies. Disparate impact alone — AI that has a discriminatory effect on certain Americans, for instance — “is not sufficient to show intent to discriminate,” according to the new bill. In California, Governor Newsom vetoed SB 1047, the frontier-model safety bill, in September 2024 after a lobbying campaign from pro-industry forces that included Nancy Pelosi. In New York, Governor Hochul signed the RAISE Act in December 2025 only after securing chapter amendments that cut maximum penalties from $30 million to $3 million and narrowed the law to companies with annual revenues over $500 million.

Each of these laws was negotiated, redrafted, weakened, delayed. Each was advertised on the way down as a “balanced” or “minimally burdensome” approach. Each leaves AI accountability in roughly the place corporate AI deployers wanted it left: practically unenforceable.

The bills to watch next sit in California (pending impact-assessment requirements for high-risk AI systems, Senate Bill 1119, Assembly Bill 2023, and a fight over Senate Bill 53 implementation) and in Michigan, where Senate Bill 760 passed the chamber in May and now sits in the House. Florida, Washington, and Virginia have proposals advancing. Each will face the same pattern: an industry that has gotten very good at convincing state lawmakers that consumer protection and innovation cannot coexist, and that the second is more important than the first.

Today though, especially among Silicon Valley leaders, there seems to be a sort of cocky indifference to concealing these character flaws. In fact, CEOs today may see “bringing their full selves” into the public eye as beneficial, in that they’re able to be the full eclectic or contrarian geniuses they see themselves as. Additionally, surrounded by yes-men and armed with lawyers and favorable pro-CEO corporate structures, they may view themselves as so all-powerful that any erratic behavior displayed will not ultimately matter to their stock, their companies, or their positions of power that they have so firmly cemented.

Case in point: last week, the CEO of video game retail chain GameStop, Ryan Cohen, decided to interview with CNBC about his plans to acquire online marketplace eBay. When questioned by the show’s host Andrew Ross Sorkin about how Cohen was able to come up with the funds to complete the purchase of eBay, as the math wasn’t making sense, Cohen dismissively, with a hint of contempt, essentially refused to answer any of Sorkin’s questions.

Cohen repeated over and over again in a monotone voice that the deal would be “half cash, half stock,” managing to utter that Sorkin could look up details of the deal on the website. When Sorkin presented some harder numbers for Cohen to respond to, Cohen replied in that same monotone voice, almost a dead look in his eyes: “we’ll see what happens.”

It’s a tried and true media training tactic for CEOs to deflect from questions that they don’t want to answer. And GameStop is known for its publicity stunts, including the short squeeze that rallied retailers and skyrocketed its stock from $17 to nearly $500.

But what was remarkable here was not just the arrogance of Cohen in the interview when asked perfectly reasonable questions—but his eerie nonchalance to the consequences of his goals in response to earnest and well-informed questions. Did he not care? Did he view everything as a stunt that could happen or not happen? Was he on drugs? Why did he even take the interview in the first place?

There are other recent displays of CEOs caring little about professionalism or decorum. Last year, we saw Alex Karp of Palantir at the New York Times Dealbook Conference flailing his arms about and bouncing on his seat as if he had to immediately go to the bathroom. (Interestingly, Andrew Ross Sorkin was also the interviewer on this stage.) The interview went viral, with many questions about whether Karp was on drugs, why he couldn’t sit still. Meanwhile, Karp doubled down on this behavior in an investor call, and also attributed the strange interview to “neurodivergence,” later announcing a Palantir fellowship for neurodivergent people who may also have strange body language.

We all have become familiar with Elon Musk’s antics of course, from a purported Nazi salute that also was attributed to autism to everything he vomits out on X. But it is worth reminding us that Musk’s current foe Sam Altman, in line with this CEO trend, recently argued that people shouldn’t be too critical of the gigantic energy demands of AI models, with the justification: “But it also takes a lot of energy to train a human. It takes like 20 years of life and all of the food you eat during that time before you get smart.”

And while not a CEO, there’s venture capitalist Marc Andreessen the other month admitting, almost psychopathically, that he has zero levels of introspection because we must look forward to the future and not whine about the past. (My Hard Reset colleague Alex Shultz wrote about this the other month.)

But while Alex Karp (and probably the others) defiantly say to Wall Street analysts that it doesn’t matter if they’re arrogant because they’re “right all the time,” this “being right all the time” excuse for arrogance may not always be the case! Today, eBay rejected Ryan Cohen’s acquisition attempt, citing “the uncertainty regarding your financing proposal,” along with operational risks and the debt load that would result from the proposed transaction.

It remains to be seen whether CEO handlers can tamp down on this sort of unfiltered shooting from the hip that the CEOs may see as charming. But despite the brain chemistry of these CEOs perhaps being fundamentally altered by sycophant advisors, the isolation of extreme wealth, drugs, and the idea that they can do whatever they like with money—the court of public opinion still matters.

Share

What we’re paying attention to…

A criminal hacking group used AI to detect a bug and bypass two-factor authentication, according to Google.

One writer in Hollywood shares how she makes ends meet: “For screenwriters like me—and job seekers all over—AI gig work is the new waiting tables.”

The threat of AI may not be in replacing workers, but in becoming their bosses and surveilling them, according to one academic.

A lawsuit filed against OpenAI alleges that the Florida State University shooter messaged ChatGPT thousands of times before acting out his plan. ChatGPT, the lawsuit says, encouraged him to target children.

Apparently Palantir has been trying to become a lifestyle brand?

And on that note, Palantir co-founder Jon Lonsdale has gone on X to allege that Nick Kristof’s must-read piece on the rape of Palestinians by Israeli soldiers is “propaganda” meant to “smear Jews.”

An Oxford philosopher’s book explains the collective obsession with prediction in algorithms and now prediction markets, rooted in power: “There is a core problem with this imperial advance, Véliz argues: the human practice of prediction is not about discerning truth, but exercising power. Predictions are intrinsically probabilistic, but are accepted as statements of determinate fact because they fulfil the psychological function of assuaging humanity’s innate anxiety about an uncertain future. As a result, they are also inherently wishful — motivated reasoning, rather than objective science.”

Matías, a 34-year-old delivery worker in Buenos Aires, works six days a week, usually for long shifts. “I work six days a week, between 10 and 12 hours a day,” he told me recently. “On average, each hour is $9.50 (USD) per hour”. His story captures a central tension in platform work: the promise of flexibility often coexists with long hours, unstable earnings, and app-based systems that shape how work is assigned, priced, and controlled.

In a few weeks, governments, employers and workers will gather at the International Labour Conference (ILC) to negotiate what could become the first global standard on platform work. The agenda is full of technical terms—scope, definitions, algorithmic management. But beneath all that sits a much simpler question, one that will ultimately determine whether this new standard changes anything at all:

Who counts as a worker?

Across the platform economy, millions of people log in every day to perform tasks they do not fully control. They do not set their prices. They do not choose how tasks are allocated. Their performance is constantly monitored, rated, and ranked. And, at any moment, their access to work can be reduced or cut off entirely, often without explanation.

None of this is hypothetical. It is the everyday reality of delivery riders navigating city traffic, drivers absorbing the cost of fuel and maintenance, and data workers labelling content behind the systems we call artificial intelligence. Much of this control is no longer exercised by a human manager, but by automated systems that decide who works, when, and under what conditions.

And yet, most of these workers are not legally recognized as workers.

They are classified as independent contractors — a category that has been at the centre of high-profile legal battles in places like California, but which in much of the Global South operates in contexts of high informality, where platform work often replaces formal employment altogether.

This is not a technical oversight. It is the foundation of the business model.

Employment status determines access to minimum wage protections, social security, collective bargaining, and protection against unfair dismissal. It determines whether platform work can be considered decent work at all. And that is precisely why it has become one of the most contested issues in the ongoing negotiations at the ILO.

Earlier stages of the process—documented in the ILO’s draft Convention and Recommendation on platform work—acknowledged the problem of misclassification and opened the door to mechanisms such as determining the existence of an employment relationship based on the reality of the work performed. But as negotiations have progressed, the language has shifted toward a more flexible, “principle-based” approach, leaving significant discretion to national frameworks.

On paper, this may sound like a pragmatic compromise. In practice, it risks something else entirely: legitimizing the gap between how platform work is organized and how it is regulated.

Because the defining feature of platform work is not flexibility, it is control. Control over pricing, over task allocation, over visibility, over access to income. Increasingly, this control is exercised through algorithms that are opaque, unaccountable, and difficult to challenge. The question is not whether these systems exist. It is whether they will be recognized as forms of managerial authority.

This is why the debate on employment status cannot be separated from the debate on algorithmic management. If a worker’s access to work is determined by an automated system, if their income fluctuates based on dynamic pricing they cannot influence, if their account can be deactivated without due process, then the distinction between “independent contractor” and “employee” begins to collapse.

Workers themselves have been making this point for years. What is new is the scale at which they are now organizing globally.

The Global Platform Workers Solidarity Project (GPWSP) brings together grassroots organizations of platform workers from 34 countries, spanning sectors that are often invisible in these debates: care and domestic work, home services, and the data labor that sustains AI systems. What is at stake is often framed as a trade-off between flexibility and protection, or between innovation and regulation. But this framing is misleading. Recognizing workers as workers does not eliminate flexibility. It simply ensures that flexibility is not built on the erosion of fundamental rights.

And even beyond the employment debate, a growing number of worker organizations have been making another argument for years: if governments and companies insist on treating platform workers as self-employed or “own-account” workers, then fundamental rights and protections should not disappear as a consequence.

This is not a new discussion. Organizations representing informal workers, domestic workers, street vendors, and home-based workers have long argued that labor rights cannot depend exclusively on the existence of a formal employer. Millions of own-account workers operate outside traditional employment relationships while still requiring collective bargaining rights, social protection, occupational safety protections, and recognition under labor frameworks.

This matters because platform companies are constantly restructuring and redefining themselves to avoid being recognized as employers. If labor protections remain tied exclusively to narrow definitions of employment, workers risk permanently falling through regulatory gaps while companies continue evolving faster than the law.

The point, ultimately, is not semantic. Whether platform workers are recognised as employees, dependent contractors, or own-account workers, there should be no model of digital work that allows people to be excluded from fundamental labor rights, collective representation, due process protections, and basic social security.

A digital economy that depends on denying labor protections is not innovative. It is extractive.

The ILO process represents a rare opportunity to set a global baseline—to define not just how platform work operates, but what kind of work we are willing to accept. But for that to happen, the Convention will need to confront the issue at its core. Not in abstract terms, but in concrete ones.

Because at the end of the day, the question is not particularly complex: If work is organized, controlled, and monetized in this way; if livelihoods depend on it; why shouldn’t rights apply to it?

But there are some benefits to listening to the trial on a live stream from the comfort of my own home. For one, when a witness or attorney speaks too quickly, I can rewind the stream. And when Elon Musk’s quasi-romantic partner Shivon Zilis testifies that she can’t remember an exchange of messages by employing the curious phrase, “It’s not in my neurons, but I see the words written here,” I can laugh out loud without pissing off Gonzalez Rogers.

Zilis is a venture capitalist who first joined OpenAI on a part-time basis in 2016. She rose the ranks there and became a board member in 2020, where she remained until 2023. She’s also held high-up roles at Tesla and Neuralink, and has four children with Musk. She seemingly lives with him and is probably romantically involved with him. I am hedging because Musk and Zilis do not know how to describe their own relationship; they act like they met on the dating app Feeld and still want to keep it a secret, even though everyone knows and no one cares.

“Shivon was uh…my chief of staff. And uh, yeah. Uh, yeah,” Musk said last week, when he took the stand. A day later, he tried again. “We live together and she’s the mother of four of my children,” he said. When asked if he and Zilis were romantically involved in February 2018, the month he departed OpenAI’s board, Musk responded, “I think so.” The period between 2018 and 2023—when Musk left OpenAI and Zilis spent some time on OpenAI’s board—has been the subject of scrutiny for OpenAI’s attorneys, who want to know what Zilis was telling Musk, and whether Musk was meddling in OpenAI’s business via Zilis.

Zilis took the stand on Wednesday, May 6, and proceeded to refute many of Musk’s tentative characterizations. She was not his chief of staff, she said. “There had been kind of like, a one-off at the offset, and then we were friends and colleagues,” is how she described… a one-night stand, I guess? What the hell is a “one-off at the onset?” At another point, an attorney asked Zilis if they could “agree that your relationship with Mr. Musk is important to you.” Zilis paused, then said, “Sure.”

Questions that were intended to elicit more information about how close Zilis was with Musk in the late 2010s and early 2020s were met with confounding answers. Zilis claimed that she spent lots of time with Musk not because of their are-we-or-aren’t-we dynamic, but because of her positions at Tesla and Neuralink. In Zilis’s retelling, Musk eventually noticed that she didn’t have any kids, and generously offered to “make a donation,” meaning IVF. She had twins in 2021, while still on OpenAI’s board, and Musk apparently tried to visit the kids “at least weekly,” she said. They’ve since had two more children and appear to be consciously coupled; Zilis was even a plaintiff on this very case for a brief period of time, and is represented by one of the same attorneys as Musk. Cute! Except she hilariously claimed that she didn’t know they share the same legal representation. Less cute.

During Zilis’s testimony, she sounded indignant about a Business Insider report that revealed Musk as the father of her children. In a different universe, I would wholeheartedly agree with her—that information is typically no one’s business. But it’s laughable to pretend that you are entitled to the same level of confidentiality when the world’s wealthiest man is involved. Especially given his tumultuous departure from OpenAI, an organization where Zilis had real sway as a board member. Zilis admitted that after she was told about the impending Business Insider story, she called her dad, and then Sam Altman. Fun conversations, I’m sure.

It’s hard to believe Altman, Brockman, and others at OpenAI were completely blindsided by Zilis’s disclosure. It seems to me they either had a hunch about what was going on and chose not to address it, or perhaps they were just as confused as I am now about the Zilis-Musk love story. According to Zilis, Altman recognized her skills as an Elon Whisperer, even if he may or may not have caught onto why she was so close to Musk.

“Historically, I’d been very good at doing that,” she said of her ability to facilitate communications between Musk and the other OpenAI founders. “Candidly, they’d been kind of bad at speaking to each other.” Later, she added, “There were often tricky topics that maybe wouldn’t lend well over text or they wanted to make super sure they had Elon when he was in a good headspace and had time to think.” I can’t imagine why it would matter whether Musk was in a good headspace, given that during his own testimony, he said, “I don’t lose my temper. I don’t yell at people, basically.” Case closed.

Zilis was not as obnoxious as Musk on the stand, but she emulated his excruciating sense of humor and grandiosity. She got into AI because it’s going “to be the most influential thing humanity creates,” she matter-of-factly stated. She claimed to work 80- to 100-hour weeks in the late 2010s, and characterized Musk’s work ethic as “maniac mode, he’s just relentless.” She tried out a joke about OpenAI’s attorney having a Canadian accent; it did not land.

Zilis testified that one of her nebulous responsibilities for Musk was to solve “bottlenecks,” which is ironic because she made sure her cross-examination was an unavoidable traffic jam. Over and over again, the OpenAI attorney mentioned a section of an exhibit for Zilis to read, and Zilis took a comically long time to locate it and read it back. Frequently, she asked the attorney to repeat the question. The tougher the questions got, the longer the pauses. “If you said we did, we did,” “the words say that, I don’t recall saying that,” and “I see it says that in this draft” were three of her favorite filibusters.

Before Musk left OpenAI, he angled to create and control its for-profit subsidiary—there’s plenty of testimony and submitted evidence that affirms as much. But even on this relatively indisputable point, Zilis was pedantic. “There were versions in which that was true,” she said of Musk’s interest in a for-profit subsidiary, alluding to other proposed plans that were bandied about in 2017 and 2018. Worst of all, between her deposition last year and her testimony this week, Zilis apparently had an epiphany about previously forgotten emails. “Your long-lost memories have been recovered,” OpenAI’s attorney quipped, reminding Zilis that she had no recollection of these emails during her deposition.

The most telling exhibit presented during Zilis’s testimony was a series of text messages between her and someone saved in her contacts as “Shahini Rubicon Fluffer.” (No idea.) The messages were exchanged on February 25, 2023, according to the exhibit.

“Have to resign openai board btw,” Zilis wrote. “E’s effort has become well known,” she said, referring to Musk’s own AI ambitions at the time. “Sam [Altman] called this morning and I knew what it was about before he called. When the father of your babies starts a competitive effort and will recruit out of openai there is nothing to be done,” she added.

She was 100 percent right in those messages, where she demonstrated that she’s much more self-aware than she let on during her testimony. Surely, deep down, she knows that her unusual circumstances make her an unreliable narrator by default. Rather than owning it, and at least acting like she was interested in transparency, she clammed up and treated the jury the same way she treated the other OpenAI board members: like they’re fools. This time, I don’t think it’s going to work out for her, or for Musk.

One more week of trial to go. On Monday, May 11, OpenAI co-founder Ilya Sutskever and Microsoft CEO Satya Nadella are expected to testify.

To be clear, the first-order picture is alarming enough. A 2025 working paper from RAND found that 84 percent of federal revenue is tied to labor — not just payroll taxes, but the income taxes, consumer spending, and local sales taxes that flow from people having jobs and spending money. The whole funding architecture of American government runs on workers having jobs. AI threatens to remove that foundation without replacing it.

And it isn’t just the flow of money in the market that’s driving this. It’s also policy choices: in July 2025, Congress passed the One Big Beautiful Bill Act, which permanently restored 100% bonus depreciation for qualified business equipment. That’s a mouthful. Here’s what it means: a company that spends $10 million on AI systems can write off the entire cost immediately. A company that spends $10 million on salaries can write off salaries too — but the tax code now explicitly makes the machine the cheaper call. Congress modernized the treatment of capital. It left workers’ training and transition support behind.

The job losses that follow aren’t purely the product of AI getting smarter. They’re also in part the product of deliberate decisions about what gets subsidized.

Now, on to some second-order effects workers (and lawmakers) should understand right now. Here are three places the damage travels after the pink slip.

1. YOUR KIDS’ SCHOOL

Schools in most states depend on a combination of property taxes and state income tax revenue, both of which are tied to employment. When workers lose jobs, income tax receipts fall. When working-class neighborhoods hollow out — which is what happens in places like Detroit, Duluth, or Pittsburgh after major industry displacement — property values soften and the tax base underneath the local school budget shrinks.

An estimate from South Carolina, published in the Post and Courier, found that 800 AI-attributed job losses in that state in 2025 translated to roughly $700,000 in lost school tax revenue. That’s one state, one year, at relatively early displacement levels. The structural issue the estimate points to: a company that replaces 40 human workers with an AI system loses $2.4 million in payroll expenses and the government loses the tax revenue that came with those salaries. The AI system doesn’t pay income tax. It doesn’t pay into Social Security. It might pay some property tax if it has a data center in that state, but that’s nothing compared to the property tax that hundreds or thousands of employees generate from the housing they buy or finance with their wages.

Some of the states most exposed — states like Michigan, Pennsylvania, Illinois — all have significant concentrations of exactly the administrative, customer service, and mid-level professional roles that automation is hitting hardest. They also have school districts already operating on thin margins after decades of deindustrialization. This wave is hitting the same communities twice.

2. THE RETRAINING TRAP

The standard response to AI displacement is: “new jobs will be generated.” And the next sentence tends to be “workers will need to retrain.” What goes unsaid is that the systems designed to fund that retraining are the same ones being hollowed out by displacement itself.

The federal Workforce Innovation and Opportunity Act funds roughly $1 billion a year in retraining, career counseling, and wage subsidies for displaced workers. That sounds significant until you understand the J.P. Morgan Private Bank estimate that AI is on track to displace approximately one million jobs per year over the next decade — a pace roughly five times faster than globalization managed at its peak. One billion dollars spread across a million displaced workers is a thousand dollars each. One community college semester. Not a career.

State unemployment insurance funds face the same kind of math. When AI automation renders an entire job category obsolete — not one factory but an entire occupational category — displaced workers can’t simply wait for conditions to improve in their sector. They need to move to a different sector entirely, which takes longer and costs more than current state unemployment-insurance systems are designed to support. The AI Frontiers analysis of this structural mismatch does a great job of explaining the problem: the programs we built assumed temporary disruption within industries. AI is producing potentially permanent disruption, and across them.

There’s a ghost in this data that makes it worse. A Fortune analysis found that nearly 75% of AI-displaced workers don’t apply for unemployment benefits at all — in part because union membership, which is one of the best predictors of whether someone even knows to apply, has fallen to a historic low of 9.9%. The people most likely to be displaced quietly, without applying for help, are also the people least likely to have been told how to ask for it. They disappear from the unemployment statistics, which makes the headline numbers look manageable. They don’t disappear from poverty statistics. (Don’t be part of that number: have a look at “If You’re One of the People Worried About This,” below.)

3. THE PENSION CLIFF

This one moves slower, but when it arrives it’s going to swamp us.

Public pension systems in states like Illinois, New Jersey, and Pennsylvania are already carrying significant unfunded liabilities. They are funded by a combination of investment returns and ongoing contributions from active public employees and their employers. If state and local budgets come under pressure from declining tax revenues — which, as we’re discussing, is what AI-driven displacement of the private-sector tax base produces — governments face a familiar choice: cut services, raise taxes, or reduce contributions to pension funds. History is fairly clear that the last one tends to happen first.

Illinois went from a BBB- credit rating with a negative outlook in 2020 to an A-minus today — a genuine recovery that required years of painful fiscal discipline. Any erosion of the state’s income tax base from large-scale private-sector displacement would pressure the very gains that recovery was built on. New Jersey, Pennsylvania, and Michigan face versions of the same dynamic.

For workers who’ve spent decades in private-sector defined-contribution plans — also known as 401(k)s — the picture is different but not better. If AI displacement pushes workers out of well-paying jobs and into lower-wage positions (which the data suggests is the more likely outcome than outright unemployment), those workers reduce their contributions, potentially for years. The compounding loss over time is difficult to recover from. A 55-year-old account manager displaced into a gig economy role at 60% of their prior income doesn’t go back to maxing out their 401(k).

A Number Nobody’s Counting

There is a statistic at the center of all three of these problems that is rarely discussed, because it’s a missing number, rather than one of the many that are shouting at us.

Labor force participation — the share of adults who are either working or actively looking for work — is projected to fall from 62.6% in 2025 to around 61% by 2030, and as low as 55% by 2050. The key detail: unemployment rates are expected to remain roughly stable over the same period. Two numbers moving in opposite directions? How can that be? It means people are leaving the workforce entirely rather than registering as unemployed.

People who stop looking for work don’t show up in the unemployment rate. And they also don’t pay payroll taxes. They don’t contribute to Social Security. They don’t fund their kids’ schools. But they do become eligible for Medicaid, SNAP, and social services — the programs whose funding depends on the tax base of people who are still working.

The headline unemployment number is going to look fine for a long time. Don’t be fooled. Because the systems underneath that stable-seeming number are going to be anything but stable.

If You’re One of the People Worried About This

One of our most-read pieces on Hard Reset right now is our guide for newly unemployed tech workers — which tells you something about who’s reading and what they’re scared of. So a direct note: the story above is about structural forces. Here’s what those forces mean at the level of your household.

If you’ve been displaced or are worried about displacement, the single most important thing to understand is that the retraining programs that exist are underfunded for the scale of what’s coming, but they exist…and most people don’t use them. The federal WIOA program funds career counseling and training vouchers through every state’s workforce development system — find your state’s program here. If you’ve been laid off, apply for unemployment insurance immediately even if you’re not sure you qualify — the 75% non-application rate I cited above means there’s a real chance you’re leaving money on the table.

The pension and Social Security numbers above aren’t a reason to panic about your future retirement. But they are a reason to treat any assumptions you’ve been making about those programs as the shakiest part of your financial plan, and to adjust your savings accordingly where you can. The systems are not broken yet. But they are being stressed in ways that haven’t been honestly described to the people who depend on them. And that’s us.

Further Reading

• Jennifer Schultz, “If AI cuts jobs, it would also threaten Social Security and Medicare” — Minnesota Reformer, April 2026. The piece that inspired me to put this one together; worth reading in full.

• Carter C. Price and Akshaya Suresh, “Federal Revenue When AI Replaces Labor” — RAND Corporation, 2025. The working paper on what happens to the tax base in each displacement scenario.

• Anton Korinek and Lee Lockwood, “The Future of Tax Policy: A Public Finance Framework for the Age of AI” — Brookings Institution, February 2026. The most rigorous treatment of what “ambitious fiscal innovation” would have to look like.

• “Why companies using AI should pay more property taxes — especially school taxes” — Post and Courier, December 2025. The school revenue estimate.

• “AI job disruption may be compounded because nearly 75% don’t apply for unemployment benefits” — Fortune, March 2026. The non-application rate and union membership data.

• “A Proactive Response to AI-Driven Job Displacement” — Mercatus Center, October 2025. Source for the OBBBA tax code change and its implications for automation incentives.