Think of Meta’s AI strategy like a shitty DJ set at a rave: lots of turning dials up and down while seeking a positive crowd reaction. Do you like our slightly cheaper AI models? Ehh. Should we sell cloud infrastructure? Mehhhh. Do you like our pervert glasses? Muted cheers from pockets of ravers who aspire to harangue strangers at the park. What about creating AI-generated images based on other people’s public Instagram accounts? Boooo. (That obviously invasive feature lasted all of three days before Meta yanked it.)

As Meta sorts through a pile of haphazard consumer products, it’s also trying to differentiate itself and get its AI branding straight. Somewhat understandable, given how everyone else has obvious, exploitable baggage, and Nvidia’s Jensen Huang is lobbying for open-source models that threaten the long-term viability of Anthropic and OpenAI.

OpenAI is the pioneer of the industry, for better and mostly for worse—co-founder Sam Altman is regularly drawing unwanted attention to the organization’s increasingly iffy IPO valuation while struggling to explain away its endless spending. Anthropic is taking fire from every direction for being woke, for being closed-source and expensive, and also because CEO Dario Amodei is far from the level-headed moralist he often claims to be. Google and Microsoft are the behemoths force-feeding their models into search engines and apps; best-case, they are hoping that your annoyance at Gemini and Copilot turns to indifference followed by dependency. SpaceXAI is an absolute mess, but it does have a stranglehold on gamblers who treat the stock market like it’s one big DraftKings parlay. Chinese models, meanwhile, are rapidly gaining prominence because they’re much cheaper and more efficient.

Where does that leave Meta? It appears Zuckerberg and the company’s public relations team are turning up the dial on AI Optimism.

Last week, Zuckerberg released an advertisement purporting to represent Meta’s buoyant AI ethos. “We’re focused on giving every person the tools to reach your full potential and making sure the benefits of technology are distributed to everyone,” he wrote. In the ad, a plain-speaking fella’s voice-over decries how “some people will have you believe AI will make us less connected. That it’s gonna leave us behind.” Good news, he says: “We couldn’t disagree more. Call us optimists. Call us dreamers. Just as we’ve always done, we’re betting on people.”

In a follow-up Wall Street Journal op-ed titled “The AI Future Is for Everyone,” Zuckerberg doubled down on a message of positivity and inclusion. The sixth-wealthiest person on earth, whose corporation controls platforms with roughly 4 billion monthly users, says he believes in “delivering personal superintelligence to everyone.” The op-ed was also a perfect platform to take some potshots at Altman, Amodei, and their respective organizations.

“It is surprising that the discourse from many of those who are developing artificial intelligence is so filled with doom,” Zuckerberg wrote. “I don’t understand why anyone who believes that AI will eliminate most jobs and much of humanity’s relevance would rush to build that future. The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems dangerous.”

Zuckerberg, like Palantir’s Alex Karp earlier this month, is going knives out against Anthropic and OpenAI, whose AI-will-kill-us-all-unless-its-me-controlling-everything shtick is absolutely, transparently self-serving. Altman ruminates about best- and worst-case scenarios for superintelligence approximately once per week; it’s a way for him to build hype about AI products that otherwise are not dazzling consumers. Amodei swears he only pushes closed-source models because otherwise, dastardly China will beat us in the AI race and “achieve permanent military superiority or perpetrate incredibly deep repression of their own people.” Left unsaid in his spooky “China wins the AI race” hypothetical is that Anthropic falters from its frontrunner status.

It does not take a genius, or a megalomaniac, to point out these discrepancies. Zuckerberg might even be less credible than Karp—though I don’t care to compare them in great detail—when he claims in his op-ed that he’s driven by “more people sharing greater prosperity, health and freedom.” That certainly sounds nice if you don’t think about it for more than five seconds.

Okay, I thought about it for six seconds. Zuckerberg believes that human prosperity will result from decentralized superintelligence, an outcome that requires OpenAI and Anthropic to implode. Do you know who stands to benefit in that scenario? Meta and Mark Zuckerberg.

Did Zuckerberg think about “more people sharing greater prosperity, health, and freedom” at his own company when it installed mandatory employee tracking software? What about when Meta reassigned thousands of its engineers to do monotonous AI tasks? What about two months ago, when Meta laid off 10% of its workforce? What about when he turned his Palo Alto compound into an allegedly illegal school so his kids wouldn’t have to be taught around Bay Area riff-raff? Or when he reportedly became one of Hawaii’s biggest landowners, so he could construct a compound complete with a bunker?

“Putting power in people’s hands to pursue their own aspirations is how humanity has made the most progress,” Zuckerberg wrote. Is that how he’d describe the content moderation jobs Facebook and its subcontractors doled out to workers who developed PTSD from categorizing the most disturbing images and videos on the internet? Is that how Zuckerberg’s buddy Alexandr Wang, now Meta’s chief AI officer, would talk about the “jobs” he “provided” in some of the world’s poorest communities? Check out this empowering passage about AI data annotation company Scale, which was cofounded by Wang, from Karen Hao’s book Empire of AI:

“One group of eight workers in North Africa said Scale reduced their pay by more than a third in a matter of months. At least one worker was left with negative pending payments, suggesting that he owed Scale money. When the group attempted to organize against the changes, the company threatened to ban anyone engaging in ‘revolutions and protests.’”

Scale eventually blacklisted contractors in Kenya, and did so via a cold, blanket email that their paltry income source was kaput, effective immediately. Not a great outcome for workers who rely on money to “pursue their aspirations,” but don’t worry: Zuckerberg says more jobs are coming from Meta and other benevolent AI providers. “[If] superintelligence is widely distributed, then I believe we will see more jobs in the future, not fewer,” Zuckerberg wrote.

PAID PROMOTION

This year, the Roosevelt Institute is developing a bold vision for the good life: a policy agenda that builds economic security and cultivates the institutions and leaders needed to deliver it. Through research, residencies, and events, the Institute will propose ways to address the affordability crisis by lowering costs and increasing incomes, as well as give people the time and agency needed to live full and meaningful lives.

The Good Life Residents Program is an initiative focused on developing new leaders and advancing people-centered ideas across several of the Institute’s economic policy priorities.

Residents in the AI and Markets program will produce short, original policy research and writing, and engage in structured dialogue with peers and senior advisors on what policy levers the government could use to ensure that private investments drive a diversified AI sector and a deconcentrated economy that serves all Americans. Each residency will culminate in a public event that amplifies these new ideas and voices in a dialogue to discuss and debate various national- and state-level policy options that enable the good life.

See more information and application links on our website: https://rooseveltinstitute.org/careers/good-life-resident-program/

The “AI optimism” spiel might prove to be relatively effective in the short-term. I can easily see Altman and Amodei taking the bait and saying something apocalyptic that proves Zuckerberg’s point. But aside from landing a few barbs, Zuckerberg is going to run into the same problems as his competitors. He might even be creating more glaring issues for Meta. Not only is he just as self-serving as his peers—he’s also trying to occupy a deeply undesirable lane. The public is not “pro AI” or optimistic about its outcomes, which is part of why other tech leaders oscillate between “best technology ever” and “AI might destroy humankind.” They have selfish aims, but they’re loosely aware they can’t sing AI’s praises all the time. No one wants to hear that crap. Zuckerberg, it seems, is going to try for a nonstop positivity campaign anyway.

Funnily enough, Zuckerberg is well-aware that AI is unpopular. Otherwise, he and Meta wouldn’t have spent nine months secretly plotting with Louisiana lawmakers about the construction of a $50 billion data center; instead, Meta would have negotiated a potential deal in the public eye. As the New York Times recently uncovered, this gargantuan data center project has caused skyrocketing rent prices in nearby communities, where construction workers are swooping in and looking for housing. A 41-year-old woman was evicted from her trailer park home with a month’s notice when her rent jumped from $250 to $1,495 per month, the Times reported. “That was my home for over a decade,” she told the Times. “Not everyone is making Meta money.”

She, too, will be happy to learn that Meta will someday provide her with personalized superintelligence. Unless, of course, Zuckerberg’s conveniently rosy projections don’t materialize. I’d be curious to hear Mark “AI optimist” Zuckerberg’s explanation for why, if anything goes awry with Meta’s Louisiana data center, “everyone else took on most of the potential downside,” the Times reported. Zuckerberg “gave his company an escape hatch to get out of it years before its partners,” according to the same investigation.

That sure doesn’t strike me as an optimistic outlook on the AI industry. More so, it reads like a desperate, power-hungry hypocrite who’s still turning dials up and down, hoping something hits.

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

• Sam Altman was asked whether other companies—not just Hugging Face—could’ve been hacked by OpenAI. He shrugged and said, “I mean, there could be, yeah.”

• DoorDash is beginning to hype up a drone delivery service called “DoorDash Air.” It does not sound imminent, but it has cleared some important regulatory barriers. I for one am excited for the first time I’m hit from above by an unknown substance, and cannot ascertain whether it’s bird poop or someone’s aioli dip.

• eBay issued a statement apologizing to Ina and David Steiner, the founders of the site EcommerceBytes, for subjecting them to a terrifying harassment campaign in the late 2010s. The Steiners wrote critical posts about eBay, which did not go over well among former eBay executives and employees. “It began with anonymous, threatening Twitter DMs to the Steiners, then quickly escalated to deliveries to the couple’s home,” Wired reported. “These deliveries included a bloody pig mask, pornography, a package of fly larvae and live spiders, and another package containing live cockroaches. Other deliveries were themed around death, including a copy of a book about grieving a spouse and a funeral wreath from a local florist.” In addition to the public apology, the Steiners reportedly secured a $55.7 million settlement from eBay.

• Every few months, a startup debuts a personal AI assistant that will remind you to brush your teeth, drink water, and eat food, but with personality. “Hey dummy, have you gone to the bathroom today?” That kind of thing. The latest example is Orchid, which released an ad where a 20-something couple who clearly should not be together has separate chats with their AI assistant about their anniversary. The girlfriend in the ad commands the AI assistant to remind her stupid boyfriend that he forgot their anniversary. The AI assistant does so, and then books a reservation for the stupid boyfriend. Then the couple has a nice anniversary dinner, despite the girlfriend knowing that her boyfriend didn’t remember their anniversary in the first place.

Y Combinator says they hand-selected the most promising technical builders in the world for the two-day program that featured world-class panelists, demos of products by YC companies, and access to over $25,000 in AI compute credits. The program offers opportunities for professional capital and legitimacy. The side events and afterparties are an introduction to Bay Area Tech’s social circles and ranks.

Speakers reinforced throughout the weekend that proximity to this event is proximity to success. Jensen Huang reportedly told attendees, “If you’re here, you’re gonna make it.”

At one afterparty that took place atop the famous tech workspace Frontier Tower, very few attendees were actual YC School graduates. Most were other young Tech hopefuls looking for proximity to the event if they couldn’t gain entrance. One guest who did have the Startup School badge — one that had gotten him into more exclusive and luxurious afterparties that required them — told me he’d gotten more out of the networking and social aspects of the weekend than the headlining talks. He was in the Chase Center for Sam Altman and Garry Tan’s closing conversation. He described it as “nothing revolutionary.”

YC has since posted their talk online. Following their flashy entrance, they discussed tech as community, Altman at one point posing its “spirit of helping each other” as an upside to “all of the negatives of the culture of the Bay Area.” Tan reinforced that attendees are a “most rarefied set of people,” and promised that room would make up overlapping professional and social circles for a long time.

They then talked AI safety, and just days after an OpenAI test model went rogue in an unprecedented autonomous hack, said that anyone who isn’t a little worried “is not taking this seriously enough.”

Altman acknowledged his company has made some massive errors, and said, “It’s really important we don’t have a loss of control accident with AI; it’s really important we don’t have power be too concentrated in a small number of models or companies, but diffused throughout the economy so people can defend themselves.”

He added, “You can help on the concentration of power issue simply by starting a successful startup. That’s the only thing you can do.”

There’s obvious irony in Sam Altman speaking against the consolidation of power in tech at an event that brought together power in tech’s present and future. The YC Startup School’s objective is elite access for new names, hand-picked by the top names. That’s not inherently malicious and exclusionary; institutions that drive the cultivation of networks and access to Silicon Valley have always done that. It is a reminder, albeit from Altman, of what concentrated power in AI means in sociopolitical contexts.

A few miles south, another institution actively shaping the future of Tech is encouraging critical scrutinization of the effects of AI in tandem with its development. Stanford Critical AI, an interdisciplinary coalition of academics from around the world, recently released the opening statement of their ongoing effort to match AI development with critical inquiry into the technology’s effects on power, ethics, and governance.

The “Manyfesto” acknowledges Stanford University’s longstanding status and power in Tech and in academia. Stanford’s role in AI’s development has been crucial, and the group says they have a responsibility to critically analyze the impacts and dangers of AI.

“There is no Silicon Valley without Stanford, and no Stanford without Silicon Valley,” the opening statement says. “We write from a pivotal time of political and social unrest. In the United States and in the world, democratic institutions are eroding. Polarized discourses fuel anti-democratic movements. Tech manifestos recently released from Silicon Valley intend to mainstream techno-fascist and white supremacist ideologies, justified on the basis of the AI arms race and imperialist claims. Again, we can course-correct. This is not the only trajectory.”

Share

The progression, safety, and distribution of power in AI is more than a matter of entrepreneurship. Researchers are bluntly telling us so. AI is a sociopolitical technology, accompanied by a sociopolitical work culture. Acknowledging that is a crucial first step to safety in the age of AI.

The Manyfesto isn’t a means to an end — more is to come from the project. I’m currently in the process of interviewing co-founders of Stanford Critical AI and co-signers of its opening statement. I look forward to sharing my conversations with them next week. So far, they’ve been refreshingly hopeful.

Musk, who has not had any on-record interviews in some time, sat face-to-face with Zanny Minton Beddoes, The Economist’s Editor-in-Chief, for what he likely expected to be a cordial conversation. The interview naturally turned antagonistic the moment that Musk was confronted with information that didn’t adhere to his reality—for example, the idea that he was sowing discord on X, or that USAID cuts at his behest had led to deaths in Africa.

In order to psychologically grasp the idea that people dislike him, Musk parroted the talking point that even if he was disliked, the mainstream media was “hated” and “loathed” far more than he was. “You” are more loathed than I am, he said, pointing to the journalist as an emblem of the mainstream media; she responded that most people probably didn’t know who she was.

He admitted that while some people may not like him, more people like him than hate him, pointing to his 250 million X followers (neglecting to mention that he owns the platform and that the platform may point people to his account).

Beddoes interrogated a few of his assertions that he stated as objective fact—for instance, that money will not matter to humans in ten years. How might humans live? Musk sort of brushes off the concern, retorting that the Treasury will simply cut checks for people as a form of universal basic income. Musk said that he may have gotten a little too involved with politics, but then predictably blamed much of DOGE’s problems on the DOGE workers themselves.

I posted a video criticizing Musk’s interview on Instagram to a deluge of comments from his swooning followers. Whether they were bots or miserable people or both, it was striking just how many people were genuinely angry about a critique of Musk, especially compared to similar critiques of other billionaires in tech (Zuckerberg, Bezos, and Ellison don’t seem to have the same blind cult support). People lobbed insults about my lack of intelligence compared to Musk; one person claimed that I was jealous because I would never contribute anything close to society compared to what Musk has. Even the CEO of Juice Press was in my comments telling me I needed to “stop” with the dramatic pauses, and that he vehemently disagreed with me.

Meanwhile, everyday investors who sunk money into Musk’s SpaceX are losing money as the share value plummets, bringing Tesla’s value down with it. xAI is suing Minnesota over legislation that would fine a user $500,000 each time they create explicit deepfakes—which probably shouldn’t be a controversial law. People are indeed dying from USAID cuts.

Why are people so emotional? The only way to square the support, in my mind, is that Musk has a Trump-like hold on people in a greater-than-god sort of way. The idea of a “misunderstood genius” still grips the public consciousness, because people are desperate to cling on to the idea that some “brilliant” guy has it under control.

What I find fascinating is that despite loathing the mainstream media—and claiming that we all do as well—Musk still wants to sit down for an interview.

What we’re paying attention to this week…

A deep-dive into Trump’s World Liberty Fund, which one crypto magnate—once aligned with the Trumps—is alleging as fraud.

AI companies are increasing their spend in Washington to deter regulation: “OpenAI nearly doubled its federal lobbying expenditure to a record $2.22mn in the first half of 2026, compared to last year, while Anthropic nearly tripled its spending to $3.53mn, according to federal disclosures. Google, along with Microsoft, spent as much as in any quarter since 2020.”

What happens when 12,000 tech employees at companies like SpaceX and OpenAI become multimillionaires overnight. (It concentrates in fewer places, not more.)

We wrote a few weeks ago about the voluntary retirement of the white collar worker. Here’s why more companies are opting for voluntary buyouts as part of layoffs.

A Netflix employee opened up about his ketamine usage at a “trust retreat.” He was subsequently fired.

The Trump administration is banning some of China’s humanoid robots.

And yet these days, it’s not just an over-zealous government that’s watching us (though they are). It’s us watching each other. Recording each other. Watching ourselves. And even now filming ourselves being watched.

In the last few weeks, amid the launch of the Meta smartglasses, I’ve been struck by how culturally acceptable it’s become for us to contribute to our own surveillance—oftentimes without knowing exactly where and how our documentation and data is being stored, consumed, or analyzed. Because the recordings of us seem so inevitable—on city streets, in stadiums, at airports, on our phones—some have decided to embrace it to the full extent it can be embraced. Myself, in certain ways, included.

Last week, an English influencer and model posted an Instagram carousel of herself recorded through the lens of her home security camera. The fish eye angle of the photos makes us feel like we are watching her rather than seeing her (dubbed “surveillance chic” by a writer trying to make sense of this cultural phenomenon). Another influencer is using her building’s elevator security camera to document daily outfits; others posting Ring camera images say they are subversively winking to the surveillance state.

Workplace culture has accepted surveillance, too. Companies are adopting products that claim to track and hack productivity, while developing entire databases of workers’ digital behaviors. Job interviews are being recorded by both candidates and employers—and firing decisions are made based on the aforementioned recordings of people’s daily keyboard strokes. (This is not lost on the workers developing these surveillance products, given they are surveilled at the very jobs where they develop the surveillance products.)

Meanwhile, the companies that are surveilling us are trying as hard as they can to upend their own past historical identities as spy-like, opaque, sinister, in bed with some deep state. These days, they are trying to be something else altogether: cool. Cute, even.

Palantir, the company which helps ICE track the location of immigrants, launched a bright blue “chore coat,” emulating a “workman’s” coat that may affiliate Palantir with something more industrial, more down-to-earth. The Palantir employee who developed the coat actually describes himself as being responsible for “vibes,” and can be seen promoting the “elite” nature of his merch on X.

One woman, a tech worker, even advocated for the recording of others as a psychological imperative of some kind. She told the Wall Street Journal that she was “triggered” because someone asked her to stop recording them—also sharing that she has taken to recording job interviews and even first dates as part of her daily routine.

Online stalking is not new, and neither is the recording of people or online behavior analysis. But there is a fusion of forces today that feels unprecedented.

And that fusion is between the proliferation of surveillance cameras feeding our locations and actions into an increasingly authoritarian state—and a social and digital culture that is encouraging everyday citizens to document everything. It’s allowing us, paradoxically, to conflate cultural trends of expression with the acceptance of surveillance.

Because although elevator selfies can feel cute in the moment, the intentions of the technology underlying all the networks we feed are nebulous. We actually don’t know what is happening necessarily with our information snapped by cameras that are not ours. Is the goal for us to record every single interaction and movement of our day so we can better train AI or some future Metaverse (given the first one succeeded so spectacularly)? Is it to allow the government and advertisers to keep dossiers on our locations, sentiments, and personalities? Is it merely to show public company shareholders that they are innovating by developing gadgets that we may not actually need?

And even more questions abound about the future of consent. How will two-party consent laws apply when Meta glasses record someone who is inebriated and confessing something private during an intimate conversation? What will companies do when they possess footage of sexual interaction that was not meant to be recorded? What employees will have access to information that was not meant to be preserved or accessed for play-by-play documentation?

Candidly, it’s psychologically uncomfortable to ask these questions. And the companies know this. It’s awkward to ask why we (myself included!) are exposing ourselves to such a degree—or why we are documenting things that may have previously required a subpoena or sophisticated technology to access.

To be clear, I’m not saying that all documentation or expression for online consumption is bad. But I also think that in a world where surveillance no longer seems like a choice, perhaps surveilling ourselves and others more and more is an impulse to retain some semblance of control or agency over it all.

We’ll start to see the answers to the previous questions play out, sometimes in very creepy ways. The other week, Waymo called the police on two teenagers who were drinking and playing with a toy gun in one of their driverless vehicles. The teens were locked in the car, observed by a remote Waymo employee, until the police arrived. And last week, it was reported that a Milwaukee police officer searched the license plate number of someone he was dating over 200 times. He also searched for his partner’s ex.

These companies’ products feel so ingrained in our daily movements that they seem all but inevitable as part of our act of living and working. But ultimately, the idea that we’re taking back our personal agency by surveilling ourselves and others 24/7 is not true. The only agency we may be definitely contributing to at the end of the day is law enforcement.

Share

What else we’re paying attention to…

A rural New York school district is deploying a humanoid teacher this fall—manufactured, bizarrely, by a crypto sex doll company.

The surge of AI in workplaces may be the catalyst that knowledge workers needed to unionize.

The CEOs of the two largest prediction markets appear to despise one another.

US Army employees are burning through their AI tokens.

The FIRE movement seemed to this journalist like a fringe movement of people who didn’t want to work. Turns out, it’s more of an intentional group that wants to reject hustle culture.

AI embellishes images of birds, leading to confusion among birdwatchers. On that point: AI is also duping homebuyers into thinking kitchens are renovated.

I was amused by this headline! Meta is testing an AI bedtime story app for people who have no imagination

OpenAI’s model escaped containment, hacking into the AI research platform Hugging Face in order to steal answers to a test they were being graded on.

Share

Tech founders have no real need for a third party to get the word out about their companies. It’s faster, easier, and cheaper to self-produce and promote on social media. And, you get to control your own narrative.

That doesn’t make legacy publications obsolete. I’ve found it actually makes their appeal stronger.

Share

We were all reminded of this last week with a viral post by Vice President of 1X Dar Sleeper. Their humanoid home task robots scored a feature on Wired, where the journalist discussed the bizarre and suggestive marketing tactics the company was using to promote the launch; the article has since been updated to tell readers 1X had taken down aspects the author originally referenced. Sleeper and other executives were unhappy.

He was disappointed and regretful, but not surprised, since he’s found “dishonesty and malice” rampant in journalism. His call to action to let “the technological revolution… inspire a journalism renaissance” instead of letting it “fall into further decay” made me wonder, as outbursts with this motif always do, if there’s nuance to his preferred solution to the industry beyond us writing whatever our subjects want (which is neither ethical nor realistic).

I read the Wired article and noted that 1X had, in its aftermath, scrubbed its marketing before airing its grievances to Wired privately and publicly. Sleeper then posited the article as journalism in microcosm, and spoke to journalists in a tone that was condescending about letting the need for clicks and inflammatory stories get in the way of human brilliance being shared.

Clicks driving profit in digital journalism is a problem, one I promise journalists are also frustrated by and wary of. It is, ironically, a symptom of how tech has changed our industry to be less human, and gets in the way of the celebration of brilliance tech executives say they want. I understand the frustration that coverage of a critical moment wasn’t what they were expecting, and the Wired article was admittedly a little strange — but so were the relevant aspects of 1X’s marketing. It wasn’t the perfect profile piece on their product, but condemning journalism as clickbait felt unearned to me. The most inflammatory part of that saga, and the post that garnered two and a half million views, was the company’s executive making it publicly known he didn’t get what he wanted.

That brings me to a workshop in San Francisco this weekend for early-stage founders to learn how to get similar press coverage without putting thousands of dollars a month into PR. Press Club, the hosts, showed off their AI agent that scans thousands of tech articles, catalogs reporters who write them, and gives a personalized ranking of the top hundred journalists who may write the story they want about their company. It generates “human-sounding” cold emails, inserting personalized praise and references to the journalist’s recent work before pitching the user’s company as their next story. Even without using their AI, they walked attendees through the process of cold emailing tech journalists.

It was fascinating to watch the presentation of an algorithm designed to win my own attention, and even in the wake of the 1X debacle, it was fascinating to see what journalism still signifies. With journalism’s reputation in a volatile place these days, what still incentivizes founders’ hopes of winning a feature slot on Wired, Forbes, or TechCrunch? If there is at least some consensus of distrust for journalists in this space, why not just rely on social media for distribution and reach?

One of the co-founders of Press Club answered this plainly: credibility. The opportunity for self-promotion that social platforms made possible quickly oversaturated them. Reputable publications, in the words of Kathy Wang, “create hype and FOMO a LinkedIn post can’t.”

That explanation makes the 1X outburst make more sense. They felt attacked by an outlet with credibility. Their response was to attack that credibility. If we’re putting journalism’s relationship with the public in microcosm, that reaction is a great place to start.

One thing the event emphasized about looking for press coverage: sell yourself and your company as a dual product. The mentality — the Founder’s Mindset, as I keep hearing — has to be “I’m crazy, and that’s why I’m uniquely capable of building this company.” Wang then laughed and added, “It needs to actually be true, guys. Journalists will fact-check you. You can’t lie.”

Credibility aside, I wonder more than ever what role journalism now plays. The life cycle laid out in front of me is mostly artificial. A possible future of journalism production looks like an AI agent reading the articles, creating and quantifying profiles of journalists, and writing personalized emails that play to egos, with the end goal of a brief one-on-one that becomes an article the AI that incited the interaction can algorithmically predict the outcome of. It makes me wonder who, or what, is actually reading our pieces that are seen as prizes of social capital.

How journalism shifts and maintains footing on the moving fault lines of the digital space has always interested me. Social media and the democratization of access to an audience has morphed journalists and outlets from arbiters of information distribution into… something else. Hostile and opportunistic intermediaries, or independent watchdogs over private and public power. I’ve found it depends who you ask and depends on the day.

The humanity in journalism that is clearly here to stay is the power dynamic it disrupts by observing. The scarcity of social credibility in an internet where anyone can promote themselves is what makes founders still hope to land in Wired, however the story ends up.

Despite substantial media coverage, only one world record was ultimately “broken” over the course of the event. Aron D’Souza, an Australian businessman and an Enhanced Games founder, nevertheless expressed a grandiose vision for the athletic competition. Year one, he told The New Yorker, was a proof of concept, rather than a one-off gimmick. His multi-pronged plan? “It works, we generate revenue, we can push that into scientific development, better drugs come out, and eventually, through capitalism, we cure death.”

I reached out to D’Souza when he was traveling to Las Vegas for the Enhanced Games, though I was inquiring about another exceedingly ambitious project of his called Objection AI. I’m referring to it in the past tense, because the platform launched in April, and by late May, had already been taken (mostly) offline. This struck me as an unsurprising development for the venture, which was allegedly funded in part by Peter Thiel—a good friend of D’Souza’s.

The Intercept described Objection as “a mishmash between Snopes.com for right-wing culture war issues and a private defamation arbitration service marketed to the everyman.” I agree with that assessment. I wrote about Objection too, for Hard Reset.

A selection from the piece:

D’Souza describes Objection as an “AI tribunal” that “gives everyone a fast, affordable, evidence-based way to dispute statements in the media.” I would say “everyone” and “affordable” are a stretch; it costs at least $2,000 to lodge a media complaint with Objection…

Objection is for wealthy people who hope to spin up a counter narrative as quickly as possible. The swiftness of the AI tribunal is one of D’Souza’s selling points; he claims his platform issues a “verdict” (my scare quotes) within a couple of days, as opposed to the years it can take to sue for defamation.

When I emailed D’Souza, it was to ask if Objection was dead. D’Souza responded, “Since launching, we’ve found that our customers want to adjudicate much more complex cases (with commensurate willingness to pay). As such, we’re re-engineering the product and will be re-launching it in a few weeks.” He offered to do an interview closer to the re-launch.

Nearly two months later, D’Souza wrote to announce “Primary,” an index of journalists who are assigned a score and ranking based on the “rigor of their reporting.” No mention of adjudicating cases for a fee. Instead, Primary is, according to D’Souza, an IMDb for journalists. Each article that a journalist writes is supposed to be assessed by an LLM and rated 0 to 1,000 based on seven combined metrics; you can read the methodology here.

D’Souza joined me for a video interview late last week, a few days after Primary went live. The full interview is above.

When I asked D’Souza about whether Primary is a new product or a relaunch, he told me it’s a relaunch with the “same team,” and that Objection will return “in due course.” (As of Tuesday afternoon, Primary’s CTO, Kyle C Grant-Talbot, still hasn’t updated his LinkedIn to reflect that he’s no longer the CTO at Objection.)

I asked if Thiel, or other venture capital firms, were funding Primary. (D’Souza previously stated that Thiel and “leading” VCs funded Objection.) D’Souza responded, “Peter’s been a friend and advisor on all of my projects for 20 years, and we discuss things very frequently, but I’m the principal funder of Primary.” I also asked D’Souza how many people paid a fee to file an Objection; he told me “a few dozen started the process, but we learned that without engagement from the other side, it wasn’t worth even taking their money.” They were refunded, he confirmed.

Alex Shultz: Okay, because the way it was framed to me, you wrote that there was overwhelming interest, that “customers want to adjudicate much more complex cases with commensurate willingness to pay.” That implied something different.

Aron D’Souza: The people who wanted to pay were some of the wealthiest people in the world. Billionaires. They said, “I’d pay $100,000, $200,000 to have this article analyzed.” But unless the journalist on the other side was going to actually participate, it was pointless to take on their case.

In D’Souza’s retelling, Primary serves as a “permanent record,” an “IMDb for journalists.” Objection wasn’t a failed product, it was simply ahead of its time. “There’s no point giving out a grade unless there’s a report card,” is how he put it. But even after our conversation, I’m confused by D’Souza’s “report card” system—how he thinks journalists cite sources, and how he thinks he can make the sourcing process more transparent. Take this (lightly edited) exchange about a piece written by New York Times reporter Maggie Haberman. (A video excerpt is also available below.)

Alex Shultz: One of [Haberman’s] most recently indexed articles was about Trump’s plans to fire the FDA commissioner, from around May. The article has a 496 out of 1000 rating, and she seems to have been dinged on sourcing—a 182 [score] out of 1000. I assume this is an AI summary, correct me if I’m wrong. It says there’s one named source—President Trump, quoted dismissing the firing report—and two anonymous sources, “a person briefed on the matter” and “a person close to the conversations.”

I agree that naming a source is a stronger attribution method than anonymous sourcing. It carries more legitimacy, and when I work on stories I always encourage people to go on the record. But there are lots of reasons why they can’t. In this case, the on-the-record source is Donald Trump, who has an enormous history of changing his mind and flat-out lying about things.

Aron D’Souza: But that’s also the thing: Donald Trump, because of the intense scrutiny placed on him, has more or less a permanent, well-documented record in the public consciousness.

Alex Shultz: Right, but you wouldn’t call him a very trustworthy source, would you? It seems your LLM assigned Trump a certain level of—to be clear, on this story, Haberman got a 182 sourcing score. Two days after it was published, the FDA commissioner was fired, or left. So she was correct.

Aron D’Souza: She was correct, and that’s actually really powerful. You’re anticipating a version-two feature: predicting the future. If Maggie Haberman says the FDA commissioner is about to step down and predicts that with anonymous sources, her rating in a future version would dramatically increase once it’s proven true. That’s a very powerful indicator of quality journalism.

Now, the philosophical difference I think we can discuss is the use of anonymous sources. It creates a powerful asymmetry: the person being attacked by the anonymous source is named, but there’s no opportunity to rebut or assess the credibility of the anonymous source, because it sits behind a claim made by the journalist or the editorial board. What we don’t know, with anonymous sources used by major outlets, is the level of verification undertaken. I think that’s a major structural gap, and one of the primary reasons public trust in media has declined so much. In a better world, Haberman wouldn’t say “a source close to the president said X.” She’d say, “A source independently verified by Primary said X.” How would that verification happen? It’s a cryptographically complex problem. It’s actually what my engineering team and I spend most of our time on.

I really don’t know what to make of that response, or some of the other sentiments D’Souza expressed throughout our conversation. I earnestly tried to hear D’Souza out. You might watch and find that I was too easy on him—a fair critique—but my goal was to see if he could comprehensibly explain his point of view. He speaks for figures like Thiel and Sam Altman (another reported friend): billionaire moguls who foster deeply hostile feelings about how reporters and editors do their jobs.

My takeaway is that underneath Primary’s more conciliatory veneer, the same hostility remains. And Primary’s founder, D’Souza, is prone to bold pronouncements—though he doesn’t seem to have a great grasp of how journalism works. That was my opinion going into our conversation, and it remains my opinion now.

It’s easy to see why people think generative AI is a revolutionary technology. After all, lots of jobs involve writing, and ChatGPT, Claude, and other large-language-model-based systems can write faster than any human can. As a result, and in addition to the anecdotes from people saying, “I have 10Xed my productivity using AI,” there are empirical studies showing that people can complete certain tasks, such as coding, significantly faster with AI. Because AI so clearly makes lots of individual people more productive, it seems like a foregone conclusion that it will do the same for the economy as a whole.

So why is it that, four years after ChatGPT’s release, generative AI has not yet improved productivity at either the level of individual companies or across the economy as a whole?

Some have referred to this apparent disconnect as a “productivity paradox,” a phrase that economist Erik Brynjolfsson (building on an observation by Robert Solow) coined to refer to the period in the 1970s and 1980s when information technology was advancing rapidly but economy-wide productivity statistics barely budged. Once companies restructured their organizations and workflows to center the computer and Internet, productivity growth did indeed pick up (although less dramatically than it did during the Industrial Revolution and after World War II). Brynjolfsson and others think the same thing is happening with AI. Once companies rewire themselves for the AI era, they say, real productivity gains will come.

But there may be a much simpler explanation: the mere fact that AI allows people to work faster does not mean it is increasing productivity, at least not in the ways that truly matter.

************************

In casual conversation, when someone says that they have increased their “productivity,” they typically mean they are producing more stuff in less time. In this context, productivity means:

But in the economic sense, which is the sense that matters in determining how a technology will affect businesses and economies, “productivity” (specifically, total factor productivity) is not about how much you produce or how quickly you produce it. It’s about the economic value of your inputs and outputs, rather like a large-scale version of return-on-investment:

To be sure, the colloquial and economic meanings of productivity overlap quite a bit. Looking at the numerators, more stuff typically translates to more economic value. A manufacturer can typically get more money by selling 1,000 widgets than it can by selling 500. Likewise, looking at the denominators, cutting down the time it takes to produce a good or service tends to increase economic productivity because time is a key economic input, especially if human labor is part of the production process. If you only need 10 hours to complete a $500 task that used to take you 20 hours, you have doubled both your casual/personal productivity and your true/economic productivity.

There is much more to economic productivity, however, than just quantity and speed. And that is where the disconnect between productivity in the more casual personal sense and productivity in the technical economic sense lies.

************************

For starters, the economic value of a good or service depends on its quality, not just its quantity. Quantity certainly matters; the Industrial Revolution began with textiles in large part because new machines made it possible to produce vastly more clothing and other textiles in vastly less time by automating the various tasks involved in textile production. But, crucially, automating those tasks did not reduce the quality of the textiles. In fact, it usually improved it.

A scene in the musical Fiddler on the Roof contains a fictional-but-memorable illustration of this. In the scene, the tailor Motel gets a new sewing machine. Motel excitedly explains to his neighbors that the sewing machine “works twice as fast.” Holding up a swath of fabric, he points out “how close and even the stitches are” and exclaims, “from now on, my clothes will be perfect—made by machine!”

It was not just the machine’s speed that he valued; it was its precision and reliability. Regardless of how quickly it allowed him to work, Motel presumably would have preferred to stick with manual tailoring if his sewing machine frequently went rogue and produced shirts that were the wrong size, contained a third sleeve, or unraveled in his customers’ hands.1

************************

Which, of course, brings us to generative AI. Start with the denominator of the economic productivity equation. The AI infrastructure build-out has already cost $1.5 trillion, with spending expected to accelerate further next year, in addition to the perpetual costs associated with powering and cooling the chips. That means AI will have to generate a lot of value just to break even when it comes to economic productivity.

Supposedly, the way it will do so is by automating tasks in so-called “knowledge work” fields like software engineering, law, and medicine. But generative AI is notoriously and perhaps incurably prone to serious errors, such as fabricating case law, deleting code databases, and inserting incorrect patient diagnoses. The result — not unlike what Mickey encountered when he tried to get an enchanted broom to do his chores for him in The Sorcerer’s Apprentice scene from Fantasia — is often that professionals must spend more effort trying to clean up the AI’s messes than they would have spent if they had simply done the task themselves from the beginning.

Credit: IFA Film.

Such error-proneness hurts both parts of the economic productivity equation. In the numerator, output that contains serious errors or is otherwise low-quality will have little or no value. In the denominator, knowing that serious errors are a regular occurrence means that a significant amount of time must be devoted to quality control, which drives up input costs.

The assumption seems to be that such errors are, or soon will be, so rare that their negative impact on productivity will be swamped by the gains from increased output. But that is not how technological progress usually works. Despite tech bro promises that generative AI will get exponentially better, there usually are diminishing returns when it comes to squeezing additional improvements out of a given technology.

AI boosters are fond of saying that today’s AI is the worst AI you will ever use. But it’s just as likely that today’s generative AI is about as good as it will ever get.

************************

Even if generative AI’s reliability issues were miraculously solved, its outputs simply may not be economically valuable enough to significantly increase productivity. The Industrial Revolution was transformative because it made it much easier for people to get things they truly needed, sometimes quite desperately. Even if the production processes introduced some inefficiencies and flaws, the demand for clothing (and food, faster transportation, construction materials, etc) was so overwhelming that there was no question that dramatically increasing output would unlock immense economic value.

Will the same hold true for the things generative AI produces? The corporate world has been aggressively trying to automate, outsource, gigify, or eliminate office-based work for years, and much of the low-hanging fruit has already been plucked. Thus far, there is little evidence that AI is creating new economic value from the what’s left — even for tasks that generative AI is supposedly great at.

Recall this Financial Times chart from my last column:

AI has nearly doubled the number of app releases released per month. But overall, the number of apps with significant use has started to decline. The mere production of more stuff is not creating new economic value when it comes to app development. We shouldn’t be surprised if that pattern repeats in other white-collar fields where generative AI is still struggling to catch on.

Another force to bear in mind: many of the use cases where generative AI truly shines have a steeply negative effect on productivity. The rise of generative AI will almost certainly hurt the productivity of cybersecurity specialists because generative AI is great at launching cyberattacks but terrible at writing the secure code necessary to stop them. People are also using AI to hoodwink companies into hiring people for jobs they cannot perform. More subtly, low-quality AI “workslop” destroys productivity in many of the settings where it is used because the proliferation of such slop wastes time and makes it harder to identify work product that is fit-for-purpose.

While people talk about hallucinations and slop as if they are mere nuisances, they actually go to the core of why it’s doubtful AI can ever spur a new Industrial Revolution. Whatever gains generative AI unlocks from completing tasks faster may be swamped, or at least offset, by all the ways in which its unreliability destroys economic value.

With generative AI, there may not be a productivity paradox at all — just a productivity illusion.

Does this make me a nerd? Yes. Is it a good use of my time? Typically, no. Every once in a while, though, I spot an unusual or newsworthy donation. Which is…a slightly better use of my time.

July 15 was the FEC’s Q2 filing deadline. We’re halfway through the year, and the midterms are only four months away. It’s an ideal moment to take stock of who’s been spending big of late, and who’s keeping their powder dry. I perused thousands of filings today, keeping an eye out for Big Tech and tech-adjacent figures. For OpenAI co-founders Greg Brockman and Sam Altman, it was a quiet quarter. Anthropic’s Dario Amodei, on the other hand, dropped what appears to be his first seven-figure political donations. He wasn’t the only person at Anthropic to aggressively enter the political fray.

Below are some of the most intriguing, and largest, donations from the tech world that were received by campaigns and PACs between April 1 and June 30. The list below is nowhere near exhaustive, and in some cases, the people named have given even more money to the same campaigns or PACs—they just did so in 2025 or the first quarter of 2026. This is a snapshot meant to provide insights into what’s currently animating billionaires and other influential tech figures.

The Q2 Multi-Million-Dollar Donation Club

Reid Hoffman

Who are they? Billionaire venture capitalist, co-founder of LinkedIn, formerly chummy friend of Jeffrey Epstein.

What was their most notable donation? How about $10 million to a super PAC supporting squeaky-clean James Talarico, the Democratic Senate candidate in Texas. Talarico is running against Republican nominee Ken Paxton, whose wife announced she was divorcing him on “biblical grounds” last year. Paxton makes most other scandal-plagued politicians look like Mother Theresa, or at least James Talarico, by comparison.

Miguel “Mike” Bezos

Who are they? Jeff Bezos’s adoptive father. Inspiring: He provided Jeff with a six-figure loan so Jeff could start Amazon.com. Mike is a billionaire now.

What was their most notable donation? Bezos gave $5 million to With Honor Fund II, a nonpartisan super PAC that tries to get veterans elected to Congress.

Michael Bloomberg

Who are they? Billionaire former mayor of New York, center-right kingmaker who frequently parachutes into races to rescue uninspiring centrists who might lose to a progressive or leftist, namesake of the Bloomberg Terminal, short man who will always be known for stop-and-frisk.

What was their most notable donation? At least $6 million to a super PAC backing Democrat Micah Lasher, who recently won a tight (if underwhelming) primary to represent New York’s 12th district in Congress. More on that race below.

Chris Larsen

Who are they? Billionaire angel investor, crypto pioneer (take that however you wish), Gavin Newsom superfan, proponent of San Francisco being a surveillance state.

What was their most notable donation? At least $3.5 million towards You Can Push Back, a super PAC for Alex Bores, who was Lasher’s toughest competition in NY-12 (and relatively pro-AI regulation). For those of you keep track at home: Bloomberg 1, Larsen 0.

Larsen also gave $500,000 to Talarico’s super PAC.

Saikat Chakrabarti

Who are they? Unsuccessfully ran for Congress to replace Nancy Pelosi as San Francisco’s representative. (Chakrabarti finished a distant third behind Connie Chan and Scott Wiener.) Former AOC chief of staff who could not secure her endorsement during his primary run. Has a lot of money from his time as a software engineer and Silicon Valley investor.

What was their most notable donation? $1.5 million to the Justice Democrats PAC. Also converted his campaign committee into a super PAC for Chan, who is more left-leaning than Wiener. (Chan and Wiener are competing against each other in the general election.)

Vivek Ramaswamy Fan Club

For those who are unaware: Failed 2024 presidential candidate Vivek Ramaswamy is running for governor of Ohio. Normally, the GOP candidate would be a heavy favorite to win this race, even during midterms that will favor the Democrats. But Ramaswamy is in a closer-than-expected battle against Democratic candidate Amy Acton for two reasons:

• He’s bad at politics and campaigning.

• A not-insignificant portion of his base is racist and doesn’t want to vote for him.

Two conservative billionaires to the rescue!

Ratmir Timashev

Who are they? Billionaire founder of Veeam, a data management company. He recently told Forbes he wants to make Columbus, Ohio, “the next Silicon Valley—America’s innovation hub.” Yes, he attended Ohio State.

What was their most notable donation? $5 million to a super PAC supporting Ramaswamy’s gubernatorial campaign.

Elon Musk

Who are they? Le epic wealthiest man alive who’s gonna get us to Mars any decade now.

What was their most notable donation? $5 million to the same Ramaswamy super PAC. Quite a turn of events: In early 2025, Musk reportedly booted Ramaswamy from co-leading DOGE because they didn’t get along.

Big Republican Party Donors

Jan Koum

Who are they? Right-wing billionaire co-founder of WhatsApp.

What was their most notable donation? $1 million for MAGA Inc.

Bijan Tehrani

Who are they? Co-founder of two cursed platforms: Stake and Kick. Stake is a casino that runs on crypto. Kick is a streaming service where some of the world’s most self-absorbed Zoomers harass people on camera and/or record themselves gambling on Stake.

What was their most notable donation? Wouldn’t you know it, he gave $1 million to the MAGA Inc. super PAC.

Cameron and Tyler Winklevoss

Who are they? Billionaire crypto bros, literal identical bros, tertiary characters in Mark Zuckerberg’s come-up story.

What was their most notable donation? At least $750,000 each to the National Republican Senatorial Committee. It’s cute that they gave the same amount on the same day.

Anthony Wood

Who are they? Founder and CEO of Roku. He’s the reason you have a TV remote that never totally works properly.

What was their most notable donation? $500,000 to the Texans for a Conservative Majority super PAC.

Big Democratic Party Donors

Ron Conway

Who are they? An angel investor and the “Godfather of Silicon Valley.”

What was their most notable donation? $250,000 to the Senate Majority PAC, which supports Democrats in the U.S. Senate.

Irwin M. Jacobs

Who are they? Co-founder of the chipmaker Qualcomm. And 92 years old!

What was their most notable donation? $300,000 to Serving CA, a super PAC that tried to uplift Ammar Campa-Najjar, a progressive who ran in California’s 48th congressional district. Campa-Najjar finished third in the primary.

Michael Novogratz

Who are they? Billionaire tech investor and founder of Galaxy, yet another crypto company.

What was their most notable donation? $500,000 to Next Generation Action, a new-ish PAC that seems to have a center-left bent.

Patty Quillin

Who are they? A philanthropist for left-leaning causes and wife of Reid Hoffman.

What was their most notable donation? $500,000 to the Working Families Party PAC. It’s interesting timing—the WFP has been getting its lunch eaten by the DSA in recent congressional races that have pivoted progressives against democratic socialists.

Anthropic Gets (More) Political

I spotted a six-figure super PAC donation from a Google DeepMind engineer, and a six-figure donation from an OpenAI researcher, but neither AI company produced as many large donations during Q2 as Anthropic.

Dario Amodei

Who are they? CEO and founder of Anthropic. Purportedly concerned citizen about artificial intelligence while also running America’s most successful artificial intelligence startup.

What was their most notable donation? $1 million to Public First, a super PAC that claims to advocate for AI safeguards and candidates who are pro-AI regulation. Amodei’s check likely made its way to another PAC that unsuccessfully tried to bolster Alex Bores’s congressional run in NY-12.

Peter Lofgren

Who are they? A research engineer at Anthropic.

What was their most notable donation? Normally I’d avoid including someone who isn’t an obvious public figure, but Lofgren donated $999,900 to Public First. Lotta money.

Shauna Kravec

Who are they? A research engineer at Anthropic.

What was their most notable donation? A cool $500,000 to Public First. Same reasoning as above for Kravec’s inclusion here.

Nicholas Joseph

Who are they? Technical staff at Anthropic

What was their most notable donation? $250,000 for Public First.

Jan Leike

Who are they? AI research at Anthropic.

What was their most notable donation? $400,000 to Public First.

Daniel Ziegler

Who are they? AI safety research at Anthropic.

What was their most notable donation? $500,000 to Dream NYC, another (unsuccessful) super PAC for Bores.

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

• This week alone, I have read stories about how Kalshi:

  • Partnered with OpenAI to subtly show World Cup odds in ChatGPT

  • Wants to offer “contracts” (my scare quotes, because this is all gambling no matter how much Kalshi polishes its turds) for predicting what percentage of flights will be cancelled at an airport during a specific period of time.

  • Is now accepting bets on drug trials—whether they succeed and whether they’re granted regulatory approvals.

  • Had to alert federal regulators to irregular bets placed by President Trump’s teleprompter operator, who allegedly used his advanced knowledge of what the president was about to say in order to pocket more than $100,000, according to an ABC News report.

• New York just became the first state to institute a moratorium on new data centers for up to a year. Meanwhile, in Wyoming, relatives of U.S. Sen. Cynthia Lummis reportedly struck a deal to sell thousands of acres of land to Microsoft so the Big Tech company can construct a data center. Lummis happens to be one of the staunchest Big Tech allies in Congress.

• I am not in the business of making predictions about Mr. Market—my assumption was/is that SpaceX stock will inexplicably balloon just like Tesla’s stock—but so far, Elon Musk’s second major IPO isn’t going very well. SpaceX stock dipped below its initial $135 IPO on Wednesday.

• On Bluesky, I started a thread about a lawsuit filed by 26 Meta workers. They allege that after taking or requesting medical leave, they were included in mass layoffs at the company because of internal AI systems that track “productivity and output metrics” and “AI-token consumption.” Meta has denied the allegations in the lawsuit. I’ll be keeping an eye on this one.

At a quarterfinals watch party hosted by and for tech workers (at the immigration law firm Alma in downtown San Francisco), while chatting about AI over pizza, beer, and Norway vs. England, the recurring ‘Powered by Lenovo AI’ lower third caught my attention. I asked the group of engineers I was talking with if they knew what Lenovo AI was, or why it was the ‘Official Technology Partner.’ There were shrugs and mild interest. Obviously AI is involved in the World Cup; it’s 2026.

When I looked into it later, I found grandiose advertising for FIFA AI Pro, a specialized model developed with Lenovo. It’s touted by FIFA as the first of its kind technology, promising democratization of access to the most complete set of analytics and data available.

The Miami FIFA offices, more or less the command center for the 2026 Cup, was reportedly set up as an AI command center. The journalist FIFA allowed inside on a short leash described the presence of Lenovo’s AI as prominent, and the nature of its implementation as unclear. A Lenovo representative told him the agent allowed FIFA to cut down on necessary data analysis staffers three times over.

FIFA made the agent available for free to all 48 participating teams, and posit equity as an upside: one of the abilities of the system, hypothetically, would be the coaching staff of a smaller team, like Cape Verde, entering intricate questions in plain language about powerhouses like Argentina, and generating charts, 3D visuals, and historical context and game tape.

Tech is deeply embedded in this World Cup — the ball itself has sensors detecting touch, movement, and position. Every eligible player underwent a full-body scan to create a digital avatar “accurate to within millimeters.” The avatars have been used throughout the tournament in stadium replays, global broadcasts, and “semi-automated” refereeing decisions. Video Assistant Referees (VARs) have been used throughout the month to double check nearly every call by the refs on the field. It’s been used to overturn goals that likely would have stood a decade or two ago. AI has even been used to protect players from online abuse. 30,000 keywords, if left on social media platforms of teams and players, trigger instant hiding of the comment and investigation by FIFA, who says online abuse can result in bans from purchasing FIFA tickets and attending matches. (Side note: the anti-abuse AI works on Facebook, Instagram, Youtube, Tiktok, and Threads — but Elon’s X opted out. Didn’t need AI to predict that one.)

FIFA was incredibly proud to debut their proprietary AI, constantly framing it as a tool of equity and fairness, and said their efforts reflect a “broader change in how sports technology is packaged.” This World Cup is also a World’s Fair of AI, or at least how technological innovation now all leads back to it.

Even the watch party full of AI engineers themselves were uninterested in the technology as spectacle. The room had divided support for Norway and England and a split whiteboard of final score predictions. England’s first goal of the match may or may not have been the result of the ball hitting a camera cable overhead, which FIFA rules dictate a stop in play to correct. FIFA let the goal stand, saying the sensors inside the ball did not detect a collision with the wire. Many players on the Norwegian bench, and ESPN (human) analysts, say the ball hit the wire. FIFA sided with their AI system over human eyes. Norway was eliminated.

Norway’s manager Ståle Solbakken later told ESPN that though he fully believes the ball hit the wire, “I can’t say anything about that because FIFA… If there’s been no sound or there has been no [reading] in the chip, what can I say against that?”

AI now seemingly runs every layer and aspect of experiencing the World Cup. It’s the ball, the referee, the assistant coach, the social media moderator, the graphics, the analysis, and (at least) every other commercial. Whether or not it has improved the game, in true reflection of the moment, AI is everywhere. Big Tech, courtesy of FIFA, has had a massive global stage to push the forced inevitability of its prominence. Letting AI run the World Cup didn’t really fix any existing problems — increasing precision of offsides calls by millimeters won’t make people suddenly trust FIFA.

The next World Cup could take AI further into every aspect of the game — or, like the 2022 “Crypto Bowl” that saw its paramount cryptocurrency sponsors in legal and financial ruin a year later, the 2026 World Cup could be the overconfident presage of a crash.

And it’s not just something journalists have to grapple with, either—just today, I read a story about a man who rented a house for the weekend, only to get an AI-generated image from the homeowner claiming $6,000 of tabletop damage. (No doubt, such a con has sadly probably worked in the past.)

One system has been helping the layperson identify where images are coming from for years—an attempt to combat misinformation and disinformation. Called C2PA (short for the Coalition for Content Provenance and Authenticity), information—like images and videos—can be “credentialed” through verification of technologists, software companies, and hardware companies. A Canon or Leica camera, for example, can sign off on the fact that that camera with a certain serial number took the photo. An editing platform like Pixel Studio can then sign off on its role in altering the image. All of this can be communicated to social dissemination platforms, like Instagram or TikTok, where the information is shared.

Tech companies like OpenAI, Meta, YouTube, LinkedIn, and TikTok are part of the coalition to participate in this network—as well as phone manufacturers, editing platforms, camera manufacturers and some AI companies. It’s not completely fool-proof, of course—in part because some organizations don’t participate in the coalition (Anthropic and Apple are not, and neither is X, after the Elon Musk takeover…naturally.) Plus, while the metadata attached to images is true, C2PA isn’t necessarily saying the content itself is true.

In theory and in practice, C2PA has been an important tool in tracing the origins of digital information. But as with all innovations, there may be unintended consequences—and that’s why human rights nonprofit Witness published a report calling out the scenarios in which a beneficial tool could be weaponized to dox journalists or human rights defenders.

I spoke earlier today with Jacobo Castellanos, a trust and safety researcher at Witness. He told me that while C2PA is critically important, work needs to be done to ensure that journalists, photographers, and human rights advocates involved in sensitive information dissemination are not being traced if they don’t want to be.

While content provenance inherently cannot allow for identity to be included, different organizations wanting to legitimize their information with identity attachment—including a news organization, let’s say—might want to have an extension that allows for their journalists’ names to be attached. That individual journalist, however, might want to separate their name for safety reasons.

The report calls out an example scenario of a hypothetical journalist in this conundrum named “Amara” — someone who wants to share information while also protecting herself:

“She strips the personal information from the manifest using the redaction function in her editing software, and publishes the footage anonymously through a smaller, unlicensed platform that does not require credential verification. She cannot use the outlet she normally works with. That outlet operates under a platform distribution license that requires credentials on the state trust list. Publishing without her credential is not an option.

The economic consequences are immediate. The outlet that wanted the story no longer has it. The relationship with that editor — built over years of reliable work — is strained. The fees she would have earned are gone. Content published on unlicensed platforms is easier to dismiss:

the government’s communications apparatus is practiced at labeling inconvenient journalism as unverified or fabricated. Without her credential and without the institutional backing of a licensed outlet, the footage is more vulnerable to exactly that framing.

She publishes anyway, because the alternative is not publishing at all. She thinks that is the end of it, but what Amara does not know is that the manifest, stripped of her name, still carries the fingerprint of her working practice: the device identifier from her camera, the editing software signature, the specific sequence of adjustments she applies, with minor variations, to almost everything she shoots.”

Someone outside of a news organization could also be susceptible, the report calls out—such as “a human rights documenter who redacts their name from a manifest before publishing, unaware that a watermark embedded in the image still points to the original signed file.” Even a war photographer who is very careful to trace his tracks in documenting military operations can be traced back to his camera and a mobile application—even if his organization and personal identity are anonymous.

These are hypotheticals, reiterates Castellanos, not actual problems occurring in real time yet. But he still urges regulators to become familiar with the technical architecture of C2PA, and proposes an increased presence of independent, public-interest watchdogs to ensure that this origin-tracing power is not too centralized in the hands of one state government or tech company.

What else we’re paying attention to…

A few months ago, a speaker came to a panel conversation with an unusual tact: she sat on stage without uttering a word. The speaker was a woman named Sarah Wynn-Williams, a former policy director at Meta, then-Facebook. Meta sued her for breaching her non-disparagement agreement. She has since sued them for violating her free speech, claiming she was “under duress” when she signed the agreement.

Meta is using AI to target employees who have disabilities, according to a lawsuit. ⁠The company relied on factors such as ​productivity and AI token usage, which disadvantaged people who missed work because of medical conditions.

When AI is a member of your family: how one single mother, seemingly primed by a dependency on Amazon’s Alexa, developed a relationship with an AI named Sapphire.

Passengers are falling asleep in Waymo vehicles leading to frantic 911 calls.

Nearly two hundred economists, Nobel laureates, and executives have signed on to a statement warning of the harms of AI, titled “We Must Act Now.” Signatories include Jack Clark, a co-founder of Anthropic; Eric Schmidt, the former chief executive of Google; and Vinod Khosla, a prominent venture capitalist; however there were no specific policy proposals included in the letter.

Even Vogue thinks Meta’s smart glasses have a surveillance problem. From Dr. Sarah Saska, who we interviewed a few months ago about Meta’s 2016 “trend”: “The immediate risks are covert and sexualized recording, filming people in vulnerable moments, capturing children, harassment, stalking, and footage that can fuel extortion or deepfakes. The ‘creep’ framing is provocative, and it points to something real … But I worry that it shrinks the problem to one creep on a train taking a picture, when the structural danger is always-on cameras becoming ordinary, human workers reviewing footage, and data pooling in an ecosystem owned by one company.”

New York is the first state to impose a one-year ban on the development of data centers, specifically those using 50 megawatts or more.

Nurses at Kaiser, the largest private employer in the state of California, are protesting the hospital system’s emphasis on AI to evaluate its performance. One commenter on a post about the story writes: “As a bedside RN, what strikes me is that Kaiser inverted the correct design. The nurse should be the judge of the AI’s output, not the other way around. When an algorithm scores a nurse’s empathy, professional judgment becomes something to defend against a machine. When a nurse evaluates what an algorithm produces, judgment stays where it belongs.”

Much of her first term in the Senate was spent pinning America’s ills on China. In late 2020, she posted that China “has a 5,000 year history of cheating and stealing.” In 2023, she accused Barbie movie producers of intentionally displaying a map that “supports Communist China’s territorial claims to the South China Sea.” (The map in the scene she referenced was supposed to be “a child-like crayon drawing,” a Warner Bros. spokesperson rebutted.)

When Blackburn ran for re-election in 2024, her campaign circulated an advertisement called “Breaking China,” where she threw down plates with the Chinese flag on them. “China stole our jobs, sent us a virus, they’re buying up our land and spying on us,” she said.

Now 74 years old, Blackburn is running for Tennessee governor. She is very likely to win, though lately she’s been trying and failing to explain why she’s been evading debates in the lead-up to next month’s GOP gubernatorial primary. Rather than appearing on stage with her main opponent, Rep. John Rose (an equally thirsty MAGA acolyte), Blackburn is hoping that an extra obnoxious and xenophobic campaign ad about—you guessed it, China—will carry her home. Take it away, senator:

Blackburn: “How hard am I gonna crack down on China? Well, here’s a clue.”

Narrator: “Marsha Blackburn worked with President Trump to take on communist China. As governor, Marsha will fight to protect Tennessee land from Chinese front companies, close loopholes, and hunt down every communist who tries to defy us.”

Blackburn: “It doesn’t take a fortune cookie to figure it out. Here in Tennessee, we’re going to stop communist China and protect Tennessee land.”

As others have clocked, fortune cookies have little to do with China. That part of the ad is racist. The rest is much ado about nothing. There is no evidence that “Chinese front companies” are buying up land in Tennessee; Blackburn admitted as much to a local reporter. Recent reporting indicates that 2% of agricultural land in the United States is foreign-owned, and out of that segment, less than 1% is estimated to be owned by Chinese entities. Tennessee has its own laws on the books that restrict foreign investment in agriculture and real estate. Like many conservative policy planks, this is Blackburn treating her constituents like they’re cats drawn to a laser pointer. Oooh, look over here!

If we presuppose that Blackburn’s first ad of this election cycle was inevitably going to blame China for something or other, then the question becomes: why did she make “farmland” her bogeyman?

I ask because typically, when Blackburn invokes China, she finishes her sentence with menacing mentions of artificial intelligence. “Our adversaries like China are not slowing down on AI, and we cannot give them any advantage in the deployment of this emerging technology,” Blackburn said in September 2023. She’s delivered variations of this line dozens of times. “AI is going to be our biggest development,” she said in 2025. “If China wins that, they’re going to be able to break a lot of encryptions that we rely on… China would like to dominate this market by the time we get to 2030.” This sort of language is common among executives at OpenAI, Anthropic, and Palantir, too. America must win the AI race over China for the sake of our economy and for national security reasons. (Reasoning that conveniently implies the financial success of AI providers is our collective patriotic duty.)

Blackburn is not exactly a strategic mastermind, but she does know more about the AI industry than most of her Senate colleagues. She’s the chairperson of two tech-oriented subcommittees, and presents as having a relatively antagonistic relationship with Big Tech: she’s floated purported “child safety” bills about social media and AI that would mostly be a disaster. On the whole, she’s roughly as deferential to AI providers as anyone else in Congress. She’s previewed how she’ll govern around AI, which is to say, she’ll eventually tell Tennesseans to stuff it and go with it. “Let me be direct: Data centers are going to be a good and important part of Tennessee’s economic growth, but we’ve got to be thoughtful about their placement,” she said last month, when she cautiously recommended against a data center next to a zoo.

Americans do not want to hear half-assed data center solutions from their governor. Nobody is requesting data centers in their state, let alone their neighborhood! My theory: Blackburn is generally aware that sustained AI chatter isn’t to her benefit outside of Capitol Hill. It’s one thing to saber-rattle about winning a nebulous AI war against the “new axis of evil” (her phrase) in the Senate; it’s another thing for a gubernatorial candidate to remind voters that you’re ultimately bullish on a job-killing technology viewed negatively by the public. Blackburn doesn’t shy away from discussion about AI in her current job, but her campaign site doesn’t mention AI (or at least, AI isn’t listed as an “issue” on her site). She penned a Breitbart op-ed on Thursday entitled, “Tennessee Must Show America How to Stand Up to Communist China.” It’s almost entirely about the made-up farmland conspiracy. No AI talk.

It’s a macabre reflection of conservative politics in 2026. Bigoted, trollish advertisements that fearmonger about Chinese conspiracies can garner media attention and help you cinch the governor nomination in a deep-red state. Expanding your red scare routine to rile up your base about AI, though? Even Blackburn seems to believe that might be a step too far—and she’s right. Just…not about literally anything else.

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

• XBOX CEO Asha Sharma announced plans to lay off approximately 3,200 workers by the end of 2027; 1,600 of those workers were laid off this week. Sharma wrote that “our business today is not healthy” and “we must reset XBOX.” Bethesda Game Studios Union, which was affected by the cuts, posted a thread about their impact.

• Sony tried to sneak in its own not-good video game news: no more physical discs for PlayStation games after January 2028. People are not happy about it!

• Getty photojournalist Kevin Dietsch smartly posted up outside of the Smart Valley Lodge in Idaho, site of a somewhat notorious billionaire summer camp. Among those in attendance this year: Palantir CEO Alex Karp, Uber CEO Dara Khosrowshahi, OpenAI Chairman Bret Taylor, and Huge Fan of Mergers and Acquisitions David Zaslav.

• From the Financial Times, Meta is testing-driving novel ways for people to be creepy. The tech company is prototyping AI glasses “that would use cameras and audio recordings to capture a wearer’s every moment.” More, from the report: “With Meta’s current AI smart glasses, an LED in the corner of the frame lights up to signal to others when a wearer is taking photographs or filming. However, executives are planning not to activate the LED when the super-sensing features are being used, according to multiple people familiar with the matter.”

The film itself is less than half an hour, and was funded by Automattic, the parent company of Tumblr and WordPress, and directed by Bao Nguyen, known for feature documentaries about icons of culture like Saturday Night Live, BTS, and Bruce Lee (producers called his filmography a “throughline” with open source code in terms of cultural impact).

It urges the public to rally behind open-source code as principle, and refers to itself as a “Manifesto for the open web.” Its main human subject is Automattic CEO and founder Matt Mullenweg, present at the screening.

Mullenweg describes widespread platform decay, eloquently referred to as “enshittification,” as a cycle internet users are probably already accustomed to: a platform that initially offers free services evolving to restrict free features behind a paywall, flooding the platform with AI slop, and prioritizing corporate monetization. Major social media platforms are blatantly susceptible to this; my first thought was how drastically Twitter has changed in the last five years, name aside. It’s a foolproof method of building organic usership to exploit it for commercial interests.

The film uses creation platforms like social media as a leading example for how the open internet decentralized and democratized access to almost everything — removing the “barrier to entry,” as they put it. There’s a reason “content creator” is a job now, and a profitable one. The enshittification of it all is one piece of a more existential problem, as unregulated closed-source generative AI (like OpenAI, deeply ironically) is trained on data and content from individual users who didn’t opt in. When your work is posted on closed-source platforms, you don’t own it. It will inevitably be used in ways you can’t consent to and won’t see a profit from. It can also go away at the will of the platform’s owner.

It concludes with a relatively vague call to action and sense of urgency. It asks the public to vote with their wallet and their choice of platforms. It frames a user’s choice of open source versus closed source as “tenant vs. landowner,” and an internet without open source as dystopian.

In the succeeding panel, producer Michael Dinh summed it up as a question for our future selves: “Did we contribute, did we hold people accountable, or did we watch the world burn?” (The room made sounds of startled amusement at this.)

I am more or less the target audience for the doc. I’m 23, which is old enough to have watched platform decay of some major social networks in real time. I’m technically a content creator who posts my original work on a closed-source platform (Substack), with an organization that wouldn’t exist the same way without the open internet’s removal of a barrier to entry (Hard Reset). The stakes they describe apply to me, and the context they provide is consistent with my own experiences. I came away from it understanding what the problem was, and skeptical of the optimistic proposed solution.

I spoke with another attendee afterwards, a former tech reporter of over two decades, who reflected my own feelings: the message of the doc is important. The messenger, being the head of WordPress, makes the piece itself feel a little like an ad, at times. She interpreted the call to action as a call for people to create their own websites, conveniently the service WordPress provides. “I agree with the ethos,” she reiterated, “but I wish someone else had made this.”

Later, I chatted with an older gentleman, a retired career-long software engineer, who had a similar takeaway. He told me he’s a fan of open source and its principles, but felt that the stated incentive of social contracts and giving back by using open-source platforms felt overly idealistic. He used Tinder as a quick example. “Nobody gives a shit if it’s open source or not,” he said. “They care if they can get what they want out of it.”

They both, separately and unprompted, told me that a discussion of government regulation felt missing from the story, besides a brief reference in the in-person panel following the screening. Mullenweg said that open source is “constantly under attack in ways you wouldn’t expect,” referring to government regulation and legislation like age verification. He called them well-intentioned slippery slopes, and that’s where that conversation ended. A seemingly more viable solution, and worthwhile discussion, is his own company.

The creators and funders of “Code For The People” reiterated constantly that this is an optimistic, human-driven story about technology — in many ways, it is. One of my takeaways from its premiere was the value of open source itself, and I didn’t speak with anyone who disagreed. The internet has a problem the doc presented a viable solution for. The more prominent conflict I saw was the people asking them to fight for it, and how.

The documentary came out publicly today at noon. You can watch it for free here.

What else I’m paying attention to this week:

I went to an OpenAI Codex event hosted by HackerSquad expecting to see young AI founder hopefuls modeling themselves after Musk and Zuckerberg, and instead found a surprisingly human, collaborative atmosphere. It reminded me of the gap between the people making the tools and the tech leaders who have made an authoritarian swing.

Madison Square Garden was in the news this week for a wedding — have you heard more about that, or the apparent database of celebrities they’re keeping, noting their sexual orientation and quantifying them from “low to high risk”? After news that TSA-style face scanners were used around San Francisco during Pride celebrations last month, what does word of A-listers being tracked by sexuality mean for the rest of us?

Finally, there was a rumor going around that YC HQ had an AI simulator of Garry Tan. We were intrigued and tried to track it down to no fruition — it was just a Twitter joke — but at face value, it felt hilariously possible.

I was skeptical; there are abundant scams in the Instagram DMs these days, from all-natural deodorant offers to your prototypical reply guy. But after some preliminary research, I found that the startup putting on the salon had a Crunchbase profile, and the founder had a LinkedIn profile with a reputable university attached. So although it was entirely feasible that this still could have been fabricated, I decided to RSVP to the Partiful link in the DM.

A few days before the event, I received a push notification from the event asking us to come to the salon having read an article entitled The Rival Theologies of Artificial Intelligence. The piece, which covers “the deep tension between the humanism of the Church, with two millennia of teaching founded on the unique dignity of a finite, limited mankind, and the Promethean humanism of Silicon Valley” was published on a site called Palladium—which, according to this 2019 Jezebel article, launched in 2018 as “an online-only publication dedicated to finding a way out of the postwar liberal order’s current quagmire.”

Despite the fact that coverage of Peter Thiel’s intellectual gatherings had come out that same week, I was already intrigued. When I arrived at the apartment address, I was tempted to turn back—a vision of some sort of abduction flashing before my eyes. But then I saw a man gingerly buzzing up to an apartment, and I approached him to confirm that he was also attending the salon. We walked up four flights of stairs to a snug apartment, where a couple of guests were seated on a couch and a few stools in a living room. There were no refreshments beyond a table of LaCroix sparkling waters, and the no-frills environment seemed a far cry from that of a billionaire-funded secret society. (I later found out that this cozy Upper East Side one-bedroom was not the event organizer’s apartment, but the apartment of a friend of hers who was in Bali.)

Subscribe now

The organizer was a mid-20s-something woman with a posh British accent and an unflinching gaze and posture. Given the link to the Palladium article, I thought I might be in the minority in terms of my opinions on AI. But interestingly, the other attendees were all pretty negative about AI, and about centralized power in general. The man who had buzzed up ahead of me was a former data engineer-contractor at Anthropic, and reflected on the culture of eerie deference and reverence towards AI, Claude, and Dario Amodei at Anthropic while he was there. He was a self-described anarchist living in Bushwick.

Another woman, who was very vocal in the conversation with her concerns, was a professor of ethics at a New York college. The man to my left seemed ambivalent about AI, but was there for the conversation about religion. He was somewhat attached to Catholicism but repellent to cults, he said, especially after one tried to recruit him. The two remaining attendees were doctors who appeared to be a couple, a radiologist—interestingly, one of the first medical jobs that could be taken over by AI—and an orthodontist. Those two were probably the most bullish about the future of AI, but then became a little more subdued when they realized that the more talkative of us were probably more well-researched. The orthodontist admitted that she did not read, and that TikTok was her main source of information.

As the conversation continued on, limited to a tight hour, I saw the organizer very subtly and expertly guiding the conversation in a direction that left room for the possibility that technology could improve the quality of our lives. Although AI CEOs were not to be listened to, neither was the government. Even though the private sector can abuse its power, so can organized religion.

All of that is true, of course, and it’s why the Bushwick data engineer admitted to us that he was an anarchist. “I suppose you don’t have to believe in anything,” said the TikTok-watching orthodontist, to which several responded: “Well, not believing in anything, is, in fact, a belief.”

The ethics professor, despite her usage of AI for her more mundane life tasks, was generally concerned about it. Admitting her contradictions, she went on to lament about screens in stores and the labor being shifted onto the consumer, citing the work we all have to do and time we have to spend when there isn’t a cashier, barista, or person familiar with the the products we’re purchasing to guide us.

The organizer listened calmly and intently before responding. “I understand,” she reassured. “But what about a situation where you have to interact with a barista who just went through a break up. They have completely botched your coffee order and because they are in a bad mood, they put you in a bad mood. Wouldn’t you, in that case, want a screen?” She implored us to think about what kind of better future technology could enable, even if we disagreed with her barista comment.

At that moment, the organizer’s comment about the barista break-up reminded me viscerally of an interview Joe Rogan facilitated with venture capitalist Marc Andreesson. In the interview, Andreesson talks about how easy it will be when workers are replaced by AI. AI is not drunk, he said, and they haven’t just gone through a break-up either.

Currently, I don’t know if the salon was entirely sinister—or if this young woman is some secret missionary of Marc Andreesson or Peter Thiel capable of abduction. I still need to do a bit more research to figure out how they are funded, and what her mission is.

But by the end of the discussion, even the concerned professor remarked that she thoroughly enjoyed the Palladium article. The anarchist data engineer admitted that we need more positive visions of the future with technology, emphasizing that we can’t just complain about what’s to come, but that we must be architects of the ideal world we want to live in.

It’s true that we must do that. The real question is, who’s hosting the salons.

Share

What else we’re paying attention to this week…

The “depravity market” is real: prediction markets are letting you bet on whether a wildfire will burn your town down. From the piece: “One major concern stemming from wildfire prediction markets is arson… Theoretically, making a bet could give someone the perverse incentive to start a fire or help one grow.”

The effective altruist movement—which believes that there is a specific and mathematical way for wealth to be dispersed for the net benefit of society—is courting people who’ve made money from Anthropic to invest in their efforts.

This is a must-read profile about Rachel Whetstone, the former comms boss at Facebook, Uber, Netflix, and Google—and wife of the Trump-supporting gubernatorial-hopeful Steve Hilton. From the piece: “In interviews with more than a dozen friends, associates, and former coworkers — most of whom spoke anonymously — Whetstone was described as ‘brilliant,’ ‘fearless,’ and ‘almost iconic,’ but also ‘transactional,’ ‘tyrannical,’ and ‘one of the meanest people around.’ (Several described her, essentially, as all of the above.)”

AI startups are turning down younger entry-level talent for Silicon Valley men with advanced degrees, research shows.

Share

If you have read any media coverage of AI, there’s a pretty good chance that you’ve seen breathless quotes comparing AI to the most transformative developments in the history of humankind. According to AI enthusiasts, AI will remake our lives in the same way that controlling fire, the Industrial Revolution, and electrification remade those of our ancestors.

But what if, instead of initiating a technological revolution, today’s AI systems end up being like…the Wankel engine?

During the 1950s and 60s, the automobile was in many ways the centerpiece of the United States’ society and economy. The Beach Boys sang hit songs about muscle cars, the new Interstate Highway System made commuting and long-distance road travel easier than ever, and GM, Exxon, and Ford spent nearly the entirety of the 1950s and 1960s occupying the top three spots in the Fortune 500.

Right as the automobile was reaching the peak of its social and economic power, a German engineer named Felix Wankel developed a new engine that was billed as having the potential to be much more efficient than the nearly century-old internal combustion engine. The Wankel engine had a rotary design, which, in theory, should have made it dramatically more efficient and reliable than complex piston-driven engines. The Wankel was also smaller, quieter, and had fewer moving parts.

The Wankel engine design. Attribution: WikiMedia Commons.

Automakers and investors alike salivated, as described in Brent Goldfarb and David Kirsch’s 2019 book, Bubbles and Crashes. When the technology was first introduced in Europe in the mid-1960s, including a Mercedes that could go from zero to 60 in 4.3 seconds, a speculative bubble developed around the technology. Investors and the media alike ignored occasional reports of design flaws that could render the Wankel engine incapable of living up to its hype. When sales started off slow, Wankel boosters instead blamed the technology’s rocky rollout on over-regulation.

As it turned out, the engine’s technical limitations made it commercially impractical. GM licensed the technology but never brought it to market after it discovered that difficulties in sealing the rotating engine made Wankels less efficient and reliable than the clunkers it was supposed to replace. Investors in the companies that had promised a Wankel revolution ended up losing their shirts.

“Speculators,” as Goldfarb and Kirsch put it, had “underestimated the likelihood of technological failure.”

Fortunately for the U.S. economy, neither GM nor any other major corporation had bet the company on the Wankel engine. No auto manufacturer had fired their old engineering experts or shut down production lines for cars with traditional powertrains. Had they done so, the United States’ most important industry could easily have experienced a crash big enough to send the whole country into recession.

The story of the Wankel engine raises important questions that we should be asking ourselves today: What would happen if the country’s largest industry did go all-in on a technology that was fundamentally incapable of achieving its hype? To go one step further, what if the entire economy somehow transformed into a giant bet on the success of such a technology?

We may be about to find out.

The tech industry dominates the U.S. economy today to an extent the auto industry never did. The ten most valuable U.S. corporations are all tech companies that (with the partial but notable exception of Apple) are pinning their futures on the success of generative AI, especially large language models like those that underpin ChatGPT and its competitors. The AI industry is in the midst of an $8 trillion spending splurge to build the data centers and buy the specialized chips needed to train and run generative AI models. To recoup that investment, which is increasingly fueled by risky debt, AI companies will have to generate trillions of dollars in additional revenue by 2030.

The hope appears to be that AI will return this value by automating vast swaths of white-collar work, a prospect that SpaceX raised in its prospectus, which suggested that white-collar work was a $26 trillion market that AI companies could seize. SpaceX’s arithmetic is dubious — as economist Aswath Damodaran noted in a recent podcast, $26 trillion is more than the employee expenses of all publicly traded companies combined. But even if there were $26 trillion worth of work to automate, generative AI doesn’t look like it’s up to the task.

Over the past year, studies and surveys have repeatedly found that most companies are not seeing any return on their AI investments. This should not be surprising; AI systems frequently make mistakes, including an alarming number of catastrophic failures. Such workslop means that much of the human labor saved at the front-end by delegating a task to AI is lost at the back-end when a human has to come in and clean up the bot’s mess.

The use case where AI has shown its greatest commercial potential is computer programming. As NYU professor Zeynep Tufekci wrote in a recent New York Times column, the output of coding “is either verifiably right or wrong, functional or not functional, and can be definitively checked through an automated process.” That makes coding a task for which generative AI should be uniquely suited.

But even in coding, it is far from clear that AI is actually adding value. According to the enterprise data company Entelligence, for each dollar that companies spend on AI coding tools, 44 cents goes towards fixing AI-generated bugs, 27 cents goes towards rewriting the code, and 11 cents is lost to other inefficiencies from the code review process. In other words, the vast majority of AI coding dollars are wiped out by the effort needed to fix AI’s coding mistakes. Only 18 cents of each dollar actually reaches users as a shipped product.

Attribution: Entelligence.ai

Even this may overstate the value of AI coding. While the number of iOS app releases has shot up since AI coding took off last year, the number of apps with significant use and the number of app reviews have actually declined, suggesting that even the “shipped product” is often worthless.

And remember, coding is something that generative AI is supposed to be especially good at. In other settings, particularly those where the cost of errors is high, AI can be close to unusable. AI-driven legal tools frequently generate inaccurate or incomplete case summaries, non-existent quotes, and even entire opinions that simply do not exist. Such errors can result not only in embarrassment and wasted time; they also can lead to judicial sanctions and (most important to corporate lawyers) having to write off billable time and even pay opposing counsel’s fees.

AI boosters try to dismiss AI’s failures to generate — and their alarming tendency to destroy — economic value by claiming that AI is merely experiencing growing pains, often by citing dubious models favored by corporate consultants like the “productivity J-curve” and the Gartner hype cycle. Those models suggest that while productivity may initially fall with new technologies, they eventually rise after companies figure out how best to use it. Under the Gartner hype cycle, innovative new technologies enter a “trough of disillusionment” after an initial surge of interest, but the trough eventually gives way to a “slope of enlightenment” that ends with a “plateau of productivity.”

Attribution: Jeremykemp at English Wikipedia

But many technologies never recover from their initial growing pains. The Economist found that six of ten technologies that enter the trough of disillusionment never rise again. As the Wankel engine shows, sometimes that trough deepens into a depression.

A depression could, sadly, be exactly where all of us are headed. Unlike with the Wankel engine, the world’s largest companies are going all-in on generative AI, racking up trillions of dollars in debt and cannibalizing key resources from the rest of the economy in the process. Tech companies are firing software engineers and companies are replacing administrative workers with AI, often with predictably awful results.

Right now, fears surrounding AI tend to focus on what will happen if the technology does live up to its hype and put tens of millions of people out of work. But AI failing to achieve its supposed potential could bring about a different kind of economic catastrophe. There simply is no precedent for what happens when an industry that dominates the global economy bets the proverbial farm on a fundamentally flawed technology. If AI turns out to be a dud, it could be the economy that goes up in flames.

The manifesto was inflammatory and got lots of attention, which I presume was Karp’s goal as he promoted his book. This week, Palantir released another manifesto about “AI sovereignty.” The “AI sovereignty” manifesto is only nine points (much appreciated). It arrives at an especially perilous moment for Anthropic and OpenAI. Both companies would like to IPO, but are frantically adjusting to unexpected regulatory scrutiny from the Trump administration, blowback from enterprise clients who have “tokenmaxxing” buyer’s remorse, and the public’s distaste for how AI is a job-killer that’s making everything more expensive.

Wouldn’t you know it, Palantir’s latest manifesto accompanied an announcement about an expanded partnership with NVIDIA; they teamed up to offer an open-weight AI model to U.S. government agencies and “critical U.S. companies.” The pitch is basically that NVIDIA can supply a cheaper LLM than closed-sourced LLMs provided by Anthropic, OpenAI, and Alphabet—and on top of that, Palantir grants clients an extra layer of security and customization around how they deploy and train their models. The implicit promise of “AI sovereignty” is that, unlike Anthropic CEO Dario Amodei, Palantir and NVIDIA will not engage in any pesky meddling or moralizing. How the U.S. government chooses to utilize the NVIDIA-Palantir stack is simply none of Palantir’s business.

My sense is that the “AI sovereignty” manifesto didn’t hit like the last one—it didn’t sufficiently creep people out to the point that it went viral. So on Wednesday, Karp went on CNBC and did his best to reiterate the main sentiments of the manifesto while drawing attention to his new business venture with NVIDIA.

Karp successfully achieved his viral moment. Video Clips Guy Aaron Rupar referred to Karp’s 20-minute “Squawk Box” interview as a “televised nervous breakdown.” To the untrained eye, it certainly appeared that way. I strongly suspect that Karp does not actually care how he presents, and in fact probably lays it on thick on purpose. The people pointing and laughing at his, uh, eccentric behavior are ultimately amplifying his message to a larger audience, which includes prospective clients who are not concerned with whatever Karp has going on, or that Palantir’s headquarters are based in the third circle of hell.

I will caveat that Karp’s nearly uninterrupted “interview” was certainly strange. Roughly halfway through his ramblings, he was possessed by the spirit of Ayn Rand and proclaimed that the “voice of American business [is] being channeled through me.” He spent a considerable amount of time defending the state of Israel, anointing himself “the most publicly supportive CEO” of the country. He complained about how potential wealth taxes would “punish” billionaires. He decried the “far left” and “far right,” though he failed to define what he actually objected to about each “side” of the political spectrum.

Karp also hilariously declared that “American enterprises are run by the shrewdest, most wily intelligent people on the planet.” (They are not.) He continued his longstanding “joke,” or whatever you would like to call it, that although his parents would like for him to become a faculty member at Berkeley, the university is too woke to consider him. (He repeated this “joke” three times.) And he claimed that he does not do drugs—he is simply a “neurodivergent crazy person.”

If you can look past…all of the above, the underlying message that Karp delivered is fascinating. He wants to attach himself to an increasingly mainstream view: OpenAI, Anthropic, and other closed-source LLM providers are overpromising, overcharging, and underdelivering. “Who controls the models?” he rhetorically asked. “Who controls the weights? Who controls the value of your business?”

These are legitimate questions. And this is a legitimate point: “I’m telling you, in this country, at every single enterprise I deal with, these people are livid,” Karp said of CEOs he knows. “They’re like, ‘I am paying for tokens that create no value.”’ He encouraged investors watching the CNBC segment to call up a random CEO and ask for their private opinion about the utility they’re receiving from closed-sourced, token-based providers.

The rest of Karp’s ranting was duplicitous—a naked attempt at getting one over on his newfound AI competitors. He went after Amodei over and over again, poorly softening his insults with assurances that he was not actually “throwing shade.” An example:

“These people are—Sam and Dario—there’s nothing more fun than debating Dario in private. So I’m not throwing shade at them, but something has gone completely wrong. And the basic view among enterprises in this country is I’m going to chillax and waste my time with tokens. I’m going to get no value and they’re going to get my IP.”

One more:

“I’m not throwing shade. Dario is a literally historic figure. And he came from behind, and he’s now number one. I’ve been in business, whatever you want to call I do, for a long time. I’ve never seen anything like this.”

When Karp circled around for a third barb, he finally dropped all pretext and overtly criticized Amodei for trying to establish red lines around the Department of War’s Claude usage. “I would say it is a loser to restrict something from the government because you don’t agree with how the government fights war, and then open it up to the world, including our adversaries,” Karp said. “Go sell that to the American people where they already don’t like you.”

Amodei’s red lines are indeed arbitrary and insufficient, given that Anthropic has otherwise enthusiastically lent its tools to the American military, which likely killed 120 Iranian schoolchildren. But Karp is untethered from whatever he deems to be the “American people.” He’s just trying to score brownie points with the U.S. military, which is why he waxed on and on about America’s alliance with Israel, and how China is our main adversary in an AI race that we must win at any cost. (“It’s binary: they will win or we will win,” he said.)

Ironically, Amodei has talked about China in similarly simplistic terms. He tried to have it both ways with the Pentagon, but publicly second-guessing a contract with the U.S. military is a fool’s errand. Karp recognizes the moment and is pouncing. He’s a more effective messenger for faux-patriotism because he fully aligns with a techno-fascistic vision. He’s already working with U.S. government and law enforcement agencies, and he sees an opportunity to expand those relationships at the expense of Anthropic and OpenAI. He also sees an opportunity to expand Palantir’s commercial ties, which are becoming a real revenue-driver for the company.

Those ties are fickle, though. Palantir is not a popular “brand.” It is prone to backlash because of security concerns and Karp’s horrifying commentary. European countries like France, Spain, and Germany have ended or are winding down their contracts with Palantir; the United Kingdom is rumored to be next.

So yes, on “Squawk Box,” Karp made a persuasive case against closed-model LLMs that could potentially cache or transfer client data. That does not mean he’s operating from a position of undeniable strength. Disregard his claims that Palantir products are “completely agnostic,” meaning the company has zero interest in which AI models it stacks onto. Try to look past his distracting, extraneous musings, and his incessant bragging, like: “In my business, we have much more demand than we can supply,” Karp said on CNBC. Don’t believe him? “Ask the Ukrainians. Ask the Israelis. Ask our Department of War. Ask the enterprises that are working. We do not have to oversell what we have.”

Except, as the interview concluded, Karp—who didn’t seem to realize he was still on camera—said he’d love to appear on “Squawk Box” again soon. A funny offer, given that Palantir supposedly doesn’t need to “oversell” anything. In truth, Palantir is engaged in the same AI rat race as the companies that Karp is impugning. Congrats to him on the free infomercial about his partnership with NVIDIA. I’ll believe Palantir isn’t just as antsy as its competitors when Karp stops releasing a tech manifesto every time he’s promoting a new product.

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

• The New York Times dug into President Trump’s ludicrously corrupt financial dealings since his return to office. According to Trump’s 927-page financial disclosure form, his administration released an “AI action plan” full of potential deregulatory measures on the same day that the president purchased up to $5 million each in “Broadcom, Meta, Amazon, Apple, Microsoft and Nvidia.” Stock prices for most of those companies have increased. Incredible foresight from our brilliant leader.

• Meta’s endless string of embarrassing updates and news reports continued this week. My three favorites:

  • Meta is instituting rate limits for the “conversation focus” feature of its AI glasses, meaning you have to spend an additional $20 a month to unlock more hours per month where you can “dynamically amplify the voice of the person you’re talking to,” as Meta defines it. The Verge has an interesting breakdown about why “conversation focus” doesn’t seem like the sort of thing that needs to be rate limited at all.

  • Meta is reportedly gearing up to start a cloud computing business—basically, renting out its computing power to companies that have at least a little bit of traction with their AI models. You may recall xAI announcing a similar plan a few months ago. Not a good sign about interest in Meta’s own AI models.

  • Mark Zuckerberg considered purchasing cursed predictions market Kalshi, according to NPR. Reportedly, negotiations didn’t go very far, and so now Zuckerberg wants to launch his own prediction market that (allegedly) will not use real money. We’ll see how long that lasts, or if the play money predictions market even launches.

• Amazon and Google’s newest greenhouse gas emissions reports are bleak: Amazon’s emissions increased by 16%, and Google’s emissions increased by 18%. Both Big Tech companies cited data center construction as a contributor to the uptick.

Over 150 parades happened across the country this past weekend. In major cities, especially tech hubs like New York, Austin, Seattle, and San Francisco, Pride can be a sociopolitical wind sock for prominent industries. Last year, we saw overall corporate participation and sponsorships of major Pride celebrations drop. Organizers, and businesses themselves, attributed this to Trump’s return to office and intense attacks on DEI.

Subscribe now

Tech wasn’t an outlier. In 2025 Amazon and Google were among major companies to noticeably scale down their acknowledgements and sponsorships of Pride. Meta did not acknowledge Pride at all last year. Their absence from their HQ city’s parade was very noted in San Francisco.

Like many young people flocking to SF for jobs in tech’s new gold rush, I got to experience San Francisco’s Pride parade for the first time this past weekend, and kept my eye out for rainbow tech logos. Amazon, Apple, Salesforce, and Zoox (a subsidiary of Amazon) participated. Meta, X, and Google, in a rare occurrence for the Bay Area, did not make their presence known.

In the best of times, we can dismiss corporate participation in Pride as rainbow capitalism. In the worst, I get to see a meaningful look on my friend’s face when they see their employer, Salesforce — home to a grueling entry-level culture I’ve written about — publicly endorse the validity of their identity when they have something to lose.

It’s far from altruistic, but it’s not nothing.

I was interested to see if any of the Frontier AI companies that have made homes in San Francisco would participate. They did not. OpenAI, Anthropic, and NVIDIA, to name a few, haven’t been around long enough to publicly fluctuate their support for diversity and equity with changing political climates, and possibly learned from their tech boom elders who have. There is little to no scrutiny of their absence in comparison with companies like Meta, who had a public falling out with San Francisco Pride. Most conversation in relation to AI at Pride concerned gay bars installing facial recognition devices that tracked the information and behavior of patrons.

After laying low throughout the weekend, AI was back to cultural preeminence with a duo of major conferences.

The fourth annual AI Engineer World’s Fair is back in town. Over six thousand engineers and researchers from around the world are attending in person. The schedule is overwhelming. There are over 400 official conference sessions. Themes range from security to tokenmaxxing. The cheapest tickets available were over a thousand dollars each. The only thing more exclusive than the conference are its side events and afterparties — freeform discussions about AI over ping pong, pickleball, World Cup watch parties, bar trivia, and run clubs. They require applications upon RSVPing that are reviewed by the host.

The Laude Institute held a one-day Open Frontier conference Tuesday in San Francisco with global leaders in AI research. The entirety of the seven hours of panel discussions were livestreamed on YouTube, then inexplicably made private early Wednesday morning. Before it was taken down, the description billed it as “A day of talks, panels, and conversation about what it means to keep AI research open at the moment it matters most.”

Opening speaker and moderator Andy Konwinski, co-founder of Laude, Databricks, and Perplexity, said openness and open access has historically been one of America’s competitive advantages in innovation.

“Open needs a heavyweight contender in the ring with the big labs, and we are it,” he said. “If we don’t do it, I don’t think anybody is going to.”

‘Open’ in this context does not refer to OpenAI, the company; as defined by TechTarget: “The practice of open AI entails openly sharing AI models, the provenance of training data and the underlying code. Closed AI obscures or protects one or more of these things.” To some high-profile participants, it also seemed to mean anti-regulation.

“The voices that are calling for AI to be controlled, and basically closed, are coming at it from a position that it is somehow worse for us to be free to explore our ideas with AI,” Catanzaro said. “The history of technology shows that when things are open, people use them to build, to defend, and to explore in ways that are just really fundamental.”

Nathan Lambert, post-training lead at the Allen Institute for AI, said that AI will be the most powerful technology of our lifetimes, and therefore, one of the most closely regulated. He expressed concern of “[T]he top open models being banned in the next three to six months.”

Share

Panels brainstormed aloud methods for raising the American public’s confidence and comfort with AI models, out of fear that public opinion correlates with regulation. Ion Stoica, professor of computer science at the University of California, compared American public sentiment on AI to China. He thinks the positive sides of AI are not being pushed enough by researchers.

“If you look at the sentiment, [China] is much more positive about AI than the US,” Stoica said. “We [researchers] are not enough, because the people are going to drive the politics. If 70% or 80% of the American public is going to be against AI, it’s going to impact regulatory framework.”

A coincidentally-timed poll this week illustrated bipartisan support for tightening AI regulation. More than two in three respondents were in favor of the federal government establishing “a formal review process for the most advanced AI models before they can be widely released.”

A participant summarized: “The talk of the town is mistrust.”

If a better relationship with the public is the goal, especially after a celebration of visibility they were notably absent from, Frontier AI may need to decide if openness and plurality are philosophies, or the theme of the week.

But what comes next? Will it be a dystopian hellscape where full-time jobs disappear, and we are all stitching together ad-hoc gigs to train AI models? Or will there be universal basic income, where we are freed up as human beings to self-actualize by reveling in art and music and nature?

The reality is probably somewhere in the middle, but one thing is true: despite all the chatter about doom and displacement, we may not actually be prepared for how to tangibly handle the fallout across the workforce. We know displacement is coming, but we can and should take action now to redirect people to different earnings opportunities, skillsets, or generally ways to bide their time and create value in their communities.

While AI companies and private equity firms are gleefully rolling out data centers to meet the demands of the AI companies’ outsized valuations, the people who are no longer employable (or who may need to adapt to work alongside or in parallel to automation) may be left behind.

A new report from the Brookings Institution calls out the speed at which everything is moving, and suggests some policy solutions to help workers—from recent college graduates to older workers in the trades—ready themselves for the impact.

I spoke with the author of the report, behavioral scientist Xavier Briggs, who suggested that we look to countries like Denmark that are directing workers towards apprenticeships in the trades. Trades training is not new of course, but Briggs says that these pathways must be inherent to education systems, and present themselves early to people in high school who might otherwise have thought that office jobs were inevitable or preferred.

“We need to make it possible for young people to have experiences in types of jobs outside of the office,” Briggs said, “And adolescence is a good window. Kids are old enough to be interested, but young enough to be open-minded and flexible.”

This happened with Indiana schools, where the state restructured the high school year and learning weeks to help people understand there’s a pathway into construction jobs, for example—not to dictate their careers, but to blend class-based and work-based learning. “[Individual] teachers and labor unions can’t do it alone,” he added. “We need to restructure the high school experience.”

While it’s feasible to direct teenagers into the trades professions early on, it is still unclear what mid-career white-collar or knowledge workers might do if their labor is no longer seen as valuable to the economy. Here, Briggs points to early retirement (still begging the question, in my mind, what people who previously sought identity through work might do with all this free time, and whether employers will fund this early retirement) or the state.

Perhaps one solution, the report calls out, are “human in the loop” roles to work in tandem with AI systems—to stop AI from making decisions that require a lot of context, nuance, and human judgment. I asked Briggs if AI companies might push back against such mechanisms that could be perceived as trying to slow down their growth.

“To win public trust, [the AI companies] need to respect society’s need to feel in control,” Briggs responded. He emphasizes that these companies do realize it will backfire for AI to unilaterally decide high stakes situations—like public benefits, parole, and child welfare— without human oversight.

Briggs referenced technologies that have operated alongside human labor since the dawn of manufacturing. Decades ago, automation made the manufacturing process of cars safer both for workers and consumers; today, AI may free up time for nurses and doctors to spend more time with patients, and teachers to spend more time with students.

The day before I spoke with Briggs, a $500 million bipartisan initiative launched to retrain workers displaced by the AI boom, championed by former New Hampshire Governor Gina Raimondo and including the participation of Anthropic, OpenAI, Amazon, Microsoft and Bank of America. The program “service year” opportunities for young people in healthcare or education and updating unemployment insurance programs to help laid-off workers start businesses with AI.

Briggs is positive about such initiatives, but believes that there remains critical work that humans will need to do, emphasizing the need for people who use “judgment and emotional intelligence with customers, clients, and citizens.”

Subscribe now

Here’s what else we’re paying attention to this week…

The CFTC is investigating Donald Trump Jr.’s connection to prediction marketplace and betting platform Polymarket. Meanwhile, Meta is urging his lieutenants to develop partnerships with Polymarket and Kalshi, 18- to 34-year-olds, the employees added, and Meta is aiming to reach at least 100 million monthly active “predictors” for the app.

Anthropic’s new chief economist published this delightful thought about humanity in 2023: “...​​it is optimal to take a 1 in 3 chance of ending human existence in exchange for a 2/3 chance of dramatically raising living standards by a factor of 55.” Meanwhile, philosopher jobs at AI companies are in high demand these days.

TikTok’s AI is censoring creators who are making videos about the news.

Software startup Domo was once valued at $2.8 billion. Its market capitalization on Friday stood at $133 million. The story of what happened here.

Anthropic thinks they’re the only AI company who can save us. WIRED digs into CEO Dario Amodei’s blind spots, and the company’s “good guy” problem.

Former Meta policy lead Sarah Wynn-Williams is suing Meta for blocking her from promoting her memoir Careless People. The complaint cites a blatant violation of the first amendment, and also accuses the company of “coercive surveillance.”.

After months of tense negotiations, the U.S. government is bringing Anthropic’s model Mythos back online.

A bill from California Assemblymember Josh Lowenthal threatens fines of up to $1 million per child for platforms found liable for harming kids through negligent product design. Meta is fighting for exemptions.

A writer for The Atlantic wore Palantir’s workman coat for a week. Apparently, the inside of the coat reads: Ask yourself constantly, Am I winning? If the answer is yes, nothing else matters. Chaos is tolerable; pain is tolerable. The only thing that matters is to win.

Inside cyberscam compounds across Laos, Myanmar and Cambodia, women are beaten and raped, forced to work 18-hour days chatting with 100 online scam victims at a time. Known as “pig butchering,” trafficked workers—who initially thought they’d have jobs as social media managers—are forcibly trained to convince online users to invest their money in a fund that eventually reveals itself to be fraudulent.

Share

Sundberg doesn’t explicitly call herself a journalist, so I’ll keep my media critique to a single sentence: it’s in your best interest to convince the unlikable billionaire that he should address at least one of his controversies, so you don’t look like a complete sellout.

More interesting than whatever Sundberg was thinking is that Meta’s press team, presumably in conjunction with Zuckerberg, felt it was the right time for him to debut his upgraded personality model. Under Zuck’s leadership, Meta is having an abysmal three-month stretch—one of the most embarrassing quarters of any tech company, ever.

Meta executives have been sloppily apologizing for, and retracting, initiatives that were poorly rolled out and immediately protested by workers. Zuckerberg, meanwhile, has yet to actually say “I am sorry” or “I apologize” for anything. The best he’s offered, in a leaked memo obtained by Reuters, is that “we’ve made mistakes and will almost certainly make more.”

I do not think that response would fly at most companies. It’s the sort of thing Zuckerberg can say without (immediate) repercussions, because he has majority voting control of Meta. Zuckerberg’s disrespectful, aimless leadership style isn’t consequence-free, though. His string of fuck-ups, and the internal revulsion they’ve caused, remind me of Marc Benioff’s foot-in-mouth syndrome at Salesforce—just on a much larger scale.

I don’t know what comes next for Meta or its workers—I’m not sure anyone does, Zuckerberg included. That’s exactly the problem. Here’s a timeline of Meta’s last three months, which, written out, certainly reads like a company in peril.

Late March

Meta is found liable in two landmark cases centered on whether its apps are unsafe and designed to be addictive for younger users. The monetary damages are negligible, but open the door to other, potentially costlier lawsuits.

April

April 8: Meta announces its new AI model, Muse Spark. It’s fine, but doesn’t blow anyone away.

April 12: Mark Zuckerberg is eagerly participating in a project to build an AI version of himself that Meta employees can chat up in their actual boss’s absence. “Zuckerberg has become increasingly hands-on as he oversees Meta’s AI push,” the Financial Times reports.

April 17: Reuters reports that Meta is going to lay off upwards of 8,000 employees in May, which represents roughly 10% of the company’s workforce.

April 21: Reuters obtains another big scoop: Meta is going to install mandatory tracking software on the computers of U.S.-based employees via a tool called Model Capability Initiative (MCI). The tracking software is marketed internally by Meta as a net-positive way to train AI models on how humans interact with their computers. Actual humans at Meta are disturbed by the privacy implications of MCI, and are not enthused about helping AI eliminate their own jobs, especially with layoffs pending.

According to communications reviewed by the New York Times, many Meta workers immediately raised concerns about MCI. “This makes me super uncomfortable,” one worker wrote. “How do we opt out?” Andrew Bosworth, Meta’s CTO, callously responded, “There is no option to opt-out on your corporate laptop.” A Meta spokesperson assures the Times that “there are safeguards in place to protect sensitive content” picked up by MCI.

April 29: Meta’s first quarter earnings report looks great, raising the question: why is the company about to lay off 10% of its workforce?

April 30: During an all-hands meeting, Zuckerberg tries to flatter Meta workers by telling them that they’re geniuses; their collective brainpower is uniquely valuable for the MCI tool and the company’s AI models. “The AI models learn from watching really smart people do things,” Zuckerberg said, according to audio later obtained by More Perfect Union. “In general, 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 [contractors].”

May

May 12: Employees distribute flyers at multiple Meta offices, asking colleagues to sign an online petition against the MCI tool. “Don’t want to work at the Employee Data ​Extraction Factory?” the flyer reportedly asks.

May 14: The online petition, which demands that Zuckerberg not “collect employee ‘computer-use’ data for the purposes of training AI Models,” surpasses 1,000 signatures. The signatories are confirmed Meta employees, not outsiders.

May 18: Approximately 7,000 Meta workers are reassigned to new initiatives where they will be tasked with AI development. This work will be “more rewarding,” Meta’s head of HR assures workers in a memo. It is not. Ensuing interviews with anonymous workers are brutal. “It’s literally the gulag,” one worker relayed to Wired. “You have zero purpose in life all of a sudden, you barely interact with anyone, you just have these tasks every week.”

May 20: Meta officially lays off 10% of its workforce. (This is in addition to the reassignments.)

June

June 2: Meta CTO Andrew Bosworth reports for cleanup duty. During an internal chat, he admits that morale at the company is “maybe not the worst it’s ever been in 20 years here, but it’s probably up there. It’s definitely up there,” he said, according to Business Insider. Separately, Meta announces that it will allow for 30-minute tracking pauses to its MCI tool, as well as opportunities for workers to apply for MCI “exemptions.”

June 3: Online petitions against MCI surpass 1,600 signatures. Organizers object to Meta’s half-hearted MCI compromise, and reiterate their demand for a full ban on the tool.

June 12: A VP on Meta’s Applied AI team tells workers who were required to move to the company’s AI unit that they can now apply for other internal roles if they so choose. Reuters reports on a memo written by Zuckerberg, where he says the company has made mistakes and will make more. “I don’t want to overpromise because the world is changing in ways that are out of our control,” he says of what the future may hold for Meta workers.

June 14: Zuckerberg attends the UFC Freedom 250 event on the White House lawn.

June 15: Bosworth’s apology tour continues. According to Wired, he characterizes Meta’s enormous AI team restructuring as “atrocious,” and pledges to improve Meta’s snack offerings and travel budgets.

June 22: Meta workers’ months-long concerns about MCI are proven correct; Wired reports that “Meta left potentially sensitive information collected from employee laptops accessible to anyone inside the company.” The exact timing of when Meta became aware of the potential data leak is unclear, which Wired hinted at in its report:

“Meta spokesperson Tracy Clayton initially confirmed to Wired that the company is investigating the security issue. As this story was being published, he added that Meta is pausing the data collection program indefinitely. ‘We have carefully designed this program with privacy safeguards and while we have no indication at this time that any data was improperly accessed by Meta employees, we’re pausing it while we investigate,’ Clayton says.”

June 23: Unlike other major AI developers, Meta has yet to submit its AI models for voluntary review by the federal government, the New York Times reports. Also, Zuckerberg bravely tells the Feed Me newsletter that he’s a fan of chicken tenders.

June 24: Meta pulls back even further on its AI team restructuring. In a memo obtained by Business Insider, the company says it will now “defer to each individual’s choice,” rather than forcing workers to remain on the AI training unit.

June 25: A former Meta executive named Sarah Wynn-Williams sues the company, alleging it attempted to stop her from promoting her book about her time with the organization. The title of the book is “Careless People.”

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

• Alex Bores, who proposed some interesting AI policies, lost his House primary race to Micah Lasher on Tuesday. I’ve seen takes about how Bores’s narrow defeat is an ominous sign—the theory is that ideologues like Greg Brockman will be emboldened to meddle even more than they did in this race in order to defeat political candidates who suggest AI regulation. I don’t buy that theory.

  • New York’s 12th Congressional District includes Manhattan’s Upper West Side, Upper East Side, and Midtown; its voter base is older, wealthy, and unusually attuned to traditional endorsements and mainstream media coverage. The district’s outgoing representative, Jerry Nadler, endorsed Lasher, who scored a bunch of other noteworthy endorsements and ran as the “safe” option. Bores used to work at Palantir and was unable to fully secure the district’s left-flank of voters because he’s unabashedly pro-Israel. If anything, his AI stance may have helped him stay within striking distance, given his otherwise uninspiring campaign and background.

• Apple announced that it’s raising the prices of its Mac computers and iPads. “We have never seen a component price increase this much, this quickly,” Apple said in a statement. Apple is referring to AI hyperscalers—the companies that run data centers—and their insatiable thirst for more memory and storage chips, which have become significantly more expensive. As a result, the same component parts are causing a price increase for consumer goods like laptops and phones, as the Wall Street Journal reported. Yet another reason for people to be annoyed by AI!

• Anthropic is still trying to convince the White House to lift an export control directive that effectively banned usage of the company’s Claude Fable 5 AI model. Apparently, one of Anthropic’s new tactics is to keep CEO Dario Amodei as far away from negotiations as possible. Cofounder Tom Brown has been enlisted in Amodei’s place, which seems like a goo idea. “Tom Brown is not being a weirdo like Dario and can actually engage,” a source told Wired.

I followed one suggestion Monday night to an Irish Pub in San Francisco’s Mission neighborhood. A beer garden with an insistent amount of posters for a Catherine O’Hara movie marathon was not the space I expected tech-focused business casual networking to thrive in, but it was packed over an hour before its start time.

The event brought in a somewhat even ratio of women to men, from some still in college to some nearing retirement age. The disarming setting and confident regulars made conversation easy and frequent. A representative of the organization hosting — The Founder’s Club — told me this tradition was formed in the UK, and now takes place monthly in major cities across the United States. They host regular meetings in Miami, Austin, Chicago, LA, Houston, and New York. They focus on AI, health and wellness, or women in tech. Their San Francisco events are catered to start-up founders.

I talked for a while with the founder of a seven-year-old company that designs reusable hardware for satellites and space missions. He was visiting the Bay from Argentina to attend events like this one. He told me fundraising has become very different than it was less than a decade ago; now, every potential investor asks where AI fits into the operation. His company builds physical computing systems, but his software engineers, as is now standard, check and fix code that AI writes rather than coding themselves.

Share

He asked me in turn how AI has affected my world as a writer, and asked if I use it in my work. I told him that I never publish AI-generated writing, and the most its cultural prevalence has affected my work is having to scrutinize my love of em-dashes. He was surprised, and asked if I worry that my work is being used to train AI models (of course I am). That made me wonder how that concern translates to a company that designs physical hardware. AI has been trained well enough to write initial code for his company; does he worry it could absorb and dispel their proprietary information?

Not AI itself, he told me; he’s considered this frequently. Hackers have always been willing and capable to steal and sell information, and the advantage of AI makes that threat more acute. It’s made cybersecurity a greater focus than ever.

A couple from France, also visiting the Bay for meetings like this one, is seeking investors for research about mental health and anxiety in children. They hope to develop software that can help adolescents and their families understand triggers and patterns of their child’s anxiety. They told me raising money in France for this project has been difficult, where the medical research space tends to be insular. This was their first event in San Francisco, and they were excited to be in a space they described as more peer-focused than academically structured.

One of them gestured around the patio at the groups of people in engaged conversations, and observed, “There’s a founder’s mindset here.”

A young man with The Founder’s Club stood up on a bar stool — no microphone, just a projecting voice — and began to call names who had signed up to give their minute-long pitch to the room. An iPhone timer was started when the pitch did. Attendees gathered around whoever was speaking, some of whom also opted to stand on the stool. The group applauded when each pitch ended. They were supportive and responsive. Just as I was wondering how effective an event this may be, members of the crowd started calling out that they wanted to meet with the founders ending their spiels. I watched two young founders exchange information and plans to form a partnership as two start-ups competing with Stripe.

Sustainability was a notable throughline. A young man, from atop the bar stool, showed the room his company’s software tracking materials used in the fast fashion industry with the goal of incentivizing recycling and upcycling fabric. An Australian man told me about his company’s goal to invest in and advance infrastructure for electric, renewable aviation and marine technology. A young woman from Ethiopia told me about Kapsule’s work to expand the infrastructure for healthcare data across Africa.

Leave a comment

I spoke with another of the event’s staffers as folks began to exchange LinkedIns and close their tabs. He said this format of informal bar networking draws new people every week and regulars who enjoy the social aspect, interesting conversation, and hearing new, young entrepreneurs speak.

He asked why I hadn’t pitched from the bar stool. I reminded him that I’m a reporter, not a founder.

“You should still get up next time,” he told me. “Everyone could use a good chance to practice their spiel.”

A new report from Mijente, Just Futures Law, and Surveillance Resistance Lab—the second report of its kind, the first being published in 2018—finds an alarming amount of development in the surveillance economy. This development is propelled by circular deals between tech and the federal government, all focused on enriching surveillance- and defense-tech companies, while keeping tabs on dissidents, immigrants, and pretty much anyone who has a presence or footprint online.

Lobbying feels quaint at a time when oligarchs like Elon Musk and Peter Thiel and venture capitalists like Marc Andreessen feel married, or at least always in bed with, federal policymakers. Social media scrapers, predictive analytics, and real-time location tracking are all tools in the arsenal of the government now. The One Beautiful Bill increased ICE’s budget 1000%, and Customs and Border Patrol spending surged sevenfold in the second half of 2025 alone.

According to the report, tools like Mobile Fortify, which conducts facial recognition, has been used 100,000+ times on streets in Minnesota, Chicago, and Portland. Palantir’s ImmigrationOS are fusing IRS, Social Security, and ALPR data into deportation dossiers. And social media scrapers like Clearview AI and ShadowDragon are already being used to monitor and create profiles on journalists, legal observers, and protesters.

Meanwhile, even people who are relatively offline are being caught by varied surveillance mechanisms. Drones can fly around to scan the faces of people attending protests, while social media scrapers are profiling us across platforms to try and monitor. The power of the subpoena has been weakened, allowing companies to get access to data more rapidly than what a typical process for accessing private communication or private social media activity might look like.

The thesis of “we have tech just in case we have to use it!” has been weaponized, says Jacinta Gonzalez, Mijente member and founder of the No Tech for ICE campaign. The tech is not being used “just in case,” but rather daily, as a means to implement authoritarian control, and as a mechanism to create profitable businesses to take advantage of crises.

“This is incredibly dangerous when paired with desire for control,” Gonzalez told me. “It’s not only a desire to control the economy and government, but a desire to control how we think and how we move and organize—and whether or not democracy is possible. [They are] trying to create conditions so that democracy is not possible, also enabling a police force that is rogue and unaccountable.”

The silver lining is that there are fights in local jurisdictions that are stalling local police departments’ access here. Organizers, according to Gonzalez, are naming and shaming contractors and policymakers—and also becoming familiar with the local police budgets, curbing the usage of these technologies at the procurement level.

“For those who feel like they have safety, I urge them to speak out,” Gonzalez concluded. “Immigrants have been the testing ground for authoritarian policing. But this is a warning of what’s to come if we don’t wake up to how these power structures and systems are impacting our freedom.

What else we’re paying attention to this week…

Meta has prioritized the build of a prediction market similar to Polymarket and Kalshi. Users bet with points, but they haven’t ruled out allowing users to bet with real money. What could go wrong? This comes the same week Meta was forced to roll back a decision to track staffers’ clicks. Wired has covered this extensively. Over 1,500 staffers signed a letter demanding Zuckerberg stall his click tracking plans, facing immense internal pressure, it sounds like he relented for now.

Google is sinking $75 million into film production company A24, claiming that they’ll bring more “AI tools” into filmmaking.

Vermont has passed a bill banning AI chatbot therapy, requiring therapeutic services to be provided by a licensed mental health professional.

Sam Altman’s eyeball scanning company is being investigated for bribery of foreign officials in Thailand, an alleged violation of the Foreign Corrupt Practices Act (FCPA).

Humans will never colonize Mars, and Musk knows it.