Meta’s AI Push Turns Into an Internal Crisis as Thiel’s Secretive Society and SBF’s Pardon Hunt Add to Tech’s Turbulent Week

Meta’s all-out effort to dominate the artificial intelligence race is generating more than technical setbacks and expensive hiring sprees. It is also producing a morale crisis inside the company, where employees shuffled into newly created AI teams say they feel sidelined, overmanaged and increasingly disconnected from the work they were hired to do.

That internal turmoil was one of several high-profile topics discussed this week on WIRED’s Uncanny Valley podcast, alongside a newly exposed trove of documents tied to Peter Thiel’s invitation-only society, Dialog, and Sam Bankman-Fried’s latest attempt to win political favor through a possible pardon from the Trump administration. The episode also touched on broader questions around the state of the AI industry, including Meta’s uneven progress and Anthropic’s ongoing negotiations with the government.

What ties the stories together is a familiar theme in the tech world in 2026: power is concentrating, institutions are reorganizing around artificial intelligence, and the people inside those institutions are often the last to know how disruptive the changes will be.

Meta’s AI reorganization is colliding with employee frustration

Meta has spent the past year trying to close the gap with rivals such as OpenAI and Anthropic, pouring money into AI infrastructure, recruiting expensive talent and reshaping its workforce around the company’s superintelligence ambitions. But the internal fallout from that transformation is now impossible to ignore.

As part of a major restructuring last month, Meta moved roughly 7,000 employees into AI-focused roles. The company’s newly expanded AI push includes Meta Superintelligence Labs, an elite research effort that sits at the center of CEO Mark Zuckerberg’s plan to make the company a leading force in frontier AI.

One of the teams absorbing those employees is Meta’s applied AI engineering group, which serves as a practical support layer for the more prestigious lab. Employees who spoke about the situation described the move as demoralizing, saying they were pushed into work they did not choose and that now feels repetitive, narrow and lacking in ownership.

The sense of disillusionment has become severe enough that an internal meeting last week erupted into open hostility. During a company-only call for the applied AI unit, someone interrupted the discussion with an expletive-laced attack on the team’s leadership and urged attendees to pass along the insult to a Meta AI executive. The recording, which has circulated internally, was notable not only for the outburst but for the silence that followed it.

For many employees, the resentment appears to be rooted in more than just a rough reorg. It reflects a deeper anxiety that Meta is asking workers to become auxiliary labor for machines, even as those same machines may eventually replace some of their jobs.

Why Meta employees are so angry

Employees describe the new assignments as far less intellectually rewarding than the roles they held before the restructuring. Instead of building products or features with visible impact, many say they are now fine-tuning systems, responding to AI failures and completing tasks that feel like cleanup work for a model rather than original engineering.

That may sound routine to AI insiders, but for the staff caught in the transition, the issue is one of status and purpose. Workers who previously saw themselves as building consumer products or supporting ambitious product lines now say they are stuck performing work they did not ask for, under a structure they did not help shape.

The situation is made worse by the fact that Meta’s broader business remains financially strong. The company has reported standout quarters and continues to generate enormous revenue from its core advertising business. That means employees are watching the company do well overall while being told their teams need to absorb layoffs, rearrangements and new reporting lines to support a costly AI strategy that has yet to prove itself.

There is also a structural tension at the heart of the company’s AI pivot. Meta is spending heavily to catch up in a market where it is not clearly leading. That makes the shift feel to many employees less like a confident evolution and more like a scramble.

In practical terms, that scramble has also come with other changes that workers say have deepened the sense of surveillance. Meta has been expanding monitoring of laptop activity in ways the company says are tied to AI development and productivity. For employees already uneasy about their new roles, that added layer of oversight has sharpened the feeling that they are being treated less like creative contributors and more like training data with badges.

What workers say they lost in the shuffle

• Choice over whether to join the AI unit
• Control over the type of work they do
• A sense of mission beyond model support
• Confidence that management understands their concerns

Leadership’s morale fix did not land the way Meta hoped

Meta’s leadership has not been silent. Zuckerberg and other top executives have publicly embraced the AI transition as a once-in-a-generation opportunity, framing it as a moment for the company to reinvent itself and compete at the highest level in artificial intelligence.

But one of the company’s proposed morale boosters landed poorly: a hackathon. To senior leaders, the suggestion may have sounded like a return to Meta’s old hacker culture, the kind of energetic, experimental ethos that helped define Silicon Valley’s most ambitious era. To employees in the middle of layoffs and reorganizations, it sounded like an out-of-touch distraction.

Workers reportedly pushed back quickly, arguing that they were under too much pressure to treat a hackathon like a morale-building exercise. The reaction exposed a culture gap inside the company. Executives appear to believe the AI upheaval is a chance to rally employees around a grand project. Many employees instead experience it as a top-down mandate with little room for dissent.

The contrast matters because Meta’s leadership is not just asking for patience. It is asking for buy-in to a transformation that may reshape the company’s product priorities, staffing model and future labor needs. That is a much bigger ask than a typical reorganization, and it helps explain why a seemingly upbeat idea like a hackathon became a symbol of managerial disconnect.

Andrew Bosworth’s apology and the limits of internal fixes

Meta CTO Andrew Bosworth has acknowledged that communication around the reorganization was poor, calling it “atrocious” in remarks that suggest the company understands at least part of the problem. But the remedies discussed internally have done little to calm concerns.

Among the measures floated were managerial caps, restrictions on how often employees can switch managers during restructuring and efforts to improve office amenities, including better snacks. Those kinds of changes may help at the margins, but they do not address the larger anxiety about direction, autonomy and purpose.

Senior Meta leaders are said to see the AI overhaul as a historic opportunity, while many employees describe it as a disorienting loss of agency and a sign that the company is more interested in infrastructure and output than in the people doing the work.

That gap between leadership enthusiasm and employee despair may be the central challenge Meta faces as it tries to turn AI into a durable business advantage.

Meta’s AI ambitions are still looking for a breakthrough

Despite the spending spree, Meta has not yet produced the kind of AI breakthrough that would justify the turmoil internally. The company has released models and tools, but several rollouts have been uneven, delayed or greeted with indifference by users.

Its latest model launches have not yet created the momentum Meta wants. Tools deployed across Instagram and other apps have not turned into breakout hits, and early signs suggest users are not flocking to them in the numbers the company may have hoped for. That makes the restructuring feel even more precarious for employees, because they are being asked to absorb immediate pain for gains that remain hypothetical.

Meta has, of course, made a habit of chasing giant bets. The metaverse was the last grand example, a multibillion-dollar push that reoriented the company without delivering a broadly accepted transformation of consumer behavior. AI is different in one important respect: competitors are already proving that there is real commercial demand here. That makes Meta’s lag more visible and more embarrassing.

The company’s challenge now is not simply building capable AI systems. It is convincing a skeptical workforce that the sacrifices required to catch up are worth it.

Key facts about Meta’s AI upheaval

Peter Thiel’s Dialog society emerges from the shadows

The podcast episode also examined a different kind of power network: Dialog, the invite-only society co-founded by billionaire investor Peter Thiel in 2006. For years, the group has operated with minimal transparency, declining to publish a public member list and maintaining an aura of exclusivity that has fueled speculation about who belongs and what they discuss.

That changed after a directory embedded in the society’s website code was uncovered by a Swiss hacktivist and independently reviewed by WIRED. The materials shed light on more than 200 names connected to the group, including government officials, business leaders, academics and other high-profile figures.

The list is remarkable not just for the number of powerful people involved but for the range of spheres represented. Among those tied to the 2026 retreat are sitting Trump administration officials, U.S. senators, members of the PayPal alumni network, foreign intelligence figures and ambassadors. The collection reads like a map of the modern elite, assembled around an off-the-record annual gathering.

Dialog’s existence has long fit a familiar Thiel pattern: private, ideologically charged and deeply networked. But the newly exposed records offer a glimpse into how such groups function in the age of AI, geopolitical tension and tech’s growing political influence.

The sessions sound like a parody of elite discourse

Some of the listed discussion themes are unusually blunt, even by the standards of elite retreat programming. Among the session titles reportedly planned for the 2026 event are “Money (Does?) Buy Happiness,” “Bring Back Nuclear,” “Navigating World War III” and “How’s Your Sex Life?”

Those titles may sound absurd, but they also reveal something important about the kinds of conversations happening inside these closed rooms: finance, geopolitics, energy, intimacy and existential risk all sit comfortably together when the participants are powerful enough to imagine themselves discussing civilization at scale.

The presence of a matchmaking or dating component adds another layer of surrealism. Retreat attendees reportedly could indicate whether they were single or interested in meeting others, suggesting that Dialog is part conference, part salon and part social club. That blend of intellectual ambition and personal networking is hardly unusual in elite circles, but the secrecy surrounding the group makes it feel especially theatrical.

The leaked materials suggest that Dialog is less a conventional policy forum than a tightly curated environment where tech, politics and status intersect behind closed doors.

Why the Dialog leak matters beyond gossip

At first glance, a secretive billionaire society may sound like a curiosity. In reality, it is a window into how influence operates in Silicon Valley and Washington. Groups like Dialog provide a venue for unrecorded conversations among people who shape policy, capital flows, technologies and public narratives.

That matters because modern tech power increasingly depends on relationships that are not fully visible to the public. AI policy, export controls, defense strategy, platform regulation and energy infrastructure all now depend on informal networks as much as formal institutions. When the people in those networks gather privately and off the record, the public gets only the fragments that leak out later.

The Dialog documents also underscore how closely the tech elite is tethered to political institutions. The mix of current officials, former intelligence figures and industry veterans suggests that Silicon Valley’s influence is not just economic. It is also diplomatic, ideological and sometimes strategic.

For Thiel, whose influence has long stretched from venture capital to politics, Dialog is a reminder that some of the most consequential conversations in tech do not happen on stage or in public hearings. They happen in rooms designed to keep them private.

Sam Bankman-Fried is still trying to rewrite his story

Another thread in the podcast episode centered on Sam Bankman-Fried, the disgraced crypto founder who is now serving time after his conviction. According to the discussion, he is actively seeking a pardon from the Trump administration while also exploring the possibility of some kind of comeback.

That strategy is notable less for its odds of success than for its audacity. Bankman-Fried’s public image collapsed after one of the most dramatic fraud and collapse cases in recent financial history, yet he continues to behave as if a political and reputational reset is still possible.

The idea of a pardon hunt alone would be newsworthy. The possibility that Bankman-Fried is also looking beyond that to a future role in the industry raises larger questions about how tech figures with enough money, connections or narrative discipline attempt to rehabilitate themselves after scandal.

In the age of highly polarized politics, pardon requests have become more than legal long shots. They are also public tests of influence. Bankman-Fried’s effort shows how the boundaries between law, reputation and power remain porous, especially when the subjects are wealthy and well connected.

Anthropic, the White House and the politics of model access

The episode also noted that Anthropic is still navigating a tense relationship with the Trump administration over continued access for its Claude Fable 5 models. While details of the negotiation remain fluid, the broader issue reflects a growing reality in AI: model deployment is no longer just a product decision. It is increasingly a matter of policy, security and government approval.

That is a significant development for the AI industry. A company can build a model, test it and prepare a release, but whether that model is available at scale may depend on regulatory concerns, political pressure or unresolved disputes with the government. Anthropic’s situation suggests that even the most technically advanced systems now operate inside a politically charged framework.

For AI companies, the lesson is clear. Winning in the market is only part of the battle. Winning permission to operate, especially at the frontier, is becoming just as important.

A week that shows how much power is hidden in plain sight

The common denominator across Meta, Dialog, Bankman-Fried and Anthropic is not simply controversy. It is the concentration of power in systems that are increasingly difficult for outsiders to observe and influence.

Meta is reorganizing thousands of employees around an AI future that many workers do not trust. Dialog illustrates how much elite coordination still happens through private gatherings that escape public scrutiny. Bankman-Fried’s pardon push shows how convicted tech figures continue to seek political leverage. And Anthropic’s government negotiations reveal that AI deployment is becoming inseparable from state power.

Taken together, these stories show a tech sector in a phase of transition, but not necessarily progress. Companies are moving fast, but not always in ways that produce confidence among workers, regulators or the public.

The result is a landscape defined by ambition and instability. Leaders talk about transformative technologies and civilization-scale stakes. Employees talk about exhaustion and loss of agency. Private clubs and exclusive networks continue to shape outcomes behind the scenes. And the people trying to steer the future often seem most worried about who controls the room.

What to watch next

Several developments will show whether the current turbulence is temporary or structural.

• Whether Meta can produce a major AI product that changes the company’s public narrative
• Whether more employees speak out about the applied AI unit and broader reorganization
• Whether the Dialog leak prompts any public scrutiny of Thiel’s network
• Whether Bankman-Fried’s pardon effort gains any traction
• Whether Anthropic’s negotiations affect model availability and deployment

For now, the week’s biggest story may be that the AI boom is no longer just about model benchmarks and product launches. It is also about labor, secrecy, influence and the fight to control the institutions that will shape the next phase of technology.

superintelligencenews@gmail.com