‘Tell Him He’s a Piece of Shit’: Meta’s New AI Unit Is a Total Mess
Executives and employees alike are struggling with Meta's chaotic AI strategy, according to sources and internal discussions reviewed by WIRED. The company's recent push to centralize artificial intelligence under a new unit has been plagued by infighting, unclear direction, and toxic leadership, leaving many staffers demoralized as 2026 unfolds.
Internal Turmoil
Interviews with current and former Meta employees, as well as leaked internal communications, reveal a deeply fractured organization. The new AI unit, launched with much fanfare in early 2026, was intended to streamline Meta's efforts in generative AI and large language models. Instead, it has become a battlefield of competing visions. One senior engineer described the atmosphere as “like a reality show where everyone hates the producers.” The unit’s leader, a high-profile hire from a rival tech firm, has been accused of micro-managing and publicly berating subordinates—including telling one engineer to “tell him he’s a piece of shit” in a group chat. This comment, later leaked, became a rallying cry for disgruntled staff.
Leadership Struggles
Mark Zuckerberg, Meta’s CEO, has personally intervened multiple times to settle disputes over resource allocation and technical direction. According to sources, he has oscillated between pushing for aggressive monetization of AI features and cautioning against public backlash. This indecision has trickled down, causing paralysis in product development teams. The AI unit faces constant pressure to deliver on promises of “metaverse-ready” assistants, yet internal benchmarks show these systems still lag behind competitors like OpenAI’s GPT-6 and Google’s Gemini Ultra in reasoning tasks. By mid-2026, Meta has missed two internal launch deadlines for its flagship chatbot, codenamed “Atlas.”
Cultural Clash
The turmoil reflects a larger cultural clash within Meta. The new unit was staffed partly with researchers from Meta’s Fundamental AI Research (FAIR) division and partly with product engineers from the main Facebook app. The two groups have notoriously different working styles: researchers seek long-term breakthroughs, while product teams demand rapid iteration. This conflict has escalated into passive-aggressive memos and withheld data. One internal survey from April 2026 found that only 34% of AI unit employees felt “psychologically safe” to voice dissenting opinions.
The 2026 Context
Meta’s AI struggles come at a critical time. In 2026, the AI arms race has intensified, with Apple launching a widely praised “Siri Pro” and Amazon integrating generative AI into Alexa. Meanwhile, regulatory scrutiny is mounting: the European Union’s AI Act, fully enforced since January, imposes strict transparency rules that Meta has been slow to comply with. Internally, Zuckerberg has reportedly warned that if the AI unit doesn’t show results by Q3 2026, he will consider breaking it up again—a move that would further discredit the company’s strategic direction.
What’s at Stake
For Meta, the AI unit’s failure threatens its entire ecosystem. The company relies on AI for content moderation, ad targeting, and its metaverse ambitions. If the unit remains dysfunctional, Meta risks losing ground to nimbler rivals. Employees have started leaking to the press, hoping external pressure will force change. As one manager put it: “We’re building the future—if we can stop fighting first.”
This article was published on June 12, 2026, at 5:16 PM, and is part of WIRED’s ongoing coverage of business and technology.
via Wired AI