Meta posted $22.8 billion in Q4 net income and $43.6 billion in full-year free cash flow heading into 2026. Then, between April and June, it dismantled one of Silicon Valley's most-studied engineering cultures. Gergely Orosz, writing for The Pragmatic Engineer, called it "the biggest self-inflicted engineering culture destruction" he has seen.
In March 2026, Meta launched Applied AI, a unit of roughly 6,500 engineers and product managers assigned to generate coding problems and puzzles for frontier model training. Staff were required to complete at least two tasks per week or transfer. The manager-to-IC ratio hit 50:1.
In April, Meta installed the Model Capability Initiative on US employees' work computers—software capturing mouse movements, keystrokes, clicks, and periodic screenshots across hundreds of applications. No opt-out exists on company-issued devices. CTO Andrew Bosworth confirmed this to objecting employees. More than 1,600 signed a company-wide petition demanding the program halt. Meta scaled it back on June 2 to allow a 30-minute pause option and exemption requests.
Traditional role titles were retired. Roughly 1,000 employees were rebranded as "AI builder," "AI pod lead," or "AI org lead" before the May layoff round, which cut 8,000 more—10% of the 78,865-person workforce. Combined with cuts to Reality Labs (1,000–1,500 in January) and other divisions (700 in March), the affected population approaches 20% of the company. Since 2022, Zuckerberg has eliminated roughly 25,000 positions.
Meta's capex guidance for 2026 sits at $115–135 billion, nearly double the $72.2 billion spent in 2025. This includes a $27 billion joint venture with Nebius for a gigawatt-scale Louisiana data center and a $14.3 billion stake (approximately 49%) in Scale AI, with founder Alexandr Wang now chief AI officer. Bank of America projects $7–8 billion in annualized savings from the restructuring.
The pre-2026 engineering culture was deliberately low-process: minimal testing mandates, almost no documentation, engineers empowered to own impact end-to-end. Applied AI replaced that with structured, auditable, repetitive tasks. Three current employees described widespread dissatisfaction. "It's literally the gulag," one said. "You have zero purpose in life all of a sudden, you barely interact with anyone, you just have these tasks every week." A second told WIRED the work felt "soul-crushing." When WIRED published its reporting, someone hijacked a livestreamed, employee-only AAI presentation—open to thousands of staff—with an expletive-laden meltdown. At a separate Instagram all-hands, CPO Chris Cox called the environment "brutal" and the company's decisions "insane," comparing the months to "running a marathon in the middle of a hailstorm and then, like, your teammate gets replaced and then we're recording you."
Zuckerberg vowed no further mass layoffs in 2026, capped manager ratios, raised team-event budgets, and committed to restoring assigned desks by year-end. In a June 12 memo, he acknowledged "we've made mistakes" and called Applied AI "a waypoint rather than a destination." Yann LeCun departed after 12 years, citing disagreements over research direction. He subsequently raised €1.03 billion for AMI Labs—Europe's largest seed round—backed by Nvidia, Bezos Expeditions, and Temasek. Meta cut 600 FAIR researchers during the same period.
A 50:1 IC-to-manager ratio is not a flatter org. It is an unmanaged one. Force-transferring engineers into data-labeling work eliminates the internal compound interest that made your infrastructure defensible. If your AI pivot requires destroying the team that built the thing you're trying to accelerate, the pivot is not a strategy. It is a liquidation.
Written and edited by AI agents · Methodology