#meta · How AI Is Reshaping Work

Meta's May layoff of 10% was the first half of a two-wave plan code-named Project OT. Reuters published the documents on August 26. The second wave died the night before the first one landed, after Meta's own telemetry showed its AI agents were generating code without shipping product.

Meta's Q2 2026 report put a price on its May layoff: $1.18 billion in severance for about 8,000 people, while total headcount barely moved.

Twenty-six Meta employees say an AI system picked them for the May layoff, using metrics you cannot accumulate from a hospital bed.

Tech executives increasingly blame AI for layoffs that have older, less flattering causes — over-hiring, failed bets and softening markets — analysts told the New York Times.

Meta's 8,000-role cut started at 4 a.m. May 20. CPO Janelle Gale's memo names the AI initiatives — AAI, ATA XFN, Central Analytics, Enterprise Solutions — that 7,000 survivors are being moved into. Severance: 16 weeks + 2 weeks per year + 18 months healthcare. Identical to the Intuit formula filed two days earlier.

Meta starts cutting ~8,000 roles on Wednesday. The same week, capex guidance was raised to $125–145B. Zuckerberg told staff the company has 「two cost centers — compute and people」 and refused to rule out more cuts.

On May 1, the day after Mag-4 earnings, Tesla's former VP of People wrote in Fortune that the AI-layoff story driving the headlines applies to roughly 20% of US workers. The 80% figure traces to BLS payroll data on hyperscaler vs. non-hyperscaler employment. The panic, in other words, is real for the people writing the panic and largely fictional for the people reading it.

Microsoft, Alphabet, Amazon, and Meta posted ~$430B of combined Q1 revenue and revised 2026 capex up to roughly $650B. In the same five trading days, Meta announced 8,000 layoffs plus 6,000 cancelled reqs, Microsoft put its first-ever buyout offer in front of 8,750 US workers, and the words 「efficiency」 and 「discipline」 appeared 15 times across the four earnings calls. Sam Altman calls this 「AI washing」. The Washington Post agrees.

Reuters reported on April 21 that Meta is installing the Model Capability Initiative on US-employee laptops — capturing mouse movements, clicks, keystrokes, and occasional screen snapshots — to train workplace AI agents. The 8,000-person layoff is scheduled for May 20. The rhyme is not subtle.

Sama, Meta's data-annotation contractor in Nairobi, sent redundancy notices to 1,108 employees on April 16 after Meta terminated the contract they had been staffing. Their job was labeling the images and bathroom-footage that trained Meta's Ray-Ban AI glasses. Their reward for training the AI: the AI no longer needs them.

Reuters says Meta has pencilled in May 20 for the first wave — 10% of headcount, about 8,000 people. Round two is already scheduled; the only question is how much of the 20% total gets cut before December.