If you were writing a dystopian tech thriller, you’d struggle to invent a plot more on-the-nose than what Meta announced this week. The company is cutting 10% of its workforce — roughly 8,000 people — while simultaneously rolling out software that tracks employee mouse movements, clicks, and keystrokes to train AI models.
The models that will, eventually, do their jobs.
The Tracking Tool Nobody Asked For
Reuters broke the story earlier this week. Meta is deploying an internal tool called the Model Capability Initiative (MCI) to U.S.-based employees. It captures mouse movements, clicks, keystrokes, and periodic screenshots. The goal: teach AI models the messy, real-world ways humans interact with computers. Dropdown menus, keyboard shortcuts, the digital muscle memory that machines still can’t replicate on their own.
“This is where all Meta employees can help our models get better simply by doing their daily work,” the internal memo read.
Meta CTO Andrew Bosworth framed it as part of the “Agent Transformation Accelerator” initiative. His vision: AI agents “primarily do the work” while humans “direct, review and help them improve.” The agents would “automatically see where we felt the need to intervene so they can be better next time.”
Meta says the data won’t be used for performance reviews. Just model training. Cold comfort when you’re about to be part of a 10% headcount reduction.
$135 Billion on AI, 8,000 Humans Out
The layoff memo from Meta’s Chief People Officer didn’t mention AI directly. It said the cuts would “offset the other investments we’re making.” Everyone knows what those investments are. Meta plans to spend $135 billion on AI in 2026 — roughly equal to its AI spend over the previous three years combined.
Zuckerberg has been saying this out loud for months. In January: “I think that 2026 is going to be the year that AI starts to dramatically change the way that we work.” Projects that once needed large teams were being completed by “a single very talented person” using AI tools. When Microsoft’s Satya Nadella asked how much of Meta’s coding was done by AI, Zuckerberg predicted: “maybe half the development is going to be done by AI, as opposed to people.”
Half. And then more.
The Entire Industry Is Running This Playbook
Meta isn’t an outlier. It’s the template.
On the same day, Microsoft offered voluntary retirement buyouts to 7% of its U.S. workforce — about 8,750 people. Microsoft is spending $110-120 billion on AI infrastructure this fiscal year. Its AI chief, Mustafa Suleyman, said in February that AI will replace most white-collar work within 12 to 18 months.
Amazon has trimmed 30,000 corporate employees recently — nearly 10% of its white-collar workforce. Block cut more than 4,000, nearly half its staff. Snap cut around 1,000.
The pattern: hire AI, fire humans. And no one is pretending otherwise anymore. Nearly every company has explicitly cited AI capabilities as a factor in needing fewer employees.
The Feedback Loop Problem
Here’s what makes Meta’s situation uniquely grim. The keystroke tracking creates a literal feedback loop: workers are training the system that makes them obsolete.
One Meta employee called the tracking initiative “dystopian.” Hard to argue.
And underneath the technical mechanics, there’s a philosophical question that employment law hasn’t caught up to. If your employer captures every click, every shortcut, every workflow habit you’ve developed over years — and feeds it into a model that replicates those behaviors at scale, 24/7, at a fraction of the cost — who owns that knowledge?
Your muscle memory, your workflow instincts, your accumulated expertise. Is that yours? Or does it belong to the company that employed you while you developed it?
The courts will have to answer that. Soon.
The Playbook Is Clear
Strip away the corporate language and the sequence is straightforward:
- Invest massively in AI infrastructure ($100B+ annually)
- Capture employee workflows as training data
- Build AI agents that replicate those workflows
- Cut the humans whose work you’ve automated
- Repeat
This isn’t about replacing bad workers. It’s about replacing all workers whose jobs can be decomposed into computer interactions. In a post-pandemic remote-work world, that’s most knowledge work.
Meanwhile, the Models Keep Getting Better
As if to punctuate the moment, OpenAI released GPT-5.5 the same day — calling it “a new class of intelligence.” OpenAI’s chief scientist Jakub Pachocki offered a remarkable quote: “I would say, like, I think the last two years have been surprisingly slow.”
Surprisingly slow. While tens of thousands of tech workers lose their jobs to the technology that’s apparently not even moving fast enough.
What Workers Can Actually Do
The uncomfortable truth: the workers who survive this transition will be the ones who learn to direct AI agents rather than compete with them. That’s exactly the role Bosworth described — human as supervisor, not human as executor.
But let’s not dress this up as opportunity for the 8,000 people getting layoff notices in May. For them, the AI-augmented future is cold comfort.
The real debate isn’t whether AI can do knowledge work. That’s settled. The debate is about the speed of transition and what we owe the people caught in the gap. Should companies be allowed to capture employee behavior to train their replacements? Should productivity gains from AI be shared beyond shareholders?
These aren’t hypothetical questions anymore. They’re playing out in real memos, at real companies, affecting real people — right now.