Welcome back to SavvyMonk, your one-stop for AI and tech news that actually matters.

Today's story is about Meta installing tracking software on employee computers to capture how they work, and then feeding all of that data straight into its AI models. The company says it's about making better AI agents. Employees are saying it feels like training their own replacements. And the timing couldn't be worse.

Let's get into it.

Tech moves fast, but you're still playing catch-up?

That's exactly why 200K+ engineers working at Google, Meta, and Apple read The Code twice a week.

Here's what you get:

All delivered twice a week in just 2 short emails.

Join 200K+ engineers

Meta is rolling out a new internal tool called Model Capability Initiative (MCI) on the computers of its US-based employees.

According to internal memos seen by Reuters, the tool will capture mouse movements, clicks, keystrokes, and occasional screenshots across a designated list of work-related apps and websites. All of that data will be piped directly into Meta's AI training pipeline.

The stated purpose is straightforward. Meta's AI models still struggle with basic computer tasks that humans handle effortlessly, like selecting items from dropdown menus, using keyboard shortcuts, and navigating multi-step workflows across different applications. The MCI tool is designed to capture real-world examples of how humans actually interact with software so the company's models can learn to replicate those behaviors.

The internal memo framed the initiative as an opportunity for employees to contribute by simply doing their daily work. No extra effort required. No labeling tasks. Just go about your day while the software watches.

The Bigger Picture

The MCI announcement didn't arrive in isolation. A day earlier, Meta CTO Andrew Bosworth sent employees a separate memo about ramping up internal data collection as part of the company's AI for Work initiative, which has been rebranded as the Agent Transformation Accelerator (ATA).

Bosworth laid out a vision where AI agents handle the bulk of the work while humans shift into roles of directing, reviewing, and helping the agents improve.

Meta spokesperson Andy Stone confirmed that MCI data would be among the inputs feeding the ATA effort. The company says safeguards are in place to protect sensitive content and that the data won't be used for employee performance evaluations. But the memos didn't specify exactly what counts as "sensitive content" or how the boundaries would be enforced.

This is all happening under the leadership of Alexandr Wang, who joined Meta in June 2025 as its first-ever Chief AI Officer. Wang came from Scale AI, where he served as CEO before Meta invested $14.3 billion for a 49% stake in the company. He now leads Meta Superintelligence Labs (MSL), the division overseeing MCI and the broader AI transformation effort.

The Legal Reality

Here's the part that makes this particularly uncomfortable for employees. According to Ifeoma Ajunwa, a law professor at Yale University, there is essentially no federal limit on worker surveillance in the United States. State-level laws generally only require that workers be broadly informed that monitoring is happening, which Meta has now done via these internal memos.

Ajunwa noted that this kind of real-time keystroke and screen logging pushes white-collar workers into a level of continuous surveillance that was previously reserved for delivery drivers and gig workers. The difference is that gig workers at least know the monitoring is about performance. Here, the monitoring is explicitly about capturing how you work so that an AI can learn to do it instead.

In Europe, this would be a much harder sell. Privacy and labor regulations would likely prevent this kind of blanket employee data collection for AI training purposes without much more explicit consent mechanisms. But in the US, Meta is operating within legal bounds.

The Layoff Coincidence

And then there's the timing. Meta is reportedly planning to lay off approximately 8,000 employees, roughly 10% of its global workforce, starting on May 20. Additional cuts are planned for the second half of 2026. The company has already cut thousands of positions earlier this year across Reality Labs, recruiting, sales, and global operations.

So the sequence of events reads like this. Meta installs software that records how employees use their computers. It feeds that data into AI models designed to replicate those exact behaviors. And then it begins laying off 10% of the workforce less than a month later. The company says these are unrelated initiatives. Employees can be forgiven for connecting the dots.

Meta's 2026 capital expenditure budget sits between $115 billion and $135 billion, almost entirely directed toward AI infrastructure. The company generated over $200 billion in revenue last year. These layoffs are not about financial distress. They are about strategic reallocation, moving money and headcount away from human roles and toward AI capabilities that the company believes will generate more value over time.

The Bottom Line

Meta is being more transparent about this than most companies, and that transparency is exactly what makes it unsettling. The company is telling employees, in writing, that it wants AI agents to do the work while humans supervise. It is installing tools to capture the data that makes that possible. And it is cutting the workforce at the same time. Whether you see this as honest corporate strategy or a dystopian preview of where white-collar work is headed probably depends on which side of the layoff list you're on.

Category: Workplace Automation Assessment

"Review my daily workflow across [list your tools and apps] and identify the top 5 tasks that are most repetitive and could realistically be handled by an AI agent. For each task, explain what data the agent would need, what the risks of automation would be, and what human oversight would still be required."

The uncomfortable truth in this story isn't that Meta is doing something new. It's that Meta is doing something that most companies are thinking about but won't say out loud. Every time you use a work tool with AI features baked in, some version of this data capture is happening. Meta just put it in a memo. That honesty might be the scariest part.

Hit reply, I read every response.

See you in the next one.

— Vivek

P.S. If you know a developer, founder, or tech professional who'd find this useful, forward this their way. They can subscribe at https://savvymonk.beehiiv.com/