There’s a certain poetry to being asked to train your own replacement. Meta is doing exactly that, rolling out an internal monitoring program that captures how its engineers work, then feeding that data into AI systems designed to automate those same tasks. This is happening while the company trims roughly 8,000 positions.

The initiative is called the Model Capability Initiative, or MCI. And if you’re a US-based Meta employee using approved work applications, it’s watching you. Closely.

What MCI actually captures

The MCI program monitors mouse movements, clicks, keystrokes, and screenshots across approved work apps used by Meta’s US-based employees. In English: it’s recording nearly everything an engineer does on their screen while working.

The goal isn’t surveillance for its own sake. Meta wants real-world interaction data to train AI systems that can eventually replicate human tasks. Think of it as building a digital twin of your workflow, except the twin doesn’t need a salary, health insurance, or lunch breaks.

This isn’t some skunkworks experiment buried in a research lab. It’s part of a broader company-wide push to make Meta what leadership calls an “AI-native” company. The ambition is to automate internal workflows at scale, improving efficiency while reducing reliance on human labor for routine engineering tasks.

Meta has also set aggressive internal targets for AI adoption. The company wants 65% of its engineers writing more than 75% of their code using AI tools. That’s not a suggestion. That’s a benchmark employees are expected to hit.

The job cuts add uncomfortable context

Here’s the thing. Collecting training data from your workforce looks very different when you’re simultaneously eliminating thousands of roles. The roughly 8,000 job cuts aren’t happening in a vacuum. They’re part of a pattern that’s been unfolding across Meta over the past year, as the company pivots hard toward AI and away from the bloated headcount it accumulated during the pandemic hiring boom.

The juxtaposition is hard to ignore. On one hand, Meta is asking engineers to lean into AI tools, use them daily, let the company record how they work. On the other hand, the company is laying off thousands of people. The implied message lands with the subtlety of a sledgehammer: help us build the thing that makes you expendable.

To be fair, Meta isn’t the only tech giant making this trade. The entire industry is racing to replace routine engineering work with AI-assisted or fully automated pipelines. But few companies have been quite this transparent about the mechanics of how they’re collecting the training data to do it.

The privacy implications are worth noting, too. Keystroke logging and screenshot capture are aggressive monitoring tools by any standard. Even in a corporate environment where employees use company hardware and agree to monitoring policies, the scope of MCI goes well beyond typical productivity tracking. Employees are effectively generating proprietary training datasets with every click.

Meta’s broader AI strategy

MCI doesn’t exist in isolation. Meta has been investing heavily in AI infrastructure, with significant engineering developments reported as recently as November 2023. The company also runs a program called RAISE, which offers full-time AI engineering roles based in locations like Menlo Park and New York City. The message is clear: Meta is hiring AI specialists while trimming traditional engineering and operational roles.

This is the new math of Big Tech. Fewer humans doing more work, augmented and eventually replaced by AI systems trained on the very humans they’re displacing. Meta’s 65% code-generation target for engineers isn’t aspirational fluff. It’s a roadmap for how the company expects software development to work in the near future, with humans serving as editors and supervisors rather than primary authors.

The competitive pressure is real. Google, Microsoft, Amazon, and a wave of well-funded startups are all racing to integrate AI into their core engineering pipelines. Meta can’t afford to fall behind, and Zuckerberg has made clear in multiple public statements that AI is the company’s top priority. The MCI program is one of the more concrete expressions of that priority.

For current Meta employees, the calculus is uncomfortable but straightforward. Embracing AI tools might extend your relevance. Resisting them probably won’t. And every interaction you have with those tools is being captured, analyzed, and used to build systems that could eventually perform your job without you.

For the broader tech workforce, Meta’s approach is a preview of what’s coming everywhere. If a company can record how its best engineers work and distill that into an AI model, the economic incentive to do so is overwhelming. The question isn’t whether other companies will follow. It’s how quickly they’ll catch up.

Investors watching Meta should pay attention to how effectively the company converts this training data into actual productivity gains. Collecting keystrokes is the easy part. Building AI systems that reliably automate complex engineering workflows is orders of magnitude harder. If Meta pulls it off, the cost savings from reduced headcount could be enormous. If the AI tools underperform, the company will have burned goodwill with its workforce for marginal returns.