Meta Mandates 75% AI-Written Code — Engineers Laid Off as Zuckerberg Says AI Replaces Whole Teams

What the Leaked Documents Show

Business Insider obtained internal Meta documents this week showing the company has set formal, measurable AI adoption targets for its engineering workforce. The targets are not suggestions — they are tied to performance reviews. Starting in 2026, "AI-driven impact" is a core expectation for how Meta evaluates its engineers.

The specific targets vary by division:

The AI tools being tracked include Meta's internal systems DevMate and Metamate, as well as Google's Gemini. Meta CTO Andrew Bosworth has taken direct charge of the "AI for Work" initiative, reporting its progress to Zuckerberg personally.

The 3-Step Playbook That Went Viral

A post on American Bazaar this week summarized Meta's strategy in three steps that went viral across LinkedIn, X, and tech forums — because engineers at many other companies recognized the pattern immediately:

Meta says the AI mandates and the March 2026 layoffs are unrelated. The timing — AI adoption targets announced, then hundreds cut within weeks — has prompted engineers at the company and across the industry to question that framing.

The New Org Structure: "AI Pods"

The most visible structural change is in Reality Labs, Meta's augmented and virtual reality division. Meta reorganized a 1,000-person division into small "AI pods" — cross-functional units designed to operate with minimal headcount and maximum AI assistance.

The new job titles inside AI pods tell the story clearly: "AI Builder," "AI Pod Lead," and "AI Org Lead." The previous hierarchy — senior engineers, staff engineers, principal engineers — is being replaced with a flatter structure where the primary skill evaluated is the ability to direct AI agents to produce output.

The Risk No One Is Talking About

Meta recently experienced a high-severity internal security incident that illustrates the downside of rapid AI agent adoption. An AI agent provided incorrect technical advice during a production debugging session, leading to a two-hour window during which sensitive data was exposed to unintended parties.

The incident did not become public until internal Slack discussions leaked this week — the same week as the coding mandate revelations. Meta's next-generation frontier model, internally codenamed "Avocado," has also faced delays after falling short of internal benchmarks for complex reasoning and coding. The model that is supposed to accelerate Meta's transition to AI-first engineering is itself behind schedule.

How the Big Tech AI Coding Mandates Compare

What This Means for Software Engineers

The engineers who survived Meta's March 2026 cuts share a common profile: they were already using AI tools heavily, they had clear ownership of systems that AI cannot fully replace yet (security architecture, distributed systems design, product judgment), or they were in roles that directly manage AI agents rather than writing code themselves.

The engineers at risk are those still writing code the way they did in 2022 — line by line, without AI assistance — in areas where AI tools can now produce the same output in minutes. Meta's leaked documents make it clear: usage of AI is now a measurable performance metric, not an optional productivity choice.

The broader signal is that Meta is not an outlier. It is the most transparent company about a transition happening across the entire industry simultaneously. Google, Microsoft, Amazon, and Block have all executed significant layoffs in the past 18 months while publicly citing AI productivity as the reason. Meta has simply been the most explicit about the specific numbers.

Frequently Asked Questions