Meta moved up to 30% of engineers to data labeling and they left

Meta had a plan to replace engineers with AI. The engineers had a plan of their own.

From running systems to labeling data

New reporting from The Pragmatic Engineer fills in the human side of Project OT, the restructuring covered in late August. After the 10% layoff, Meta reassigned 20 to 30% of its infrastructure and product engineers to data labeling and AI training work. Engineers who had operated systems serving three billion people were writing training tasks for models instead.

The resignation wave

The predictable part followed. The strongest of the reassigned engineers started interviewing and left, for Anthropic, OpenAI and Google. The companies Meta is racing absorbed the talent Meta had moved onto labeling duty.

The November layoff wave stayed canceled, and the AI-native reorganization that was supposed to shrink some teams by 60% has quietly stalled.

The mechanism is worth noting for every company drafting a similar plan. A reassignment order can move an engineer to a new task list, but it cannot move their willingness to stay, and the market for senior infrastructure engineers remains liquid enough that the best ones do not have to.