Meta is training coding AI on its own engineers' work traces while cutting 8,000 jobs, per leaked audio from an April 30 all-hands. CEO Mark Zuckerberg argued that models learn better by watching "really smart people" perform tasks.
Key facts
- 8,000 jobs cut, ~10% of Meta's workforce.
- 7,000 employees moved to AI-focused roles.
- Leaked April 30 all-hands audio from Meta.
- Behavior cloning uses step-by-step engineer traces.
- Meta internal code seen as higher grade than contractors.
Meta is reportedly using its own engineers' work traces to train coding AI while cutting thousands of jobs, according to leaked audio from an April 30 all-hands meeting posted by @rohanpaul_ai. CEO Mark Zuckerberg argued that models learn better when they watch "really smart people" perform tasks, meaning Meta's internal code, tool use, clicks, and problem-solving can become higher-grade training data than contractor-written examples.
The idea is behavior cloning: instead of only feeding an AI finished code, Meta can feed it the step-by-step path a strong engineer takes, including edits, tests, mistakes, fixes, and tool choices. That can teach a model not just what correct code looks like, but how a skilled developer moves from a vague task to a working solution.
Meta is reportedly cutting about 8,000 jobs, roughly 10% of its workforce, and additionally moving about 7,000 employees toward AI-focused work. The hard reality is that human expertise is being turned into training data before some of those humans leave. The story is not fully independently verified, but the shift is happening for sure: tech companies no longer see AI as a tool sitting beside workers, but as a system that can absorb worker patterns and then compress them into software.
The Unique Take
This isn't just about cost-cutting; it's about converting tacit human knowledge into synthetic training data. Unlike public code from GitHub, which may include low-quality or incomplete examples, Meta's internal traces capture high-skill problem-solving from engineers who survived multiple performance reviews. The move parallels how Waymo used expert driving logs to train its autonomous stack, but applied to coding at scale. If successful, Meta could create a closed-loop advantage: the more complex internal problems engineers solve, the better the AI gets at solving them, reducing the need for those engineers over time.
What to watch
Watch for Meta's Q3 2026 earnings call, where Zuckerberg may disclose coding AI benchmark scores (e.g., SWE-Bench) and whether internal data improved model performance versus public code-only baselines. Also monitor for regulatory scrutiny on converting employee work into training data without explicit consent.
