Meta has faced internal turmoil after a company-wide reorganization that placed a large number of engineers into a new AI unit. Meta Chief Technology Officer Andrew Bosworth said employee morale has dropped to a 20-year low and described the reorganization as "atrocious." Central to the dispute is Applied AI, a newly formed unit of roughly 6,500 people assigned to create training tasks for Meta's broader AI initiatives.
What happened on the ground
- In the span of a week multiple internal clashes surfaced; in one incident an employee hijacked a livestream to publicly insult a Meta AI executive.
- In response to the unrest, executives took a series of remedial steps: they restored office snacks and micro-kitchens, re-enabled certain travel budgets, and assigned permanent desks to employees.
- The reorganization moved engineers and other staff into a centralized unit focused on preparing and managing training workloads for large-scale AI efforts.
Why this matters
Observers and some insiders argue the problem extends beyond morale. Their critique focuses on scale and organizational design: modern breakthroughs in frontier AI increasingly come from small, focused teams rather than divisions of thousands. A large, newly assembled unit can generate internal politics, bureaucratic busywork, and coordination problems that sap productivity.
Applied AI's size — about 6,500 people — raises the question of whether such headcount can be effectively concentrated on a single, high-priority mission without devolving into friction and dilution of responsibility. The leadership's immediate fixes address employee dissatisfaction in the short term, but critics say they do not resolve the structural challenge of aligning a town-sized workforce behind fast-moving AI research goals.
Conclusion
The episode highlights a tension at major technology firms between scale as an advantage and scale as a complication. Meta's quick measures aim to calm employees, but the longer-term issue remains how to organize people and resources so that large headcount does not undermine the speed and focus required for frontier AI development.
