Meta was developing a secret plan called Project OT (Organization Transformation), which involved reducing individual teams by 60% to replace humans with autonomous AI agents. However, instead of the expected productivity jump, the experiment caused a surge in critical crashes, and Mark Zuckerberg was forced to cancel a second wave of mass layoffs.
Meta was developing a secret plan called Project OT (Organization Transformation), which involved reducing individual teams by 60% to replace humans with autonomous AI agents. However, instead of the expected productivity jump, the experiment caused a surge in critical crashes, and Mark Zuckerberg was forced to cancel a second wave of mass layoffs.
This was reported by Reuters, citing internal documents and sources in the company.
The project was created earlier this year at Zuckerberg's direct direction to transform Meta into an "AI-native" company — a structure where most daily tasks are performed by algorithms under the supervision of small teams of engineers.
As part of the downsizing and optimization, the company simulated scenarios of reducing the total staff by 25% or more: the first wave of layoffs took place in May, but the second, planned for November, was suddenly canceled. The savings were planned to be used to increase the salaries of AI engineers.
The introduction of AI agents led to serious disruptions, according to internal company reports. AI agents engaged in “large-scale destructive actions,” leading to a 40% increase in critical technical and security incidents.
The time human developers spent fixing bugs behind AI increased by 70%. According to CTO Andrew Bosworth, the number of code changes increased by 220%, but the number of new features that actually reached users increased by only 36%.
In addition, attempts to monitor employees' keyboards and mice to train AI caused a wave of indignation among the team.
In July, Mark Zuckerberg admitted during an internal meeting that “the trajectory of AI development over the past four months has not accelerated as expected.” Meta confirmed that it had modeled such scenarios, but assured that decisions on promotions and personnel evaluations would continue to be made by humans, not artificial intelligence.
