Meta has announced its most significant overhaul of artificial intelligence operations to date, with internal memos revealing its ambition to fast-track the development of “superintelligence.” In a leaked communication reported by Business Insider, Alexandr Wang, the 28-year-old head of Meta Superintelligence Labs (MSL), told employees that “superintelligence is coming” and urged them to “take it seriously,” emphasizing the need for speed in achieving this milestone.
The reorganisation will focus on four key areas: research, training, products, and infrastructure. Several senior leaders will now report directly to Wang, including Nat Friedman, investor and former GitHub CEO, who will oversee the products division.
A new project codenamed “Omni” also surfaced in Wang’s memo. While details remain limited, the TBD Lab — a specialized unit tasked with training large AI models — will spearhead efforts to explore “new directions such as an omni model.”
On the research front, Shengjia Zhao, co-creator of ChatGPT, has been named MSL’s new chief scientist. Zhao will lead research independently and is the only senior figure not reporting directly to Wang. Meanwhile, FAIR (Fundamental AI Research), Meta’s long-established research arm, will assume a more prominent role. Under Rob Fergus’s leadership, with Yann LeCun continuing as chief scientist, FAIR’s research will now directly support TBD Lab’s training initiatives.
Infrastructure has been consolidated under Aparna Ramani, a longtime Meta executive. Her team will be responsible for managing advanced data centers and GPU clusters — resources Wang described as crucial for scaling AI systems. He underscored that future advancements would rely on “advanced infrastructure, optimised GPU clusters, and developer tools.”
As part of the shake-up, Meta is also dissolving its AGI Foundations team, established just a few months ago, redistributing members across products, infrastructure, and FAIR. This is the second major AI group Meta has disbanded in 2024, following the closure of its GenAI division after lukewarm reception to the Llama 4 model.
Acknowledging the turbulence caused by repeated restructurings, Wang admitted such changes can be “disruptive” but maintained they are critical for progress. “Taking the time to get this structure right now will allow us to reach superintelligence with more velocity over the long term,” he wrote.
