On June 30, 2025, Mark Zuckerberg sent employees a memo formally launching Meta Superintelligence Labs (MSL). Bloomberg obtained the memo first, CNBC published it in full the same day, and TechCrunch followed that morning — turning a month of recruiting rumors into an official org chart.
For outside observers, the document’s significance is consolidation. The mid-June $14.3 billion investment in Scale AI that brought in Alexandr Wang, the acquisition talks reported in late June, and the steady drip of star-researcher signings all now roll up under one roof. The effort has a name and, as of Monday, named leaders.
The official launch
“We’re going to call our overall organization Meta Superintelligence Labs (MSL),” the memo reads. All of Meta’s fundamental AI research (FAIR) teams — the decade-old research organization that had long operated as its own brand — its product groups, and a new lab focused on next-generation models now sit under the MSL umbrella. Zuckerberg framed the mission as building “personal superintelligence for everyone,” and wrote that “developing superintelligence is coming into sight.” In name and in framing, this was a formal redefinition: superintelligence moved from a research talking point to the organizing goal of the company’s AI work, not a rebrand of the existing division. Bloomberg viewed the internal memo on Monday and broke the news; CNBC carried the full text, so the org structure, the leadership pairings, and the recruiting list were all public within hours.
Structure and people
Alexandr Wang leads MSL as Meta’s chief AI officer, arriving via the $14.3 billion Scale AI investment announced earlier in June; Zuckerberg called him “the most impressive founder of his generation.” Nat Friedman, the former GitHub CEO who served on Meta’s advisory group over the previous year, heads AI products and applied research. His longtime business partner Daniel Gross joins as well — Gross co-founded Safe Superintelligence, the startup led by Ilya Sutskever that Meta tried and failed to buy. The memo also names eleven researchers joining the effort, most arriving from OpenAI, Google DeepMind, and Anthropic: Trapit Bansal, Shuchao Bi, Huiwen Chang, Ji Lin, Joel Pobar, Jack Rae, Hongyu Ren, Johan Schalkwyk, Pei Sun, Jiahui Yu, and Shengjia Zhao, whom Zuckerberg credited as a co-creator of ChatGPT, GPT-4, the mini models, 4.1, and o3, and a former leader of synthetic-data work at OpenAI. The concentration of lineage in that list — from the original ChatGPT team to DeepMind’s research staff — is the part rivals will read most closely.
The frontier lab and the Llama track
MSL runs on two tracks. One continues the Llama 4.1 and 4.2 line — Meta AI already counts more than 1 billion monthly actives, so the product surface keeps scaling. The other is a deliberately small, “talent-dense” frontier-model lab aimed at reaching the frontier “in the next year or so.” The dual structure is an implicit admission that Meta needs both: the open-weight and product franchises must keep shipping, while the race to the frontier gets a dedicated team rather than relying on gradual improvement across the existing organization.
A snapshot of the talent war
The reorganization is the concrete result of an unusually aggressive recruiting campaign. OpenAI CEO Sam Altman said publicly that Meta offered signing bonuses as high as $100 million to poach researchers, and Meta CTO Andrew Bosworth described the market rates as unprecedented in his career. MSL’s launch and the eleven-name list are the first tangible outputs of that push — and, against reports of strained morale inside OpenAI, they mark the point where the June talent war moved from leaked offers to a permanent org structure.
What it means for developers
For developers and product teams, the signal is less about personnel drama than about roadmap structure. Meta split “applied and product” from “frontier research,” gave each a distinct leader, and pinned names, mission, and timeline in one memo. Over the next few quarters, Llama’s release cadence and whatever the new frontier lab produces will test the two tracks separately — and the sourcing of that eleven-person list hints at where the next round of model capabilities is likely to come from. The open question, unstated in the memo, is how the small “talent-dense” frontier lab’s output will reach developers: through the open-weight Llama channel, through Meta AI’s consumer surface, or through something not yet announced. June 30 did not answer that; it just made clear that both bets are now formally funded and staffed.
Sources
- CNBC: Mark Zuckerberg creating Meta Superintelligence Labs — read the memo
- TechCrunch: Meta restructures its AI unit under ‘Superintelligence Labs’
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.