Meta CEO Mark Zuckerberg has outlined a new strategic direction for the company’s AI ambitions, centered on a vision of “personal superintelligence” — AI designed to help individuals achieve personal goals through deeply contextual, real-time support. In a public letter released this week, Zuckerberg said that AI should be integrated into everyday life through devices like augmented reality glasses, which understand a user’s environment by seeing and hearing as they do.
Alongside this vision, Meta appears to be recalibrating its stance on open source AI. While the company has previously promoted openness as a core differentiator — particularly with its Llama model family — Zuckerberg’s latest comments suggest Meta may no longer default to open sourcing its most advanced models due to emerging safety and commercial considerations.
This shift comes as Meta ramps up investment in AI infrastructure and talent, notably acquiring a stake in Scale AI, consolidating its research efforts under Meta Superintelligence Labs, and devoting billions to data centers and model training. Recent reports indicate Meta has paused work on its latest open model, Behemoth, and is focusing instead on closed alternatives designed to power its consumer hardware ecosystem.
Though Meta maintains it will continue releasing leading open models, Zuckerberg’s letter makes clear that the company’s AI monetization strategy will center on proprietary systems embedded in Meta’s own devices — signaling a strategic pivot in how the company intends to lead in the next phase of AI computing.
