Mark Zuckerberg has introduced new language around Meta’s AI development plans that indicates a potential shift in the company’s direction. Over the last several months, Meta has invested heavily in both infrastructure and talent. At the same time, internal communications and earnings remarks have started to reflect a more cautious approach to openness, particularly as the company begins building what it calls “personal superintelligence.”
A recently published note from Zuckerberg laid out this vision. It described a future where highly capable AI systems assist individuals in their personal goals. He did not define superintelligence in technical terms or explain how such systems would be built. Instead, the memo outlined a broad idea of personalized tools that could support users across creative, social, and productivity tasks.
This approach contrasts with other companies that are developing general-purpose models aimed at automating workflows. Zuckerberg’s message suggested that Meta wants to keep the focus on individual empowerment. This vision also connects to the company’s longer-term goal of moving away from reliance on mobile devices. Meta has spent years developing smart glasses and has signaled interest in making them central to future computing experiences.
Talent Shifts and Billion-Dollar Recruiting
Meta has increased its hiring of AI experts since early this year. In one move, it brought in the founder of Scale AI through a $14.8 billion investment. That deal placed Alexandr Wang as the company’s new Chief AI Officer. The company has also recruited individuals from OpenAI and Apple, including engineers who contributed to major large language models.
Recent reporting indicated that Meta extended multiyear offers worth hundreds of millions, and in one case more than $1 billion, to staff from Thinking Machines Lab. The startup was founded by a former OpenAI executive. Meta did not confirm the financial details but acknowledged the interest in expanding the team.
Alongside those moves, Meta has committed over $72 billion to AI infrastructure. That includes compute power, model training capacity, and scaling systems. These steps suggest the company is preparing to build more advanced AI models, even as it evaluates how much of that work to make public.
Open Source Remains Unclear
For years, Meta positioned open-source AI as a safer and more inclusive approach. Company leaders argued that transparency could prevent misuse and help governments understand how models work. More recently, Zuckerberg indicated that the company may not share some of its largest models in the future.
His recent statements during Meta’s second-quarter earnings call included references to safety and practicality. According to him, some of the models now being developed are so complex that releasing them would have little benefit for outside developers. In some cases, he added, sharing might give an advantage to competing firms.
These comments followed a memo that suggested Meta would remain a leader in open-source work but would be more selective about what gets released. While this is not a reversal of past policy, it shows a growing awareness inside Meta that some advanced AI models may carry risks that make full transparency harder to justify.
Usage Gains Tied to AI Integration
Meta’s recent product performance also reflects increased use of AI to drive engagement. Time spent on Facebook rose 5 percent in the second quarter. Instagram saw a 6 percent gain. Both trends were attributed to updates in recommendation systems, which now use large language models to present more relevant content.
The company also noted that video viewership grew by 20 percent over the past year. Instagram played a large role in that growth, as Meta has focused on promoting original material and improving content ranking methods. Threads, its messaging-based app, has seen an increase in daily use following the integration of new AI tools.
All in all, Meta reported 3.4 billion family daily active people across its platforms in June. That figure included Facebook, Instagram, Messenger, and WhatsApp. It marked a 6 percent increase from the previous year and supported a 22 percent rise in revenue across those apps, reaching $47.1 billion in the quarter.
Broader Shifts in AI Positioning
Zuckerberg’s internal memo came just before Meta’s Q2 earnings report. The timing appeared aligned with the company’s efforts to frame its AI investments to investors. With delays affecting the launch of its larger Llama 4 model, internal reports suggested that Meta’s leadership had been reevaluating its approach. Some concerns were raised about the tradeoff between openness and competitive advantage.
There has also been tension around the slow progress of Meta’s generative AI roadmap. Executives inside the company have reportedly questioned whether its development pipeline can keep pace with external labs. These concerns may have shaped the more cautious stance reflected in the memo and earnings discussion.
At the same time, Meta seems to be preparing for a future in which its computing platforms are less dependent on Apple and Google. Smart glasses, which the company continues to develop, were described as key devices in future AI use. Zuckerberg pointed to this shift as an opportunity to move users toward platforms owned and controlled directly by Meta.
The company’s strategy remains focused on scaling up its AI capabilities, refining its product experience, and adjusting its messaging around transparency. While the long-term details remain unclear, the recent changes suggest that Meta is actively shaping its next phase of AI development around tighter control, personal devices, and internal platforms.
Notes: This post was edited/created using GenAI tools.
Read next: Most Adults Use AI Without Realizing, But True Power Remains Untapped