Meta is putting self-improving AI at the center of its next major artificial intelligence push. On a second-quarter earnings call on Wednesday, CEO Mark Zuckerberg described a plan for Meta Superintelligence Labs that reaches beyond models trained mainly from human examples.
The goal, as he presented it, is to build systems that can learn from themselves. That idea now sits alongside Meta’s broader product strategy, its smartglasses work, and a large investment program aimed at frontier AI models.
Meta’s new AI lab is aiming at self-improvement
Zuckerberg told investors that Meta Superintelligence Labs is focused on AI models capable of self-improvement. In plain terms, that means systems that can improve by learning from their own outputs or processes, rather than depending as heavily on human input.
He framed that shift as necessary if Meta wants to build AI that goes beyond human-level capability. As he put it: “At some level, [it’s] not just going to be learning from people, because you want to build something that is fundamentally smarter than people,” Zuckerberg said. “So … you’re going to develop a way for it to improve itself. That is a very fundamental thing that’s going to have broad implications for how we build products and how we run the company.”
That comment matters because it connects research direction to business operations. Zuckerberg was not only describing a technical milestone. He was saying that self-improving AI could influence Meta’s products and the way the company itself works.
The company is also trying to position the lab against major AI rivals. The source article says Zuckerberg’s remarks were among several comments in a 12-hour span that pointed to how Meta plans to compete with OpenAI, Google DeepMind, and Anthropic.
The message is personal superintelligence
Earlier in the morning, Zuckerberg posted a letter online and shared a video to Instagram Reels. In those materials, he said Meta’s superintelligence work would focus on “personal superintelligence.”
That phrase is doing important work in Meta’s pitch. Zuckerberg described the effort as centered on tools that help people empower themselves and improve the world. He also drew a contrast with another vision for the technology.
“This is distinct from others in the industry who believe superintelligence should be directed centrally towards automating all valuable work, and then humanity will live on a dole of its output,” he said.
Within the facts available here, Meta’s argument appears to be that superintelligence should be built into personal tools rather than treated only as a centralized system for large-scale automation. The company has not presented that merely as a research philosophy. It is tying the idea to devices, products, and the future shape of its platform.
Smartglasses are part of the AI plan
One of the clearest product links is smartglasses. Zuckerberg said he believes they “are basically going to be the ideal form factor for AI.”
Meta already has a consumer foothold in that category. Since their 2023 debut, the company has sold 2 million pairs of smart sunglasses made with Ray-Ban. The source also notes that Meta showed an augmented-reality glasses prototype last September.
That prototype is called Orion. It is not expected to ship to consumers, but Meta has said it intends to build new devices over the next few years based on Orion research.
The glasses strategy helps explain how Meta might make “personal superintelligence” feel concrete. A model that lives only inside a distant interface is one kind of product. AI attached to a wearable device is another. Based on Zuckerberg’s remarks, Meta sees smartglasses as a natural place for AI to become more present in everyday use.
Meta is separating current AI from future superintelligence
The source article notes that Meta appears to be drawing a line between two kinds of AI work. One is the AI that helps monetize core products such as Instagram and WhatsApp. The other is superintelligent AI that could eventually help power humanity’s future.
That distinction is useful because it keeps two timelines in view at once. Meta already has major products that can benefit from AI today. At the same time, Zuckerberg is presenting superintelligence as a longer-range effort that could change both products and company operations.
The earnings context also matters. Zuckerberg’s remarks came after a better-than-expected earnings report. The company has lagged behind competitors in the AI race, according to the source article, and is now spending billions of dollars to build a compact but powerful superintelligence lab focused on frontier AI models.
Recruiting and spending show the scale of the push
Meta publicly announced the lab last month and has since been recruiting aggressively. The company has brought in Silicon Valley figures including Alexandr Wang, Nat Friedman and Daniel Gross.
It has also offered AI researchers compensation packages as high as nine figures to attract them to the lab. Wang joined through an “acqu-hire” deal with Scale AI, an AI data-labeling startup he cofounded. Former OpenAI researcher Shengjia Zhao has been named head of Meta’s new research lab.
The spending is showing up in Meta’s financial outlook. The company raised its capital expenditure forecast for the year to $69 billion. Meta chief financial officer Susan Li has said employee compensation is one of the biggest drivers, along with AI infrastructure.
Even with that level of investment, Meta is forecasting a stronger-than-expected third quarter. The company anticipates revenue between $47.5 billion and $50.5 billion.
Taken together, the picture is clear: Meta is treating superintelligence as a strategic priority, not a side project. The company is hiring prominent AI talent, increasing spending, tying the effort to smartglasses, and arguing that self-improving models could reshape both its products and its internal operations.