When Mark Zuckerberg overhauled Meta’s artificial intelligence strategy earlier this year, he appointed 28-year-old entrepreneur Alexandr Wang to lead the company’s most ambitious AI effort yet. Wang, the well-connected founder of Scale AI, was brought in to accelerate Meta’s race toward frontier-model development and to recruit top researchers from competitors like OpenAI and Google.
The new group, called TBD Lab (“to be determined”), operates with unusual autonomy inside Meta’s wider AI organization. The lab was set up to give Wang a degree of separation from the traditional product-focused culture that has guided the company for years, according to people familiar with the structure. While its workspace has been described internally as distinct, the specifics of its physical location remain unconfirmed.
A Growing Divide Over Priorities
Within months of TBD Lab’s formation, philosophical differences emerged between Wang’s team and some of Meta’s longtime product and infrastructure leaders. At the center of the disagreement is how quickly Meta should apply its developing models to products like Instagram and Facebook.
Some senior product executives have pushed to use Meta’s vast social-media data to train its next frontier-model generation, arguing that AI improvements could boost feeds, recommendations, and ad performance. According to people briefed on internal conversations, Wang has been more cautious, stressing that Meta must first close the gap with leading AI labs before optimizing for specific business lines.
This tension reflects a broader cultural divide: TBD Lab researchers see themselves as building long-term, transformative AI systems, while other Meta teams are focused on near-term commercial benefits. The result has been an “us versus them” dynamic, current and former employees said.
Meta spokesman Dave Arnold disputed that characterization, saying leadership is aligned on building superintelligence while strengthening the core business.
Shifting Resources and Organizational Strain
Meta’s AI reorganization this summer placed TBD Lab at the center of a new umbrella group called Meta Superintelligence Labs. As part of the restructuring, Meta cut roughly 600 roles in older AI units while continuing to hire aggressively for TBD Lab — a signal of shifting internal priorities.
Reports of a multibillion-dollar budget shift toward Wang’s division have circulated inside the company, though Meta denies that specific figures, including the claim that $2 billion was redirected from Reality Labs, are accurate.
Tensions have also surfaced around access to computing power. Teams responsible for the social-media ranking algorithms have argued that the company’s newest compute clusters should be dedicated to improving feeds and ads — proven revenue engines — rather than solely toward training new frontier models.
At the same time, Meta continues to invest heavily in AI infrastructure, though public reporting does not support claims of a $600 billion data-center commitment.
Recruiting, Departures, and a New AI Hierarchy
After Wang joined Meta as chief AI officer, he and Zuckerberg embarked on an aggressive talent offensive, offering unusually large compensation packages to lure top researchers. The two personally courted candidates, with some recruitment efforts reportedly extending to doorstep visits and home-cooked meals.
By July, an elite roster of researchers had arrived at Meta’s Menlo Park campus with a mandate to rebuild the company’s generative-AI foundations. Meta consolidated its AI efforts into four groups: research, product, infrastructure, and TBD Lab — all operating within the new Meta Superintelligence Labs framework, with Wang overseeing the entire organization.
The reorganization also triggered notable departures, including senior AI researchers and executives who left for OpenAI, Google, and other rivals. In October, the company laid off hundreds of employees from older AI teams, though TBD Lab remained untouched.
A Push Toward “Superintelligence”
Wang and his researchers remain focused on developing what they describe internally as a pathway toward superintelligent AI systems — a vision Zuckerberg has publicly endorsed. The CEO has said that advanced AI will usher in “a new era of individual empowerment,” and Meta plans to integrate AI assistants and characters across Instagram, Facebook, and future hardware.
Meanwhile, some employees continue to debate how Meta’s frontier models should align with immediate business needs. At a recent investor call, CFO Susan Li emphasized that improving social-media recommendations with AI remains a major focus for next year.
Despite early friction, TBD Lab has stabilized. When employees hit their first share-vesting milestone in November, only two researchers departed, according to people familiar with the matter. The team now numbers around 100.
Wang is slated to host his annual AI holiday gathering in San Francisco this week, co-organized with investor Elad Gil. It remains unclear whether Meta’s top executives will attend — a small detail that reflects the larger cultural shift underway as the company reshapes itself around its most ambitious AI push yet.