Every company expects a breakthrough when it brings in a brilliant new talent. This could have been Mark Zuckerberg’s expectation when he chose Alexandr Wang as Meta’s AI chief.
But a New York Times report tell us that beneath the surface, tensions are already brewing at Meta. The team, called TBD Lab, was placed in a special workspace right beside Zuckerberg’s office in Menlo Park, a quiet glass-walled zone meant to shield it from the rest of Meta’s traditional layers. The arrival of Wang has already changed the atmosphere inside the company.
As venture capitalist Tomasz Tunguz explains to NYT, “Meta spent a huge amount of money to bring in Alexandr Wang, and that must have changed both the internal topology and the feeling of the company… You have an outsider who now has carte blanche. That probably means that a lot of previous ways of working are disregarded, which is hard.”
Only months after launch, disagreements between Wang and Meta’s long-serving executives have begun surfacing. According to people familiar with internal conversations, Wang has pushed back against Chief Product Officer Chris Cox and Chief Technology Officer Andrew Bosworth on how the new AI systems should be built.
Cox and Bosworth wanted TBD Lab’s work tied closely to Meta’s products, training the next major AI model using Instagram and Facebook data to sharpen recommendations and ads. Wang argued that Meta needed to focus first on closing the gap with OpenAI and Google’s most advanced models before turning to business features.
Difference in ideologies
Cox and Bosworth see future superintelligence primarily as a way to improve Meta’s social platforms. Wang, however, has said that mixing product-specific objectives into early model training could slow down the core scientific progress needed to build truly advanced A.I.
In one recent meeting, when Cox asked whether Instagram data could be used like YouTube data at Google, Wang argued that this kind of task-specific training would get in the way of long-term goals.
Budgeting crisis
The difference in vision is reshaping internal budgets too. People familiar with planning said Bosworth’s Reality Labs was recently asked to reduce its proposed spending for next year by $2 billion, with that amount reallocated to Wang’s group. Meta has denied this.
Access to computing power has also become a point of friction. Teams working on Instagram and Facebook ranking argue their systems generate real revenue and deserve priority. Yet TBD Lab needs massive computational resources to train the frontier models Zuckerberg believes will define Meta’s future.
For Zuckerberg, balancing these competing agendas is a delicate act. He has poured billions into accelerating Meta’s AI push, including a $14.3 billion investment into Wang’s start-up before hiring him as chief AI officer.
Meta’s team restructuring
After bringing Wang on board, Zuckerberg and his new lieutenant began a big recruitment drive, offering huge pay packages to top researchers across the industry.
By midsummer, the new specialists began arriving at Meta’s campus. Soon after, Meta reorganised its entire AI division into four groups, one for research, one for products, one for infrastructure, and TBD Lab for superintelligence. All of them now operate under the umbrella of Meta Superintelligence Labs, with Wang in charge.
However, several respected AI researchers left for competitors, and executives including Yann LeCun and John Hegeman exited. Meta also laid off 600 people across the broader AI organisation.
