Mark Zuckerberg Bets Billions On 'Startup Mode' As Meta's Superintelligence Unit Lures Alexandr Wang And Nat Friedman From Top AI Firms

Meta META is reorganizing its artificial intelligence division around compact, high-impact teams as CEO Mark Zuckerberg pushes the company deeper into what he calls "superintelligence labs," Business Insider reported.

The company has carved out a secretive group named TBD Lab, staffed with elite engineers poached from leading AI startups and headed by Alexandr Wang, the founder of Scale AI. Meta recently spent $14 billion to acquire a stake in Scale AI as part of its superintelligence push, according to media reports.

Zuckerberg said on Meta's latest earnings call, he has "gotten a little bit more convinced around the ability for small, talent-dense teams to be the optimal configuration for driving frontier research." He contrasted the new unit with the thousands of engineers who manage Facebook's news feed, pointing out a cultural move toward agility over scale.

Don't Miss:

• The same firms that backed Uber, Venmo and eBay are investing in this pre-IPO company disrupting a $1.8T market — and you can too at just $2.90/share.
• They Sold Their Last Real Estate Company for Nearly $1B — Now They're Building the Future of U.S. Industrial Growth

The Small Team Philosophy Gains Momentum

According to an internal memo sent by Wang and reviewed by Business Insider, Meta has reorganized its artificial intelligence division into four specialized groups designed to push its "superintelligence labs" initiative forward.

Wang acknowledged in his memo that "org changes can be disruptive," but argued that restructuring was critical for Meta to compete in the race toward superintelligence.

The shift mirrors a broader Silicon Valley movement toward lean operations. Kashish Gupta, co-founder of Hightouch, a San Francisco AI startup valued at $1.2 billion, told Business Insider his company relies on about 55 engineers despite raising more than $132 million.

A major AI agent launch at the company was developed by only four people, he said, emphasizing that "the people prioritizing work are the engineers themselves or the people they work with."

Trending: 7 Million Gamers Already Trust Gameflip With Their Digital Assets — Now You Can Own a Stake in the Platform

Investor Nat Friedman, the former GitHub CEO who now leads Meta's Products and Applied Research group, has echoed this sentiment in a manifesto on his website, writing, "Smaller teams are better. Faster decisions, fewer meetings, more fun."

A Fierce AI Talent War

Meta has intensified its recruitment strategy by offering nine-figure compensation packages to attract engineers from rivals such as OpenAI, Google DeepMind, and Anthropic, The Wall Street Journal reported. Such high-stakes hiring reflects a belief that a handful of elite researchers can deliver breakthroughs with outsized impact.

"You don't need too many — just a few smart, cream-of-the-crop people to have major breakthroughs and extremely disproportionate impact," AI engineer Yangshun Tay said in an interview with Business Insider.

Still, the strategy has created friction. Some longtime employees told Business Insider that the arrival of a lavishly funded unit left them feeling sidelined, with reports of resignation threats and discontent across Meta's broader AI staff.

Can Meta's Small Teams Overcome Big Company Challenges?

While compact groups have historically driven innovation, including the 2017 Google research paper "Attention Is All You Need," authored by just eight people, experts question whether the model will scale inside massive corporations like Meta.

Alloy Partners CEO Elliott Parker told Business Insider that small teams within conglomerates often deliver useful products and efficiency gains but rarely "transform their parent organization."

Meta has already reorganized its AI units several times in recent months, dissolving two divisions in four months to remove overlaps that critics likened to "slime mold," Business Insider says. Despite this turbulence, Zuckerberg remains optimistic. "For the leading research on superintelligence, you really want the smallest group that can hold the whole thing in their head," he said during Meta's earnings call.

Image: USA Today