Meta is quietly reshaping its approach to artificial intelligence with the creation of a superintelligence unit that includes a secretive group of elite hires known as TBD Lab, according to an internal memo. The unit, led by Alexandr Wang and staffed with high-profile recruits from buzzy AI startups, is charged with building Meta’s most advanced AI models yet.
While it represents only a fraction of Meta’s 70,000-plus workforce, the group reflects CEO Mark Zuckerberg’s latest management thesis: in the race for AI breakthroughs, small, talent-dense teams may outperform sprawling departments.
“I’ve just gotten a little bit more convinced around the ability for small, talent-dense teams to be the optimal configuration for driving frontier research,” Zuckerberg said during Meta’s recent earnings call.
Why Meta Is Betting on Small Teams
The philosophy mirrors Silicon Valley’s broader shift toward lean, startup-style teams, even within tech giants. Breakthroughs like the 2017 “Attention Is All You Need” paper, which underpins today’s large language models, came from just eight researchers — proof, many argue, that AI innovation thrives on small groups.
Meta is reportedly offering nine-figure compensation packages to lure top researchers, underscoring how scarce elite AI talent has become. Industry veterans argue that just a handful of “cream-of-the-crop” minds can yield disproportionately large advances.
Meta isn’t alone in this thinking.
- Hightouch, a $1.2 billion AI startup, builds products with micro-teams — sometimes just four engineers on major launches.
- Investor and former GitHub CEO Nat Friedman has championed the model, arguing that most tech companies are two to ten times overstaffed.
Risks and Tensions Inside Meta
Still, adopting a startup playbook within a 70,000-employee behemoth is far from simple. Experts warn that “special units” can create internal friction, with legacy staff resenting the high pay and prestige of new hires. Business Insider reported that the formation of Meta’s superintelligence group has already sparked tensions and even resignation threats among some researchers.
There’s also the risk of overlap and redundancy. Google has compared its own small-team proliferation to “slime mold,” while Meta itself has reorganized its AI division multiple times, dissolving two teams in the past four months alone.
“Small innovation teams inside large corporations often deliver efficiency gains or useful products — but rarely the kind of transformation that fundamentally reshapes the parent company,” said Elliott Parker, CEO of Alloy Partners, which advises corporations on startup creation.
Zuckerberg’s Bet on Superintelligence
Despite the challenges, Zuckerberg remains bullish. He argues that to push toward AI superintelligence, the optimal setup is “the smallest group that can hold the whole thing in their head.”
Whether TBD Lab can cut through Meta’s bureaucracy and deliver true frontier breakthroughs remains to be seen. But one thing is clear: as AI competition accelerates, the race isn’t just about compute power and scale — it’s increasingly about who can assemble the leanest, sharpest team to lead the way.
Other Articles
