Mark Zuckerberg is revolutionizing Meta’s approach to artificial intelligence development by embracing a startup-style model centered on small, elite teams. The company’s new superintelligence unit includes a secretive group called TBD Lab, composed of high-profile hires poached from buzzy AI startups, according to a recent internal memo.
The unit represents only a fraction of Meta’s 70,000+ employee workforce, with many members—including leader Alexandr Wang—recruited from cutting-edge AI companies. “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 explained during Meta’s latest earnings call.
This philosophy marks a significant departure from the massive engineering teams traditionally required to manage Meta’s core products like Facebook’s newsfeed. The small-teams gospel is spreading rapidly across Silicon Valley, from Big Tech giants to lean startups, as organizations seek to cut through bureaucracy and accelerate innovation.
Kashish Gupta, cofounder of Hightouch—a San Francisco AI marketing startup valued at $1.2 billion—exemplifies this approach. Despite raising over $132 million, Hightouch maintains just 55 engineers total, with a major AI agent launch staffed by only four engineers. The company expects engineers to be self-starters who build products without micromanagement.
Former GitHub CEO Nat Friedman, who joined Meta’s superintelligence effort in June, advocates this philosophy on his personal website: “Smaller teams are better. Faster decisions, fewer meetings, more fun.” Friedman argues many tech companies are two to ten times overstaffed.
The trend is supercharged by AI’s unique dynamics. The groundbreaking 2017 paper “Attention Is All You Need”—which introduced the architecture behind modern large language models—was authored by just eight researchers, primarily from Google. Meta has reportedly offered nine-figure compensation packages to poach top AI researchers, reflecting the scarcity of elite talent.
However, challenges exist. The arrival of Meta’s lavishly compensated superintelligence unit has sparked tensions and resignation threats from existing researchers. Small teams within large organizations may still face corporate bureaucracy, and potential overlap between teams could create inefficiencies—an issue Google has compared to “slime mold.” Meta has already reorganized its AI division multiple times, dissolving two units in the last four months alone.
Key Quotes
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.
Mark Zuckerberg explained this philosophy during Meta’s latest earnings call, signaling a strategic shift in how the company approaches AI development and contrasting it with the massive teams traditionally needed for products like Facebook’s newsfeed.
Smaller teams are better. Faster decisions, fewer meetings, more fun.
Nat Friedman, former GitHub CEO and current head of Meta’s AI product integration division, wrote this in a manifesto on his personal website, arguing that many tech companies are two to ten times overstaffed.
Based on my experience, AI advancement really relies on breakthroughs that come from just a few people. You don’t need too many — just a few smart, cream-of-the-crop people to have major breakthroughs and extremely disproportionate impact.
Yangshun Tay, an AI engineer, explained why the small-teams model is particularly suited to AI research, highlighting how breakthrough innovations typically emerge from compact groups of exceptional researchers rather than large teams.
For the leading research on superintelligence, you really want the smallest group that can hold the whole thing in their head, which drives, I think, some of the physics around the team size and how the dynamics around how that works.
Zuckerberg articulated this vision during Meta’s earnings call, expressing confidence that small teams represent the optimal approach for maintaining a competitive edge in the AI race and developing superintelligence.
Our Take
Meta’s pivot to small AI teams reveals a critical tension in modern tech: the conflict between scale and agility. While Zuckerberg’s approach may accelerate breakthrough research, it risks creating a two-tier system that alienates existing talent—a dangerous gamble when AI development requires both innovation and institutional knowledge.
The nine-figure compensation packages highlight an uncomfortable truth: AI has become a winner-takes-all talent market. This concentration of resources in elite teams may deliver short-term competitive advantages but could undermine long-term organizational cohesion. The reported tensions and resignation threats suggest Meta is already experiencing these growing pains.
What’s most intriguing is whether this model can truly scale. Startups succeed with small teams partly because they lack legacy systems and bureaucracy. Meta’s challenge is replicating that agility while managing products serving billions—a fundamentally different problem than building from scratch.
Why This Matters
This shift toward small, elite AI teams represents a fundamental transformation in how Big Tech approaches artificial intelligence development. As the AI race intensifies, companies are recognizing that breakthrough innovations often come from compact groups of exceptional researchers rather than massive engineering armies.
The trend has significant implications for the tech workforce. While it may accelerate AI advancement, it also signals a move toward more exclusive, highly compensated teams—potentially leaving traditional engineers feeling sidelined. The tensions at Meta illustrate how this two-tier system can create cultural friction within organizations.
For the broader AI industry, this validates the startup model’s effectiveness even at enterprise scale. If Meta succeeds, other tech giants will likely follow suit, reshaping hiring practices and organizational structures across Silicon Valley. However, the approach’s sustainability remains uncertain—small teams may deliver incremental innovations without achieving the transformative breakthroughs corporations seek. The outcome of Meta’s experiment will influence how the entire industry balances agility with scale in the critical race to develop superintelligence.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- Sam Altman Addresses OpenAI Executive Exodus at Italian Tech Week
- OpenAI’s $157B Valuation: Can It Win the Brutal AI Race?
- Tech Workers Are the Real Winners in the AI Talent War, With Pay Set to Soar by 2024
- Google Requests Update to Immigration Rules to Hire Top AI Talent by 2024
- Microsoft Pay Data Reveals Significant Salary Premiums for AI Workers
Source: https://www.businessinsider.com/mark-zuckerberg-startup-mode-meta-small-ai-teams-2025-8