Mark Zuckerberg said there are very few 'seats on the boat' of his AI lab, which doesn't have top-down deadlines (www.businessinsider.com)

Meta CEO Mark Zuckerberg said his new superintelligence lab is deliberately tiny and talent-dense — “seats on the boat are precious” — and described a group of roughly 50–100 researchers rather than “many, many hundreds.” He said the team is highly selective because poor hires or bloat would disproportionately harm progress. The lab operates with flattened management, fewer non-technical layers, no top-down deadlines, and generous resources: multimillion-dollar pay packages, substantial GPU capital, and a major strategic move—a roughly $15 billion investment to acquire nearly half of Scale AI and bring its CEO Alexandr Wang into leadership after Zuckerberg judged Llama‑4 “not on the right trajectory.” Meta has actively poached talent from OpenAI and DeepMind to staff the effort. For the AI/ML community this underscores a strategic shift toward small, autonomous research units that prioritize deep technical expertise and long-horizon experimentation over conventional execution metrics. The model has trade-offs: talent density and flexible timelines can accelerate frontier breakthroughs by protecting research from short-term business KPIs, but it also raises risk that one weak hire or coordination failure could impede progress. Flattened orgs reduce managerial overhead and keep researchers technical, but they must actively counteract knowledge decay and maintain rigorous peer review and infrastructure (e.g., GPU provisioning, dataset quality) to sustain high-impact outcomes.