Scale AI Founder Alexandr Wang: Establish Internal Guidelines to Combat Noise
Alexandr Wang, founder of Scale AI and head of Meta's Superintelligence Lab, offered advice to his 18-year-old self during a discussion at YC Startup School 2026: establish an internal compass for how to navigate the future and maintain faith to combat noise.
He discussed with Garry Tan the rebuilding of cutting-edge AI labs from scratch, how talent density compounds, and how to identify worthwhile bets on entire exponential curves in one's twenties, while covering topics such as personal superintelligence forms, the necessity for models to be inexpensive, and the vision being more important than intelligence.
Event-driven AI talent and computing power are concentrating in high-density cutting-edge labs, with funding flowing to teams that can provide independent research and high computing power. Meta's Superintelligence Lab benefits from talent aggregation, while low-density or slow-executing labs are under pressure.
Source: Public Information
ABAB AI Insight
Alexandr Wang founded Scale AI at 19 (YC S16), focusing on AI data labeling and infrastructure, turning it into a modern AI data backbone, and becoming a young self-made billionaire at the peak of its valuation; in 2025, Meta invested about $14.3 billion for nearly half of Scale and recruited him to lead the Superintelligence Lab, where he subsequently built core teams like TBD Lab from scratch, launched the Muse Spark series of models, and completed organizational restructuring.
On the capital path, Meta exchanged large-scale equity investments and high computing resources for control over data infrastructure and core talent, with the strategic motive of quickly addressing the lagging situation of the Llama series and rebuilding the closed loop from training to product in the superintelligence race, concentrating resources on small, high-density research groups rather than traditional large organizations.
This is similar to the early talent density-first strategies of OpenAI and Anthropic, as well as Google's acquisition path for DeepMind; the current phase of the AI industry is shifting from model capability expansion to organizational control and execution density as the key to victory.
Essentially, this is a restructuring of the industrial chain driven by capital concentration and talent density compounding: high computing power and high freedom resources are gathering at a few nodes that can continuously bet on the longest exponential curves, as execution efficiency and feedback loop speed become the real bottlenecks in the superintelligence race.
ABAB News · Cognitive Laws
- Belief density determines the ownership of exponential curves.
- Noise is a byproduct of consensus; the compass is a tool for survivors.
- Talent density compounding > capital scale itself.