Surge AI achieved $1B in revenue with fewer than 100 employees by prioritizing high-quality data over scale, treating data curation as a machine learning problem akin to search ranking. The company’s success underscores a broader shift: lean, mission-driven teams can outperform larger organizations by focusing on real-world effectiveness rather than benchmark metrics. Chen critiques the overreliance on flawed AI benchmarks, advocating instead for human-led evaluations across domains like coding and physics to measure true progress. He believes AGI is still over a decade away and warns that current incentives misalign model development with societal benefit. Reinforcement learning environments—simulating complex, real-world tasks—are the next frontier, enabling AI to learn through interaction like humans. Surge emphasizes 'dream objective functions'—true goals beyond proxy metrics—and asserts that AI models will increasingly reflect the values of their creators, leading to divergent behaviors across labs. Founding Surge stemmed from Chen’s belief that shaping AI requires deep conviction, not blitzscaling.