Nathan Labenz

Is AI Slowing Down? Nathan Labenz Says We're Asking the Wrong Question

Nathan Labenz, Erik Torenberg, Cal Newport

Recent advancements in AI, exemplified by GPT-5's reasoning leaps and autonomous agents, have disproven stagnation theories by achieving benchmarks in frontier mathematics, virology, and long-horizon task execution that previously required human expertise. While productivity gains are already displacing mid-tier roles in customer service and software development, widespread adoption faces barriers related to organizational implementation and geopolitical fragmentation driven by US-China export controls. Looking toward 2027–2030 for potential AGI, the primary challenge shifts from technical capability to managing safety risks like agent deception and securing the physical infrastructure needed to support rapid, global scaling.