AI Labs are Simply Cooked | Jerry Tworek

The transition to third-generation AI labs centers on an automation-first approach to research, where AI agents streamline experimentation to accelerate insight generation. Jerry Tworek, former VP of Research at OpenAI and current CEO of Core Automation, argues that the current research bottleneck lies in manual processes, which agents can compress from months into days. He predicts that within two years, AI research will be fully automated, with models capable of independent discovery. Core Automation aims to move beyond the limitations of current transformer architectures by treating the company itself as a product, centralizing information processing to optimize decision-making. The development of reasoning models like O1 and O3 demonstrates that persistent, iterative reinforcement learning—supported by strategic compute allocation—can overcome long-standing technical barriers, fundamentally shifting the trajectory of AI capabilities and the nature of labor in the knowledge economy.