Core Automation

Core Automation is an artificial intelligence research lab focused on developing systems that automate and optimize the research process itself. The company was founded in late 2024 or early 2025 by Jerry Tworek, a former vice president of research at OpenAI. During his tenure at OpenAI, which he joined in 2019, Tworek was a key figure in the development of reasoning models like o1 and o3, and was deeply involved in the post-training of GPT-4. He departed OpenAI to pursue research he felt was difficult to conduct within the existing structure. The firm publicly emerged in April 2026, positioning itself as "building the world's most automated AI lab." Core Automation's central thesis is that the next significant advancements in AI will not come from simply scaling current methods like larger models and massive datasets. Instead, the lab is pursuing new learning algorithms and architectures that can supersede large-scale pretraining and reinforcement learning. One of its primary goals is to create models capable of continual learning from real-world experience, aiming for a 100-fold reduction in the required training data. The business model is centered on a long-term research and development strategy, attracting substantial venture capital to fund its ambitious goals, with reports indicating a fundraising target between $500 million and $1 billion. The company's approach is to use its own advanced AI agents to augment the work of its small, elite team, creating a feedback loop where automating their own work reveals new research directions and capabilities. This strategy has attracted top-tier talent from other major AI labs, including senior researchers from Google DeepMind and Anthropic. Core Automation’s ultimate vision is to empower small teams and individuals with AI systems that provide the capabilities of large organizations. This could enable a single person or a small group to tackle complex logistical, creative, or industrial challenges that would otherwise require significant capital and a large workforce. Keywords: automated AI lab, AI research, machine learning algorithms, Jerry Tworek, continual learning, transformer alternatives, frontier models, reinforcement learning, AI optimization, reasoning models, artificial intelligence systems, advanced architectures, AI agents, deepmind, anthropic, openai, automated workflows, capital-light innovation, systems optimization