NeoCognition is an AI agent lab focused on building specialized intelligence for the many professional, organizational, and software environments that shape modern work. The company’s approach centers on agents that learn the structure and rules of a particular “micro-world,” including what matters, which actions are possible, and what consequences may follow. NeoCognition argues that deeply trained, reliable agents can bring expert-level capability to more people and organizations, complement human expertise, and support new forms of creation.

The lab’s work includes research on AI agents and the NeoCognition ApprenticeBench, described as the first continual learning benchmark set in a real job. Its founding team—Yu Su, Xiang Deng, and Yu Gu—brings more than 30 years of collective research experience in agents. NeoCognition is supported by Cambium Capital, Walden Catalyst Ventures, Vista Equity Partners, and individual backers. Through its research and benchmark work, the company aims to advance an abundance of specialized agents rather than a single general-purpose agent.