NeoCognition

NeoCognition develops self-learning AI agents intended to become specialists in the environments where they work. Its approach centers on learning a world model of work: the structure, workflows, and constraints of a particular environment. The goal is to let agents acquire expertise through experience, making specialized capabilities available to more people and organizations.

Co-founded by Yu Su, Xiang Deng, and Yu Gu, the company is led by Su as CEO. Their academic research predates NeoCognition: all three co-authored Mind2Web, a dataset for developing and evaluating instruction-following web agents, with more than 2,000 tasks across 137 real websites and 31 domains. Su also co-authored SeeAct, which explored multimodal web agents and how to turn textual plans into website actions using visual information and HTML structure.

NeoCognition emerged from stealth in 2026 with $40 million in committed seed capital, in a round co-led by Cambium Capital and Walden Catalyst Ventures. It plans to sell its agent systems primarily to enterprises, including established SaaS companies, to help them build agent workers or add agent capabilities to existing software.