NeoCognition raises $40M to build self-learning AI agents that specialize like human experts

NeoCognition Emerges with $40 Million to Reimagine AI Agents

NeoCognition, an AI research lab developing self-learning AI agents, has emerged from stealth with $40 million in seed funding 1

1

. The oversubscribed round was co-led by Cambium Capital and Walden Catalyst Ventures, with participation from Vista Equity Partners 2

2

. Angel investors and founding advisors include Intel CEO Lip-Bu Tan and Databricks co-founder Ion Stoica, alongside AI researchers Dawn Song, Ruslan Salakhutdinov, and Luke Zettlemoyer 2

2

. Additional institutional participants include A&E Investments, Salience Capital Partners, Nepenthe Capital, and Frontiers Capital 2

2

.

Addressing the Unreliability of AI Agents

The Palo Alto startup was founded by Yu Su, Xiang Deng, and Yu Gu, with Su serving as an Ohio State University associate professor who has led one of the country's most established LLM-based agent research labs 2

2

. Su initially resisted pressure from venture capital to commercialize his work until he concluded that advances in foundation models had reached a point where genuinely personalized agents were feasible 2

2

. Current AI agents, whether from Claude Code, OpenClaw, or Perplexity's computer tools, successfully complete tasks as intended only about 50% of the time 1

1

. This reliability gap means agents cannot be trusted as independent workers, with every task becoming what Su describes as "a leap of faith" 1

1

.

Building Agents That Learn Like Humans Through World Models

NeoCognition's approach centers on developing agents that mirror human learning processes through domain specialization. Su argues that while human intelligence is broad, its real power lies in our ability to specialize—when we enter a new environment or profession, we rapidly master its unique rules, relationships, and consequences 1

1

. "For humans, our continued learning process is essentially the process of building a world model for any profession, any environment," Su explained. "We believe for agents to become experts, they need to learn autonomously to build a model of any given micro world" 1

1

. This mechanism for rapid specialization through experience allows agents to capture the rules, relationships, and constraints of their operating environment through use rather than through pre-training on general data 2

2

.

Enterprise Strategy Targets SaaS Companies and AI Workers

NeoCognition intends to sell its adaptable agent systems to enterprise SaaS companies, which can use them to build AI workers or enhance existing product offerings 1

1

. The pitch to software vendors is that NeoCognition's agent system can be embedded to create AI workers that improve over time within that vendor's specific operational context 2

2

. Su highlighted that investment from Vista Equity Partners is especially valuable because, as one of the largest private equity firms in the software space, Vista can provide NeoCognition with direct access to a vast portfolio of companies looking to modernize their products with AI 1

1

. The company also indicated that a deeper understanding of work environments could improve speed, cost efficiency, reliability, and safety in higher-stakes use cases 3

3

.

Academic Foundation and the Reliability Layer Focus

NeoCognition ties its commercial story to Su's academic record in AI agents. Su's Ohio State lab produced work such as Mind2Web, MMMU, and SeeAct, which the company describes as part of the foundation of the modern agent field 3

3

. Walden Catalyst's Lip-Bu Tan noted that Su's team has worked across major parts of agent development, including perception, memory, planning, evaluation, and safety 3

3

. The team currently has roughly 15 employees, most holding PhDs 2

2

. This $40 million seed funding reflects strong focus on the reliability layer of AI rather than only on larger foundation models, positioning NeoCognition to enter the race with fresh capital, academic credibility, and an enterprise pitch built around specialized agents that keep learning after deployment 3

3

.