Investors are aggressively courting AI researchers to create startups that can make AI more reliable and efficient.
Yu Suan Ohio State professor who runs an AI agent lab, said he initially resisted pressure from venture capital firms to commercialize his work. He finally took the plunge last year and turned his work into a startup when he saw that advances in fundamental models could make agents truly personalized.
NeoCognition, a startup that Su describes as a research lab developing self-learning AI agents, just came out of stealth with $40 million in seed funding. The round was co-led by Cambium Capital and Walden Catalyst Ventures, with participation from Vista Equity Partners and angel investors including Intel CEO Lip-Bu Tan and Databricks co-founder Ion Stoica.
“Today’s agents are generalists,” Su (pictured right) told TechCrunch. “Every time you ask them to do a task, you’re taking a leap of faith. »
According to Su, the problem lies in a lack of consistency. Current agents, whether from Claude Code, OpenClaw or Perplexity IT tools, successfully complete their intended tasks only about 50 percent of the time, he said.
Since agents are still very unreliable, they are not ready to become trusted independent workers, Su told TechCrunch. NeoCognition intends to change this by developing a system of agents capable of self-learning to become an expert in any field, the same way humans learn.
Su argues that although human intelligence is vast, its true power lies in our ability to specialize. When we enter a new environment or profession, we can quickly master its unique rules, relationships, and consequences.
Techcrunch event
San Francisco, California | October 13-15, 2026
NeoCognition creates agents to reflect this exact approach.
“For humans, our continuous learning process is essentially about building a global model for any profession and any environment,” Su said. “We believe that for agents to become experts, they must learn autonomously to build a model of a given micro-world.”
Su sees this ability to rapidly specialize as the essential missing link for AI to operate reliably and autonomously.
While it is possible to train agents for autonomous tasks, they should be tailor-made for a specific vertical. NeoCognition is different because they are generalist construct agents capable of self-learning and specializing in any domain.
NeoCognition intends to sell its agent systems primarily to enterprises, including established SaaS companies, who can use them to create agents or to enhance existing product offerings.
Su emphasized that an investment from Vista Equity Partners is particularly valuable for this reason. As one of the largest private equity firms in software, Vista can provide NeoCognition with direct access to a broad portfolio of companies looking to modernize their products with AI.
NeoCognition currently has around fifteen employees, the majority of whom hold doctorates.
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
Marina Temkin is a venture capital and startups reporter at TechCrunch. Before joining TechCrunch, she wrote about venture capital for PitchBook and Venture Capital Journal. Earlier in her career, Marina was a financial analyst and earned her CFA designation.
You can contact or check Marina’s outreach by sending an email marina.temkin@techcrunch.com or via encrypted message at +1 347-683-3909 on Signal.
