Deep Cogito raises $43M for self-improving AI

Deep Cogito's funding round highlights a growing niche in AI: models that can improve themselves rather than relying solely on human-generated training data. The company's approach combines two techniques. First, IDA dynamically increases computational resources during a prompt to boost output quality, then uses the performance gain to refine the model. Second, it applies process supervision, grading each step an LLM takes toward an answer instead of only the final response. The company says this reduces unnecessary or erroneous steps, lowering hardware usage.

The self-improving AI segment is attracting significant capital. Recursive Superintelligence Inc. raised $650 million in May from a consortium including Nvidia Corp. In April, Ineffable Intelligence Ltd. closed a $1.1 billion round at a $5.1 billion valuation. Deep Cogito's total funding of over $56 million is smaller, but its open-source release strategy and enterprise platform give it a distinct commercialization path, with Zscaler already among its customers.