True Intrinsics
True Intrinsics is an early-stage AI company developing continual learning technologies that enable AI agents to acquire new skills from just a few demonstrations and improve through everyday use. The company is application-first, driven by fundamental research into how models can learn from experience rather than stopping after pre-training. Its product lineup includes Kumo and Mochi, a native continual-learning companion model for AI agents.
- Artificial Intelligence
- AI Agents
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Traditional AI models stop learning after their initial pre-training phase, limiting their ability to adapt to new tasks and environments. This static approach requires extensive retraining or fine-tuning for each new use case, making AI systems inflexible and inefficient for real-world applications where conditions and requirements change over time.
Solution
True Intrinsics develops technologies that enable AI agents to continually learn, acquire new skills from minimal experience, and improve through use. The company is application-first, driven by fundamental research in continual learning, and aims to make AI systems learn from experience and get better at their work much like human professionals do. Its product lineup includes Kumo and Mochi, with Mochi described as a native continual-learning companion model for AI agents. The technology focuses on enabling agents to acquire new skills from just a few demonstrations, retain what they learn, and continuously improve toward human professional-level performance.
Target Audience
Primary customers are AI product teams and enterprises seeking AI agents that can adapt and improve through real-world use, particularly those in dynamic environments where pre-trained models are insufficient.
Features
- Continual learning architecture that allows models to acquire new skills from just a few demonstrations
- Experience accumulation and transfer across domains, reducing the cost of learning across new tasks and environments
- Semantic abstraction from interaction histories, enabling stable, reusable concepts
- Research-driven approach addressing how agents infer unknown goals and rules from limited interaction
- Focus on aligning agent semantics with human semantics for effective communication and feedback
- Investigation of scaling laws for experience-driven foundation models to understand capability growth with accumulated experience