Adaption Labs

About Adaption Labs

Adaption Labs is an artificial intelligence research startup challenging the dominant industry paradigm of scaling large language models (LLMs). Founded in October 2025 by former Cohere and Google veterans Sara Hooker and Sudip Roy, the company focuses on developing AI systems that are efficient and can continuously adapt and learn from real-world experience. Hooker, previously the VP of AI Research at Cohere, and Roy, the former senior director of inference, aim to create AI that evolves in real-time without the immense computational costs and energy consumption associated with traditional scaling methods.

The company's core thesis is that simply increasing the size and computing power of LLMs is yielding diminishing returns and is an inefficient path to more capable intelligence. Instead, Adaption Labs is building AI that learns from interaction, a concept central to true learning but not yet effectively implemented in production AI systems which remain largely static after training. Their approach is designed to be more cost-effective and powerful, potentially democratizing AI development and reducing the industry's reliance on a few major players with massive data centers.

While headquartered in San Francisco, Adaption Labs is establishing a global-first, distributed team with a presence in Canada and the UK. The startup secured a significant seed funding round reported to be between $20 million and $40 million in late 2025, and another $50 million round led by Emergence Capital Partners was announced in February 2026. The company's focus is on creating flexible, personalized, and accessible AI, with a short-term goal of enabling models to adapt to enterprise-specific workflows securely and without needing complete retraining.