AI researcher and former Cohere executive Sara Hooker has secured $50 million in seed funding for Adaption Labs, a San Francisco-based venture she co-founded with Sudip Roy. Emergence Capital Partners led the round with backing from Mozilla Ventures, Fifty Years, Threshold Ventures, Alpha Intelligence Capital, E14 Fund, and Neo. Hooker, formerly Vice President of Research at Cohere and a veteran of Google DeepMind, aims to build systems that require less computing power and adapt to tasks without expensive retraining or elaborate prompts. She said, “This is probably the most important problem that I’ve worked on.” The company challenges the prevailing assumption that bigger models always perform better, arguing that scaling alone faces limits as costs rise and performance gains slow.
Adaption Labs is developing architectures focused on adaptive data, adaptive intelligence, and adaptive interfaces, including “gradient-free learning,” which changes system behavior at inference time rather than retraining model weights. Hooker explained, “How do you update a model without touching the weights?” and added that architectural innovation can use computing resources more efficiently. Techniques such as on-the-fly adapter merging and dynamic decoding allow responses to adjust in real time, reducing reliance on fine-tuning and what she called “prompt acrobatics.” She stated the goal is “to eliminate prompt engineering.” The startup will use the funding to hire researchers, engineers, and designers as it develops alternatives to traditional chat-based AI interfaces and explores continuous-learning systems.
