The San Francisco-based firm’s co-founders are Sara Hooker, previously head of research at non-profit Cohere Labs, and Sudip Roy, Cohere’s former senior director of inference. The pair, who were based out of Cohere’s California office, left the Toronto-headquartered AI firm last month. (The Logic)

Talking point: Adaption aims to extend work that the two computer engineers launching it have been doing. At Cohere Labs, Hooker led work to build open-source large language models (LLMs) that wrote in dozens of tongues, and to come up with ways to improve their performance at a wide variety of tasks. Roy, meanwhile, worked on improving inference, the industry term for the phase when already-trained LLMs generate responses. Adaption will design its algorithms to make the process of running AI models and tools more efficient, and adjust for the hardware powering them. It’s developing technology that “enables AI to evolve real time,” Hooker said in a post.