Gradient is an R&D lab focused on decentralized AI infrastructure, building the Open Intelligence Stack (OIS). Its core products include Parallax (a distributed inference engine) and Echo (a distributed reinforcement learning framework), aiming to connect global idle compute resources (such as consumer GPUs, Apple Silicon) via P2P networks to train and run large language models in a distributed manner. Its technical mechanisms involve model sharding, heterogeneous hardware scheduling, and decentralized coordination, aiming to lower the barrier to AI compute and promote the openness and democratization of AI.