Born from one AMD GPU and a simple idea: make AI agents work together locally. We are building InnerOS, a local-first agent execution fabric that connects AMD ROCm compute, skills, and specialized agents with coordination, recovery, human approval, and verifiable evidence.