Show HN: DOS – a referee between AI agents that doesn't believe their "done"
The article introduces DOS, a kernel that acts as a referee between AI agents, designed to verify whether tasks are actually complete rather than taking agents' "done" claims at face value. It aims to reduce hallucinations and improve reliability in multi-agent AI systems.
Background
- The "DOS Kernel" is an open-source tool (node on GitHub) that acts as a referee or supervisor for AI agent workflows. It doesn't trust an AI agent's claim that a task is "done" — instead, it runs automated checks to verify completion before moving on.
- This matters because large language models (LLMs) often hallucinate or prematurely declare tasks finished, breaking multi-step automation. DOS is a lightweight orchestration layer that enforces real verification.
- It targets developers building autonomous AI agent systems (e.g., coding agents, research bots) who need reliable handoffs between steps without manual oversight.
- The tool is named after the classic "disk operating system" — a play on being a foundational layer that manages execution, but here for AI agents rather than hardware.