Ask HN: How should human overrides of AI agent outputs be recorded? (github.com)

A new protocol called the Collaborative Human-Agent Protocol (CHAP) has been introduced to streamline the collaboration between AI agents and human users in various work processes, such as code reviews and contract evaluations. CHAP aims to address the complexities involved in tracking human overrides of AI outputs, providing a structured system where each human edit is recorded in a queryable format, complete with a rationale and tags. This enhances traceability, allowing teams to efficiently analyze decisions and improve AI performance over time by utilizing past overrides for learning. The significance of CHAP lies in its ability to create a single source of truth for human-agent interactions, reducing the time spent reconstructing decisions made weeks earlier and minimizing guesswork. Key technical components include an audit trail that connects each decision through content hashes, ensuring integrity and accountability. CHAP also offers compatibility with existing tools and systems, providing a flexible framework for integrating human input into AI workflows without the need for a complete overhaul of current applications. This could lead to improved accuracy in AI responses, ultimately enhancing productivity across industries that rely heavily on collaborative AI solutions.