Teams now work alongside several AI coding agents, such as Claude Code and Codex, and each agent has its own memory, permissions and habits. Without a shared record, lessons stay in one agent's private notes, which leads to fragmented understanding of complex projects and future tech debt. This session proposes a lightweight method for humans and AI agents to collaborate and share plan, progress, and challenges.
The central idea is to maintain a set of documents: A short, one-line-per-event log gives people and agents a common timeline (a "table of contents", if you will). The details are written up and updated in a separate place so the log stays readable. A "changelog" type doc sits between the log entries and the details files, whereas an issue tracker documents the parking-lot items left to be addressed later. Habits that matter, such as recording production changes and backing up before updates, are enforced by hooks that work for any agent, instead of relying on anyone to remember. A more detailed observability appears when we link individual session notes to tasks, efforts, and other operational information.
This project is in-progress: We share what has been working and what needed tweaks (documentation drift, silent shortcuts, double-counted usage).
