AI Native
Building the Copilot Your Team Will Live In
© 2026 by Bill Cox, waywardgeert@gmail.com
This book was written by an AI workflow pipeline — an author agent and an editor agent collaborating through a structured protocol, with a human approving the final output. The pipeline itself was built using the methods described in these pages.
What this is
Your analyst has a chatbot. It runs a frontier model. It can summarize an alert, draft a template, explain a CVE. Ask it about yesterday's incident and it stares back blankly. It is a goldfish.
This book is about what it takes to turn that goldfish into a copilot. Not a better chatbot — a copilot: the other pilot. Same frontier model. The gap between the goldfish and the copilot is not intelligence. It is engineering.
What's in it
Seventeen chapters, each deriving one rule from working code and field experience with CodeRhapsody — a supervised autonomous AI coding agent built over eighteen months.
- Part I — The Gap: The goldfish vs. the copilot. Why model capability is not the bottleneck.
- Part II — Architecture: The agent as actor, real-time hints, prompt caching, ephemeral context, memory systems, skills vs. workflows.
- Part III — Trust: Security when hostile input is the workload, the counterpoint defense, and proving improvement with evidence.
Who it's for
Engineers building AI-assisted tools for security operations, DevOps, or any domain where the agent reads untrusted data and the human needs to steer in real time. The patterns are general; the examples are drawn from SecOps because that is where the author lives.