Preface: From Coding Agents to Engineering Agents
In early 2026, models stopped being tools you queried and became collaborators you delegated to. In fact, Codex and I collaborated on this guidebook in the hope that an engineering agent can use it as onboarding context for an agentic engineering workflow. With the emergence of models like Opus 4.6 and Codex 5.3, agents crossed a threshold: they could compose CLI commands fluently, plan multi-step tasks end-to-end, and navigate the messy realities of shipping products.
Together, these shifts let agents own meaningful segments of the SDLC. The term "coding agent" no longer captures what is happening. These agents write code, but they also research, plan, architect, debug, document, review, and iterate across the stack.
This guidebook is about the logical shift towards Agentic Engineering. It is built on a simple idea: humans and agents do their best work when each operates in its highest-leverage zone.
- Humans set direction, make architectural calls, apply domain knowledge, and exercise judgment.
- Agents handle boilerplate, research, first drafts, and mechanical transformations.
This is not just for engineers. Designers, product managers, QA leads, and anyone producing software artifacts can use these frameworks.