The Agentic Engineer: A Practical Guide to AI-Assisted Software Development (Paperback) (ISBN-13: 9798190805376)

Coding agents went from demo to daily tool faster than the practice around them matured. Most engineers are now operating systems they have no mental model for, judging output they cannot fully verify, and choosing tools on the strength of benchmark numbers that were designed to be quoted rather than understood.

This book fixes that. It is written for working software engineers, and it assumes you are fluent in a language, comfortable in a terminal, and uninterested in hype.

What you will be able to do

• Explain mechanically how a model generates code, and predict the failure modes that follow from it
• Read a capability claim critically, and know which question deflates it
• Write specifications an agent can actually execute, and manage context as the finite budget it is
• Build verification that scales with how much you delegated, instead of trusting code that merely looks right
• Prepare a codebase, a CI pipeline, and a team for agent work without opening holes in any of them

Built to age honestly. Half of this field will look different in a year, so the book keeps the two halves apart. Durable mechanics are argued from how the systems work. Market and product detail is marked as a dated snapshot, sourced to a primary reference, and quarantined in an appendix you can check against reality in a minute. No price, score, or version number is hidden in the prose.

Opinionated, with the evidence attached. It takes positions on multi-agent swarms, on what parallelism actually buys, on the measured gap between how fast engineers feel and how fast they are, and on which security problems remain unsolved rather than merely mitigated. Where a judgment is the author's rather than the field's, it says so.

Twenty-four chapters, four appendices, and a set of templates you can copy into your own repository.