New Book — March 2026

The Agentic Crew

How software engineers learn to build with agents

AI agents don't just autocomplete your code. They read your codebase, reason about it, make changes across files, run your tests, and iterate. This book is your practical guide to the transition — from writing software to engineering it.

The Agentic Crew book cover

The ground is shifting

01

The craft is evolving

We're moving from writing every line by hand to directing, orchestrating, and collaborating. Less typing, more thinking. Less editor, more engineering.

02

No playbook exists

The tools showed up fast — Copilot, then Claude, then agents that run entire tasks autonomously — and we're all figuring it out in real time.

03

Your skills matter more

Systems thinking, taste, judgement, knowing what to build and why — those matter more now, not less. This book shows you how to wield them.

Why does this book exist? Because I have a genuine, slightly irrational desire to share everything I've learned about agentic engineering. Because I want to shine in the eyes of my kids. And — let's be honest — because having written a book is technically a conversation starter, even if it's a terrible one.

I've been a software engineer for a long time. I know what it feels like to have a codebase in your head — to open an editor, navigate to the right file, and just write the thing. That muscle memory, built over years of keystrokes and debugging sessions, is real. It's earned.

And now the ground is shifting under our feet.

— From the Foreword

What's inside

17 chapters covering the shift, the tools, and the engineering practices that make agentic work actually work.

01

Introduction

The fundamental loop of software engineering is breaking. What it means to go from writing software to engineering it — with a crew of agents at your side.

02

Context

The single most important skill in agentic engineering isn't prompting — it's context management. Learn to curate what your agents see.

03

What Is an Agent?

Placing AI tools on a spectrum from autocomplete to autonomous systems. What makes something truly agentic, how agents fail, and the right mental model for working with them.

04

Principles from Building My Own Agentic Tooling

Parallel visibility, structured handovers, configurable trust, and version control as infrastructure — principles that emerged from building a multi-agent control plane.

05

Guardrails

The trust gradient, permission scoping, approval gates, and environment-specific rules. How to give agents power without giving them the keys to production.

06

Git as Agent Infrastructure

Small commits, descriptive messages, branch-per-task, and worktrees for parallel agents. Git as your undo button and review interface.

07

Sandboxes

Giving agents the freedom to be wrong. Worktrees, containers, and ephemeral environments that make failure cheap and experimentation free.

08

Testing as the Feedback Loop

Tests become the agent's eyes. TDD as a superpower, fast test suites as agent infrastructure, and the virtuous cycle of better tests and more autonomous agents.

09

Convention Over Configuration

Why consistent file naming, formatters, project layouts, and CLAUDE.md files make agents dramatically more effective.

10

Local LLMs vs. Commercial LLMs

Privacy, cost, capability trade-offs, team cost management, and getting started without paying the farm. The practical hybrid approach of matching the model to the task.

11

Prompting as Engineering

How to write effective agent prompts — task decomposition, constraint specification, the prompt-as-spec pattern, and anti-patterns to avoid.

12

Multi-Agent Orchestration

Running agents in parallel, the handover pattern for sequential work, branch-per-agent isolation, and knowing when more agents isn't better.

13

Agents in the Pipeline

Using agents in CI/CD — PR pre-screening, overnight agents, cost control in automation, and the trust question when nobody's watching.

14

When Agents Get It Wrong

War stories from the field — the eager refactorer, the hallucinated library, the infinite loop — plus a diagnostic playbook for systematically debugging agent failures.

15

When Not to Use Agents

The overhead tax, architecture decisions, security-critical code, working with legacy codebases, and the craft argument. Knowing when to put the agent away.

16

Agentic Teams

How agent-assisted development changes code review, the junior engineer question, knowledge distribution, compliance and audit trails, hiring, and team rhythms.

17

Final Words

The craft isn't dying — it's shedding its skin. What endures, what changes, and why the captain still matters.

About the author

Rasmus Bornhøft Schlünsen

With over 15 years in the software industry, Rasmus has built everything from Django database backends and embedded systems to Rust CLI tools and open-source libraries used by thousands. He's an active open-source contributor with projects like django-timescaledb and rana. Currently, he works as an AI R&D Engineer at Clovrlabs.com.

He knows what a production incident feels like at 2am and what it takes to ship real products. Now he's navigating the shift to agentic engineering — and writing the playbook he wished he had.

Connect on LinkedIn →

"To my kids — you're the best crew I've ever had. Always."