Agentic Programming

Software development is changing. For the first time, developers are working alongside systems that actively participate in building software—and the most important shift isn't that AI writes code faster. It's that it changes what you can delegate.

Agentic Programming is a practical roadmap for that transition: from writing software directly to designing, guiding, verifying, and eventually orchestrating systems that create it.

At its center is the AI Fluency Ladder—five levels of delegation, from prompting and task delegation to agentic workflows, verified execution, and autonomous execution. The book shows how developers climb it, why a new bottleneck appears at each level, and how the developer's role changes along the way.

The first half builds the foundation: how AI systems actually behave, the mental models agentic programming depends on, and the failure modes of probabilistic systems. The second half is hands-on—it walks through building an agentic harness, the execution architecture that coordinates AI agents through workflows, verification, retries, and recovery, and makes higher levels of delegation reliable enough to trust.

It's written for software developers, architects, engineering leaders, and technical founders who want to understand how AI is changing software development. No machine-learning background required—the focus is the practical implications of working with AI systems, not the mathematics behind them.

This book is being published in progress. The core narrative is already here—the foundations, the AI Fluency Ladder, and the full harness-building material—while the final part is still being written and the diagrams remain in draft form (text placeholders that will be replaced with polished graphics). Early readers get every update free, including the finished diagrams and remaining chapters, and lock in today's price, which rises as the book is completed.