Simon Willison Launches Guide to Agentic Engineering Patterns
The new project establishes professional coding standards for engineers using autonomous AI agents to generate and execute code.
Software developer Simon Willison has launched "Agentic Engineering Patterns," a framework designed to document coding practices for professional engineers utilizing coding agents. The initiative aims to move beyond simple AI completions toward a disciplined approach to autonomous software development.
Agentic Engineering is defined as the process of building software using coding agents—such as Claude Code or OpenAI Codex—that possess the capability to both generate and execute code to iterate independently. Unlike "vibe coding," where users may write software without close attention to the underlying code, Agentic Engineering is positioned as the professional counterpart. According to Willison, it represents a way for professional software engineers to improve and accelerate their work by amplifying their existing expertise.
A Living Guide to AI Development
Rather than publishing static blog posts, Willison has structured the project as a "guide" composed of "chapters" that are intended to be updated over time. This iterative format reflects the rapidly evolving nature of AI tools and the patterns that emerge from their use.
The first two published patterns focus on the fundamental shifts in the development lifecycle. The first, "Writing code is cheap now," addresses the plummeting cost of generating initial working code. The second, "Red/green TDD," explores the utility of Test-Driven Development (TDD) as a mechanism for agents to verify their own output through a cycle of failing and passing tests.
The Shift from Writing to Verifying
This transition marks a significant pivot in the role of the software engineer. As the cost of producing initial code drops toward zero, the primary challenge for developers is shifting away from the manual act of writing syntax. Instead, the professional focus is moving toward directing, verifying, and architecting complex systems.
By establishing formal patterns, the project seeks to help developers maintain technical rigor and reliability. In an era of high-velocity AI generation, the ability to architect a system and verify its correctness becomes more critical than the ability to write the code itself.
The Future of Autonomous Iteration
As more professional engineers adopt these tools, the industry is moving toward a model where agents handle the repetitive cycle of implementation and self-correction. The "Agentic Engineering Patterns" guide serves as an early blueprint for this transition, providing a structured way to integrate autonomous agents into a professional workflow without sacrificing quality.
Future updates to the guide are expected to expand on these initial patterns as more use cases for autonomous execution and verification are documented in the field.