Writing about Agentic Engineering Patterns (simonwillison.net)

A new project has emerged focusing on "Agentic Engineering Patterns," which aims to document coding practices that optimize the use of coding agents like Claude Code and OpenAI Codex. These tools not only generate but also execute code, allowing for independent testing and iteration without constant human oversight. This initiative seeks to bridge the gap between casual coding—often associated with non-programmers—and the enhanced capabilities for professional software engineers who leverage these agents to amplify their expertise. The first two "chapters" of this ongoing project have already been published, addressing critical aspects of agentic engineering. The initial chapter, "Writing code is cheap now," explores the implications of drastically reduced costs for producing working code and its effects on developers' workflows. The second chapter, "Red/green TDD," discusses how test-driven development can help coding agents create more concise and reliable code with minimal assistance. The project will continue to evolve, with additional chapters anticipated weekly, aiming to consolidate knowledge in this rapidly changing domain of AI-assisted programming.