Overview
Simon Willison has launched a new project to document Agentic Engineering Patterns - coding practices for using AI agents that can both generate and execute code. This represents professional software engineering enhanced by AI agents, distinct from casual “vibe coding” where non-programmers generate code without understanding it.
Key Points
- Agentic Engineering focuses on professional developers amplifying their expertise with coding agents like Claude Code and OpenAI Codex, not casual AI-generated coding
- The central challenge is that code generation costs have dropped to nearly zero - requiring new approaches to development workflows and team dynamics
- Test-driven development (TDD) proves especially effective because agents write more reliable code when given clear test specifications to work toward
- The content will be published as living ‘guides’ - evergreen content designed to be updated over time rather than frozen blog posts
- All content is human-written with AI assistance for editing only - maintaining authenticity while leveraging AI for supporting tasks