Agentic Engineering Patterns is a collection of practices for getting the best results out of coding agents such as Claude Code and OpenAI Codex, published on Simon Willison’s blog from February 2026. It is organised as a sequence of chapters, each a self-contained pattern. The post introducing it states the goal as producing something that helps answer the question “how do I get good results out of this stuff” all in one place, against a body of the author’s own existing writing on AI-assisted programming that he describes as relatively unstructured.
Its form is the part its author treats as novel. He describes the collection as not exactly a book but “kind of book-shaped”, and publishes it using a shape of content he calls a guide: a collection of chapters, where a chapter is effectively a blog post with a less prominent date, designed to be updated over time rather than frozen at the point of first publication. He presents guides and chapters as his answer to publishing evergreen content on a blog, a problem he reports having been trying to solve for a while. That format — rather than any single publication event — is why the collection is recorded here as a series.
Its stated debt is to the chapter-shaped pattern format popularized by the software design-patterns literature of the 1990s, which the introduction post describes as a loose inspiration rather than a model followed closely. See Writing about Agentic Engineering Patterns for the post that introduced the project.
Scope & Structure
The guide’s chapters are grouped under six headings. Principles holds “What is agentic engineering?”, “Writing code is cheap now”, “Hoard things you know how to do”, “AI should help us produce better code”, and “Anti-patterns: things to avoid”. Working with coding agents holds “How coding agents work”, “Using Git with coding agents”, and “Subagents”. Testing and QA holds “Red/green TDD”, “First run the tests”, and “Agentic manual testing”. Understanding code holds “Linear walkthroughs” and “Interactive explanations”. Annotated prompts holds two worked examples, and an Appendix collects prompts the author reuses.
The project began with the two chapters published alongside its announcement — “Writing code is cheap now” and “Red/green TDD” — with a stated intended cadence of one to two chapters a week. The author says he does not really know when he will stop, because there is a lot to cover.
Status
The guide’s opening chapter describes the project as very much a work in progress, like the field it attempts to cover, and states that no chapter should be considered finished: the author says he will continue adding chapters as new techniques emerge and will update existing ones as understanding of the patterns evolves. The editorial goal stated there is to identify and describe patterns for working with these tools that demonstrably get results and that are unlikely to become outdated as the tools advance.
Its stated editorial constraint is that the prose is written by the author rather than generated: he holds a strong personal policy against publishing AI-generated writing under his own name, and says it will hold for this project, while using LLMs for proofreading, fleshing out example code, and other side tasks.
Related
Agentic Engineering, AI Coding Agent, Vibe Coding, Writing about Agentic Engineering Patterns