Simon Willison's agentic engineering guide defines the discipline and its core patterns¶
Insight: Simon Willison defines agentic engineering as "the practice of developing software with the assistance of coding agents" — tools that can both write and execute code, running tools in a loop to achieve a goal. He argues that code execution is the defining capability: without the ability to directly run code, LLM output is of limited value. Rather than accepting quality degradation, developers should leverage agents to improve code quality by tackling technical debt that was previously too time-consuming.
Detail: The guide covers multiple chapters: principles, working with agents, testing and QA, understanding code, and annotated prompts. On better code, Willison identifies refactoring tasks (API redesigns, nomenclature cleanup, code duplication elimination, large file splitting) as ideal agent applications: "The cost of these code improvements has dropped so low that we can afford a zero tolerance attitude to minor code smells." He advocates asynchronous agent workflows for refactoring via background agents and a compound engineering loop where retrospectives from completed projects create reusable instructions.
Sources
- Simon Willison — "What is agentic engineering?" (2026-03-15)
- Simon Willison — "AI should help us produce better code" (2026-03-10)
- Simon Willison — "My fireside chat about agentic engineering at the Pragmatic Summit" (2026-03-14)
- Simon Willison — "Fireside chat about agentic engineering at the Pragmatic Summit" (2026-03-17)
Related: existing entry "Simon Willison's agentic engineering patterns guide" in external/claude-code.md — COMPLEMENTS