# Agentic Engineering Agentic engineering is the practice of developing software with the assistance of coding agents: tools that can both write and execute code in a loop. The key word is *execute*. Without code execution, LLM output is just text you have to manually verify. With it, agents can iterate toward working software by running tests, checking outputs, and fixing their own mistakes. The discipline is distinct from "vibe coding" (accepting AI output without review). Agentic engineering demands that the human remains responsible for what gets shipped: deciding what to build, verifying that it works, reviewing the code, and ensuring quality standards are met. [[Simon Willison]] puts it well: writing code has never been the sole activity of a software engineer. The typical workflow involves an [[AI Agent Harness]] (like [[Claude Code]], [[Aider]], or [[Cursor.com]]) wrapping a [[Large Language Models (LLMs)|Large Language Model]] with tools (file system, shell, browser, APIs) and running them in a loop. The human provides direction, context, and judgment. The agent provides speed, breadth, and tireless iteration. See [[How coding agents work]] for the mechanics. Key patterns that have emerged in agentic engineering include: - [[Agentic TDD]]: using red-green test cycles to validate agent output - [[AI Subagents]]: dispatching fresh agent instances for parallel or scoped work - [[AI-assisted code comprehension]]: using agents to explain and visualize code - [[Hoard things you know how to do]]: building a personal knowledge base of working solutions that agents can recombine - [[Code is cheap, quality is not]]: the economic shift that makes all of this viable ## References - https://simonwillison.net/guides/agentic-engineering-patterns/ ## Related - [[AI Agents]] - [[AI Agent Harness]] - [[How coding agents work]] - [[AI Subagents]] - [[AI Agent Orchestration]] - [[Agentic TDD]] - [[Agent Development Lifecycle (ADLC)]] - [[Agent-Native Product Decomposition]] - [[AI Verifiability]] - [[AI Verifiability as a Capability Ceiling]] - [[Markdown-based Installation (MD Scripts)]] - [[Code is cheap, quality is not]] - [[Hoard things you know how to do]] - [[Cognitive debt]] - [[AI-assisted code comprehension]] - [[Unreviewed AI code anti-pattern]] - [[Context Engineering]] - [[Claude Code]] - [[Simon Willison]]