What is agentic engineering?

Simon Willison introduces "agentic engineering" as the practice of developing software with coding agents (Claude Code, Codex, Gemini CLI), defining agents as LLMs that run tools in a loop — with code execution as the key differentiator. He argues the human role shifts to specifying problems well, verifying results, and updating agent instructions based on accumulated learning. He also draws a sharp line between agentic engineering and "vibe coding" (Karpathy's term for unreviewed prototype-quality LLM output), positioning the former as a rigorous, production-quality discipline.