Mise en Place for Agentic Coding: Deliberate Preparation as Context Engineering Methodology

The paper 'Mise en Place for Agentic Coding' by Andrew Zigler (accepted at VibeX 2026) argues that the current 'vibe coding' workflow—prioritizing speed over deliberate preparation—creates a systematic alignment problem. Insufficient context forces AI coding agents to produce code that requires extensive debugging and refactoring, consuming large amounts of development time. Drawing from the culinary concept of mise en place ('everything in its place'), Zigler proposes a three-phase methodology: contextual grounding (externalizing tacit knowledge into structured documents), collaborative specification (human-agent dialogue to produce detailed design artifacts), and task decomposition (converting specs into dependency-aware task records).

Application of MEP in a competitive hackathon showed that roughly two hours of preparation enabled a rapid parallel implementation of a full-stack educational platform by concurrent AI agents—a dramatic improvement over typical vibe coding outcomes. The paper also coins 'context fluency' as an emerging developer skill: the ability to craft rich, structured context that AI agents can act on. This connects to backward design and tacit knowledge externalization frameworks. Zigler concludes with a research agenda for empirically validating preparation-phase methodologies in AI-assisted software development, signaling a shift from speed-first to context-first coding practices.