Simon Willison says “vibe coding” and responsible AI-assisted engineering are increasingly blurring in real-world software work

Simon Willison reflects on a podcast discussion where he realized his earlier distinction between casual “vibe coding” and more disciplined agentic engineering is becoming less clear. He argues vibe coding can be useful for personal tools, but remains risky when building software for others without understanding or reviewing the code.

The piece highlights a broader shift as AI coding tools move from autocomplete into autonomous development workflows. As these practices converge, teams may need stronger norms around code review, testing, and accountability to keep productivity gains from undermining software quality.

Vibe coding and agentic engineering are getting closer than I'd like