Simon Willison identifies key anti-patterns in agentic engineering, focusing on the problematic practice of submitting unreviewed AI-generated code in pull requests. He emphasizes that developers must take responsibility for reviewing agent-produced code before sharing it with collaborators, ensuring it's functional, properly scoped, and well-documented. The article provides practical guidelines for effective agentic engineering workflows that maintain code quality and team collaboration standards.
Background
Agentic engineering involves using AI agents to assist with programming tasks, but this introduces new challenges in code quality and team collaboration. As AI coding assistants become more prevalent, establishing best practices for their responsible use is crucial.
- Source
- Simon Willison
- Published
- Mar 5, 2026 at 01:34 AM
- Score
- 6.0 / 10