What is agentic engineering?
simonwillison.net
March 16, 2026
3 min read
🔥🔥🔥🔥🔥
55/100
Summary
Agentic engineering involves developing software with the help of coding agents that can write and execute code. Examples of coding agents include Claude Code, OpenAI Codex, and Gemini CLI.
Key Takeaways
- Agentic engineering involves developing software with the assistance of coding agents that can write and execute code, such as Claude Code and OpenAI Codex.
- Coding agents operate by defining a goal, generating and executing code in a loop until the goal is achieved, leveraging code execution as a key capability.
- Effective use of coding agents requires humans to specify problems in detail, provide necessary tools, and iteratively verify and refine results.
- The term "vibe coding" refers to unreviewed, prototype-quality code generated by LLMs, distinguishing it from production-ready code.
Community Sentiment
Positives
- The distinction of 'agentic engineering' emphasizes the evolving role of engineers in managing AI-generated code, highlighting the importance of accountability in AI applications.
Concerns
- The term 'agentic engineering' feels misaligned with traditional engineering disciplines, suggesting a lack of clarity in defining its significance within the AI field.
- Concerns about accountability in AI-generated code underscore the limitations of AI agents, indicating that human oversight remains crucial in the development process.
