Agentic Engineering
Agentic Engineering is a software engineering methodology that emphasizes the role of engineers as system orchestrators and decision-makers in AI agent-assisted development environments, rather than as direct code authors. Its core focus is on directing autonomous AI agents to perform tasks such as code generation, testing, and integration by designing goals, constraints, and task decompositions, while maintaining overall control over system reliability, maintainability, and output quality.
Timeline (last 90 days)
Karpathy coined the term Agentic Engineering to describe the disciplined end of the spectrum.
Production APIs handling money flows should use Agentic Engineering, while weekend prototyping can use pure Vibe Coding.
Agentic engineering refers to engineering practices centered on autonomous agents, building systems capable of independently planning, executing multi-step tasks, invoking tools, and even self-iterating.
The agentic engineering field currently lacks unified best practices, and prompting strategies, context window utilization methods, and multi-agent collaboration architectures are still rapidly iterating.
All Facts (5)
卡帕西正式提出了'Agentic Engineering'（智能体工程）的概念，以及'软件演进三阶段论'：软件1.0（传统命令式编程）、软件2.0（深度学习）、软件3.0（大语言模型智能体编程范式）
85%UnverifiedKarpathy coined the term Agentic Engineering to describe the disciplined end of the spectrum.
50%UnverifiedProduction APIs handling money flows should use Agentic Engineering, while weekend prototyping can use pure Vibe Coding.
50%UnverifiedAgentic engineering refers to engineering practices centered on autonomous agents, building systems capable of independently planning, executing multi-step tasks, invoking tools, and even self-iterating.
50%UnverifiedThe agentic engineering field currently lacks unified best practices, and prompting strategies, context window utilization methods, and multi-agent collaboration architectures are still rapidly iterating.
50%