Vibe Coding：迈向一种 AI 原生的语义与意图驱动编程范式
Vibe Coding: Toward an AI-Native Paradigm for Semantic and Intent-Driven Programming
AI总结：
本文提出 AI 原生的 vibe coding 范式，通过意图、语义与反馈机制让智能体生成软件，并分析其架构、收益、风险与研究方向。
AI中文摘要：
大型语言模型的最新进展使开发者能够通过与人工智能系统对话来生成软件，而不是直接编写代码。本文介绍 vibe coding，这是一种新兴的 AI 原生编程范式：开发者指定高层功能意图，以及对所期望“vibe”（基调、风格或情感共鸣）的定性描述；随后，智能体将这些规约转换为可执行软件。我们形式化了 vibe coding 的定义，并提出一种参考架构，其中包括意图解析器、语义嵌入引擎、智能体代码生成器和交互式反馈回路。文中描述了一个假设性实现。我们将 vibe coding 与声明式、函数式和基于提示的编程进行比较，并讨论其对软件工程、人机协作和负责任 AI 实践的影响。最后，我们考察了已有报告中的生产率提升和民主化效应，回顾了强调漏洞与潜在减速的近期研究，识别对齐、可复现性、偏见、可解释性、可维护性和安全性等关键挑战，并概述未来方向与开放研究问题。
英文摘要：
Recent advances in large language models have enabled developers to generate software by conversing with artificial intelligence systems rather than writing code directly. This paper introduces vibe coding, an emerging AI-native programming paradigm in which a developer specifies high-level functional intent along with qualitative descriptors of the desired "vibe" (tone, style, or emotional resonance). An intelligent agent then transforms those specifications into executable software. We formalize the definition of vibe coding and propose a reference architecture that includes an intent parser, a semantic embedding engine, an agentic code generator, and an interactive feedback loop. A hypothetical implementation is described. We compare vibe coding with declarative, functional, and prompt-based programming, and we discuss its implications for software engineering, human-AI collaboration, and responsible AI practice. Finally, we examine reported productivity gains and democratizing effects, review recent studies that highlight vulnerabilities and potential slowdowns, identify key challenges such as alignment, reproducibility, bias, explainability, maintainability, and security, and outline future directions and open research questions.
