一句话：结合大规模实证与 SASE 框架，描绘从 AI 辅助开发到自主智能体协作的软件工程 3.0 路线图。
软件工程正从 AI 辅助开发（SE 2.0）转向 Agentic 软件工程（SE 3.0），即由自主 AI 智能体充当协作者。本文是一份技术简报，旨在为规范化的 SE 3.0 实践奠定基础。作者综合了三方面工作：其一，基于 AIDev 数据集对真实环境中智能体活动进行大规模刻画；其二，对来自 157 个开源软件项目的 567 个 Claude Code 拉取请求开展实证研究；其三，提出结构化 Agentic 软件工程（SASE）框架，重新界定参与者、过程、制品与工具四个维度。其目标是为参与者提供有证据支撑的心智模型、针对智能体撰写 PR 的评估模式，以及一套用于设计可信人机工作流的实用术语（如 agent command environment 与 agent execution environment）。
Software Engineering (SE) is shifting from AI-assisted development (SE 2.0) to Agentic Software Engineering (SE 3.0), where autonomous AI agents act as collaborators. This technical briefing establishes the foundations of a disciplined SE 3.0 practice. We synthesize: (1) a large-scale characterization of agent activity in the wild based on the AIDev dataset, (2) an empirical study of 567 Claude Code pull requests across 157 open-source software projects, and (3) the Structured Agentic Software Engineering (SASE) framework that reframes actors, processes, artifacts, and tools. Attendees will acquire evidence-backed mental models, evaluation patterns for agent-authored PRs, and a practical vocabulary (e.g., agent command environment and agent execution environment) for designing trustworthy human-agent workflows.
@inproceedings{paperbot11797,
title = {Agentic Software Engineering: A Roadmap to Software Engineering 3.0},
author = {Ahmed E. Hassan and Hao Li and Dayi Lin and Bram Adams and Tse-Hsun (Peter) Chen and Yutaro Kashiwa and Dong Qiu and Haoxiang Zhang and Miku Watanabe},
booktitle = {ICSE},
year = {2026},
doi = {10.1145/3774748.3776601}
}