⌘K
Change language Switch ThemeSign In
Narrow Mode
Agentic AI 时代，软件工程如何被重塑｜QCon 北京 2026 即将开幕
I InfoQ 中文 @InfoQ 中文
One Sentence Summary
QCon Beijing 2026, themed 'Reshaping Software Engineering in the Era of Agentic AI,' delves into cutting-edge topics such as autonomous system architecture, consumer-grade agents, AI engineering, and the evolution of R&D paradigms.
Summary
This article is an in-depth preview of QCon Beijing 2026. It highlights that AI is evolving from an auxiliary tool into an 'agent' capable of planning and execution, driving the transformation of software engineering from code-driven to intent-driven. The content covers insights on autonomous system architecture from Ed Huang of TiDB, practical experience in consumer-grade agents from Fliggy, and Bosch's exploration of model limitations and world models. Furthermore, the conference focuses on core engineering challenges such as Agent Infra, long-term memory, observability, Coding Agents (e.g., Vibe Coding), and embodied AI, demonstrating how AI is deeply reshaping the entire chain from underlying infrastructure to upper-level R&D models.
Main Points
* 1. The role of AI is evolving from a 'conversational tool' to an 'acting agent'.Agents possess planning, execution, and self-feedback capabilities, driving software systems from passive instruction response to proactive collaboration, ushering in a new paradigm for software engineering. * 2. System architecture is evolving towards 'autonomous system architecture'.Agents are becoming new computational entities, shifting the focus of system design from API calls to long-term state management, multi-agent collaboration, and infrastructure that supports continuous operation. * 3. The R&D paradigm is shifting from Code First to Intent First.Software behavior is increasingly generated by models at runtime; developers are shifting from precisely defining code to defining intent, with AI deeply involved throughout the entire R&D lifecycle. * 4. Agent engineering faces a leap in complexity.The key to moving from demo to production systems lies in building agent runtimes, memory middleware, observability assessment engineering, and compute optimization for AI workloads.
Metadata
AI Score
86
Website mp.weixin.qq.com
Published At Today
Length 2603 words (about 11 min)
Sign in to use highlight and note-taking features for a better reading experience. Sign in now
原创 QCon 2026-03-23 14:31 浙江
汇聚来自一线科技企业与创新团队的技术专家，围绕 AI 工程化、系统架构与研发模式演进展开深入探讨。
策划｜QCon 全球软件开发大会
编辑｜Kitty & 燕珊
当 AI 从“对话工具”进化为“行动主体”，软件工程正迎来新一轮系统性重构。
在大模型能力持续突破之后，AI 正从“辅助工具”走向“行动主体”。以 Agent 为代表的新一代 AI 系统，不再只是响应指令，而是具备规划、执行与自我反馈能力，能够在复杂环境中完成多步骤任务。这一变化，正将软件系统从“被动执行”推向“主动协作”，也由此开启了软件工程的新一轮范式演进。
在这样的背景下，QCon 全球软件开发大会·2026 北京站将于今年 4 月正式举办。本届大会以“Agentic AI 时代的软件工程重塑”为主题，汇聚来自一线科技企业与创新团队的技术专家，围绕 AI 工程化、系统架构与研发模式演进展开深入探讨。
主题演讲：多元视角，
共探 Agentic AI 的边界与可能
大会首日上午的主题演讲，将拉开思想碰撞的帷幕。
TiDB 联合创始人兼 CTO 黄东旭将带来《The Age of Autonomous Systems 自主系统的时代》的深刻洞察，Agent 不是一个功能，而是一种新的计算实体，很明显已经开始吞噬一切软件，我们的系统架构也面临从应用架构到“自主系统架构”的转变。 围绕这一趋势，本次演讲将聚焦几个关键问题：当软件成为“主动行动者”，系统架构如何演化；为什么 AI 的真正挑战不是模型能力，而是长期状态管理；从 API 到 CLI 到状态驱动，边界在哪里；多 Agent 协作将如何重塑软件系统、如何有效实现协同；以及什么样的基础设施能够支撑持续运行的智能体。
阿里巴巴中国电商事业群飞猪 CTO 陈烨博士将以《消费级智能体的演进：重塑交易范式与增量未来》为题，分享智能体如何在大消费场景中创造全新价值与交互体验。他将复盘飞猪在春节期间的实战数据，观察 AI Native 模式在转化率上展现出的代际优势：AI 引导的 D2O（发现即下单）在部分类 SKU 品类中表现尤为突出。同时，深入拆解在复杂旅行场景下，如何依托 Agentic Coding 与 Post-training 构建高可用 Agent 的核心逻辑。希望通过飞猪的实践案例，探讨如何依托私有数据资产与服务基因，在 AI 下半场探索高价值增量市场，进而探寻下一代 OTA 核心竞争力的新范式。
Bosch 集团首席 AI 科学家阎栋博士将发表题为《模型之外》的演讲。他将从大模型的底层机制出发，探讨模型在记忆与持久化能力上的根本局限，并援引人类智能与模型智能的对比研究，引出世界模型（World Model）作为突破现有能力桎梏的关键方向。最后，他将重返维特根斯坦的哲学论断，以更宏大的视角回应“模型之外还有什么”这一行业根本问题，为与会者带来对 AI 能力边界的深度反思。
专题演讲：十八般武艺，
解码 Agentic AI 工程化核心
从“LLM 应用”到“Agent 系统”：工程复杂度跃迁
如果说过去两年，行业的核心问题是“如何用好大模型”，那么今天，问题已经演变为：如何构建一个真正可用、可靠、可演进的 Agent 系统？
本届 QCon 将系统呈现 Agent 工程的关键议题。在 “Agent Infra 架构设计” 专题中，来自百度、字节跳动、小米的专家将解析智能体运行时、记忆中间件与中枢网关的构建；“智能体记忆觉醒” 专题则聚焦长期记忆如何成为智能体进化的“第二大脑”；而 “Agent 可观测性与评估工程” 专题，将直面大规模 Agent 在生产环境中的“不可知”困境，分享从“盲目调优”到“数据驱动”的评估飞轮建设。这些议题，正在成为 AI 应用从“Demo”走向“生产系统”的关键分水岭。
“OpenClaw 生态实践”专场将聚焦一线实践与踩坑复盘，分享企业如何构建私有 Skills、制定安全护栏、搭建审计与回放机制、建立质量 / 效率指标体系，最终把自托管 Agent 从可用的 Demo 升级为可靠的生产系统。
AI 工程体系重构：从推理优化到系统级能力建设
随着 AI 应用规模化落地，企业开始面对一系列“工程真实问题”：
* 如何在有限算力下实现高吞吐、低延迟推理？
* 多模态模型如何统一调度与高效运行？
* 如何保障 AI 系统在高并发与复杂场景下的稳定性？
* 在金融、工业等场景中，如何实现合规、安全与可控？
围绕这些问题，本届大会设置多个 AI 核心专题，覆盖从底层推理到上层应用的完整技术链路。“大模型算力优化” 专题将聚合来自清华大学、无问芯穹、京东的专家，深入探讨推理加速、KV Cache 压缩与生成式推荐的高性能架构；“AI 原生基础设施” 专题则关注面向 AI 工作负载的调度与运行时系统，腾讯、华为、蚂蚁集团的实践者将分享从 K8s 与 Ray 的融合，到面向 AI 原生负载的统一弹性调度体系。
这些内容将重点呈现：AI 发展不只是能力问题，更关系着基础设施的演进。
软件工程范式迁移：从“代码驱动”到“意图驱动”
Agentic AI 的出现，也正在改变软件的构建方式。
在传统软件工程中，开发者通过代码精确定义系统行为；而在 AI 时代，越来越多的行为由模型与 Agent 在运行时生成与决策。
这带来了几个关键转变：
* 从 Code First 走向 Intent First
* 从 API 调用 走向 能力编排
* 从 静态系统设计 走向 动态系统演化
与此同时，AI 正在深度参与研发流程本身。“Coding Agent 驱动的研发新范式” 专题将呈现这一变革的最前沿：从蚂蚁集团的 Vibe Coding 平台实践经验，到平安科技的金融级 AI Coding 落地，再到网易、京东科技、百度、淘宝闪购等呈现深度企业级 AI Coding 实践 。软件工程，正在从“人构建系统”，走向“人与 AI 共同构建系统”。
从数字世界到物理世界：AI 边界的进一步拓展
除了软件系统内部的演进，本届 QCon 也将关注 AI 向物理世界的延伸。
围绕 “具身智能与物理世界交互” 专题，大会将探讨：
* AI 如何在真实环境中进行感知、决策与执行
* 工业场景中具身智能落地的工程挑战
* 空间建模与仿真的规模化实现路径
来自地瓜机器人、乐享科技、小雨智造的一线专家，将带来从机器人操作系统设计、VLA 模型落地到工业具身规模化实践的完整分享。北京科技大学副教授将分享「空间智能初探：重建与生成的双路径实践」，这些议题标志着一个重要趋势：AI 的工程问题，正在从“软件复杂性”，走向“系统复杂性”。
一线实践与真实经验：回到工程现场
作为全球知名的技术大会，QCon 始终坚持“源于实践”的内容标准。
本届北京站的讲师，来自阿里巴巴、腾讯、字节跳动、小米、蚂蚁集团、百度、快手、美团、网易等一线科技企业，以及多家 AI 创新机构。他们将分享：
* 大规模 AI 系统的架构设计与演进经验
* Agent 在真实业务中的落地实践与反思
* 工程效率提升的具体方法与路径
* 技术决策背后的权衡与取舍
这些内容，将为技术团队提供可借鉴的实践参考，也为行业提供更清晰的演进路径。
结 语
当 AI 从“工具”进化为“Agent”，
软件工程的边界，正在被重新定义。
Agentic AI 时代已经到来。
QCon 北京站，期待与你一起，理解变化，构建未来
I InfoQ 中文 @InfoQ 中文
One Sentence Summary
QCon Beijing 2026, themed 'Reshaping Software Engineering in the Era of Agentic AI,' delves into cutting-edge topics such as autonomous system architecture, consumer-grade agents, AI engineering, and the evolution of R&D paradigms.
Summary
This article is an in-depth preview of QCon Beijing 2026. It highlights that AI is evolving from an auxiliary tool into an 'agent' capable of planning and execution, driving the transformation of software engineering from code-driven to intent-driven. The content covers insights on autonomous system architecture from Ed Huang of TiDB, practical experience in consumer-grade agents from Fliggy, and Bosch's exploration of model limitations and world models. Furthermore, the conference focuses on core engineering challenges such as Agent Infra, long-term memory, observability, Coding Agents (e.g., Vibe Coding), and embodied AI, demonstrating how AI is deeply reshaping the entire chain from underlying infrastructure to upper-level R&D models.
Main Points
* 1. The role of AI is evolving from a 'conversational tool' to an 'acting agent'.
Agents possess planning, execution, and self-feedback capabilities, driving software systems from passive instruction response to proactive collaboration, ushering in a new paradigm for software engineering.
* 2. System architecture is evolving towards 'autonomous system architecture'.
Agents are becoming new computational entities, shifting the focus of system design from API calls to long-term state management, multi-agent collaboration, and infrastructure that supports continuous operation.
* 3. The R&D paradigm is shifting from Code First to Intent First.
Software behavior is increasingly generated by models at runtime; developers are shifting from precisely defining code to defining intent, with AI deeply involved throughout the entire R&D lifecycle.
* 4. Agent engineering faces a leap in complexity.
The key to moving from demo to production systems lies in building agent runtimes, memory middleware, observability assessment engineering, and compute optimization for AI workloads.
Key Quotes
* As AI evolves from a 'conversational tool' to an 'acting agent,' software engineering is undergoing a new round of systematic refactoring. * An agent is not just a feature, but a new computational entity that is clearly beginning to eat all software. * Software engineering is shifting from 'humans building systems' to 'humans and AI co-building systems'. * The real challenge of AI is not model capability, but long-term state management.
AI Score
86
Website mp.weixin.qq.com
Published At Today
Length 2603 words (about 11 min)
Tags
Agentic AI
Software Engineering
Autonomous Systems
Coding Agent
AI Engineering
Related Articles
* From Context to Long-Term Memory: Architectural Design and Practice of LLM Memory Engineering\" architecture.") * Practices and Reflections on Vibe Coding in Code Generation and Collaboration * OpenAI Frontline Development Observations: Those Who Can Manage 10-20 Agents Simultaneously and Run Hour-Long Tasks Are Leaving Other Engineers Far Behind * From Clawdbot to the 2026 AI Coding Explosion | A Conversation with PingCAP CTO Dongxu * What You Don't Know About Claude Code: Architecture, Governance, and Engineering Practices - Tw93 * Top-level Thinking and Methodology for AI Coding Products: From Paradigm Revolution to Spec Coding * “AI on the Front Lines: How Developers are Reshaping the Software Development Process” | Roundtable Discussion * From 7.9% to 54% Adoption: The Three-Stage Evolution of Kuaishou's Intelligent Code Review * Cognitive Reconstruction: After Three Months with Speckit, I Abandoned It - Escaping the Dilemma of Powerful Tools That Are Hard to Use Well * 1,500 PRs, 0 Human Coders: Building a Million-Line Internal Product Driven by Codex HomeArticlesPodcastsVideosTweets