从「构建 AI」到「驾驭 AI」，100+ 实战案例拆解 AI Native 时代的工程新实践！
2026 年 QCon 全球软件开发大会大会 · 上海站将于 10 月 22 日-24 日举办，来自阿里巴巴、腾讯、字节跳动、美团、小红书、蚂蚁集团、无问芯穹、阶跃星辰、快手、百度、网易、微软、华为、携程、平安等企业的一线技术实践者，将从 AI 应用、Agent 工程、AI Coding、基础设施、数据系统、质量、安全、SRE 以及具身智能等方向，分享 AI 进入真实生产环境之后正在发生的变化。
阿里巴巴中国电商事业群、飞猪 CTO，陈烨博士已确认出席主题演讲，并发表题为《从 Coding 到消费：AI 的下一场革命》的主题分享。
AI 为什么先学会写代码，而不是先学会点外卖？一个关键原因是，Coding 天然拥有编译器、单元测试、CI 等“Verifier”：反馈密集、即时且客观，任务又完全发生在数字世界里。相比之下，消费场景的反馈更加稀疏、延迟且主观，交易与履约还要穿透商家、库存、配送等真实的物理世界。
这意味着，AI 从 Coding 走向消费，并不只是把一个更强的模型接入消费应用，而是要重新构造一套面向真实世界的验证与执行闭环。本次演讲将从“可验证性”出发，分析 AI 为什么沿着 Coding → Agent → 消费的路径演进，以及消费级 AI 真正的难点究竟在哪里。
Frontier Lab 的“Infra × Data × Compute”配方翻译到消费场景：如何构建 AI-Native 的供给与服务网络，如何将静态与实时消费数据转化为训练资产，以及如何通过后训练建立持续迭代的数据飞轮。核心判断是：下一阶段 AI 的关键，不只是让模型更聪明，而是把物理世界的履约过程变成可验证、可学习的训练信号。
最终，下半场真正的机会，不在于再造一个 AI Chat，而在于让 AI 真正进入交易、服务与履约，让消费成为通向更强通用智能的重要训练场。
陈烨，飞猪 CTO，负责飞猪 AI 战略的规划与实施。威斯康星大学麦迪逊分校信息系统与计算机博士，发表 30 余篇 NeurIPS、SIGIR、KDD、WWW 等顶会论文，三次获得 KDD、SIGIR 最佳论文奖，部分成果被斯坦福大学、加利福尼亚大学伯克利分校纳入研究生课程；持有 10 项美国专利。曾创立人工智能公司虎博科技，其多模态大语言模型 TigerBot 入选首批国家大模型备案名单；曾任美团点评资深副总裁，主导构建搜索与效果广告平台。
除此之外，上海交通大学人工智能学院副教授/无问芯穹联合创始人兼首席科学家戴国浩博士、小影科技创始人 & CEO 韩晟等多位来自学界与产业界的专家也将在 Keynote 主题演讲中，分享 AI 从前沿洞察到技术创新再到产业应用过程中的不同视角与关键探索。
本次大会还策划了Loop Engineering、千行百业 Agent 创新实践、Agent 自主进化：从记忆到持续学习、Agent as a Service、Vibe Coding 时代的新质量债、理性驾驭 AI 的 SRE 可靠性工程、金融 AI Native工程实践：从研发提效到业务破局、AI Infra：算力效率决定规模化落地、AI Native 架构等 20 个专题论坛，届时将有来自不同行业、不同领域、不同企业的 100+资深专家在现场带来前沿技术洞察和一线实践经验。
查看更多详情可扫码或联系票务经理 18514549229 进行咨询。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
From "Build AI" to "Manipulated AI", 100+ battle cases to break down new engineering practices of the AI Native era!
2026 QCon Global Software Development Congress Shanghai Station will be held on 22-24 October to share the changes that AI applications, Agent Engineering, AI Coding, Infrastructure, Data Systems, Quality, Safety, SRE, and smartness are making after entering the real production environment.
Alibaba China Electric Enterprise, Flying Pig CTO, Dr. Chen Zheng, has confirmed to attend the keynote speech and presented a theme-sharing paper entitled " From Coding to Consumption: The Next Revolution in AI ".
AI, why did you learn to write code instead of taking out? One key reason is that Coding naturally owns a compiler, a unit test, an “Verifier” like CI: feedback is intense, immediate and objective, and the task is entirely in the digital world. By contrast, the feedback from the consumer scene is more thinner, delayed and subjective, and transactions and compliance go through the real physical world such as merchants, stockpiles, distributions, etc.
This means that AI is going from Coding to consumption, not just to connect a stronger model to consumption applications, but to re-engineer a set of authentication and enforcement closed loops for the real world. The presentation will start with "valitableness" to analyze why AI has evolved along the path of consumption and where the real difficulties of consumer AI are.
The "Infra x Data x Company" formula for Frontier Lab translates into a consumption scenario: how to build an AI-Native supply and service network, how to translate static and real-time consumption data into training assets, and how to train to establish continuous iterative data flyers. The core judgement is that the key to the next phase of AI is not just to make models smarter, but to turn the physical world ' s compliance process into a verifiable, learningable training signal.
In the end, the real opportunity in the second half is not to create another AI Chat, but to actually get AI into transactions, services and compliance, and to make consumption an important training ground leading to stronger universal intelligence.
Chen Shui, flying pig CTO, responsible for planning and implementing the flying pig ' s AI strategy. Doctorate in Information Systems and Computers, University of Wisconsin Madison, published more than 30 NeurIPS, SIGIR, KDD, WWW, and received three awards for the best dissertations from KDD, SIGIR, some of the results of which were integrated into postgraduate studies at Stanford University, University of California Berkeley; held 10 United States patents. An artificial intelligence company, TigerBob, was created, with a multi-modular large-linguistic model, TigerBot, on the first national LLM filing list; and was a senior vice president of the Mission Point Review, leading the construction of a search and impact advertising platform.
In addition, a number of experts from the academic and industry communities, such as Dr. Dai Kuo Ho, Co-founder and Chief Scientist, and CEO Han Xuan, Associate Professor/Unquestioned Dome at the College of Artificial Intelligence, Shanghai University of Transportation, will share different perspectives and key explorations of AI from the frontier to technological innovation to industrial applications.
The Congress also designed Loop Engineering, Agent Innovations, Agent Automation: From Memory to Continuous Learning, Agent as a Service, Vibe Coding's New Quality Debt, the SRE Reliability Project for Rational Management AI, Financial AI Native Works: From Research and Development to Business Fragmentation, AI Infra: Accomplishing Efficiency Decision Scales, AI Nature Architecture, and so forth.
More details can be found by searching or contacting the ticket manager for advice.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.