AsiaInfo reports 14.6% per-developer code output gain from company-wide AI Coding
A TMTPost article excerpted from Alibaba Cloud's Yunqi Strategy Reference describes how AsiaInfo put nearly 10,000 developers on company-wide AI Coding with support from Alibaba Cloud's full-stack AI R&D tools, using Qoder to extend from code completion to requirements, design, testing and defect fixing. According to AsiaInfo data-intelligence chief architect Zhu Jun, AI covers about 80% of its core BSS systems, monthly active use among a thousand-person operations team stays above 90%, per-developer code output rose about 14.6% and end-to-end efficiency from requirements to testing improved 8.6% on average — figures supplied by the company and not independently verified in the article. The piece also notes Qoder's newly launched Qoder Wake, which runs defect-locating, fixing and final-testing agents on separate machines that share context, an early form of what AsiaInfo calls a digital employee.
It is one of the few public accounts of a near-10,000-developer organization shifting to agent-assisted development, though the efficiency figures are self-reported and lack independent measurement.
本文摘自《云栖战略参考》，这本刊物由阿里云与钛媒体联合策划。目的是为了把各个行业先行者的技术探索、业务实践呈现出来，与思考同样问题的“数智先行者”共同探讨、碰撞，希望这些内容能让你有所启发。
当AI Coding 从程序员的辅助工具演变为重构研发体系的核心力量，当“人均 AI 助手”从口号变为研发团队的日常，中国软件产业的 AI 化进程，正在悄然越过一个新的分水岭。
但 AI 带来的改变，不仅仅是写代码更快一点。代码层面的提效只是序章，真正的深层变革，是软件研发与应用范式的整体性重构，是向 AI Native 的代际跃迁。这意味着改变不是在原有软件架构上叠加 AI 功能的局部优化，而是以通用人工智能为核心根基，重新定义软件的基础设施、开发方式与应用形态——软件产业正在从 “人驱动代码”的旧时代，步入“智能体驱动系统”的新阶段。