Coding Agent 为何仍难用
推荐指数 61.0 NO. 005 · 2026.10.10
发布2026/10/09Score52Comments24
为什么值得看
作者实测发现 Coding Agent 在文件编辑、错误修复等基础功能上体验糟糕，频繁崩溃且误判任务完成度。这揭示了当前 AI 编程工具的核心瓶颈不在模型能力，而在工程化交互设计。
编辑判断
这篇文章戳中了一个被回避的行业现实：GPT-4 级别的模型能力并没有自动转化为 GPT-4 级别的产品体验。Cursor 和 Windsurf 的火爆让很多人误以为 Coding Agent 已经成熟，但实际上多数团队卡在上下文管理、工具调用可靠性和长任务状态追踪这三座山上。
更值得警惕的是，这个问题可能被结构性低估——模型厂商在卷 benchmark，IDE 插件团队在卷功能清单，但中间层的"可靠执行层"缺乏标准定义和评估体系。如果你正在做 AI 编程工具，不要只测模型准确率，要测 10 轮连续编辑后的任务完成率；如果你在用现有工具，建议把 Agent 限制在单文件、短周期的重构任务，而非多文件架构设计。
社区反馈
意见分歧 24 条评论
核心争论：瓶颈在模型能力还是工程化设计？社区经验两极分化
Enjoyed this article, lots I can relate to. One thing that seems to help for me is to do the docs before plans (collaboratively edit with agent). Then I understand what this change is going to look like from the user's perspective before we start implementation. This seems to help keep things on tra
It's true! It feels like we've been talking about harness optimisation and 'cool features' available in the cli tools for months at this point, but ostensibly there has not really been any significant upgrades to these harnesses since at least Claude Code imo. It does feel like a contrived way to ha
I used to relate to this article quite a bit. In the last 3-4 months, not so much. I've found that the latest models -- Opus 5.5, Astra, etc. juggle multiple tasks, delegate exceptionally well and are very good at working independently. I still occasionally have issues with open-weight models, but t