AI编程代理为何仍离不开人类
推荐指数 70.0 NO. 010 · 2026.09.18
发布2026/09/17Score64Comments61
为什么值得看
当前AI代理虽能完成游戏开发和代码迁移等单次 impressive 的任务，但无法自主构建完整软件系统。作者指出上半年大规模部署代理的尝试结果令人失望，呼吁行业重新思考所需的基础原语。
编辑判断
这篇文章的尖锐之处在于戳破了一个行业幻觉：把现有LLM堆成"代理大军"并不能自动产生可靠工程产出。上半年很多团队用Cursor+Claude循环生成代码，结果得到的是表面能跑、内部脆弱的"纸牌屋"系统。
真正卡住的不是模型能力，而是缺乏"可验证的中间抽象"——人类工程师的价值恰恰在于知道在哪里设置检查点、如何定义完成标准。作者暗示的下一个突破点可能是形式化验证与LLM的结合，而非更聪明的单轮生成。
如果你在用AI代理做大型重构，现在最该补的不是prompt技巧，而是把测试覆盖率、类型系统和架构约束做成代理能消费的机器可读规范。
社区反馈
意见分歧 59 条评论
核心争论：AI代理无法自主构建完整系统的根源是环境复杂度还是缺乏自我纠错的学习机制
My bet is that we'll see a second layer of harness emerge, as self-driving codebases become the target. There will be an application facing harness, orthogonal to the agent facing harness. The app harness will represent the software factory that is emergent for the specific application being develop
I suspect it won't be a harness, but just a more specific LLM trained in the universe of user-selected context of vetted resources. Why? Because LLMs are always going to be dumb when they're trained at scale. Their ability to speak software diverges from their friendly user input layer. A harness wo
fine-tuning may be a more scalable approach to LLM personalization than sending all the same context to two LLMs I'm working towards both in my homelab to see which works better with little qwen