lv outlines an AI-native approach to product delivery
In the second installment of his “Becoming an AI Native Builder” series, lv defines an AI Native Builder as one person working with AI to cover the full loop from requirements and design through development, testing, deployment and post-launch feedback. He says agents such as Claude Code, Codex and OpenCode handle execution, while people decide what is correct, approve key decisions and remain accountable for the outcome. lv also says he built CanvasAgents from zero to one using this methodology and uses it as a main tool that he plans to keep improving.
The essay presents an author-defined workflow for combining AI execution with human judgment across the entire product-delivery lifecycle and uses CanvasAgents as his own example of applying that workflow.
这是「成为 AI Native Builder」连载的第 2 篇，全系列目录见文末。
哈喽，我是 lv。
上一篇开篇，我提到了一个词：AI Native Builder。
这个词，是我给自己现在这个角色起的名字。
它不是一下子冒出来的，背后是我这几年一点点被"推"过来的路：
- 几年前我是前端，开始琢磨 AI 能不能进入前端开发的日常，还写了一本小册《AI 赋能前端研发从 0 ～ 1》；
- 后来和团队的设计师合作，把 AI 驱动的 Design To Code 落到团队流程里，我也从前端走成了 Design Engineer；
- 再后来开始做 Flutter 多端应用，把 AI 带进 Flutter 的日常开发；
- 现在做需求交付工程师，需求梳理、设计、前端、移动端、后端、测试……一个人身兼多个角色。