让几个 AI 同时开发,我会先拆任务之间的依赖。
Worktrunk 的介绍是管理独立工作目录,方便多个 AI 并行干活。假设要做一个预约页面,页面样式和确认邮件可以分开写;但预约字段还没定,我会先把字段定下来,再让两边开工。
目录分开了,两边仍可能对同一个预约理解不同。我还没亲测这个工具,这只是我的拆法,不确定能省多少时间。
https://t.co/BpYNyLiiNh
@melvindvivas I'd start with one small change that already has a clear test, then check the diff and run that test before expanding the task. Do you find Spark more useful for those bounded edits or for exploring an unfamiliar codebase? I haven't compared the two yet.