AI 應從用戶理解的位置開始
AI Should Meet You Where You Are
AI 協助理解陌生程式碼時,應從用戶理解受阻的位置開始,跟進一個具體例子,而非只提供更多名稱。文章以註冊流程為例,從按下按鈕開始逐步展示相關檔案,並在每一步停下回答問題;圖示則串連表單、請求、帳戶、佇列和電郵。
A detailed explanation can leave us with more names and no clearer picture.
Turtleand opened an unfamiliar codebase to change a welcome email. The AI explained routes, services, background jobs, and configuration. He still did not know where the email came from.
So he changed the request:
“Follow one signup with me. Start when someone presses the button. Show me the relevant file at each step, and pause so I can ask questions.”
A small diagram gave the filenames a place: form, request, account, queue, email. At the queue, he paused. Was this where the message was sent? No. This part added a job; another picked it up.
Better help starts where understanding stops. Sometimes that means a conversation. Sometimes a diagram. Sometimes one concrete example, followed slowly enough to notice what is missing.
Before editing, he explained the journey aloud. He still had questions, but now he knew where to ask them.
Try this next: Ask for one concrete example to be followed from beginning to end. Pause at the first step you cannot explain.
Based on AI Should Meet You Where You Are on Turtleand Growth. For keeping practice and judgment in assisted work, see AI Is an Electric Bicycle. Keep Pedaling..
來源:DEV Community · dev.to