Interactive Rubber Ducking With GenAI
Interactive Rubber Ducking is a technique that uses a generative AI as a questioning partner. The AI asks the developer one question at a time, helping them explain their design ideas step by step.
The process works like this:
The developer starts with a prompt that tells the AI to ask a single question.
Each question and answer is saved in qa.md.
When the session is finished, the AI combines all the Q&A into a spec.md document.
With models such as Claude Opus, the AI can also look through the codebase, pull in relevant documentation, and suggest cache‑related changes for the Pongo library.
The purpose is not to replace the developer’s creativity. Instead, the AI helps to:
Spot things the developer might have missed.
Check the assumptions being made.
Create a clear, repeatable specification.
Throughout the whole process, the developer stays in control.