Rubber duck debugging in VS Code & GitHub Copilot CLI

Burke Holland and Reynald Adolphe demonstrate how “rubber duck debugging” can be applied inside Visual Studio Code using GitHub Copilot CLI, by bringing in a second model family to review and challenge the first at key boundaries in the workflow.

Overview

The video explains the classic idea of rubber duck debugging (talking through your code to surface mistakes) and maps it to an AI-assisted workflow:

• Use GitHub Copilot CLI in VS Code as a structured way to:
  • Plan work
  • Implement changes
  • Validate with tests
• Introduce a second model family to review and challenge the first model’s output at key checkpoints.

What “combining models” means here

The approach described is to avoid relying on a single model’s output end-to-end:

• One model produces an initial plan or implementation.
• A second model family is used to:
  • Review the plan for gaps
  • Challenge assumptions
  • Spot edge cases
  • Push for test coverage and verification

Demo focus: boundaries to challenge

The demo is framed around inserting review points where mistakes commonly slip through:

• Planning boundary: sanity-check the approach before writing code
• Implementation boundary: challenge correctness and completeness while coding
• Testing boundary: verify behavior and look for missing tests or scenarios

Chapters (from the video description)

• 00:00 Intro to rubber duck debugging
• 01:04 Combining models
• 04:00 Rubber duck debugging: what it is
• 06:20 Demo
• 16:20 Closing