Tackling your tech debt with Copilot coding agent
Quick Overview
Brittany Elich from GitHub details how to effectively tackle technical debt using GitHub Copilot Coding Agent by employing the WRAP methodology: Write effective issues, Repository instructions, Atomic tasks, and Pairing with the agent, which ultimately transforms overwhelming tech debt mountains into manageable terrain.
Key Points: GitHub's Brittany Elich presented a strategy for managing tech debt using Copilot Coding Agent to turn 'Tech Debt Mountains' into 'Managed Terrain'. The WRAP methodology for backlog management includes: Write effective issues, add Repository instructions, break work into Atomic tasks, and Pair with the coding agent. Effective issues should have descriptive titles and include ample context, such as examples of desired changes, to guide Copilot. Repository instructions allow engineers to specify how Copilot should interact with the application context, which is often too large to fit into issue prompts. The speaker notes that humans excel at understanding 'why,' navigating ambiguity, and cross-system thinking, while Copilot excels at expanding patterns, tireless execution, and exploring possibilities. An example demonstrated creating comprehensive tests for a button component, which took Copilot 8-10 minutes, saving the engineer significant time. The long-term goal is to intentionally schedule large blocks of development time for tech debt work, allowing engineers to focus on strategic planning instead of constantly reacting to inherited debt.
Context: The presentation by Brittany Elich, Senior Software Engineer at GitHub, focuses on strategies for proactively managing technical debt, which often gets neglected in favor of new feature development. Elich frames the challenge as overcoming 'Tech Debt Mountains' that slow velocity and demotivate engineers. She introduces a structured approach, leveraging GitHub Copilot Coding Agent as a powerful tool to accelerate the remediation of this debt by handling repetitive and detailed tasks.