Rubber Duck Thursdays - Build for the love of code

Key Concepts

• GitHub Actions: Cloud-based CI/CD service for automating software workflows.
• GitHub Issues: System for tracking tasks, bugs, and feature requests within a repository.
• GitHub Copilot: AI pair programmer that offers code suggestions and autocompletion.
• GitHub Spark: A new interface for building applications using natural language, currently in public preview for Copilot Pro Plus subscribers.
• Model Context Protocol (MCP): A protocol that allows AI agents to interact with external tools and services.
• MCP Server: A server that implements the Model Context Protocol, providing access to specific functionalities.
• Copilot Coding Agent: A feature that allows Copilot to perform tasks autonomously in the background.
• Custom Instructions: Instructions that can be provided to Copilot to influence its behavior and code generation.
• Dependabot: A service that automatically creates pull requests to update dependencies in a repository.
• Dev Container: A containerized development environment that provides a consistent and reproducible environment for coding.
• SAR (Static Analysis Results) Files: Files containing the results of static code analysis, used by code scanning tools.

GitHub Updates and Announcements

GitHub Actions: M2 Pro Powered Hosted Runners

• GitHub Actions now offers M2 Pro-powered macOS hosted runners in public preview.
• These runners provide performance gains and can reduce build times.
• The runners are equipped with a 5-core CPU, 8-core GPU, 14 GB of RAM, and 14 GB of storage.

GitHub Issues Updates

• Users can now easily duplicate GitHub issues.
• The destination repository can be changed before creating an issue.
• Issues can be created from anywhere in the GitHub UI using a new modal in the top right corner.
• API support updates and additional improvements have been made.

Deprecation of Azure Endpoints for GitHub Models

• Azure endpoints for GitHub models are being deprecated.
• The recommendation is to use the models.github.ai endpoint.
• The old endpoints will be temporarily supported until October 17th.

Pull Request Files Changed Public Preview Updates

• Commenting on individual files, including binary, deleted, and renamed files, is now supported.
• Performance has been improved by reducing unnecessary re-renders and improving responsiveness.
• Additional quality-of-life updates have been implemented.

Upcoming Applications and Changes to Copilot Code Review

• The "coding guidelines" setting is being replaced with Copilot instructions.
• The automatic Copilot code review checkbox is now easier to find and use in repository rule sets.
• A dedicated enterprise organizational policy is being introduced.

New GitHub Copilot Activity Report

• A new GitHub Copilot activity report provides enhanced authentication and usage insights.
• The report includes authentication tracking, details on the last surface used, IDE version, and data freshness.
• The report refreshes every 30 minutes instead of daily.
• Comprehensive coverage across Copilot features (IDE, g.com, CLI, and mobile) is included.

GraphQL API Rate Limits

• Timeouts are now included as part of the primary read limit for the GraphQL API.
• Normal usage is unlikely to be affected, but integrations may hit timeouts earlier.

GitHub Copilot App Modernization for .NET

• GitHub Copilot App Modernization for .NET is now in public preview.
• It uses AI agents to remediate code and modernize applications in days and weeks instead of years.
• It supports migrating between different databases, storage options, and identity providers.

Code Scanning Updates

• Code scanning will stop combining multiple SAR runs uploaded in the same SAR file.
• The concept of "category" is being introduced for SAR files.
• These changes will be available in Enterprise Server 3.18 and on GitHub.com from July 22nd.

Centralized Private Registry Configuration for Dependabot

• Centralized private registry configuration for Dependabot is now generally available.
• Organization admins can centrally manage private registry access.
• Organization-level configuration serves as defaults, while preserving flexibility for repository-specific needs.
• The feature is available for all public repositories, and private repository support is available for Advanced Security customers.

Adding a "Not Set" Option for GitHub Code Security Features

• Organization security configurations now have a "Not Set" option in addition to "Enabled" and "Disabled."
• This allows repository admins to decide whether to enable or disable specific features.

GitHub Copilot in Eclipse Updates

• GitHub Copilot in Eclipse has received several updates, including:
  • Smarter commit message generation
  • Enhanced context understanding
  • Revamped getting started experience
  • Support for Eclipse 2024-03 and 2024-06
  • One-click commit message generation
  • Workspace and context in chat
  • MCP logs in console view

Secret Scanning Adds Validity Checks

• Secret scanning now includes validity checks for over 40 secret detectors.
• Validity check support has been added for 45 additional token types across over 30 providers.

GitHub Apps Can Now Review Secret Scanning Bypass Requests

• GitHub Apps can now review secret scanning push protection bypasses and alert dismissal requests.

GitHub Spark Public Preview

• GitHub Spark is now in public preview for Copilot Pro Plus subscribers.
• It provides a native experience for building apps using natural language.

Copilot Coding Agent Supports Instructions.md

• Copilot Coding Agent now supports instructions.md files for custom instructions.
• This allows for modularizing and segmenting instructions based on file paths.
• github/copilot-instructions file is the global one.
• Separate instruction MD files can be brought in manually or automatically based on file extensions or folders.

Copilot Coding Agent Base Branch Selection

• The Copilot Coding Agent now allows users to select the base branch for tasks.
• This enables users to delegate tasks on branches other than the main branch.

Blog Post Recap: Debugging UI with AI

• The blog post recaps a previous Rubber Duck Thursdays episode where GitHub Copilot agent mode and the Playwright MCP server were used to debug UI issues in the Octo Arcade app.
• The post includes the stream recording, useful links, and a walkthrough of the debugging process.
• It highlights the importance of clear requirements and well-formed prompts when using AI tools.
• The post provides practical tips and examples of prompts used during the debugging session.

Love of Code Hackathon

• GitHub is hosting a hackathon called "Love of Code" from July 16th to September 22nd.
• The hackathon has six categories:
  • Buttons, Beeps, and Blinking Lights
  • Agents of Change (AI-powered experiences)
  • Terminal Talents
  • Game On
  • Worldwide Wonders (web apps)
  • Everything But the Kitchen Sink (niche projects)
• Participants can win prizes.
• The stream brainstormed ideas for projects in each category, focusing on fun and whimsical concepts.

Building an MCP Server: Rubber Duck Debugging Adventure

• The stream attempted to build a rubber duck debugging MCP server in the form of an adventure game.
• The goal was to create a gamified experience for solving bugs, with the user as the hero of a quest.
• The stream used Copilot to scaffold out the project, generate code, and create user stories.
• The MCP server was designed to provide guided questions and encouragement to help users solve their own problems.
• The stream explored the use of prompts, resources, and tools within the MCP server.
• The stream also experimented with GitHub Spark to create a UI for the rubber duck debugging experience.

Key Takeaways

• GitHub is continuously releasing new features and updates to improve developer productivity and collaboration.
• GitHub Copilot is becoming increasingly powerful and versatile, with new features like Copilot Coding Agent and GitHub Spark.
• The Model Context Protocol (MCP) provides a way to extend the capabilities of AI agents by connecting them to external tools and services.
• The Love of Code hackathon offers a fun and creative outlet for developers to build interesting projects.
• Clear requirements and well-formed prompts are essential for successful AI-assisted development.