Rubber Duck Thursdays - Let's build

Key Concepts

• GitHub Updates: Recent changes and features released by GitHub.
• Copilot: AI pair programmer offering code suggestions, code completion, and code review.
• Copilot Coding Agents: AI agents that can be assigned tasks and create pull requests.
• GitHub Models: A platform for trying out and using AI models in applications.
• GitHub MCP (Model Completion Provider) Server: A server that provides access to AI models for code completion and other tasks.
• Chat Modes: Different modes in Visual Studio Code chat, including ask, edit, agent, and custom modes.
• Internationalization (i18n): Designing and developing applications to be adaptable to different languages and regions.
• Localization (l10n): Adapting an internationalized application for a specific language or region.
• Prompt Engineering: Crafting effective prompts for AI models to achieve desired results.
• Dev Containers: A containerized development environment that can be used with VS Code.
• GitHub Actions Runner Controller: A tool for running self-hosted runners on Kubernetes.
• CodeQL: A code analysis engine for identifying security vulnerabilities.
• Secret Scanning: A feature that detects secrets in code and alerts users.
• Artifact Attestations: A mechanism for verifying the integrity of software artifacts.
• Open Policy Agent (OPA) Gatekeeper: A policy engine for Kubernetes.

GitHub Updates (June 5th - Present)

• Model Evaluations with Models CLI: Allows evaluating prompts against criteria on the command line, enabling automation and integration with CI. Results can be outputted in JSON format.
• GitHub App Manager Granted to Teams: Enables assigning permissions for GitHub App Manager to teams.
• Private Registries for Go CodeQL Scans: CodeQL can now access private dependencies stored in private registries for Go projects.
• CodeQL Adds Swift Support: CodeQL 2214 adds Swift 611 support, and CodeQL 22220 improves coverage for Go and adds support for Swift 612.
• Delegated Alert Dismissal for Secret Protection: Admins can restrict who can dismiss secret scanning alerts, requiring requests for dismissal from others.
• Scheduled Reminders for Pull Requests: Schedule reminders on Slack and Microsoft Teams for pull requests.
• Remote GitHub MCP Server in Public Preview: GitHub hosts a version of the MCP server, eliminating the need for local Docker instances. Supports OAuth 2 for authentication.
• AI Prompt Editor and Evaluations Tooling Supports Multi-Turn Conversations: Allows evaluating prompts in the context of a conversation.
• New User Dashboard Refreshed for Enhanced Onboarding: New users will discover get for beginners playlist getting started tasks actional do actionable docs covering some fundamental concepts there as well.
• Actions Runner Controller N20 120 Release: New release with Red Hat Open Shift public preview ma support for arc secrets updates to DND, fail pod retry mechanism rolling updates and metrics.
• Copilot Code Review Supports Custom Instructions: Copilot can be steered in the review process using a custom instructions file.
• VS Code May Release for GitHub Copilot: Includes prompts and resources support, authentication support for remote MCP servers, development mode debugging capabilities for Node and Python MCP servers, extension APIs to publish and manage collections on MCP servers, agent mode and chat enhancements, smarter coding assistance, AI-powered settings search, source control and Z workflow improvements, copilot coding agent integration.
• Organization Rule Sets Now Available for GitHub Team: Repo rule sets are now available in GitHub team plans.
• Mean Time to Remediate Metric for Code Pull Request Alerts: Included on the security overview dashboard.
• Latency Graph Now Defaults to Off: Newly created public repos will fault to off gun disabling and removing store data for inactive repositories for inactive blah blah blah blah blah.
• Agent Modes Generally Available with MCP Tools Support in Visual Studio: Use multiple custom variables in AI evaluations tooling great.
• Updates to GitHub Copilot Consumptive Billing Experience: Monthly allowance of premium requests per user copilot pro pro plus business and enterprise plans now enforced.
• Getting Started with Copilot Coding Agent is Now Easier: Users can delegate tasks by assigning issues etc making it easier to get started.
• Copilot Spaces Now Support Issues and Pull Requests: I really like copilot spaces actually.
• Upcoming Deprecation of 01 GPT45 03 mini and GPT40: GitHub copilot will deprecate the following models in github.com/copilot and the IDE model pickers GT45 open 03 mini and GPT40 the recommendations there use GT41 use 03 04 mini and GPT41 so those older models getting phased out there by the looks of it.
• Dependable Compute Migration to GitHub Actions: That has now I think happened.
• Enterprise Owners Can Now Discover and Manage Repositories Owned by Enterprise Managed Users: Okay great that kind of discoverability piece there.
• Enforce Mission Policies with Artifact Access Stations in Kubernetes Using the Open Policy Agent Gatekeeper: So again if you're in the Kubernetes space this is a great update for you.
• GitHub Desktop 35 Copilot Commit Message Generation Now Generally Available: Filter based rule set targeting let's take a look at that ah okay nice the visibility private props semi custom properties security tiers high security not including PHP languages so now we can go and uh yeah match by filters so we previously could only select by uh matching a name specifically selecting repos or selecting all repos but now we've got this additional layer of um flexibility to go and uh match by filters so great good to see that.
• Metrics Page Helps GitHub Advanced Security Code Security Users Prioritize Their Vulnerabilities for Remediation: Great yeah I really like this actually so you can go and uh look at this graph now so you can see right how many alerts have we got how many actually have a patch how many have a patch but the severity is high and by the way um the exploit prediction score is also um high as well and then that allows you to start prioritizing some of that technical debt there so um really nice to see that one.
• Cleansbot Dependabot Support for Gradal Lock Files is Now Generally Available: Secret scanning rest API responses include first location detected and had more locations now G uh security updates for apps and API access bear with me my uh.
• Security Updates for Apps and API Access: GitHubs enforcing two new security improvements into GitHub apps no off apps that are private or owned by enterprise managed user accounts now restricted unless you are a member of the organization or enterprise that owns the application makes sense second technical definition of internal visibility and calculating access an SSO authorized token has within a multi-organization enterprise but if you're um you know using GitHub apps and an or enterprise that is probably one that you want to go and take a look at yes.
• Copilot Coding Agent is Now Available for C-Pilot Business Users: So uh was previously only available to Copilot enterprise users um now they have opened it up to Copilot business subscribers once an administrator has enabled the Copilot coding agent policy users can assign issues to Copilot just as they would another developer or ask Copilot to open a pull request from chat there we go and yeah the agent consumes GitHub minutes and copilot premium requests so yeah if you're on C-Pilot business um ask your admin to go and take a look at that policy.
• Github Models Now Has Support Moving Beyond Free Limits: This is what I mentioned a bit earlier um so whereas before every user had like an allocation which is still there um now you can go beyond that either with a bring your own key which is kind of cool um or you can do pay as you go billing uh with GitHub then as well so yeah this is now something that you can start uh using uh to kind of scale and start using that in your apps so good to see that oops.
• Configuring Which Secret Scanning P Patterns are Included in Push Protection: So if you need that kind of granularity and you're using secret scanning push protection uh or well yeah then that's good.
• Anthropic Claude Sonic 4 and Claude Opus 4 are Now Generally Available in Git C-pilot: Good to see that then improved attachments and larger context in co-pilot chats in public preview oh that was a lot that was a lot the last uh two weeks have clearly uh been busy with uh with the team there.

Internationalization of Copilot Airways

• Goal: Add internationalization support to the Copilot Airways web and mobile apps, supporting English, French, and Spanish.
• Web App (Next.js):
  • Issue Creation: Used GitHub Copilot in github.com to create a GitHub issue for adding internationalization.
  • Copilot Coding Agent: Assigned the issue to Copilot Coding Agent, which created a pull request with initial implementation.
  • Implementation Details:
    • Added a language selector component to the profile page.
    • Implemented a useTranslation hook for managing translations.
    • Created JSON files for English, French, and Spanish translations.
    • Updated the readme with instructions for adding new languages.
  • Limitations:
    • Not all pages were fully translated.
    • Some hardcoded strings were not updated.
    • Live refresh of translations was not working correctly.
  • Iteration: Prompted Copilot to translate additional pages, such as the flight search and reservation pages.
• Mobile App (Swift/iOS):
  • Copilot Agent Mode in Xcode: Used Copilot Agent Mode in Xcode to implement internationalization.
  • Implementation Details:
    • Added a language selector to the profile view.
    • Updated the main tab view to support internationalization.
  • Bug: The main tab was not updating correctly.

Custom Chat Modes in Visual Studio Code

• Exploration: Explored the new custom chat modes feature in Visual Studio Code.
• Planning Mode: Created a custom chat mode called "Planning" for generating implementation plans.
• Configuration: Configured the Planning mode to use the GitHub MCP server and other relevant tools.
• Prompt Engineering: Iterated on prompts to effectively use the MCP tool for creating GitHub issues.
• Challenges: Encountered issues with the prompt not being clear enough for Copilot to use the MCP tool.
• Resolution: Refined the prompt and reloaded the window to ensure the changes were applied.

Examples and Applications

• Smart Home Solution: Mentioned a user working on a smart home solution and suggested looking into Home Assistant.
• Smart City Solution: Discussed a user working on a smart city solution and the potential benefits of connected traffic systems.
• AI in Applications: Highlighted the use of GitHub Models for integrating AI into applications and the ability to evaluate prompts.
• Kubernetes: Discussed the actions runner controller and artifact attestations in Kubernetes.

Step-by-Step Processes

1. Creating a GitHub Issue with Copilot:
   • Open the Copilot chat in github.com.
   • Provide a detailed description of the desired feature or change.
   • Specify requirements, acceptance criteria, and relevant files.
   • Assign the issue to Copilot.
   • Review the generated pull request.
2. Creating a Custom Chat Mode in Visual Studio Code:
   • Run the "Chat: New Mode File" command in the command palette.
   • Choose to create the file in the workspace or user profile.
   • Enter a name for the chat mode.
   • Configure the tools to be used by the chat mode.
   • Write the description and instructions in the chat mode file.
3. Implementing Internationalization:
   • Add a language selector component.
   • Implement a translation hook or library.
   • Create translation files for each supported language.
   • Update the UI to use the translated strings.

Key Arguments and Perspectives

• Importance of Prompt Engineering: Emphasized the need for clear and iterative prompt engineering to effectively use AI tools.
• Benefits of Copilot Coding Agents: Highlighted the ability to delegate tasks to Copilot Coding Agents and parallelize work.
• Value of Custom Chat Modes: Showcased the potential of custom chat modes for tailoring the chat experience to specific scenarios.
• Continuous Learning: Stressed the importance of continuous learning and adapting to new technologies.

Notable Quotes

• "Every day is a school day."
• "So important to think about how you write your prompts and how you're wording things."

Technical Terms and Concepts

• CI/CD (Continuous Integration/Continuous Delivery): A set of practices for automating the software development process.
• MVP (Minimum Viable Product): A version of a product with just enough features to satisfy early customers and provide feedback for future product development.
• OAUTH 2: An authorization framework that enables applications to obtain limited access to user accounts on an HTTP service.
• JSON (JavaScript Object Notation): A lightweight data-interchange format.
• Kubernetes: An open-source container orchestration system.
• Docker: A platform for developing, shipping, and running applications in containers.
• React: A JavaScript library for building user interfaces.
• TypeScript: A superset of JavaScript that adds static typing.
• Swift: A programming language developed by Apple for iOS, macOS, watchOS, and tvOS.
• Xcode: An integrated development environment (IDE) for macOS.

Logical Connections

• The discussion of GitHub updates led to the exploration of new features like custom chat modes and Copilot Coding Agents.
• The internationalization project served as a practical example for demonstrating the use of Copilot and custom chat modes.
• The challenges encountered during the internationalization project highlighted the importance of prompt engineering and iterative development.

Data, Research Findings, or Statistics

• Mentioned that the VS Code release 1.101 introduced custom chat modes.

Synthesis/Conclusion

The session provided a comprehensive overview of recent GitHub updates and demonstrated the practical application of new features like custom chat modes and Copilot Coding Agents. The internationalization project showcased the potential of these tools for accelerating development and improving code quality. The session also emphasized the importance of prompt engineering and continuous learning in the ever-evolving landscape of software development.