How to chat with your repo & create PRs with Copilot | GitHub Checkout

Key Concepts

• GitHub Copilot on github.com: An immersive AI-native experience within the GitHub website.
• Issue and Pull Request Creation: Using Copilot to generate issues and pull requests directly from screenshots and descriptions.
• Copilot Spaces: Accessing Spark and Agents within the Copilot interface.
• Agentic Dimensions: Assigning Copilot to solve issues autonomously.
• Prototyping: Quickly creating and iterating on code snippets and website components for learning or production purposes.
• Model Selection: Choosing different AI models (e.g., GPT-4.1, Claude Sonnet 3, Opus 4, o3-mini) within Copilot Chat for varied responses.
• Threads (Multiverse): Maintaining separate conversation contexts with different models to avoid pollution and improve results.
• Integration with GitHub Platform: Leveraging existing GitHub features and entities (issues, pull requests, discussions) within Copilot.
• AI-Native Orchestration: Using natural language to manage and iterate on GitHub resources.

1. GitHub Copilot on github.com: An Immersive Experience

• Daniel introduces github.com/copilot as a comprehensive AI-native experience within the GitHub website.
• It allows users to chat with Copilot, interact with files, issues, and pull requests, and create new issues and pull requests.

2. Issue Creation Example

• Problem: An overlapping arrow icon in a screenshot.
• Process:
  1. Upload the screenshot to Copilot.
  2. Provide a brief description: "Please create an issue... and describe what the issue is."
  3. Copilot automatically generates an issue with a title and description based on the image and context.
• Details: Copilot can utilize existing issue templates if available in the repository.

3. Copilot Spaces and Agentic Dimensions

• Copilot interface includes access to Spaces (including Spark) and Agents.
• Agentic Scenario: Assigning Copilot to an issue to solve it autonomously.
• This represents a new dimension of Chat and agentic capabilities within GitHub.

4. Control Center for Copilot

• github.com/copilot acts as a central control panel for all Copilot-related activities.
• Users can retrieve information from their account, such as a list of open issues assigned to them.

5. Prototyping and Learning

• Copilot can create quick prototypes for small components or website sections.
• Example: Using Copilot to understand CSS variables by creating an interactive prototype.
• Users can tweak the code, preview changes, and share the prototype with colleagues for collaboration.

6. Iterating on Text and Markdown

• Copilot supports iterating on larger blocks of text.
• It can create markdown files and render mermaid diagrams.
• All existing GitHub capabilities are integrated into Copilot Chat.

7. Model Selection and Threads (Multiverse)

• Copilot allows users to choose from various AI models (GPT-4.1, Claude Sonnet 3, Opus 4, o3-mini).
• Different models may provide different answers based on their interpretation of the data.
• Threads: Enables maintaining separate conversation contexts (universes) with different models.
• This prevents context pollution and allows users to explore different solutions without interference.
• Users can switch between threads to compare results and continue conversations with the most relevant context.

8. IDE vs. github.com Chat

• IDE: More context-aware due to the open repository, focuses on the current codebase, and offers powerful agentic mode.
• github.com Chat: More open-ended, has broader context from the GitHub platform, and provides easy access to GitHub entities (issues, pull requests, discussions).
• The choice depends on the user's current workflow and needs.

9. Advantages of AI-Native Experience on GitHub

• Seamless integration with the existing GitHub platform.
• Future enhancements will deepen the connection between AI and GitHub.
• The goal is to enable users to orchestrate everything on GitHub using natural language in a chat interface.
• Capabilities will expand beyond basic issue and pull request creation to more complex tasks.

10. Conclusion

• GitHub Copilot on github.com offers a powerful AI-native experience that integrates seamlessly with the GitHub platform.
• It provides tools for issue and pull request creation, prototyping, collaboration, and model selection.
• The platform is continuously evolving to enable users to manage and iterate on GitHub resources using natural language, making it an invaluable tool for developers.