I Let GitHub Copilot Code FOR ME (It Fixed Its Own Bugs!)
GitHub Copilot is evolving—now with an innovative Agent Mode that’s currently in preview. This new feature goes beyond simply generating code by enabling Copilot to iterate on its own output and even fix its own code. When bugs occur, Copilot automatically refines its code, addressing runtime errors and completing any necessary subtasks. It can also suggest terminal commands, much like other coding assistants such as Cursor.
Key Features
• Self-Healing Code:
Agent Mode allows Copilot to continuously evaluate its code. When an error or bug is detected, it iterates on its output until the issue is resolved.
• Iterative Improvements:
Beyond the initial code generation, Copilot now refines its work based on runtime feedback and terminal output. It infers additional tasks necessary to fulfill your prompt completely.
• Terminal Command Suggestions:
Similar to Cursor, Copilot can recommend terminal commands, streamlining your workflow by integrating code suggestions with command-line actions.
Additional Enhancements
GitHub has rolled out several other features alongside Agent Mode:
• Copilot Edits:
A new interface supports multi-file editing, making it easier to apply changes across a project.
• Next Edit Suggestions:
This feature offers code suggestions based on your recent edits, similar to Cursor’s tab-based navigation.
• Prompt Files:
Teams can now build, store, and share reusable prompts. These files help maintain consistency and streamline workflows by predefining instructions for Copilot.
• Vision Capabilities:
In the Visual Studio Code Insiders edition, you can attach images to your chats—a feature still in preview.
Getting Started with VS Code Insiders
To take full advantage of these new features, you’ll need to install Visual Studio Code Insiders along with the GitHub Copilot extensions (Copilot Chat and GitHub Copilot). Once installed, sign in with your GitHub account, and you’re ready to explore both the chat and Agent modes.
A Hands-On Look
In a live demonstration, a developer prompted Copilot to create a Next.js app integrated with the OpenAI SDK. Copilot executed terminal commands, set up the project, and even corrected errors along the way. The tool also supports voice commands via a VS Code speech extension, providing an alternative to typing. Although some aspects—such as the lack of a detailed code history and auto-acceptance options—could be improved, the overall speed and efficiency of Agent Mode were impressive.
Pricing and Final Thoughts
GitHub Copilot offers a free tier that includes up to 50 AI interactions per month and real-time code suggestions for approximately 2000 lines of code. The Pro version is available at a modest cost (roughly the price of two coffees or sandwiches) and is free for verified students, teachers, and maintainers of popular open source projects.
While there is room for enhancements (like more robust history functionality), the new Agent Mode is a significant step forward in AI-assisted coding. It not only increases productivity but also reduces errors by iterating until the code fully meets your requirements.
Have you tried GitHub Copilot’s Agent Mode yet? Share your experience in the comments, and stay tuned for more updates on AI-powered development tools!
00:00 github copilot
00:15 Agent Mode: Self-Healing and Iterative Improvements
00:54 Demo: GitHub Copilot in Action
00:57 New Features in GitHub Copilot
02:17 Project Padawan: A Sneak Peek
03:24 Setting Up Visual Studio Code Insiders
04:05 Hands-On with GitHub Copilot
04:10 demonstration
09:29 Pricing and Final Thoughts
10:34 Conclusion and Subscrib