In this episode of the VS Code Insiders podcast, hosts James Monttoagno and Pierce discuss the recent GitHub Universe 2025 conference and delve deeply into the evolution of Visual Studio Code (VS Code) as an AI-native editor. Pierce shares his experience attending GitHub Universe for the first time, highlighting the developer-centric energy and passion at the event, which contrasts with more corporate-feeling Microsoft events. They emphasize the excitement around VS Code, GitHub Copilot, and the ongoing improvements to github.com, underscoring the importance of the conference for developers and their plans to attend future events.
A key topic explored is the concept of VS Code as an AI-native editor. Pierce explains that VS Code has evolved from a simple code editor into a platform deeply integrated with AI capabilities, such as GitHub Copilot and various AI-powered features embedded throughout the editor, terminal, and source control views. This integration is designed to empower developers by enhancing productivity and embracing an abundance mindset in coding. The team has also open-sourced significant parts of the AI functionality, including the GitHub Copilot Chat extension, to foster community collaboration and transparency.
The podcast dives into the behind-the-scenes development of new AI features, particularly focusing on the planning agent mode. This mode helps developers plan features more effectively by breaking down complex tasks into manageable sub-agents that perform specialized research and return concise, relevant information. This approach improves response quality and context management within the editor. The hosts discuss how this planning mode fits between high-level project planning and in-turn task management, enabling smoother workflows and better integration with source control and implementation phases.
Another major highlight is the introduction of custom agents, such as the Test-Driven Development (TDD) agent, which exemplifies how specialized AI agents can be tailored to specific workflows. These agents come with defined prompts, tools, and models optimized for particular tasks, improving consistency and quality across development teams. Custom agents can be version-controlled alongside code, ensuring repeatability and shared best practices. The conversation also touches on the seamless integration of these agents across different platforms, including VS Code, GitHub.com, and the CLI, emphasizing a unified developer experience regardless of the environment.
Finally, the hosts discuss the broader implications of AI agents in development workflows, including the new agent sessions view in VS Code that centralizes management of multiple AI tasks across local and cloud environments. They highlight the importance of choice and flexibility, allowing developers to use different AI models and tools under a single subscription. The episode closes with reflections on the rapid progress made in AI-powered development tools over the past year and the ongoing commitment of the VS Code team to listen to community feedback, continuously improve the product, and foster an open-source ecosystem that supports developers worldwide.
