GitHub unveils new AI-powered developer tools platform

GitHub has launched its enhanced Copilot-powered developer platform, introducing new features and partnerships aimed at providing developers with more flexibility and control over their tools.

Among the significant announcements is GitHub Spark, a tool designed to allow users to create micro apps using natural language, even without coding experience, alongside collaborations with Anthropic, Google, and OpenAI. These collaborations will enable developers to incorporate models into their workflows seamlessly.

Enhancements to GitHub Copilot in the Visual Studio Code (VS Code) environment include a multi-file edit feature. This allows users to utilise Copilot Chat in VS Code to execute complex edits across multiple files at once via natural language prompts. This change is intended to streamline the editing process within comprehensive projects.

GitHub has also introduced Copilot Extensions, available to all users, allowing developers to query integrated tools from well-known services like Atlassian Rovo, Docker, Sentry, and Stack Overflow. Additionally, these extensions permit developers to create private tools within their own environments. Copilot Extensions are set to become generally available in early 2025.

In a bid to extend support across different platforms, the code completion capabilities of Copilot are now in public preview for Xcode, benefiting those developing applications for Apple platforms.

Another feature, offering fast feedback on code submitted for review, has also been unveiled. This feature allows users to receive initial feedback on their code within 30 seconds, which can help refine code toward being "ready to merge" even as they await feedback from human reviewers. Users can request reviews from Copilot through Visual Studio Code or the GitHub website when creating a pull request.

Further customisation is available to users who can now provide custom instructions, allowing Copilot Chat responses to be tailored according to individual preferences, tools, and organisational coding conventions. This personalisation is further enhanced by incorporating context from repositories, pull requests, and discussions, supported by Bing integration for a comprehensive AI-native experience.

GitHub has also introduced the latest iteration of Copilot Workspace. Used by over 55,000 developers, this space has facilitated more than 10,000 pull requests. GitHub has implemented over 100 updates to the workspace based on developer feedback, including a build and repair agent and features to detect and correct necessary changes. New additions include brainstorming modes, VS Code integrations, iterative feedback mechanisms, and enhanced AI support.

A further development involves expanded GitHub Models, now available in public preview. More than 70,000 developers have experimented with various AI models through GitHub. With the recent updates, users can now compare models side-by-side, support multi-modal models, save and share prompts, and access new cookbooks and SDKs in GitHub Codespaces.

Updates also feature Copilot Autofix, aimed at improving code vulnerability fixes at scale. This feature is shown to resolve vulnerabilities over three times faster than previous methods and includes security campaigns to help developers manage issues efficiently. The integration with tools like ESLint, JFrog SAST, and Black Duck's Polaris platform enables developers to streamline their security workflows. These security campaigns and partner tool integrations are currently available in public preview.