GitHub, the world’s largest software development collaboration platform, announced the launch of GitHub Copilot X, the company’s vision for the future of AI-powered software development.
GitHub Copilot started a new age of software development as an AI pair programmer that keeps developers in the flow by auto-completing comments and code. And less than two years since its launch, GitHub Copilot is already writing 46% of code and helps developers code up to 55% faster.
AI-powered auto-completion is just the starting point. GitHub’s R&D team at GitHub Next has been working to move past the editor and evolve GitHub Copilot into a readily accessible AI assistant throughout the entire development lifecycle. This is GitHub Copilot X. GitHub is adopting OpenAI’s new GPT-4 model, and is introducing chat and voice for Copilot, and bringing Copilot to pull requests, the command line, and docs to answer questions on developers’ projects.
Thomas Dohmke, CEO at GitHub, said: “With AI available at every step, we can fundamentally redefine developer productivity. We are reducing boilerplate and manual tasks and making complex work easier across the developer lifecycle. By doing so, we’re enabling every developer to focus all their creativity on the big picture: building the innovation of tomorrow and accelerating human progress, today.
“From reading docs to writing code to submitting pull requests and beyond, we’re working to personalise GitHub Copilot for every team, project, and repository it’s used in, creating a radically improved software development lifecycle. Together with Microsoft’s knowledge model, we will harness the reservoir of data and insights that lies in every organisation, to strengthen the connection between all workers and developers, so every idea can go from code to reality without friction. At the same time, we will continue to innovate and update the heart of GitHub Copilot — the AI pair programmer that started it all.
“GitHub Copilot X is on the horizon, and with it a new generation of more productive, fulfilled, and happy developers who will ship better software for everyone. So — let’s build from here,” Dohmke added.
A ChatGPT-like experience in your editor with Copilot chat
- GitHub is bringing a chat interface to the editor that’s focused on developer scenarios and natively integrates with VS Code and Visual Studio. This does far more than suggest code. Copilot chat is not just a chat window. It recognises what code a developer has typed, what error messages are shown, and it’s deeply embedded into the IDE. A developer can get in-depth analysis and explanations of what code blocks are intended to do, generate unit tests, and even get proposed fixes to bugs.
- Copilot chat builds upon the work that OpenAI and Microsoft have done with ChatGPT and the new Bing. It will also join GitHub’s voice-to-code AI technology extension the company previously demoed, which it’s now calling Copilot voice, where developers can verbally give natural language prompts.
Copilot for pull requests
- Developers can now sign up for a technical preview of the first AI-generated descriptions for pull requests on GitHub. This new functionality is powered by OpenAI’s new GPT-4 model and adds support for AI-powered tags in pull request descriptions through a GitHub app that organisation admins and individual repository owners can install. These tags are automatically filled out by GitHub Copilot based on the changed code. Developers can then review or modify the suggested description. This is just the first step GitHub is taking to rethink how pull requests work on GitHub.The company is testing new capabilities internally where GitHub Copilot will automatically suggest sentences and paragraphs as developers create pull requests by dynamically pulling in information about code changes.
- GitHub is also preparing a new feature where GitHub Copilot will automatically warn developers if they’re missing sufficient testing for a pull request and then suggest potential tests that can be edited, accepted, or rejected based on a project’s needs.
- This complements GitHub’s efforts with Copilot chat where developers can ask GitHub Copilot to generate tests right from their editor — so, in the event a developer may not have sufficient test coverage, GitHub Copilot will alert them once they submit a pull request. And, it will help project owners to set policies around testing, while supporting developers to meet these policies.
Get AI-generated answers about documentation
- GitHub is launching Copilot for docs, an experimental tool that uses a chat interface to provide users with AI-generated responses to questions about documentation — including questions developers have about the languages, frameworks, and technologies they’re using. GitHub is starting with documentation for React, Azure Docs, and MDN, so we can learn and iterate quickly with the developers and users of these projects.
- GitHub is also working to bring this functionality to any organisation’s repositories and internal documentation — so any developer can ask questions via a ChatGPT-like interface about documentation, idiomatic code, or in-house software in their organisation and get instant answers.
- The benefits of a conversational interface are immense, and GitHub is working to enable semantic understanding of the entirety of GitHub across public and private knowledge bases so it can personalise GitHub Copilot’s answers for organisations, teams, companies, and individual developers alike based on their codebase and documentation.
- Moving forward, GitHub is exploring the best ways to index resources beyond documentation such as issues, pull requests, discussions, and wikis to give developers everything they need to answer technical questions.
- GitHub’s work to rethink pull requests and documentation is powered by OpenAI’s newly released GPT-4 AI model. Even though this model was just released, GitHub is already seeing significant gains in logical reasoning and code generation. With GPT-4, the state of AI is beginning to catch up with GitHub’s ambition to create an AI pair programmer that assists with every development task at every point in the developer experience.
- Moreover, it’s helping GitHub Copilot understand more of a developer’s codebase to offer more tailored suggestions in PRs and better summations of documentation.
Copilot for the command line interface (CLI):
- Next to the editor and pull request, the terminal is the place where developers spend the most time. But even the most proficient developers need to scroll through many pages to remember the precise syntax of many commands. This is why GitHub is launching GitHub Copilot CLI. It can compose commands and loops, and throw around obscure find flags to satisfy your query.