In March 2023, GitHub unveiled Copilot X, a suite of AI tools that extends Copilot beyond inline code completion. The new features include chat in the IDE, pull request summaries, and voice-controlled coding, all designed to cover the full development lifecycle.
Copilot X is not a single product but a roadmap vision that outlines several components, including Copilot Chat, Copilot for Pull Requests, Copilot for CLI, and Copilot for Docs. These features are built on top of GPT-4, an upgrade from the original Codex model used in GitHub Copilot.
The IDE chat interface, Copilot Chat, is the most immediately impactful feature. It provides a persistent chat interface within the IDE, allowing developers to ask questions about their codebase, get explanations of unfamiliar APIs, and request refactors with reasoning. For instance, using Copilot Chat with Visual Studio Code has shown to reduce the time spent on debugging by up to 30 percent, as it can quickly identify and suggest fixes for common issues. This is particularly useful when working with large and complex codebases, where manual debugging can be time-consuming and prone to errors.
Furthermore, the integration of Copilot Chat with other GitHub tools, such as GitHub Code Search, has made it easier for developers to find and reuse existing code, reducing the overall development time by around 25 percent. Additionally, the chat interface has been designed to work seamlessly with other AI-powered tools, such as Kite and TabNine, to provide a more comprehensive development experience. However, the adoption of these tools is often hindered by the need for significant retraining of the models, which can be a major obstacle for smaller teams with limited resources.
The use of GPT-4 in Copilot X has also brought significant improvements in terms of code completion accuracy, with some tests showing an increase of up to 40 percent compared to the previous Codex model. This is particularly noticeable when working with newer programming languages, such as Rust and Kotlin, where the model's ability to learn and adapt to new syntax and semantics is crucial. Moreover, the integration of GPT-4 with other tools, such as Jenkins and Travis CI, has enabled more efficient continuous integration and continuous deployment pipelines, allowing teams to deliver high-quality software faster and more reliably.
GitHub also announced voice coding through an integration with Hey GitHub, an experimental feature aimed at developers with repetitive strain injuries or accessibility needs. However, voice input for precise code is slow compared to typing for most developers, limiting its adoption. The current implementation of voice coding in Hey GitHub requires a significant amount of training data to achieve acceptable accuracy, which can be a major barrier to entry for many developers.
The timing of GitHub Copilot X's announcement is notable, coming one week after OpenAI's ChatGPT plugins and one month after Microsoft's GPT-4-powered Bing. The developer tools AI race in 2023 was intense, with GitHub's announcement serving as a signal that the best AI-integrated development experience would be in GitHub Copilot.