GitHub is announcing a new service called GitHub Copilot Workspace, which aims to streamline the process of starting a new coding project. Traditionally, developers have spent a significant amount of time reading through code and documentation to understand how to begin a project. However, with Copilot Workspace, this process can be greatly simplified.

The new service will be integrated into GitHub repositories or libraries, allowing developers to describe their project goals through prompts. Copilot Workspace will then provide suggestions on how to start the project, offering a step-by-step process. Users can then edit the suggestions to their liking, run the code, and even use Copilot to help complete the remaining tasks.

According to Jonathan Carter, the head of GitHub Next, the company has heard from customers about the helpfulness of Copilot during the middle stages of a project, but its limitations in the initial stages. By integrating Copilot into the very beginning of a project, developers can spend more time actually coding, rather than reading through code and documentation to figure out how to start.

Copilot Workspace is also expected to be particularly useful for reviewing older code, as the AI can scan the codebase and understand its inner workings faster than a human can read through written documentation.

Since the launch of GitHub’s Copilot, the use of AI in coding tasks has become increasingly prevalent. Lightweight models and large language models alike are being optimized for a variety of coding-related tasks, from simple summarization and writing to more complex code completion.

Microsoft, for example, has released a small model called Phi-3 that is capable of code-writing tasks. Meta has also released versions of their large language models specifically for coding projects.

It will be exciting to see how GitHub Copilot Workspace and other AI-powered tools continue to shape the future of software development in the years to come.