Is the future of software development an AI-powered IDE? GitHub is launching the idea.
At its annual GitHub Universe conference in San Francisco on Monday, GitHub announced Copilot Workspace, a development environment that leverages what GitHub describes as “Copilot-powered agents” to help developers think, plan, build, test and run natural language code.
Jonathan Carter, head of GitHub Next, GitHub’s software R&D team, presents Workspace as a sort of evolution of Copilot, GitHub’s AI-based coding assistant, into a more general tool, leveraging recently introduced features such as Copilot Chat, which allows developers to ask questions about code. in natural language.
“Through our research, we found that for many tasks, the biggest sticking point for developers was getting started, and in particular knowing how to approach a (coding) problem, knowing which files to edit, and know how to consider multiple options and solutions. their compromises,” Carter said. “So we wanted to create an AI assistant that could meet with developers at the start of an idea or task, reduce the activation energy needed to get started, and then collaborate with them to make the necessary changes on the ‘whole basic base.’
At last count, Copilot had more than 1.8 million paying individual customers and 50,000 business customers. But Carter envisions a much broader base, attracted by highly attractive feature expansions, like Workspace.
“Since developers spend a lot of time working on (coding problems), we believe we can help developers every day through ‘thoughtful partnership’ with AI,” Carter said. “You can think of Copilot Workspace as a complementary experience and development environment that complements existing tools and workflows and helps simplify a class of development tasks… We think there’s a lot of value that can be delivered in an AI-native development environment that isn’t limited by existing workflows.
There is certainly internal pressure to make Copilot profitable.
Copilot loses an average of $20 per month per user, according to a Wall Street Journal report, with some customers costing GitHub as much as $80 per month. And the number of competing services continues to grow. There’s Amazon’s CodeWhisperer, which the company made free to individual developers late last year. There are also startups, like Magic, Tabnine, Codegen and Laredo.
Given a GitHub repository or a specific bug in a repository, Workspace — powered by OpenAI’s GPT-4 Turbo model — can develop a plan to (attempt to) eliminate the bug or implement a new feature, leveraging on an understanding of the repository comments, responses to issues and a larger code base. Developers receive a code suggestion for the bug fix or new feature, along with a list of items they need to validate and test that code, as well as controls to edit it, save it, refactor it, or cancel it.
Image credits: GitHub
Suggested code can be run directly in Workspace and shared among team members via an external link. These team members, once in Workspace, can refine and modify the code as they see fit.
Perhaps the most obvious way to launch Workspace is to use the new “Open in Workspace” button to the left of issues and pull requests in GitHub repositories. Clicking on it opens a field to describe the software engineering task to be accomplished in natural language, such as “Add documentation for changes to this pull request”, which, once submitted, is added to a list of “sessions » in the new dedicated workspace view.
Image credits: GitHub
Workspace systematically executes requests step by step, creating a specification, generating a plan and then implementing that plan. Developers can dive into any of these steps to get a granular view of suggested code and changes and remove, rerun, or rearrange steps as necessary.
“If you ask a developer where they tend to get stuck with a new project, you’ll often hear them say it’s a matter of knowing where to start,” Carter said. “Copilot Workspace alleviates this burden and gives developers a blueprint from which to start iterating. »
Image credits: GitHub
Workspace enters technical preview on Monday, optimized for a range of devices, including mobile.
It’s important to note that because it is in pre-release, Workspace is not covered by GitHub’s IP indemnification policy, which promises to help cover legal costs for customers facing IP claims. third parties alleging that the AI-generated code they use infringes intellectual property. (Generative AI models notoriously regurgitate their training data sets, and GPT-4 Turbo was trained in part on copyrighted code.)
GitHub says it hasn’t determined how it will produce Workspace, but will use the preview to “learn more about the value it offers and how developers are using it.”
I think the more important question is: Will Workspace solve the existential problems surrounding Copilot and other AI-powered coding tools?
An analysis of more than 150 million lines of code committed to project repositories over the past few years by GitClear, the developer of the code analysis tool of the same name, found that Copilot caused the transfer of a more bad codes to the code bases and more codes. re-added instead of reused and streamlined, creating headaches for code maintainers.
Elsewhere, security researchers have warned that Copilot and similar tools can amplify existing bugs and security issues in software projects. And Stanford researchers found that developers who accept suggestions from AI-based coding assistants tend to produce less secure code. (GitHub pointed out to me that it uses an AI-based vulnerability prevention system to attempt to block insecure code in addition to an optional code duplication filter to detect regurgitations of public code.)
However, developers are not afraid of AI.
In a June 2023 StackOverflow survey, 44% of developers said they now use AI tools in their development process, and 26% plan to do so soon. Gartner predicts that 75% of enterprise software engineers will employ AI code assistants by 2028.
By emphasizing human review, perhaps Workspace can actually help clean up some of the mess introduced by AI-generated code. We’ll find out soon enough when Workspace is in developers’ hands.
“Our primary goal with Copilot Workspace is to leverage AI to reduce complexity so developers can express their creativity and explore more freely,” Carter said. “We sincerely believe that the combination of humans and AI will always be superior to either alone, and that’s what we’re betting on with Copilot Workspace. »
techcrunch
