Is the future of software development an AI-powered IDE? GitHub is spreading the idea.
At its annual GitHub Universe conference in San Francisco on Monday, GitHub announced Copilot Workspace, a development environment that uses what GitHub calls “Copilot-powered agents” to help developers brainstorm, plan, build, test and run code to help in natural language.
Jonathan Carter, head of GitHub Next, GitHub’s software research and development team, calls Workspace something of an evolution from GitHub’s AI-powered coding assistant Copilot into a more general-purpose tool that builds on recently introduced features like Copilot Chat, which lets developers ask questions to code in natural language.
“Through research, we found that for many tasks, the biggest point of friction for developers is getting started and, in particular, knowing how to approach a task [coding] “They know which files need to be worked on and how to consider multiple solutions and their trade-offs,” Carter said. “That’s why we wanted to create an AI assistant that could meet developers at the start of an idea or task, reduce the activation energy required at the start, and then work with them to make the necessary changes across the core base.”
At last count, Copilot had over 1.8 million paying residential customers and 50,000 corporate customers. But Carter envisions a much larger base attracted by feature expansions with broad appeal like Workspace.
“As developers spend a lot of time working on it [coding issues]”We believe we can help empower developers every day through a ‘thoughtful partnership’ with AI,” said Carter. “You can think of Copilot Workspace as a companion experience and development environment that complements existing tools and workflows and enables the simplification of a class of developer tasks… We believe there is a lot of value in an AI-native developer environment, which is not the case “It is not restricted by existing workflows.”
There is certainly internal pressure to make Copilot profitable.
According to a Wall Street Journal report, Copilot loses an average of $20 per month per user, 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 available to individual developers for free late last year. There are also startups like Magic, Tabnine, Codegen and Laredo.
Given a GitHub repo or a specific bug in a repo, Workspace – powered by OpenAI’s GPT-4 Turbo model – can create a plan to (attempt to) fix the bug or implement a new feature by relies on understanding the repo’s comments. Problem answers and larger code base. Developers receive code suggestions for the bug fix or new feature, along with a list of things they need to validate and test that code, as well as controls for editing, saving, refactoring, or undoing.
Photo credit: GitHub
The proposed code can be executed directly in Workspace and shared with team members via an external link. Once in Workspace, these team members can refine and tinker with the code as they see fit.
Perhaps the most obvious way to launch Workspace is the new Open in Workspace button to the left of Issues and Pull Requests in GitHub repos. Clicking on it will open a box describing the software development task to be performed in natural language, such as: B. “Add documentation for the changes in this pull request” which, once submitted, is added to a list of “sessions” within the new pull request’s dedicated workspace view.
Photo credit: GitHub
Workspace systematically executes requirements step by step, creates a specification, generates a plan, and then implements that plan. Developers can dive into each of these steps to get a detailed view of the proposed code and changes, and delete, re-execute, or rearrange the steps as needed.
“If you ask a developer where they get stuck on a new project, you will often hear that they don’t know where to start,” Carter said. “Copilot Workspace takes that burden away and gives developers a roadmap to start iterating from.”
Photo credit: GitHub
Workspace enters technical preview on Monday and is optimized for a range of devices, including mobile.
Importantly, because of the preview release, Workspace is not covered by GitHub’s IP indemnification policy, which promises to cover the legal fees of customers faced with third-party claims alleging that the AI-generated code they use infringes IP. (Generative AI models are known to regurgitate their training data sets, and GPT-4 Turbo was partially trained using proprietary code.)
GitHub says it hasn’t yet determined how Workspace will be produced, but that it will use the preview to “learn more about the value it provides and how developers use it.”
I think the bigger question is: Will Workspace solve the existential problems surrounding Copilot and other AI-powered coding tools?
An analysis of over 150 million lines of code committed to project repos over the past few years by GitClear, developer of the code analysis tool of the same name, found that Copilot led to more broken code being pushed into codebases and more code being created re-added instead of reused and optimized, causing headaches for code maintainers.
Elsewhere, security researchers have warned that Copilot and similar tools can exacerbate existing bugs and security issues in software projects. And Stanford researchers have found that developers who accept suggestions from AI-powered coding assistants tend to produce less secure code. (GitHub emphasized to me that it uses an AI-based vulnerability prevention system to attempt to block unsafe code, in addition to an optional code duplication filter to detect public code regurgitation.)
Nevertheless, developers do not shy away from 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 by 2028, 75% of enterprise software engineers will be using AI code assistants.
By emphasizing human review, Workspace may actually be able to help eliminate the clutter caused by AI-generated code. We’ll find out soon enough when Workspace gets into developers’ hands.
“Our main goal with Copilot Workspace is to use AI to reduce complexity so developers can unleash their creativity and experiment more freely,” said Carter. “We firmly believe that the combination of humans and AI will always be superior to one or the other alone, and that is exactly what we are committed to at Copilot Workspace.”