Windows Subsystem, Copilot Both Heavily Open Sourced: What Surprises Did Microsoft Bring Us in Its Late Night Event?

Image

Organized by | Tu Min

Produced by | CSDN (ID: CSDNnews)

Every early summer, the tech world welcomes a wave of "new product shows," especially May and June, which have almost become a lively time like a "mini Spring Festival Gala" for developers – Microsoft Build, Google I/O, Apple WWDC take turns appearing, bringing a large wave of new technologies and tools, trying every possible way to attract developers' attention. This year, Microsoft took the lead, with the Build 2025 conference debuting at 12:05 AM on May 20th.

At this conference, Microsoft CEO Satya Nadella and CTO Kevin Scott took the stage in person. Surprisingly, OpenAI CEO Sam Altman and Tesla CEO Elon Musk, who have not always seen eye-to-eye, as well as Nvidia CEO Jensen Huang, also "appeared" at the conference, albeit via online video connections, having brief conversations with Nadella on topics such as collaboration, large models, and chips.

Overall, AI is undoubtedly Microsoft's most important strategic direction. However, this year "open source" also became another keyword that ran throughout the event. Not only were the core functionalities of Copilot on VS Code opened up, but even the Windows Subsystem for Linux (WSL) was significantly open-sourced, which was truly unexpected!

Image

Of course, in addition to these major releases, Microsoft also had many other noteworthy technological innovations at this conference. From product experience to underlying tools, what new changes has AI brought to the entire industry? Let's take a look at what this conference was all about.

Image

Nadella opens the conference

There was a small interlude - just a few minutes after Nadella took the stage, there seemed to be some protestors trying to disrupt the event, but the conference quickly returned to normal.

"We are in the 'middle game' of a platform shift - things are starting to scale and accelerate rapidly," Nadella described the current technological wave in his opening remarks. "This is somewhat similar to the early 1990s when Web technology was just starting, or like the rapid rise of cloud computing and mobile internet in the 2000s."

He pointed out that in 2025, we are ushering in a new era of Web expansion. "Starting from the integrated technology stack products initially, we have moved to a cloud-supported, more open and extensible Web architecture phase."

In terms of developer tools for building software, data shows that the Visual Studio product family currently has over 15 million users, GitHub has over 150 million developers, and GitHub Copilot users have also surpassed 15 million. Nadella is full of confidence about this: "This is just the beginning."

Image

Image

Microsoft is constantly refining developer tools

Currently, Microsoft is continuously upgrading these tools: Visual Studio is receiving multiple feature updates, including support for .NET 10, introduction of live preview and smoother design-time experience, enhanced Git tooling, and a new debugger built for cross-platform development. The update rhythm will also be adjusted to a monthly stable release, allowing developers to get new features more promptly. VS Code just released its 100th open source version, adding multi-window support, and developers can now directly view and manage staged content in the editor. On the GitHub front, Microsoft is promoting the deep integration of AI and open source. It announced that it will open source the Copilot extension in VS Code (https://github.com/microsoft/vscode) and integrate these AI-driven features directly into VS Code's core codebase, making AI part of the development experience and laying the foundation for the continuous evolution of GitHub Copilot.ImageRegarding GitHub Copilot, Microsoft stated that it can now help developers with version migration tasks such as upgrading Java 8 to Java 21, or .NET Framework to .NET 9. Copilot automatically handles dependency updates, suggests fixes, and learns from developers' modifications, making the entire migration process smoother and more automated. Microsoft also announced the launch of an autonomous agent specifically designed for Site Reliability Engineering (SRE) scenarios - the Azure SRE Agent. This SRE agent automatically starts when an online failure occurs, performs initial troubleshooting, identifies the root cause, and attempts to mitigate the issue. It then logs the incident report as a GitHub Issue with complete suggested fixes. Developers can further assign these fix tasks to GitHub Copilot to continue processing, achieving closed-loop automated operations.

More notably, Microsoft has launched the first complete Coding Agent, upgrading Copilot from a "conversational programming assistant" to a true "collaborative development partner."

Image

Using the new Coding Agent is very simple: just like assigning tasks to a teammate, assign one or more GitHub Issues to Copilot. You can do this on the GitHub website, mobile app, or command line. Additionally, you can issue instructions directly through Copilot Chat in GitHub or VS Code, for example:

@github Open a pull request to refactor this query generator into its own class

Once a task is received, the Coding Agent responds with an 👀 emoji and starts the workflow in the background: it launches a virtual machine, clones the repository, configures the environment, and analyzes the code using enhanced retrieval (RAG) technology driven by GitHub Code Search.

During the process, the agent continuously pushes modifications as Git commits to a draft Pull Request and updates the description. At the same time, you can see its reasoning and validation steps in the session log, making it easy to track its thinking and identify issues.

With Model Context Protocol (MCP), you can also connect external data and capabilities to the agent. You can configure an MCP server in your repository settings, or directly call the official GitHub MCP Server to get data. Moreover, thanks to the support of visual models, the agent can also "understand" images in GitHub Issues, including bug screenshots or feature sketches.

When the task is completed, Copilot flags you for review. You can leave suggested modifications, and the agent will automatically read the comments and propose corresponding code updates. It will also combine relevant Issue or PR discussion content, as well as custom instructions in the project, to ensure it understands your intent and follows project conventions.

Image

At the conference, Nadella himself performed a live demo, deeply experiencing this Coding Agent:

Image

Currently, this feature is available to Copilot Enterprise and Copilot Pro Plus users.

When discussing the expansion of the "Agent ecosystem," Nadella specifically mentioned OpenAI's Codex Agent, which was just released last Friday, and also invited OpenAI CEO Sam Altman via remote connection to join the conversation and discuss how software engineering can continue to evolve with the participation of various intelligent agents.

Image

Nadella: Sam, you've been tracking the evolution of different interaction forms in software development, from the command line, to chat interfaces, and now coding agents. Can you talk to us about your vision for the future direction of software engineering? And how can developers freely switch and collaborate efficiently between different interaction methods to fully leverage these intelligent tools?

Sam Altman: This was actually the path we envisioned back in 2021 when we launched the initial version of Codex. At that time, we collaborated with the GitHub team, and one of the initial goals was to hope to build a multimodal, collaborative development platform one day, making AI a true "virtual partner" for every developer.

Today, we finally took this step, and being able to achieve such a coding agent experience is still somewhat incredible to me. I think this is one of the most transformative programming methods I have ever seen. Developers can now truly have a "virtual teammate" that can receive tasks - you can let it continue working on what you were just doing, or you can have it take on more complex tasks. Perhaps one day in the future, you can say to it: "I have a bold idea, go spend a few days implementing it."

Main Tag:Microsoft Build

Sub Tags:AIDeveloper ToolsWindowsOpen Source


Previous:When Thinking Becomes a Burden: Unveiling the "Thinking Traps" of Large Language Models

Next:Truth in Light vs. Algorithmic Illusion: Scientific Photography in the Age of AI

Share Short URL