Table of Contents
Why are Windows apps using so much RAM despite AI coding tools?
Confusion recently erupted regarding the future of the Windows codebase. A LinkedIn post by a high-ranking Microsoft engineer suggested a massive, AI-driven rewrite of the operating system. Microsoft has since issued a formal denial. The company clarified that there are no current plans to rewrite the Windows 11 kernel using AI to convert C++ to Rust. This situation highlights the tension between experimental research and actual product strategy.
The Source of the Controversy
The alarm began when Galen Hunt, a Distinguished Engineer at Microsoft, published a post outlining an aggressive objective: “Eliminate every line of C and C++ from Microsoft by 2030.”
Hunt described a strategy using AI and algorithms to rewrite massive codebases. He cited a specific metric: “1 engineer, 1 month, 1 million lines of code.”
For technical stakeholders, this metric is concerning. Converting one million lines of code in a single month suggests a speed that precludes meaningful human review. Since Windows is built largely on C and C++, the post implied that the operating system’s core stability was being handed over to automated generation tools. The usage of “our strategy” and “North Star” indicated this was company-wide policy rather than a siloed experiment.
Microsoft’s Official Stance
Following significant backlash from the developer community, Microsoft leadership corrected the narrative.
- Frank X. Shaw, head of communications, confirmed the company is not rewriting Windows 11 with AI.
- Galen Hunt edited his post to clarify the context. He stated the initiative is a research project designed to facilitate language migration, not a roadmap for Windows 11.
While Rust offers superior memory safety compared to C++, the concern was never about the destination language. The concern lies in the method. Using AI to modify legacy code at the scale of an operating system introduces significant risk regarding bugs, security vulnerabilities, and stability.
The Reality of AI in Microsoft’s Code
Despite the denial regarding the Windows kernel, AI is already deeply embedded in Microsoft’s development lifecycle. The distinction between “research” and “production” is narrowing.
CEO Satya Nadella recently noted that AI generates approximately 30% of the code in Microsoft repositories. Furthermore, the company’s CTO has projected that AI could generate up to 95% of code by 2030. While Windows 11 is not currently undergoing a total AI rewrite, the company is actively normalizing AI-generated software.
The Immediate Problem: WebView2 and Resource Management
While the debate over C++ versus Rust is theoretical, users face a tangible performance issue today: the proliferation of web-based technologies in desktop apps.
The user experience in Windows 11 degrades not because of the kernel language, but because of WebView2 and Electron. These frameworks allow developers to build desktop applications using web technologies. While efficient for development, they are resource-intensive for the end user.
Key examples of this inefficiency include:
- Microsoft Teams: Built on WebView2, this application often consumes 1–2GB of RAM while idling.
- Discord: An Electron-based app that can spike to 4GB of RAM usage, forcing client restarts.
- WhatsApp: Previously a native UWP app that ran efficiently on 200MB of RAM, it was replaced by a WebView2 version that consumes significantly more memory.
Technical Assessment
The shift toward web-based components in the operating system itself is accelerating. Microsoft is integrating WebView2 into core Windows 11 features, such as the upcoming Outlook Agenda view in the Notification Center. This adds unnecessary memory overhead to system processes.
The concern for users and IT administrators is misplaced if focused solely on the “AI rewrite” rumor. The pressing issue is the architectural shift toward web containers for native applications. This trend burdens system hardware regardless of whether the underlying code is written by a human or an algorithm.