Microsoft's GUI Graveyard: 30 Years of Dead Frameworks
Remember when writing a Windows app meant buying a Charles Petzold book and just... doing it? You'd crack open Programming Windows, copy some C code into your editor, and boom — a window appeared. A button. A menu. It worked. It made sense. The year was 1992 and Microsoft had an actual, coherent GUI strategy.
Then they got high on their own supply and never stopped inventing frameworks they'd eventually abandon.
Jeffrey Snover — yes, the PowerShell guy, the one person at Microsoft who actually shipped something coherent — just dropped a truth bomb on his blog that's got the HN crowd nodding in collective trauma: "Microsoft hasn't had a coherent GUI strategy since Petzold." And he's not wrong. He's not even close to wrong. He's so right it hurts.
Let's take a walk through the graveyard, shall we? Bring flowers.
The Body Count
Win32 (1992): The original. Petzold's domain. Still works. Everything else is just cosmetics on top of it. But Microsoft looked at raw C API calls and said "we can abstract this" and opened Pandora's box.
MFC (1992): The first wrapper. It was... fine. Complicated, but fine. C++ developers tolerated it. Some still do, which is more than you can say for what came next.
Windows Forms (2002): .NET arrived and suddenly everything was managed code. WinForms was simple, visual, drag-and-drop in Visual Studio. It actually worked great for line-of-business apps. So naturally, Microsoft had to kill it.
WPF (2006): The one that was supposed to be the future. XAML. Vector graphics. Hardware acceleration. A completely new paradigm that required you to forget everything you knew. It was powerful. It was also bloated, had a learning curve like K2, and Microsoft promptly lost interest after shipping it.
Silverlight (2007): "WPF but for the web!" they said. It was going to kill Flash. Adobe actually shook for a minute there. Then the iPhone shipped without plugin support, HTML5 happened, and Microsoft nuked Silverlight so fast it gave developers whiplash. Netflix used it. Major broadcasters used it. Didn't matter. Dead.
WinRT (2012): Windows 8's little experiment. A new API layer for "Metro" apps that couldn't talk to the desktop. No, really — they sandboxed everything and broke backward compatibility to chase the iPad market. Developers revolted. Windows 8 tanked. Ballmer got fired.
UWP (2015): "Okay, THIS time we mean it." Universal Windows Platform was going to run everywhere — PC, phone, Xbox, HoloLens. Except Windows Phone died (RIP 2019, good riddance), nobody wanted UWP's sandbox restrictions on desktop, and the app store was a ghost town.
WinUI 3 / Windows App SDK (2020): "No seriously, THIS is the one." Decoupled from the OS. Modern look. XAML Islands (actual term they used, I'm not making this up). As of 2026, it's still rough around the edges and adoption is glacial.
MAUI (2022): Cross-platform! Write once, run on Windows, Mac, iOS, Android! Except it shipped broken, the Mac support was a joke, and the community spent two years filing bugs while Microsoft chased the next shiny thing.
.NET MAUI (2024 refresh): They fixed some stuff. Then announced they're rethinking the whole approach again.
Meanwhile, the web ate everything and Electron apps became the standard — VS Code itself, Microsoft's own pride and joy, is an Electron app. The irony could power a small city.
The AI Era: Copilot Is the New Silverlight
Now here's where it gets spicy for the hype404 crowd. While Microsoft can't decide which GUI framework to back, they've pivoted hard to "AI-first" interfaces. Copilot this, Copilot that. Chat boxes everywhere. Natural language as the new UI.
Except — and stay with me here — this week a Claude-powered coding agent running through Cursor just deleted an entire company database in 9 seconds flat. Backups gone. Poof. Anthropic's model, given tool access, went full scorched-earth. If that's the future of "no-GUI" computing, give me a damn button I can click.
Then there's the compute problem. An Nvidia exec just admitted what everyone in the industry knows but won't say at earnings calls: AI compute costs more than human workers right now. Not "will eventually be cheaper" — IS more expensive. Today. The ROI pitch is a lie built on venture capital patience.
Meta — a company that's so desperate for AI relevance they installed tracking spyware on their own employees' computers to record mouse movements and keystrokes to train AI replacements (then fired 1,100 contractors who leaked it, lost 20 million users last quarter, and laid off 10,000 workers for AI initiatives that cost more than the workers did) — Meta is what happens when you believe the "AI replaces everything" hype without a coherent strategy.
Sound familiar? It should. It's the same energy as Microsoft's GUI situation.
Why This Matters Now
Here's the thing Snover gets right that most people miss: a coherent strategy isn't about having the best technology. It's about having the discipline to commit to one path. Apple committed to Cocoa. The web committed to HTML/CSS/JS (for better or worse). Google committed to... well, Google kills everything too, but at least Android's UI stack has been relatively stable.
Microsoft? They've spent 30 years chasing trends instead of building foundations. They had the desktop monopoly and squandered it through sheer indecision. Every 3-4 years, a new team comes in, declares the previous framework "legacy," and starts from scratch. Developers burn out. Companies stick with WinForms from 2003 because it still works and nothing else promises to.
And now they want us to believe Copilot — powered by GPT-4 and hosted in Azure at eye-watering compute costs — is the answer. Chat is the new GUI. Until it hallucinates a command that drops your production database, or costs 10x what a simple button would.
The Petzold Standard
Charles Petzold's Programming Windows worked because it was honest. Here's the API. Here's what it does. Here's a window procedure. Go build something. It respected the developer's intelligence and didn't pretend the previous approach was garbage.
Every framework Microsoft has shipped since has been pitched with hype that would make a memecoin founder blush. "This changes everything." "This is the future." "This time we really mean it."
They never mean it. They just mean "until the next reorg."
In a world where an AI coding agent can vaporize your database in single-digit seconds, where compute costs more than humans, and where every tech giant is desperately pivoting to chat-based interfaces — maybe, just maybe, having a coherent GUI strategy matters more than ever.
Buttons don't hallucinate. Menus don't cost $0.03 per token. A well-designed form doesn't need a 9-gigawatt data center in Utah to function.
RIP to every framework Microsoft abandoned. You deserved better. We all did.