WinUI and MAUI should be one framework
On 20 November, David Ortinau, the product manager for .NET MAUI, published AI is Changing the Cross-Platform Decision Matrix. He took his own MAUI app, Barista Notes, about 31,000 lines of C#, and had GitHub Copilot agents rebuild it with platform UI: UIKit on iOS and Android Views on Android, still in C#. The rewrite took about three days and under four hours of his own time for the first pass.
He did that test in a year when two loud voices had already answered the question. I'm with one of them, DHH, and I think the answer means WinUI and MAUI should stop being two frameworks.
What DHH and Shopify already said
At the Rails World 2026 keynote in September, DHH said 37signals had gone "pencils down" on hand-written code. Writing code by hand there, in his words as transcribed by Sublime Coding, "is now an exceptional state. It is like seeing a bug in Sentry." The next HEY, he said, is six native apps, with a Rust core he has never read: "I don't know any Rust at all. I consider that a feature."
His wider point was that a lot of architecture existed to avoid writing the same thing twice, and once an agent writes it, that abstraction stops paying for itself. Gone are the days when sharing code was the goal.
The same month, Shopify moved its Shop app from React Native to Swift and Kotlin in 12 weeks, because "advances in coding agents changed the tradeoffs" of a shared codebase. Android cold start halved, from 4,433 ms to 2,233 ms.
That is the context for Ortinau's post. A MAUI product manager watching the industry walk away from shared UI had to test it on his own app.
What the Microsoft post actually says
It is more careful than its title. The "matrix" is one table that compares each factor before agents with what Barista Notes showed:
| Factor | Before agents | What Barista Notes showed |
|---|---|---|
| What your team knows | Pick a stack your team can write | Agents wrote UIKit and Android Views, but he still had to spot what was wrong on each platform |
| Cost to build | A second UI was a second project | About three days, under 4 hours of his time |
| Cost to maintain | Scales with the number of UIs | One feature took about 1.4 times the model time and drew more feedback |
| Capacity to verify | Part of each feature | 616 coordinator minutes against 209 for implementation |
| Runtime results | Assumed to favour platform UI | Platform UI won on startup and size |
His own numbers back DHH. Android startup went from 1,233 ms on MAUI to 615 ms with platform UI, and iOS from 569 ms to 367 ms. The Android package shrank from 45.64 MiB to 31.74 MiB, though he warns the MAUI build carries extra dependencies.
And yet his conclusion is "A single shared UI is still my starting point because it gives my team less UI code to own." The reason is verification: checking the work took three times as long as writing it.
That is a fair point, and it is also the product manager's point. The data says native is cheap to build and faster to run. The recommendation says keep MAUI as the default.
Two XAMLs for one Windows
Here is where Microsoft's position stops making sense to me. If agents write the UI and the only real cost is checking it, then the framework choice stops being about how much code you share. It comes down to how many different things a person has to know to check the result. Microsoft asks them to know two, for one operating system.
MAUI on Windows is WinUI underneath. Microsoft's own MAUI on Windows page says a MAUI app "will use WinUI 3 as its native platform", and in the same breath that MAUI "does not currently support using Windows App SDK controls."
The newer cross-framework migration guide, written for AI-assisted migration, tells you to "remove MAUI and target WinUI 3 directly" for a Windows-only app. Your ViewModels move across. Your XAML does not: Microsoft.Maui.Controls has to be rewritten as Microsoft.UI.Xaml. It even ships starter prompts so an agent can do the translation. The Windows developer FAQ asks whether you can start with WinUI 3 and add MAUI later. The answer is "Not at this time."
If an agent can translate one into the other on request, the split is not doing any work for the developer. It is an org chart.
Everyone else is already merging them
- Uno Platform 7.0, out this month, runs C# and WinUI XAML on Windows, Android, iOS, macOS, Linux and WebAssembly, ports WinUI controls from current WinUI sources, and hosts MAUI controls on all of those targets.
- Avalonia's MAUI backend adds Linux and WebAssembly, "platforms Microsoft's own cross-platform .NET framework lacks".
What convergence would look like
One .NET UI framework, with WinUI as its Windows implementation rather than a separate product, and Windows App SDK controls usable from it. Where native wins, as Shopify and Ortinau both measured, let the agent write that platform's screen natively, which MAUI already allows by embedding native views. Shared code becomes optional, the way DHH says it should be. Knowing two Microsoft dialects for the same desktop stops being required.
Agents removed the reason to share code and the reason to pick a framework by what your team can type. What is left is the cost of owning the result, and Microsoft is still making Windows developers pay it twice.
~ Mike