Microsoft needs Windows developers like never before

The last time Microsoft really needed Windows developers was when it asked them to build a new type of application that could work across phones, PCs, tablets, Xbox consoles, and headsets like the HoloLens. It was all part of a larger effort to transform Windows with a new interface for touch-friendly apps, designed to rival the iPad.

It failed miserably. Developers didn’t flock to universal Windows apps, and Microsoft eventually abandoned its touch-friendly UI in favor of a more traditional desktop in Windows 10. The cross-platform dream of the Windows 8 era didn’t last very long.

Now, Microsoft is trying something new, and it needs developers’ help again. It’s adding AI models and tools directly into Windows for this first generation of Copilot Plus PCs — laptops that have powerful neural processing unit (NPU) chips to accelerate AI tasks. I’ve written about how these new laptops will compete with the MacBook Air, transition Windows to Arm-powered chips, and set Windows up for an era of AI. The big selling points are better battery life, better performance, and the promise of AI features inside Windows and the apps you use every day.

But for all of this to work, Microsoft needs developers to adapt their apps again — and get people excited to use Windows.

AI was the talk of Microsoft Build, coming out of every speaker’s mouth and plastered on sign after sign on Microsoft’s campus. There were flashy demonstrations of Paint being used with local AI models to transform doodles into digital images and not so flashy announcements of a Windows Copilot Runtime that includes the AI frameworks and toolchains for developers to ship AI features in their apps.

A lot of these tools aren’t coming until June, when these new Copilot Plus PCs ship, but Microsoft dropped hints of how developers can start using AI capabilities in a variety of Build sessions last week.

I learned that features like OCR will be part of the Windows SDK, so developers can add text recognition into their apps, letting users easily copy text from an image just like on iOS and Android. The built-in Snipping Tool uses this same text recognition API, as does PowerToys’ new Advanced Paste feature.

Microsoft is also encouraging developers to use Hugging Face to find models to add to their apps. Hugging Face is like GitHub for AI models, where, as a developer, you can try models before you adopt them inside an app. An example of this is OpenAI’s Whisper model, which is available on Hugging Face. It provides automatic speech recognition for speech transcription and translation so it can immediately provide powerful audio tools inside Windows apps.

The upcoming Windows Copilot Library will also make it easy for developers to integrate Studio Effects like filters, portrait blur, and more into their apps. Meta is adding the Windows Studio Effects into WhatsApp, so you’ll get features like background blur and eye contact during video calls. AI-powered translation features can also be used by developers with little to no code.

Microsoft is hoping that developers will improve Windows’ new Recall feature, too. Recall lets you search through snapshots of what you’ve looked at on your PC and quickly switch back to that moment. Developers that integrate Recall will help Windows users pick up where they left off, with Microsoft selling this as a way to improve app engagement for developers. Recall works without developers adjusting their apps, but it works even better if they do.

All of these tools are designed to entice developers to bring AI-powered experiences to their apps, and if they do, it could truly transform the Windows app landscape. Microsoft and its OEM partners are hoping that this AI push kicks off renewed interest in PC purchases, both from consumers and businesses. But the AI work all feels like super early days right now. Most of the developers I spoke to at Build were at the very beginning of exploring how to integrate AI into their apps or were there because their managers had asked them to attend Build to get more information on the tools available to them to better prepare for this potential transformation.

In a Microsoft Build session about using AI “for real things in your Windows apps,” Microsoft only had a handful of developer demos that are already leveraging the NPU chip that ships in these Copilot Plus PCs. DaVinci Resolve Studio is offloading its Magic Mask tool to the NPU so it runs three times faster, the Djay app is using the NPU to separate parts of audio tracks and mix them, and CapCut is removing the background from any video clip by using the NPU.

These feel a lot less transformative than adding powerful OCR and translation tools to apps or even Microsoft’s own Recall feature. I suspect there was a lack of third-party app demos because Microsoft wanted to keep everything secret, so I’m hoping that changes soon.

Outside of the Build keynote, the busiest part of the show floor was the Surface booth, where I’ve never seen so many developers interested in Arm-based hardware before. I spent hours at this booth, downloading and running Cinebench and seeing if I could get the fans to scream. I failed to get them roaring, but I got the feeling that lots of Build attendees shared many of the same questions I have about performance and app compatibility.

There was also the shared feeling of intrigue over the AI push here, and I regularly witnessed developers questioning Microsoft employees about how these new Windows AI features work and whether they could tap into them. Microsoft’s big investment to transition Windows to Arm chips, coupled with its investments over the years to improve the dev environment for Windows, may well result in some much-needed developer love for this year’s AI push.

Microsoft has given us a taste of AI-powered features in Windows, but it’s still laying the groundwork for developers to lead the way forward. It’ll be up to developers to make Microsoft’s AI dreams come true, and the company now needs them just as much as it did with its Universal Windows Platform (UWP) push more than a decade ago.

This time around, Microsoft is making it a lot easier to overhaul an app with AI features. Instead of having to write an entirely new app, you can just weave AI models and features into an existing one. We’ve seen early signs of that so far, and it could be what sets this apart from the failed universal app dream.

I’m always interested in reader questions, so please write in with any questions you have for me or about Microsoft and I’ll try to answer as many as I can.

Q from Flavio: Isn’t Microsoft worried about having insignificant “AI market share” on mobile? Why should I use Copilot on mobile when I will have Gemini or ChatGPT fully integrated in my phone at an OS level?

A: Mobile is Microsoft’s biggest weakness in AI, especially as Google continues to integrate AI features into Android and Apple is rumored to announce an OpenAI partnership soon. Microsoft spent big on a Super Bowl commercial for Copilot earlier this year, but ChatGPT still dominates the conversations about mobile usage of generative AI. Microsoft would love to dominate AI on mobile, but without controlling the platform, it’s always going to be behind. Microsoft does win by getting mobile app developers to use its Azure OpenAI services or even when OpenAI generates revenue through its own offerings to mobile developers. After all, Microsoft gets a cut of OpenAI’s revenues, so it benefits from a potential Apple deal.

Q from Fabian: I would love to hear your thoughts on the Xbox / PC merger that Digital Foundry and others have been discussing. An Xbox Steam Deck (Snapdragon X Elite-powered?)

A: I shared some of my thoughts in a mock issue of Notepad in April. The future of Xbox looks a lot like PC, where Microsoft is about to close the gap between Xbox and PC even further. I’m still trying to figure out how that shapes the next-gen Xbox, but I think we should expect game development to be even closer between PC and Xbox, resulting in the potential to be able to play PC games on the next Xbox. On the handheld side, I think an Arm-powered device makes the most sense. That will all depend on timing, game compatibility, and how well Windows can be adapted for handhelds and transitioned to Arm. There are a lot of moving parts here, but for an Xbox handheld to be successful, I think it needs to offer both Xbox and PC games, with Windows running in the background out of the way. I personally want a handheld with the Xbox OS and UI but the ability to dig into Windows if I want to install Steam or use other apps.

Next week, I’ll be digging into what’s next for Xbox, so stay tuned for that. As always, you can reach me via email at notepad@theverge.com for any questions.

If you know anything about Project Latitude inside Xbox, or any secret project Microsoft is working on, you can reach me confidentially on Signal. I’m tomwarren.01 there.