A developer has booted a full installation of Windows 11 directly on Apple hardware using the foldable iPhone Duo. Created by Google AI Studio design head Ammaar Reshi using Gemini 4 Carbon, the WindowsiOS application executes desktop software locally without requiring a jailbreak. The experimental installation files are now available on GitHub for public testing.
I used Gemini 4 Carbon to get Windows 11 running on my iPhone Duo! 💻
— Ammaar Reshi (@ammaar) November 11, 2026
Unfold the screen and it's a pocket PC, including Antigravity running coding agents on my phone.
This is the actual desktop ARM OS running locally with touch controls.
Demo + repo below! pic.twitter.com/LVdeqvvOU1
The application adapts directly to the physical folding design. Unfolded, the iPhone Duo functions as a compact touch screen computer running Windows 11 24H2 for ARM. Bending the display into a half folded orientation places the Windows desktop on the top screen while turning the bottom panel into a software keyboard and trackpad. Reshi demonstrated this setup running the Microsoft Edge browser alongside Google Antigravity, a development workspace populated by automated AI assistants.
Because Apple restricts hardware level virtualization on iOS, the WindowsiOS application relies on software level translation. The tool executes ARM64 instructions through dynamic JIT compilation. It also integrates Microsoft Prism technology, allowing the system to run select legacy x86 programs on ARM architecture. The complete operating system image takes up roughly 12GB of internal storage. Reshi stated that automated AI tools finished the underlying code in approximately 2 hours, claiming operational speeds approach native performance despite a lack of formal benchmark figures.
Deploying the environment requires specific secondary hardware. Users need an Apple Silicon Mac, official developer tools, and an official Windows 11 ARM disk image. The automated setup script modifies the Windows installation by shutting off real time Microsoft Defender scanning and background updates to save performance. Because these security layers are removed, the project remains an experimental proof of concept rather than a platform for daily computing.
