A modern x86-64 operating system built from scratch. Native Win32 API compatibility. Linux binary support. Windows 11‑inspired desktop. 100% open source.
Countdown to First Release
Target release: August 8, 2028
One kernel for Windows apps, native Impossible OS programs and, next, Linux binaries.
The Win32 API is the kernel's native interface, not an emulation layer or a virtual machine: CreateFile() and ReadFile() go straight to native NT system calls, and PE32+ executables load and run natively. The full Win32 subsystem is still being built.
Planned: unmodified Linux ELF binaries through a built-in compatibility layer, with open(), read(), fork() and exec() translated to native kernel calls. No WSL, no containers. Not implemented yet.
Not a Linux distro. Not a Windows fork. The kernel, the UEFI bootloader and the desktop are original C and x86-64 assembly; mature libraries such as ACPICA are vendored where rewriting them would only add bugs.
1,353,056 Lines of CodeWhat works today, and how far each area is along its roadmap.
IXFS, our own filesystem, with journaling, copy-on-write and 64-bit extents, plus FAT32 read and write for USB drives.
16 roadmap sections to goAHCI with DMA and NCQ command queuing, and a native NVMe driver. A legacy ATA PIO driver remains as a fallback for older machines.
11 roadmap sections to goTaskbar, Start menu and a compositing window manager with frosted Acrylic surfaces and dirty-rectangle redraw. Mica window title bars are next.
13 roadmap sections to goBoots every CPU and schedules threads across all of them with a fair round-robin scheduler, plus mutexes, semaphores, RW locks, seqlocks and a basic RCU. Per-CPU run queues and CFS-style fair sharing are planned.
24 roadmap sections to goEthernet, ARP, IPv4, ICMP, UDP and DHCP with an RTL8139 driver work today. TCP, DNS and a VirtIO-net driver are planned.
25 roadmap sections to goA UEFI-only bootloader with a GOP framebuffer: no legacy BIOS, no VGA text mode. It reads the Secure Boot state and supports shim chain-loading and SBAT, but builds boot directly today: a shim signed under the Microsoft 2023 CA is still pending.
6 roadmap sections to goA Windows-compatible registry with HKLM and HKCU hives, RegOpenKeyEx and RegSetValueEx, saved to disk. Transactions and faster hive storage are in progress.
16 roadmap sections to goSSE2-accelerated drawing on the CPU: pre-multiplied alpha blending, gradients, blur, TrueType text and a glyph cache. GPU acceleration is a research item.
15 roadmap sections to goA minimal user-mode Win32 header set over native NT system calls today: GetCurrentProcessId, GetTickCount, CreateFileA, ReadFile and CloseHandle. The full Win32 SDK and packaged releases are planned.
19 roadmap sections to goTechnically, no. Practically, also no. But we're doing it anyway.
Absolutely not. But why let a little thing like impossible odds get in the way? We're giving it a go anyway.
Highly unlikely. We might not even have a functioning classic control panel right away, but that isn't going to stop us from trying to build a better experience from the ground up.
No. The market is completely saturated. But here it is anyway. Enjoy.
If your definition of "daily use" involves sudden kernel panics and debugging C at 3 AM, then yes, it's perfect. Otherwise, proceed with extreme caution.
Sometime between next month and the end of time. Good software takes time; impossible software takes just a little bit longer.
No. We just have a severe, incurable allergy to people telling us something is "impossible."
Impossible OS is fully open source. Star the repo, download the alpha, or contribute code.