AWE development project #2 — active, preserved, and intended for long-term development.
AWE_OS CellKernel is a Rust-first operating-system project centered on a small privileged CellKernel, a dedicated AWE boot chain, capability-oriented security, modular drivers/services, and a native userspace platform.
Status: Development project — not a finished daily-driver OS.
A historical snapshot of the repository before the 2026-08-27 project-state/documentation restructuring is preserved in
archive/pre-restructure-2026-08-27.
The goal is to turn the existing architecture into a progressively more executable, testable, and reproducible operating-system platform. Development is evidence-driven: a subsystem is not considered complete merely because an interface or design document exists.
UEFI / Firmware
|
AWE Loader
|
AWE Boot Protocol
|
CellKernel
CPU • memory • interrupts • scheduler
processes • syscalls • IPC • capabilities
|
Device / System Services
PCI • ACPI • VirtIO • storage • network • input • display
|
AWOSA Runtime
packages • manifests • sandbox • capabilities
|
AWE Capsule
|
AYUI + Native Applications
The repository already contains substantial kernel, boot, hardware-contract, storage, networking, userspace, package, update/recovery, and CI foundations. These components are being developed toward end-to-end runtime validation rather than treated as a completed product.
Important existing areas include:
- AWE loader and versioned boot information contract.
- x86_64 paging, early identity mapping, CR3 activation and interrupt primitives.
- Process, scheduler, syscall, IPC, capability and security foundations.
- PCI, ACPI, APIC/IOAPIC and VirtIO contracts/execution primitives.
- Bounded block storage, GPT, VFS, journaling and recovery foundations.
- IPv4/UDP/TCP transport foundations and bounded socket handling.
- Versioned userspace service and identity contracts.
- AWOSA runtime and native
.asd/.awospackage-admission foundations. - A/B update and recovery state-machine foundations.
- Automated formatting, compilation, tests, Clippy and boot-related CI workflows.
These are implementation foundations, not a claim that every subsystem is already production-ready.
The old repository documentation used percentage checkpoints. Those checkpoints are retained as historical engineering context, but the active project now uses runtime evidence and milestone gates as the primary measure of progress.
See PROJECT_STATUS.md for the current project state and development rules.
- PROJECT_STATUS.md — current project identity, development direction, preservation point, and next cycle.
- PRODUCT_READINESS.md — evidence-based readiness gates.
- ARCHITECTURE.md — system architecture and boundaries.
- docs/AWE_OS_100_PERCENT_MASTER_PLAN.md — long-term implementation plan.
- docs/ENGINEERING_GUIDE.md — engineering workflow and invariants.
- docs/ABI.md — boot ABI contract.
- docs/BOOT_ARCHITECTURE.md — boot flow.
- docs/DRIVER_SYSTEM.md — driver architecture.
- docs/AWE_CAPSULE.md — native application security model.
- docs/AWE_XENOSENSE.md — experimental research direction.
Historical progress/evidence files under docs/ remain useful as dated records. They should not override the current status in PROJECT_STATUS.md and PRODUCT_READINESS.md.
- Reliable x86_64 UEFI/QEMU boot.
- Deterministic CellKernel entry and early initialization.
- Executed memory/paging and interrupt validation.
- Timer/scheduler execution evidence.
- VirtIO end-to-end device exercises.
- Persistent storage runtime.
- Network runtime and recovery.
- Userspace service startup and supervision.
- AWOSA loader/runtime integration.
- Cryptographic trust for
.asdand.awospackages. - Native application lifecycle and sandbox execution.
- A/B update and recovery exercised end-to-end.
- AYUI compositor and windowing.
- Terminal and first-party applications.
- Filesystem and network managers.
- SDK/App Builder foundations.
- Security hardening.
- Fuzz/stress testing.
- Reproducible release builds.
- Hardware compatibility matrix.
- ARM64/RISC-V64 validation.
- Release certification based on evidence.
- Keep privileged code Rust-first and
no_stdwhere appropriate. - Keep the trusted computing base small.
- Keep hardware-dependent code behind explicit interfaces.
- Treat firmware, memory, process, IPC, ABI, and capability boundaries as security boundaries.
- Prefer deterministic validation and bounded resource use.
- Never describe planned work as implemented without evidence.
- Experimental intelligence/recovery concepts must remain subordinate to deterministic security and must not silently modify the trusted computing base.
AGPL-3.0.