Not a new kernel. A mediator at the syscall boundary.
AdaOS is a kernelless OS: an agent sits as a man-in-the-middle gateway at the
syscall/init level — via ptrace, seccomp, or a custom PID 1 —
observing and eventually mediating what would otherwise be direct kernel calls, without replacing
the kernel itself.
This is genuine systems-research work, not a Linux distro with a new wallpaper.
application process
│ syscall (open, read, exec, connect, …)
▼
┌─────────────────────────────┐│ AdaOS interception point │← Phase 1: observes here, read-only│ (ptrace / seccomp / init) │← Phase 2+: mediates here└─────────────────────────────┘
│ (unmodified syscall, or a mediated one)
▼
Linux kernel
Why not just fork a kernel?
Forking or patching the kernel means tracking upstream forever and trusting a research project with
ring‑0. Sitting at the syscall boundary instead keeps the real kernel in charge of memory,
scheduling, and hardware — AdaOS only gets a look at what a process asks the kernel to do,
and, later, a say in how that request is handled. Smaller attack surface, no custom kernel to maintain,
and the interception point can be swapped (ptrace today, something faster later) without touching
anything above it.
Where it actually is
Phase 1
Real interception point proving viability
A custom init process or a ptrace/seccomp-based wrapper that observes real syscalls from a real running process. Read-only interception, proven before any automation logic gets built on top of it.
Phase 2+
Active mediation scoped once Phase 1 lands
The agent actively mediates and automates specific syscalls — not yet ticketed. Scope gets written once the interception point is proven, not before.
Later
AdaDo userland
A de-Googled base, Ada integration, the app store — sits on top of a working kernelless foundation, not before it.
What this is — and isn't, yet
Is:
A real, phased systems-research roadmap, each phase independently verifiable
Part of AdaDo's distribution strategy, alongside the mobile OS and 3-tier app store