Risk Register

Risk Register

Purpose

This page tracks project risks that could make XAI OS hard to build, hard to benchmark, or unsafe to operate.

Risk Table

ID	Risk	Impact	Mitigation	Related page
R-001	Overclaiming Linux/BSD speedups	Damages credibility	Treat 10-45%, 3-18%, and 2-12% as design targets until measured	[[Performance Targets
R-002	Losing documentation during wiki moves	Future Codex sessions lose context	Keep canonical wiki pages and run link/page checks before docs cleanup	Home
R-003	Driver scope explosion	Bootable prototype stalls	QEMU serial, VirtIO block, and VirtIO net first; defer broad hardware	[[Driver Model
R-004	ARM/NVIDIA firmware opacity	AArch64 port blocked or delayed	Prefer UEFI/SBSA where available and track board-specific gaps	[[ARM NVIDIA Port
R-005	AI agent unsafe code modification	App or host compromise	Review-before-apply default, scoped permissions, rollback, audit logs	[[Security Model
R-006	Weak benchmark methodology	Invalid performance claims	Require tuned Linux/BSD baseline and JSON result artifacts	[[Testing and Benchmarking
R-007	Memory gains too small vs tuned Linux	Core target not proven	Measure raw bandwidth separately from effective AI bandwidth	[[Memory System
R-008	TCP/UDP gains small vs kernel-bypass Linux	Network target overstated	Compare against kernel TCP and relevant kernel-bypass baselines	Networking
R-009	Full Swift support too early	Runtime scope expands before boot stability	Keep Swift as later optional runtime work	[[Open Decisions
R-010	Project scope creep	OS fails to reach bootable prototype	Keep QEMU on macOS first and one work package per commit	[[Codex Work Packages

Review Cadence

Review this page at the end of each implementation phase. Add new risks when acceptance criteria fail for architectural reasons.

Related Pages

• Open Decisions
• Implementation Plan
• Testing and Benchmarking