Local AI
Testing what can actually run well on consumer hardware, with practical evaluations instead of benchmark hype.
Practical Linux, self-hosting, and local AI.
Real hardware. Real testing. No hype.
A personal technology lab for local AI, Linux, homelabs, self-hosting, open source, software development, and practical experimentation.
Current Focus
Linux Systems
Arch, Debian, Ubuntu, Wayland desktops, terminal workflows, systemd services, and conservative system maintenance.
Homelab Infrastructure
Docker services, ZFS storage, DNS, VPN-routed workloads, dashboards, and documentation-first operations.
Featured Projects
Monolith
A local LLM workbench for model testing, prompt suites, eval imports, and context-scaling experiments.
Quant Lab
A Markdown-first evaluation workflow for testing local models against practical technical prompts.
Wumbo Core
A Docker-based homelab with ZFS storage, DNS, VPN-routed workloads, media services, and documentation.
Operating Philosophy
Stability before novelty
Working systems beat fragile experiments.
Documentation before memory
If a system matters, it should be recoverable.
Evidence before hype
Benchmarks need context. Claims need testing.