I break things until I understand how they work. Currently learning kernel-level debugging, container orchestration, and what it takes to run unsupported hardware on macOS.
The work here is not polished output. It is the slow part: reading failures, removing bad assumptions, and learning what each patch actually changes.
Started with a GTX 970 on Ventura using OpenCore Legacy Patcher. What followed was over a year of CPU power management issues, GPU artifacts, kernel panics, missing boot logos, and no hardware acceleration.
The breakthrough came when I stopped layering fixes and cleared every OCLP request, then rebuilt the config from scratch: WhateverGreen patches, framebuffer edits, AGPM injection. It finally booted clean with acceleration. Still learning why each patch works.
Older device, checkra1n then palera1n. Multiple failed attempts: apps would not install, random reboots, boot loops.
Learned to read panic logs, understand the boot chain, and not trust every tutorial. Eventually got a stable jailbreak, but the real lesson was debugging a system I could not just reinstall.
The stack is practical, but the point is not the logo list. It is learning how each layer behaves when it becomes the machine I rely on.
Main development machine. Chose Arch to learn the stack from the bottom up: no hand-holding. Running Wayland on KDE, still smoothing out the rough edges with fractional scaling and GPU offload. Using it for code, containers, and learning how a modern Linux desktop actually fits together.
Running a few VPS instances with Docker Compose. Plex for media, CasaOS as a dashboard, and a rotating cast of self-hosted services I am experimenting with: monitoring, backups, auth. Learning to operate infrastructure, not just deploy it.
Not training models: running them. Experimenting with local LLMs and tooling around them: RAG pipelines, function calling, agent loops. Building small projects to understand where local inference fits and where it does not.