Jacob Wiseberg

I finally made the jump I’d been putting off since I wrote about keeping WSL alive: the Beelink is now running Omarchy instead of dual-booting its way around Windows. Paired with that, I bit the bullet and picked up a Claude subscription instead of running everything through pay-as-you-go API credits. Both changes were overdue.

Why I kept avoiding Linux on the shared machine

The Beelink has always been a weird hybrid for me. It’s powerful, cheap for what it does, and shared with my fiancée as our household’s backup storage and document processor. I was hesitant to touch the OS because I didn’t want to turn a “safe space” appliance into something that broke every time I wanted to tinker. WSL was my compromise: Fedora in a box on top of Windows, good enough to scratch the Linux itch without risking the stuff my fiancée actually depends on.

Omarchy changed my calculus. It’s an opinionated, pre-configured Arch and Hyprland setup, which means most of the config work I’d normally spend a weekend on (window manager, keybindings, theming, the usual rabbit hole) comes already decided for me. That mattered a lot after the whole .config Ship of Theseus exercise. I didn’t want a fourth from-scratch dotfiles project; I wanted somebody else’s well-reasoned defaults that I could deviate from selectively.

What actually changed on the machine

I kept the household-critical stuff isolated and let Omarchy take the rest:

• Documents and shared storage moved to a dedicated drive mount, untouched by the OS swap.
• Hyprland replaced whatever window juggling I was doing between WSL and Windows natively.
• My existing neovim and tmux configs from the .config repo dropped in almost unchanged, since I’d already done the work of stripping them down to just what I control.
• Tailscale and mosh carried over directly, so the “access my Linux box from my laptop” setup from the WSL days kept working without modification.

The net result is that the machine my fiancée uses for photos and paperwork looks and feels exactly the same, while the terminal environment underneath it is finally native Linux instead of a VM pretending to be one.

Leveling up with Claude instead of staying API-only

On the AI side, I’d been running OpenCode against whichever provider had the best price-to-performance that week. That flexibility is still valuable, but it has a cost: constant plan-switching, re-reading pricing pages, and never quite settling into a workflow long enough to get good at it.

A Claude subscription flips that tradeoff. Instead of metering every agent run against a bill, I get a flat cost and Claude Code as a first-class terminal tool instead of something I had to wire in myself. It’s the first time my AI tooling cost has been predictable month to month, and predictability turns out to matter more to my actual usage than squeezing out the cheapest token.

I didn’t drop OpenCode. It’s still my daily driver for anything where I want full control over which model answers, and it’s still how I run the local Ollama setup I wrote about. But Claude Code on Omarchy has become where I reach for anything that benefits from a tighter, more integrated terminal experience and a plan I don’t have to think about.

Where this leaves me

Between this and the Hermes/Ollama setup, I’ve landed somewhere I’m actually happy with: a Linux machine that’s mine to tinker with but doesn’t put my fiancée’s files at risk, a subscription that removes the cost anxiety from the AI tools I lean on daily, and a local fallback for anything I don’t want leaving my network. None of this is final, it never is, but it’s the first setup in a while that feels like it’s built to last rather than something I’ll scrap in three months out of boredom.