Series · 13 parts
The Quiet Machine
The log of turning an idle gaming PC into a private, always-on server the Mac drives from anywhere: local AI and a whole dev environment, on hardware I own, nothing rented back to me. The Mac is the brain, the box is the muscle — and half the work is turning things off on purpose.
- 01 Turning an Idle Gaming PC Into a Private AI Server A 2021 gaming PC sat under my desk for two years doing nothing. Decent GPU, wired Ethernet, powered on maybe once a month so Windows could patch itself…
- 02 Reclaiming 27.6 GB From an Old Dev Machine Android Studio, Flutter, Unity, JetBrains, the whole JVM zoo: all wiped months ago. And yet D:\ was still choking under 27.6 GB nobody was using.
- 03 Passwordless SSH Into Windows, Then Lock It Down The box sits in the corner, no monitor, no keyboard, set to auto-power-on after an outage. It comes back to the Windows lock screen, waiting for a PIN…
- 04 Reaching the Home Server From Anywhere With Tailscale Tailscale made the box reachable from the Mac: no port forward, no password ever crossing the internet.
- 05 The First Real Reboot, and the RDP Rabbit Hole I killed the power to the box for the first real reboot, a full shutdown and cold boot, and didn't touch a key again: no PIN entered, no console login…
- 06 Making the Box and Mac a Symmetric Server, No macFUSE Finder showed me exactly one shared folder from the box, the share I'd set up weeks earlier. Not the D:\ drive underneath it, where every project and…
- 07 Mirroring a Dev Machine Safely, on a Windows Box A tool that installs cleanly today becomes a "why doesn't it work when I log in as the other account" bug three weeks from now, reported by me, to me…
- 08 Claude Code Lands, and the Memory Framework Gets a Cleanup Installing Claude Code on the Windows box was one line:
- 09 GitHub Actions Self-Hosted Runner on a Windows Gaming PC Every CI run for this project (backend, Android, web) cold-started a fresh GitHub VM and re-downloaded half of NuGet and Gradle's internet, every single…
- 10 Local LLM Benchmarks on an RTX 3060: 35 Tokens a Second By this point the box has SSH, Tailscale, a self-hosted CI runner, and a memory framework that indexes thousands of files. None of that answers the…
- 11 A Difficulty Router That Never Calls an LLM to Decide The gateway had been up for a while, still without a router: a local gateway with a few local models and optional cloud slots, answering behind one…
- 12 Cold-Reboot Self-Recovery on an Unattended Windows Server Part 5 proved the box could survive a cold reboot: sshd answered at the lock screen, no PIN typed, nobody at the keyboard. That was real progress, and it…
- 13 QLoRA Fine-Tuning a Coding Assistant on an RTX 3060 The box already ran a local coding model wired to a RAG and memory layer: retrieval keeps it grounded in facts that change constantly, a long system…