TerminalBytes
A blog by a backend dev who keeps over-engineering things in the basement and writing them down. Solar-powered iPhone OCR servers, Kubernetes clusters on N100 mini PCs, Kindle 7th-gens turned into homelab status displays, an iOS audiobook client built from scratch because none of the existing ones do what I want. If you’ve ever asked “why would you even do this?” about a project, you’re in the right place.
Also: I absolutely love my iPod Classic. Still the best music player ever made.
The “best mini PC” posts are the ones I actually bought from. The setup guides include real electricity-bill numbers and real Amazon prices. Recent rabbit holes: Strix Halo mini PCs doubling in price in six months , the Mac Mini being the wrong default for OpenClaw , and running Minecraft servers on a $400 mini PC for two years .
Recent Posts
-
Back up Hugging Face models to a NAS before they vanish
Models do disappear from Hugging Face. How I keep a local copy I can verify: pinned hf download, SHA256 manifests, Ollama's blob store, and a NAS or big drive.
-
PHP2Shell and Plex2Shell: lock down your homelab in 2026
Plex2Shell and PHP2Shell hit home servers in one week. What 11 Docker images really ship, which ports to close, and how to go VPN-only.
-
My mini PC costs $3 a month. The VPS math for 2026
Two months of smart-plug data from my home mini PC: 22 kWh and CA$3.08 in August, priced against DigitalOcean, Hetzner and Oracle's free tier in 2026.
-
Ollama vs llama.cpp vs MLX on a Mac Studio, measured
Qwen3.5 9B through Ollama (GGUF and MLX), llama-bench, mlx-lm and rapid-mlx on an M3 Ultra. Measured tokens per second, memory, and what MTP changes.
-
Mac mini alternatives for local LLMs: M6, M5 and Strix Halo
Apple's M6 Mac mini and M5 Ultra Mac Studio versus 128GB Strix Halo mini PCs for local LLMs, from a 256GB M3 Ultra owner. Bandwidth, price, measured t/s.
-
SpiderCat: jailbreak a Kindle on firmware 5.19.5
SpiderCat jailbreaks almost any Kindle on firmware 5.16.3 to 5.19.5 by opening a book. Patched in 5.19.6. Which tool fits your firmware, and how to run it.
-
Jellyfin 12 on a mini PC: the spy-free TV setup for 2026
LG TVs got caught logging audio and scanning your LAN. How I run Jellyfin 12 on a $280 N150 mini PC as an offline-TV media box, with real upgrade numbers.
-
Run Qwen3.8 27B locally: real numbers from my Mac Studio
Qwen3.8 27B benchmarked on a Mac Studio M3 Ultra: Ollama and llama.cpp numbers, the 1-bit quant tested, RAM math, and the hardware that actually fits.
-
How to run DeepSeek V4 Flash locally (Mac, Linux, Windows)
How to run DeepSeek V4 Flash locally on Mac, Linux, or Windows: hardware that fits 167GB, llama.cpp setup, and realistic speed expectations.
-
RomM 5: self-host your retro game library on a mini PC
RomM 5 turns a mini PC into a self-hosted retro game library with browser play and save sync. Setup guide plus the hardware tiers that make sense.