🖥️ Tech

Homelab, self-hosting, local AI, and home-server gear — written from a box that actually runs it.

What running NixOS is actually like day to day, on a server and a laptop

NixOS describes a whole machine in config files and keeps every previous version bootable. Here is how that works, what the learning curve looks like, and why rollback changes how you experiment.

Fixing my OnePlus 12 myself: back glass, charging port and battery

My OnePlus 12 had a cracked back, a charging port that quit, and a tired battery. I replaced all three at home in a few hours. Here's what I bought and what was actually hard.

The best mini-PC for a home LLM server (and when a mini-PC is the wrong call)

What actually matters when you buy a small, quiet box to run local LLMs — memory bandwidth over core count — plus the three setups worth buying in 2026 and the one case where a tower still wins.

2.5GbE at home: the switch and adapters worth buying

2.5GbE is the rare homelab upgrade that's actually cheap now: it runs on the Cat5e/Cat6 you probably already have. Here's how to pick between unmanaged and managed switches, and which adapters are worth it.

The best cheap network cable testers for a homelab

You don't need a pro-grade certifier to catch the Ethernet faults that actually bite a homelab. Here's what a cable tester needs to do, which cheap ones are worth it, and when to reach for something bigger.

Cat5e vs Cat6 vs Cat6a: what actually matters at home

The Cat5e/Cat6/Cat6a decision matters a lot less than the cable aisle wants you to believe. Here's what each rating actually buys you at homelab distances, and where your money is better spent.

Git-push-to-deploy for your homelab: stop SSHing in to change things

If the only record of how your homelab is configured is your shell history, you don't have a homelab, you have a pet. Put every stack in a git repo and let a push be the deploy.

How to crimp your own Ethernet cables (and the three tools you need)

Crimping your own Ethernet cables pays for itself the first time you need a 14-foot patch cable and the drawer only has 3-foot and 25-foot ones. Here's the three tools that make it easy and the mistakes that ruin a crimp.

A 10-inch mini rack: tidy a small homelab without a full-size rack

A full 19-inch rack is overkill for a switch, a mini PC, and a patch panel. Here's why a 10-inch mini rack fits a small homelab better, and the shelves, patch panel, and cable management worth adding to it.

Pin your Docker image tags: `latest` is not a version

Running `:latest` in your homelab means every restart is a surprise upgrade you didn't choose. Pin versions, upgrade on purpose, and let a bot open the pull request.

What to actually buy for a Raspberry Pi homelab node

A Raspberry Pi is cheap; the accessories that make it a reliable homelab node are where people get it wrong. Here's the power supply, storage, case, and PoE decisions that actually matter, and the ones that don't.

3-2-1 backups for a homelab, on a budget

RAID is not a backup. Here's the 3-2-1 rule explained for real homelabs, a cheap way to actually implement it, and the one step everyone skips that turns a backup into a false sense of security.

The Bambu Lab A1 Mini + AMS Lite: the multicolor 3D printer that earns its bench space in a homelab

Why the A1 Mini is the printer I recommend to homelabbers who want parts, not a project — what the AMS Lite actually does, where the 180mm build volume bites, and how to think about it before you buy.

The best drives for a home NAS: what to buy, what to avoid, and why

Not all hard drives belong in a NAS. Here's how to choose drives that survive running 24/7 in a vibrating multi-bay enclosure, the technology to avoid, and how to buy capacity without overpaying.

The best mini PCs for an always-on homelab (and what to ignore on the spec sheet)

A mini PC is the cheapest way to run services 24/7 without a power bill that hurts. Here's what actually matters when choosing one for a homelab, the specs that are marketing, and how to pick by workload.

The best UPS for a homelab (and how to size one without overpaying)

A UPS isn't about running through a blackout — it's about surviving the two-second blip that corrupts a database mid-write. Here's how to size one for a homelab, what features actually matter, and what to skip.

The best used GPUs for a home LLM server (2026): buy VRAM, not hype

For running local LLMs at home, the used market beats new almost every time — if you buy for VRAM. Here are the cards worth hunting for at each budget, what to avoid, and how to check a used GPU before you pay.

Cheap smart plugs to measure what your homelab actually costs to run

You can't optimize a power bill you can't see. Here's how a $10 energy-monitoring smart plug tells you exactly what your homelab costs per month — and pays for itself the first time it catches an idle power hog.

How to reach a homelab service from anywhere without opening a single port

Port forwarding puts your home network on the public internet's target list. Here are the safer ways to reach your services remotely — overlay networks, tunnels, and reverse proxies — and when each one fits.

Give every self-hosted service its own subdomain (reverse proxy 101)

Stop remembering port numbers. Here's how a reverse proxy plus wildcard DNS turns http://192.168.1.10:8096 into photos.yourdomain.net — with real HTTPS — and why it's the single best quality-of-life upgrade for a homelab.

A budget home-recording setup for documenting your homelab (the mic and headphones I actually use)

If you want to make videos, voiceovers, or calls about your homelab that don't sound like a laptop mic in a shoebox, you need two things: a real microphone and closed-back headphones. Here's the affordable pair I use and why.

How much VRAM do you actually need to run a local LLM?

A plain-English guide to sizing GPU memory for local models: how to estimate VRAM from parameter count and quantization, why context length eats more than you expect, and the tiers that actually matter when buying.

I built a 9-in-1 USB cable tester from a kit — and it fixed a problem I didn't know I had

Assembling a cheap solder-it-yourself USB cable tester taught me why so many cables in the drawer were quietly failing, and gave me a 10-second way to catch them. Here's the build and what it's good for.

Nginx Proxy Manager vs Traefik vs Caddy: which reverse proxy for your homelab?

The reverse proxy is what turns a pile of containers into service.yourdomain.net with automatic HTTPS. Here's the honest trade-off between the three most popular choices, and which to pick for how you work.

Ollama vs llama.cpp vs vLLM: which one should run your home LLM server?

Three ways to serve local models, three different priorities: easy, lean, or fast-at-scale. Here's the honest comparison for a homelab, and how to pick without over-engineering.

Self-hosting behind Cloudflare: the caching gotchas nobody warns you about

Cloudflare in front of a home server is free and worth it, but its cache will happily serve your visitors a stale site for hours. Here's what it caches, why a fix can look like a failed deploy, and the cache rules that make updates land instantly.

Tailscale vs WireGuard vs a self-hosted VPN: which one for your homelab?

They all use the same fast, modern encryption — the real difference is who does the hard part. A plain-English comparison to help you pick how to reach your homelab from anywhere.

Why half your USB cables are 'charge-only' (and how to prove it in 10 seconds)

That cable that charges your phone but won't sync it isn't broken — it was built that way. Here's why charge-only USB cables exist, how to spot one, and the fastest ways to test any cable in your drawer.

Running a local LLM on a home GPU — and reaching it from anywhere with Tailscale

A practical walkthrough for self-hosting Ollama on a desktop GPU, the context-length setting that quietly wrecks performance, and how to reach it securely from your laptop without opening a port.