WhiskeyLab

Homelab builds, machine learning, local AI, and honest tech and gadget reviews.

Follow @whiskeylab.io Media inquiries

The blueprint

Tinkering, testing, and building it myself

I'm a tech enthusiast who spends most of my free time in the homelab: building servers, wiring up networks and seeing how much I can run on my own hardware. Lately that means machine learning and local AI, with models running privately at home instead of in someone else's cloud.

WhiskeyLab is where I share it all. You'll find the builds, the experiments that worked and the ones that didn't, and straight-talking reviews of the tech and gadgets I get my hands on.

Homelab

Servers, networking, self-hosting and small, efficient builds that run around the clock.

Machine learning

Training, tuning and benchmarking models, and finding out what everyday hardware can really do.

Local AI

Private assistants and models running at home: no subscriptions, no data leaving the house.

Tech & gadget reviews

Hands-on, honest takes on hardware and gadgets, tested in real use rather than on a spec sheet.

From the lab

Where it all happens

A few shots from around the WhiskeyLab: the desk, the rack and whatever just came off the printer.

Triple-monitor desk setup with matching orange and red wallpapers, an orange desk mat and a mechanical keyboard
The command centerThree screens, one color scheme.
A bright green honeycomb-pattern 3D-printed ball held in front of a Bambu Lab P2S printer
Fresh off the printerBambu Lab P2S
A small wooden homelab rack with a Palo Alto firewall, a network switch, an HP mini PC, four-bay storage and a touchscreen on top
The homelab rackFirewall, switch, mini PC and storage
A small retro-style monitor showing The Great Wave off Kanagawa, sitting under the main monitors
Retro cornerA tiny vintage-style screen on the desk.

Featured project

The Ark: an offline library and private AI for the house

The Ark is one small computer that keeps working when the internet doesn't. Any phone, tablet or laptop on the home Wi-Fi opens one dashboard and gets all of Wikipedia, medical and repair guides, full courses, books, music, an offline map and a private AI chat. Nothing ever leaves the house.

It's free and open source, so anyone can build one on their own hardware.

Inspired by Project N.O.M.A.D. The idea started with Project N.O.M.A.D. from Crosstalk Solutions: offline knowledge, courses, maps and local AI behind one dashboard. I loved the concept and rebuilt it my way, with the tools I wanted, extra features and a lot of tuning for a small, quiet mini PC.

What I customized

  • Ask the ArkA local AI that can search the whole library, look at photos, and listen and talk back with offline voice. The models get plain names like "Quick Answers" and "Deep Thinker".
  • Ark MechanicA repair assistant that knows your own vehicles and equipment and reads the service manuals you add.
  • More on the shelfA bookshelf for books and PDFs, a music server for your own library, shared notes and a data toolkit, all on the same dashboard.
  • Regional map and mesh radioA street-level offline map of the area you pick, plus Meshtastic and MeshCore clients for off-grid LoRa messaging.
  • Tuned for small hardwareFits on a single 512 GB drive. The AI runs fully on a 6 GB GPU at about 37 words a second, and the whole box draws 25–45 W.
  • Easy for the whole familyOpen it at ark.local, follow a printable Getting Started guide, and check live GPU and power readings on the dashboard.

A look inside

The Ark's Getting Started guide, with links to every section
The built-in Getting Started guide
The Reference Library section: Wikipedia, medical, survival, repair and learning collections
What's in the offline Reference Library
Ask the Ark: choosing a local AI model, with a reminder to double-check important answers
Ask the Ark, the private AI chat

Build your own with one command

On a fresh Ubuntu 24.04, Linux Mint 22 or Debian 12 machine, open a terminal and paste:

sudo apt install -y git && git clone https://github.com/whiskeylabbuilds/the-ark.git ~/ark && cd ~/ark && ./install.sh
  • The installer asks before each big step and is safe to run again.
  • 16 GB RAM recommended, about 500 GB of disk. A graphics card is optional (NVIDIA or AMD).
View on GitHub Full instructions, hardware notes and the MIT license are in the README.

Work with WhiskeyLab

Media & review inquiries

Have hardware or tech you'd like reviewed? I'm open to:

  • Hardware and gadget reviews
  • Homelab, networking and server gear
  • AI and machine learning hardware and software
  • Collaborations and other media requests

Email

Send an inquiry