Meet Blackbox: The Heart of My AI Playground

Welcome to my blog — a blend of thoughts, experiments, and digital tinkering. I thought I’d kick things off by introducing Blackbox, the machine that made all of this possible.

🛠️ The Origin Story

Blackbox is a Lenovo ThinkCentre M710q that I picked up secondhand. I was looking for something compact, quiet, but powerful enough to handle my growing interest in AI, local large language models, and self-hosted tools. This machine ticked all the boxes.

🔍 Specs at the Time of Acquisition

• CPU: Intel Core i5-7500T (4 cores, 4 threads @ 2.7GHz)
• RAM: 8GB DDR4 (single channel)
• Storage: 256GB NVMe SSD
• Size: Tiny. Like, fits-in-your-hand tiny.
• Noise: Practically silent — a huge plus for having it running 24/7.

🧰 Upgrades & Mods

Since getting it, I’ve made a few key upgrades:

• RAM: Added a second 8GB DDR4 SODIMM , bringing the total RAM to 16GB. This has made a huge difference when running multiple Docker containers and local LLMs.
• Storage: Upgraded to a 500GB Samsung SSD 980 NVMe
• Added a 1TB Crucial SSD, mounted at /mnt/Crucial1TB, which is now where I store most of the data, models, media, and anything that doesn’t need to hit the NVMe. There's no drive caddy, but I've got one on the way.
• Cooling: No changes needed — the thermal design on this thing is surprisingly good, even under long workloads. I did notice upon opening the chassis and cleaning it up that the thermal paste was really old, so that's something I need to address.

🚀 Future upgrades

• SSD Caddy
• Reapplication of thermal paste
• Wi-Fi card installation
• BIOS update
• 32 or 64GB RAM

🧠 What It Runs

Blackbox is my:

• 🧠 Local AI serving Ollama and OpenWeb UI:
  • hermes3:3b (default)
  • phi2.7b
  • mistral:7b-instruct
  • gemma2:b
• 🔎 Search proxy (custom Hermes-based search via CLI and web)
• 🌐 Tailscale node
• 🏴‍☠️ Torrent machine (qBittorrent-nox) to send to an N5105 mini PC running Jellyfin
• 🐳 Docker playground (Ghost, SearxNG, Portainer, etc.)