• Money for Nothing, Tokens for Free: Sovereign AI on a Budget Redux

  Posted on July 19, 2026 by pds aihardwarehomelab

  A more technical follow-up to the sovereign-AI build. I tore out llama.cpp for vLLM, swapped the Qwen 3.6 models for Ornith 1.0, and finally got all four MI100s working as one big model instead of four islands. And now I am doing the math on whether eight cards and a 397B model is money well spent, or just money for nothing.

  Continue reading →

• Champagne Taste, Beer Money: Building Sovereign AI on a Budget

  Posted on June 25, 2026 by pds aihardwarehomelab

  Sovereign AI is having a real moment, and underneath the buzzword is an old idea about owning the thing you depend on. So I pointed the same instinct that's powered my homelab since the '90s at AI, on four used datacenter GPUs and a lot of elbow grease.

  Continue reading →

• Fun and Profit with UNAS Pro 8: NVMe Cache the Hard Way

  Posted on January 31, 2026 by pds hardwarehomelab

  Otherwise known as how to void your warranty but have working stuff… standard disclaimer applies. Before you read further, this is not some glowing gushing Youtube talking head review… go there if that’s what you’re into. Yes, I’m posting this

  Continue reading →

• Claude Context at Scale: Managing Sessions Across Environments

  Posted on January 24, 2026 by pds ai

  I’ve been using Claude Code heavily for infrastructure work across multiple environments – homelab, production, various projects. The problem I kept running into: context loss between sessions. You know the drill. You’re deep into debugging something, have Claude fully up

  Continue reading →

• Home Lab (2022 Edition)

  Posted on December 23, 2022 by pds hardwarehomelab

  It’s nearing the end of 2022 and it’s time to refresh the homelab again. I recently re-allocated two of the three custom built Supermicro servers previously in the home network cabinet over to a colo cluster, creating a need to

  Continue reading →

Archives →