Start Your AI Homelab — blog series

A beginner-facing, first-person field guide to building an AI homelab, published as a GitHub Pages (Jekyll) blog. The Markdown in _posts/ is the single source of truth — Word .docx review copies are generated from it, never edited separately.

The series

#	Post	What it covers
00	Start Here	What an AI homelab is, why bother, the baseline you need
01	AI foundations	Models, tokens, context, agents — in plain English
02	Choosing your stack	Subscriptions vs APIs vs local models; model landscape & trade-offs
03	Prepare the VM	Hypervisor, base Linux VM, snapshots, first boot
04	Secure the box	SSH hardening + why physical access is the real boundary
05	Starter projects	Five things to build first, easy → ambitious

Generate the Word review copies

# Highest quality (recommended): install pandoc once
sudo apt-get install -y pandoc
python3 build_docx.py        # writes docx/*.docx

# No pandoc? The builder falls back to python-docx automatically:
pip install python-docx
python3 build_docx.py

Output lands in docx/ (gitignored). 00-MASTER-all-posts.docx is the whole series in one file for a single review pass.

Preview the site locally (optional)

GitHub Pages builds Jekyll server-side, so this is only for local preview:

bundle install
bundle exec jekyll serve   # http://localhost:4000

Publish

1. Set a real git identity (git config user.name / user.email).
2. Create the GitHub repo (<username>.github.io for a root site, or a project repo).
3. git init && git add . && git commit && git push.
4. Repo → Settings → Pages → build from the default branch.
5. Confirm the live URL renders all six posts and navigation works.

Conventions

• Each post has a generic core plus a boxed "How I run mine" sidebar — anonymized, no real IPs / internal hostnames / full service inventory.
• Before any push, run the anonymization check:

  grep -rEn '192\.168\.|10\.|your-internal-hostname|your-tailnet' _posts/ index.md && echo "LEAK — scrub before publish" || echo "clean"