Writing
A series on building an AI-enriched family photo archive — from homelab infrastructure through OCR, vision captioning, vector search, and retrieval. Grouped by the project each piece belongs to, newest first.
Project: Homelab
Building an enterprise-grade Proxmox micro-datacenter from six retired PCs — the foundation everything else runs on.
Latest
Another Piece Added to the Homelab
Have you seen the price of new GPUs?
August 29, 2026
More in this project
Building Enterprise Virtualization from Retired Hardware
Six retired PCs, a $514 budget, and the case for a resilient homelab.
August 7, 2026
Project: STRIPE
The Scan-Tag-Recall Image Pipeline Engine: ingest, OCR, vision captioning, and vector search that make a family photo archive searchable.
STRIPE: A Scan-Tag-Recall Image Pipeline for the Family Archive
Seven stages — ingest, queue, OCR, caption, cluster, index, retrieve — that turn a pile of scans into something you can search.
September 20, 2026
Project: Neanderthal
Pushing an archaic 2004 HP Pavilion a450n to its computational limits to coax modern LLM inference out of 32-bit legacy silicon.
Neanderthal: Coaxing LLM Inference from a 2004 Pentium 4
One archaic HP Pavilion a450n, 32-bit silicon, and a stubborn question — can it run a language model at all?
October 10, 2026