Old laptop runs local AI model replacing ChatGPT

An Acer Predator Helios 300 from 2017 with a GTX 1060 runs Gemma 4 E4B locally via llama.cpp on Proxmox. The setup achieves 30+ tokens per second. The author replaced ChatGPT's free tier for privacy, pairing the model with Paperless-ngx, Open Notebook, and other self-hosted tools.