I Revived the First Raspberry Pi Model B (2011) as an AI Sandbox. Most AI frameworks assume GPU farms. This one assumes you have an SD card, a coffee, and curiosity. The Raspberry Pi Model B was the first commercially successful single-board computer launched in 2011 for $35 with 256 MB RAM. Almost 15 years later, I built Datapizza-AI PHP on it — an ultra-minimal educational framework that teaches how RAG, embeddings, vector search, and AI agents work using pure PHP on vintage hardware. Why? - Prove that understanding beats horsepower - Run on the original RPi hardware from 2011 (256 MB RAM, ~3 watts) - See every API call, every cosine distance, every decision - Learn AI without Python, without Docker, without dependency hell Architecture: - API-first embeddings (OpenAI, DeepSeek, Anthropic) - Local vector store (plain JSON) - Cosine similarity in pure PHP - RAG pipeline, agents, memory, tools - Everything hand-coded, readable, hackable GitHub: https:/