Integrating AI With Retro Hardware:A Commodore 64 LLM Module
summary
Built a Commodore 64 LLM module that sits on the user port, uses a tiny ARM board with 8 GB RAM and a lightweight embedding model (<0.5 GB), and serves queries via retrieval augmented generation using a vector DB of ~100 cleaned PDFs and 368 scanned 1980s books. Hardware fits inside a 3D‑printed case, communicates over a MAX232-style serial interface, and demos show RAG answers are accurate while a vanilla small LLM hallucinates.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 2:12Commodore 64
- 3:39Commodore 64 Community
- 5:57Challenge
- 7:42First AI
- 12:21Unboxing
- 13:28Raspberry Pi Zero
- 15:50Tiny Lama
- 16:48Retrieval Augmented Generation
- 18:21How This Works
- 20:09Embedded Model
- 20:49Testing Board
- 21:27Second Attempt
- 22:00Demo Time
- 29:37Why it is not easily replicable
- 33:12Questions
related talks
50mComputer History RetroSpective: From Space Race to genAI with Live Demos
47mOwn Your LLM: Running Generative and Agentic AI Locally
01h 49mAI engineering with open access LLMs that lie, curse, and steal
38mEmpowering AI Education with Rag: Building Domain-Specific LLM Agents for Enhanced...
23mLLMs on the Edge: The Future of On-Device Intelligence
51mAvoid common LLM pitfalls
20mSelf-hosting AI LLMs - a beginners guide
01h 33mRAG How To Talk To Your Data