The article introduces NanoChat, an experimental harness for training large language models (LLMs) that can be run on a single GPU node, allowing users to train their own GPT-2 capability LLM for around $100. NanoChat is designed to be minimal, hackable, and accessible, with a focus on simplicity and ease of use. The project aims to improve the state of the art in micro models that can be worked with end-to-end on budgets of less than $1000. The article also discusses the potential applications and implications of NanoChat, including its potential for use in education and research.