AI by Hand: A Project Teaching AI Through Hand Calculations, Led by a University of Colorado Professor
A research project and learning resource that invites readers to work through the math behind AI models themselves — using Excel or paper — to transparently visualize how Transformers and LLMs work
The "AI by Hand" project is gaining attention from the tech community after being shared on Hacker News, where it received 98 points. It is a learning resource founded by Professor Tom Yeh, a faculty member in the Computer Science department at the University of Colorado Boulder.
The project's core idea is to build a deep understanding of AI models by manually working through the underlying algorithms and fundamental mathematics, often using simple tools like Excel or paper instead of relying on code or ready-made libraries. This approach makes complex AI architectures — such as Transformers, LLMs (large language models), and Neural Networks — visible and verifiable step by step.
AI by Hand content is published via the Substack platform under the name "AI by Hand ✍️", along with a members' resource archive and live seminars. There is also a GitHub repository containing Excel-based exercises and a YouTube channel for video-based instruction, allowing interested learners to choose the format that suits them best.
In an era where most people use AI through ready-made tools without understanding the mechanisms behind them, the hand-calculation approach helps Thai students and professionals truly grasp how AI models work — a fundamental skill for anyone serious about a career in AI. It is freely accessible via the website.