GitHub radar
Understand how language models work: from word counts to attention
A free interactive course in your browser walks through the whole history of text-processing AI, ending at the attention mechanism behind today's transformers. Finish it and you'll understand how models like the ones behind Claude actually work.
It's an interactive educational post with clickable diagrams and hands-on exercises covering the evolution of NLP — from early methods to transformers.
When it helps
Useful if you want to understand how language models were invented, not just use them. It's not a productivity tool — purely educational.
Pros
- Free and runs in the browser, no install needed
- 11 sections with interactive diagrams and live demos
- Exercises on real datasets with ready starter code
- Licensed CC-BY 4.0 — free to copy with attribution
Cons
- Doesn't speed up your work or replace an AI agent
- Author recommends doing exercises by hand, without AI
How to set it up — step by step
- 1Open https://huggingface.co/spaces/dlouapre/nlp-before-llms in your browser — nothing to install.
- 2Read the first section on n-gram models and explore the interactive diagrams.
- 3Send your agent the text from the block below to run the demo and explain the logic.
- 4Ask your agent to walk through the next sections the same way — bag of words, word2vec, LSTM, BERT, attention.
- 5Test your understanding with the exercises on the IMDB and WikiText-2 datasets using the provided checkpoint.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Open this interactive course https://huggingface.co/spaces/dlouapre/nlp-before-llms and walk me through the first section on n-grams: run the demo inside it and explain what it shows.
Nothing to install — it runs in the browser.
Open on Hugging Face▌ More finds