Observed arrival · 2026-09-19
How AI Works: A Visual Laboratory
An interactive guide that traces a language model from text and tokens through vectors, weights, attention, transformers, and next-token generation.
- For
- Curious beginners learning language-model mechanics
- Worth noticing
- Its 18-part journey includes interactive tokenization, matrix calculations, attention, training, hallucinations, and a knowledge check.
Field notes
The page organizes its material as a navigable 18-step sequence, from tokenization and embeddings to parameters, hallucinations, and the complete pipeline. Controls include a forward-pass demonstration, animated matrix multiplication, selectable activation functions, and attention inspection. It also links outward to the foundational Attention Is All You Need paper, OpenAI's tokenizer demo, and a 3Blue1Brown neural-network series. Examples are explicitly marked as simplified or illustrative where they do not represent production-model values.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue