transformer-explainer

Why we included this project

Explaining how a large language model works usually ends in hand-waving at a static diagram. Transformer Explainer gets past that by running a real GPT-2 model in the browser, so you can type your own prompt and watch tokens get embedded, passed through attention heads and feed-forward layers, and turned into next-word probabilities. The input and the internal state update together in real time, which makes the architecture feel concrete rather than abstract. It is a good fit for classroom teaching, self-study, or building intuition before tackling the research literature, and since everything runs client-side with no GPU or API key, it also works as a low-friction demo in workshops and talks.