✳ Inside the Answer
AN INTERACTIVE JOURNEY THROUGH A LANGUAGE MODEL
A question goes in. An answer comes out.

How does a question
become an AI answer?

See how your question becomes tokens, how a Transformer processes them, and how the AI predicts and generates an answer — one token at a time.

Our conversationJourney ready
“What is the capital of India?”
A guided, simplified view of autoregressive text generation. Select a chapter or move through the story.
↳CHAPTER 04TRANSFORMER CHILD NODES
CHAPTER 01 / 08
Progress saves automatically12% explored
CHAPTER 01 · INPUT

Before the model thinks…

Your question is not the only thing in the input. The model receives a structured conversation that includes instructions and message boundaries.

01 / 08
01 Conversation assembled
The token IDs, vector values, probabilities, and layer drawings shown here are illustrative. Real models use model-specific tokenizers, learned parameters, architectures, and decoding settings. The story preserves the core mechanics without pretending to expose a particular model's internals.