How AI Responds
The Slot Machine: next-token prediction
How does AI generate text?
The model predicts the most likely next token, one step at a time. Think of a slot machine of candidates with probabilities. Temperature reshapes those odds. This lab uses an honest teaching engine: distributions re-sum to 1.
Simulated engine
Probabilities and full replies come from a client-side simulator, not a production API. The shape of sampling and temperature is real; the vocabulary is a teaching toy.
Balanced
Move temperature, then Send or Auto: bars and wording both change.
1
Package2
Predict3
Generate4
AppendPress Send to package the request, then pick tokens by hand or run Auto for a full simulated reply.
Key Insight
Low temperature peaks on the top token. High temperature flattens the bars so unlikely words win more often. The model is not deliberating like a person: it is sampling from a learned distribution.
Optional deeper: attention
Attention lets each token look at other tokens in the window. Explore it below, or later in the Transformer lesson for the full stack.
Attention visualizer
Loading attention visualizer…
Check your understanding
1/5How does AI generate the next word?
Sources, scope & review statusVerified Jul 29, 2026
Claims this lesson makes
- Stable conceptAutoregressive models repeatedly score possible next tokens and sample or select one token before continuing.
Primary reading
Next content review: 2027-07-29. Source links establish the lesson's claims; interactive numbers remain labeled simulations unless explicitly identified as measured data.