Applied AI

Course map · S1 How LLMs work · Episode 8 of 12

Sampling

S1 E8 Explainer · 17 slides · 5 October 2026

  1. Slide 1 of 17, Cover.
  2. Slide 2 of 17, Why it matters.
  3. Slide 3 of 17, Mental model.
  4. Slide 4 of 17, Data.
  5. Slide 5 of 17, Concepts.
  6. Slide 6 of 17, Concepts.
  7. Slide 7 of 17, Concepts.
  8. Slide 8 of 17, Recap.
  9. Slide 9 of 17, Your turn.
  10. Slide 10 of 17, Data.
  11. Slide 11 of 17, Concepts.
  12. Slide 12 of 17, For engineers.
  13. Slide 13 of 17, Where it breaks.
  14. Slide 14 of 17, For engineers.
  15. Slide 15 of 17, For leaders.
  16. Slide 16 of 17, Decision rule.
  17. Slide 17 of 17, Series map.

Lesson notes

Sampling draws each token from the model's odds, so settings like temperature trade consistency for variety, and a feature must be judged by its rate across many runs.

Decision rule

Set sampling for the task, and judge a model by its rate across many runs, never by one answer.

Ideas in this lesson

  • Next-token prediction
  • Probability
  • Measuring, not quoting

Take it further