
Session 1: Mental Models & Prompt Architecture
Token prediction mechanics, probability and temperature management, context-window optimization, and the 4-pillar architecture.
Core Objectives
- •Understand how tokenizers split text and how the model predicts the next token.
- •Control output randomness with temperature, top-p, and frequency penalties.
- •Design prompts with the 4-pillar structure: Role, Context, Task, Constraints.






