Exploring Deterministic LLM Programming
1h 1mBeginner2025-10-15
Authors
Pragmatic AI Labs
Course details
Large language models (LLMs) are often seen as creative, probabilistic systems, but they can also be programmed to behave in a predictable, repeatable way. In this course, learn how to design prompts, workflows, and integration patterns that minimize randomness and produce consistent outputs.
Concepts
Introduction
- Course introduction
Foundation
- Warming strategies
- Amdahl's law and coding agents
- Jevons paradox
- Princeton's Holistic Agent Leaderboard (HAL) framework
- ELO rating system
- ELO ratings questions
- Myth developer automation
Advanced Control
- Illusion of control agents
- Increasing determinism in deep context
- Understanding AST agentic coding
- Understanding complexity
- Understanding entropy
- Understanding code churn
- Roadmap sprints and technical debt
- SATD predicts technical debt
- Provability score
Implementation
- Demo PMAT and next steps