Artificial Intelligence: Advanced Prompts Course

Lesson 1 • How Large Language Models Work

Covers token prediction, training data, and probability-based output generation. Establishes the mechanical basis that explains why prompt wording changes model behaviour.

Lesson 2 • Common Failure Modes in Prompting

Identifies hallucination, refusal, drift, and verbosity as recurring output problems. Frames each failure as a diagnosable prompt design issue with a correctable cause.

Lesson 3 • Core Prompting Vocabulary

Defines essential terms: prompt, completion, context window, system message, and role. Provides shared language used throughout the entire course.

Lesson 1 • Role and Persona Assignment

Teaches how assigning a role to the model shapes tone, expertise level, and response style. Connects persona framing to measurable improvements in output relevance.

Lesson 2 • Context and Background Injection

Explains how to supply relevant background information so the model reasons from accurate premises. Proper context injection prevents hallucination and off-topic responses.

Lesson 3 • Output Format Specification

Demonstrates how to request specific formats: lists, tables, JSON, markdown, and prose. Format control reduces post-processing effort and integrates outputs into workflows.

Lesson 4 • Examples and Demonstrations in Prompts

Introduces few-shot and one-shot example placement to anchor model behaviour. Well-chosen examples outperform lengthy verbal instructions for format and style alignment.

Lesson 1 • Decomposition and Task Chaining

Breaks complex goals into sequential sub-prompts, passing outputs forward as inputs. Chaining enables tasks that exceed single-prompt scope or context limits.

Lesson 2 • Constraint-Based and Negative Prompting

Applies explicit restrictions to eliminate unwanted content, formats, or reasoning paths. Negative constraints are often more efficient than positive instructions for precision tasks.

Lesson 3 • Chain-of-Thought Prompting

Teaches models to reason step by step before producing a final answer, improving accuracy on complex tasks. Covers explicit and implicit chain-of-thought triggers.

Lesson 4 • Role-Play and Simulation Prompting

Uses scenario-based framing to unlock specialised model behaviour for training, ideation, and analysis. Connects simulation design to practical professional use cases.

Lesson 1 • Code Generation and Debugging Prompts

Applies prompting to software tasks: writing functions, explaining code, and diagnosing bugs. Covers language-agnostic patterns and context-setting for accurate code output.

Lesson 2 • Data Analysis and Summarisation Prompts

Teaches prompts that extract insights, identify patterns, and condense large text bodies. Accurate summarisation prompts reduce information overload in research and reporting workflows.

Lesson 3 • Question Answering and Research Prompts

Designs prompts that retrieve, synthesise, and cite information accurately from provided context. Reduces hallucination risk by anchoring answers to supplied source material.

Lesson 4 • Classification and Decision Support Prompts

Uses prompts to categorise inputs, score options, and support structured decision-making. Connects classification prompt design to business logic and evaluation rubrics.

Lesson 1 • Systematic Prompt Testing

Covers A/B testing prompts, varying one variable at a time, and recording results for comparison. Structured testing replaces guesswork with evidence-based prompt improvement.

Lesson 2 • Rewriting and Restructuring Prompts

Applies targeted rewriting strategies: simplifying language, reordering components, and adding anchors. Each strategy addresses a specific class of output failure identified in diagnosis.

Lesson 3 • Feedback Loops and Self-Critique Prompts

Teaches the model to evaluate and revise its own outputs using meta-prompts and critique instructions. Self-critique loops reduce human review time and surface hidden errors.

Lesson 1 • System Message Architecture

Explains the role of system messages in setting persistent behaviour before any user turn. Covers placement, scope, and the hierarchy between system and user instructions.

Lesson 2 • Behavioural Policy Encoding

Encodes rules, ethical guardrails, and operational policies directly into system prompts. Policy-level instructions ensure consistent, compliant behaviour across all user interactions.

Lesson 3 • Dynamic System Prompt Construction

Builds system prompts programmatically by injecting variables, user data, and session context at runtime. Dynamic construction enables personalised, context-aware AI deployments.

Lesson 1 • Self-Consistency and Ensemble Prompting

Generates multiple independent responses to the same prompt and aggregates them for higher accuracy. Reduces variance and catches errors that single-pass prompting misses.

Lesson 2 • Tree-of-Thought and Multi-Path Reasoning

Extends chain-of-thought into branching reasoning trees that explore multiple solution paths simultaneously. Enables better performance on problems with no single obvious reasoning route.

Lesson 3 • Reasoning Under Uncertainty and Ambiguity

Designs prompts that elicit calibrated uncertainty, hedged claims, and explicit assumption statements. Prevents overconfident outputs in high-stakes analytical and advisory contexts.

Lesson 4 • Meta-Prompting and Prompt Generation

Uses the model to generate, critique, and optimise prompts for downstream tasks. Meta-prompting accelerates prompt development and surfaces non-obvious instruction improvements.

Lesson 1 • Latency, Cost, and Token Optimization

Reduces prompt token count and API cost without sacrificing output quality through compression and caching. Optimisation is essential for high-volume production deployments.

Lesson 2 • Prompt Libraries and Reusable Templates

Builds modular, parameterised prompt templates that teams can reuse and adapt across projects. Shared libraries reduce duplication and enforce quality standards organisation-wide.

Lesson 3 • Prompt Versioning and Documentation

Establishes practices for versioning prompts, tracking changes, and documenting intent and performance. Versioned prompts enable rollback, auditing, and team collaboration at scale.

Lesson 4 • Monitoring and Quality Assurance in Production

Implements logging, output sampling, and automated evaluation pipelines to detect prompt degradation. Continuous monitoring maintains output quality as models and user inputs evolve.