Module 1 of 3

The Agent Loop

Visualising the pulse of the agentic system.

We guide you through the 'Inference Cycle' the rhythmic pulse of the system. You will learn to see the agent loop not as a mysterious process, but as a deliberate series of design decisions.

How AI Agents Work
  • 45–60 mins
  • 4 lessons
  • foundation
Get this course

What you'll learn

  • Deconstruct the inference cycle — from tokenisation through the forward pass — and understand why statelessness is the reality you design around.
  • Explain how the orchestration layer directs a one-shot model to plan and self-correct.
  • Identify the execution boundary between the model and the environment, and articulate where responsibility sits when an action fails.

Lessons in the Module

1 The Inference Cycle 20–25 min
2 The Agent Harness 20–25 min
3 Tool Calling 20–25 min
4 Non-Determinism and the Paradigm Shift 20–25 min
Back to course overview

Further Reading

Safety

Constraint Architecture for Autonomous Agents

10 min read

Technical

The LLM-as-Judge Problem Nobody Is Talking About

7 min read

Governance

The Inference Audit Trail: Making Every AI Decision Accountable

8 min read
All insights