Advanced Concepts
Finished the learning tracks? Go deeper into the ideas behind reliable AI agents: security, human approval, evaluation, cost, and more.
These chapters build on Foundations, Intermediate, and Advanced. Each explores a topic in more depth and comes with its own lab. Choose what you want to study next, and use the preparation links if you need a refresher.
Explore the chapters
8 chapters
Prompt and program optimization
Prompt Engineering
Prompt structure, grounding, self-consistency, and ways to break complex work into stages.
Preparation: Prompting 101
Cost and performance
Token & Cost Management
Context trimming, model selection, prompt caching, and batching, with a lab on history and model size.
Preparation: Prompting 101
About 25 min · Intermediate
Read chapter : Token & Cost ManagementDownload lab ZIP: Token & Cost ManagementSecurity and authorization
Agent Security
Indirect prompt injection and tool boundaries that constrain sensitive actions.
Preparation: Your First Agent
Agent control and approval
Human-in-the-Loop Approval Gates
Saved agent state and explicit human approval, editing, or rejection before a tool runs.
Preparation: Memory
Security and authorization
RBAC (Role-Based Access Control)
Role-based permissions and per-role limits on agent tools.
Preparation: Agent Security
Reliability and fault handling
Chaos Engineering (Fault Injection)
Fault injection, incomplete tool results, and checks that detect corrupted data.
Preparation: Your First Agent
About 20 min · Intermediate
Read chapter : Chaos Engineering (Fault Injection)Download lab ZIP: Chaos Engineering (Fault Injection)Evaluation and red-teaming
Continuous Adversarial Evaluation
Attack variations, security-defense comparisons, and the limits of an evaluation harness.
Preparation: Agent Security, Guardrails and Safety
Reliability and fault handling
AI Gateways: Build the Boundary
Provider adapters, timeouts, and deliberate failover decisions at an application boundary.
Preparation: Production Concerns
Before you begin
If you are still working through the tracks, continue with Advanced. New to AI or Python? Start with Foundations and machine setup.
The time estimates come from the chapters; allow extra time for your first setup. Difficulty describes the preparation needed for each chapter.
Checkpoint
How should you choose your next chapter?
Choose a topic you want to understand more deeply, then check its preparation links. For example, Human-in-the-Loop builds on conversation memory and saved agent state. Revisit the Memory chapter if those ideas are unfamiliar before running its approval lab.