Volume 10 — AI Agents¶
Systems that take actions
Tool use, planning, memory, multi-agent patterns, and the safety and evaluation concerns that arrive the moment a model can act rather than merely answer.
Outlined, not yet written
This volume is fully outlined but the lessons are not yet authored. The chapter list below is the plan of record, not a promise that content exists.
What you will be able to do¶
- Define tools an agent can use reliably
- Design an agent loop with explicit termination and error handling
- Contain the blast radius of an agent that misbehaves
- Evaluate an agent on task completion rather than vibes
Chapters¶
- What an agent actually is
- Tool definition and function calling
- The agent loop
- Planning and decomposition
- Memory and state
- Multi-agent orchestration
- Guardrails and sandboxing
- Human-in-the-loop design
- Evaluating agents
- Cost, latency and failure containment