AI & ML
AI Agents
Part of Decision Making, level 7. About 10 hours of study across the topics below.
Agency
- The Agent LoopObserve, think, act: state plus the loop that uses it.1–2 h
- Tools and Tool UseFunction calls as actions: selection, arguments, observations, and errors.1–2 h
Cognition
- PlanningDecomposition before action: goals into steps, with search vs script honesty.1–2 h
- MemoryShort and long: working context, stored history, and retrieval as memory.1–2 h
- Reasoning-Acting Loops and Self-ReflectionInterleaving thought, tool calls, and observations; revising a plan after critiquing a failure.1–2 h
Deployment
- Workflows vs AgentsFixed graphs vs model-directed loops: where control lives, and what each risks.1–2 h
- Failure RecoveryRetries, verification, and fallbacks: what agents do when steps fail.1–2 h
- Multi-Agent SystemsSeveral agents cooperating or competing: role decomposition, communication protocols, and incentive problems.1–2 h
- Evaluating AgentsTask success, trajectories, and cost: scoring loops, not just answers.1–2 h
- Agent Safety: Permissions, Sandboxing, and Human OversightLimiting what a tool-using agent can do: least privilege, sandboxes, confirmation gates, and audit trails.1–2 h