Learning path · Agents & Orchestration · 60
AI Agents
LLM-driven systems that plan, use tools, and iterate toward goals—not just single-shot text completion.
Why it matters
- Unlocks workflows: research, ticket triage, data entry with supervision.
- Introduces loops where cost and failure modes multiply.
- Requires explicit stop conditions and human gates.
Key ideas
- Plan-act-observe loops
- Tool use
- State management
Top resources
- 01ArticleAnthropic
Building effective agents
Why this resource. The practical essay: workflows first, agents when the loop is justified.
Covers in this concept
- tool use
- orchestration
- 02PaperYao et al.
ReAct: Synergizing Reasoning and Acting in Language Models
Why this resource. Reason-then-act: the loop most agent frameworks still implement.
Covers in this concept
- ReAct
- observations
An agent is a loop: look at state, pick an action, call a tool, update memory, repeat until done or blocked. Cost and failure multiply with each step. Start with one API and a hard step limit. Tools should be idempotent. Most production incidents in this pattern begin as an unbounded retry.
Updated 2026-08-09 · Full learning path