/
Published on 27.01.2026
TLDR: Frontend Masters released "AI Agents Fundamentals, v2" taught by Scott Moss from Netflix. The course takes you from basic LLM interactions to building a production-ready CLI agent with tool calling, human-in-the-loop approvals, and observability.
Tool calling from scratch — Define tools with schemas, let the LLM decide when to use them, and execute the results. This is the foundation of agentic systems.
The agent loop — Handle tasks where the number of steps isn't known ahead of time. Stream responses and manage conversation flow. This is where agents differ from simple chat interfaces.
Evals (AI Tests) that actually help — Test non-deterministic outputs, track success rates, and use observability tools like Laminar to understand what's happening. Testing AI systems requires different approaches than traditional software.
Context management — When conversations get long, use compaction, summarization, and RAG to keep your agent focused. Context window management is critical for production agents.
Human-in-the-loop approvals — Add checkpoints so agents ask before taking sensitive actions. This is essential for any agent that interacts with real systems.
For teams building AI-powered features, this course covers the practical aspects that matter: not just making agents work, but making them reliable, observable, and safe for production use.
This article was generated from a newsletter summary. The content reflects the source material's perspective and may not represent complete analysis of all aspects of the topic.