Agent Concepts 06 — Evals, Reliability & Cost
Published:
Outcome vs trajectory evals, LLM-as-judge calibration, SWE-bench / tau-bench / WebArena, reliability patterns, guardrail layers, and cutting $/task.
← Back to Interview Hub
Agent Concepts 05 — Multi-Agent Orchestration
Published:
Supervisor vs pipeline vs handoffs, failure modes unique to multi-agent, typed handoffs, self-improving loops — and when to stay single-agent.
Agent Concepts 04 — RAG & Retrieval
Published:
End-to-end RAG, chunking tradeoffs, why hybrid search + rerank, retrieval metrics, and when agentic retrieval beats single-shot RAG.
Agent Concepts 03 — Context Engineering & Memory
Published:
Why context is the working memory, compaction strategies, episodic/semantic/procedural memory, and mitigating ‘lost in the middle’.
Agent Concepts 02 — Tool Use & MCP
Published:
Function calling mechanics, tool design principles, MCP vs bespoke integrations, error-handling contracts, and schema-constrained decoding.
Agent Concepts 01 — Agent Foundations
Published:
Agent vs chatbot vs workflow, ReAct / Plan-and-Execute / Reflexion compared, the autonomy spectrum, and the ‘when not to build an agent’ decision framework.