Build Intelligent AI Agents With Context Engineering
summary
Claims that engineering what goes into an LLM’s context—not bigger windows—cuts failures by steering retrieval, tool use, and memory. Uses graph RAG and a context graph that stores entities, metadata and decision traces so agents pull precise, explainable facts instead of drowning in documents. Recommends tool filtering, minimal viable context, memory compression and summarization. Neo4j shows a demo app and an agentic layer on its graph DB that captures traces and improves relevance and repeatable task success.
AI-written summary · about
chapters (11)
- 0:00Introduction and speaker intro
- 1:07Why context engineering matters now
- 3:53Sources of context for LLMs
- 8:27Cognitive sources and memory types
- 13:01Minimal viable context principles
- 16:06Offloading large context strategies
- 20:01Self-modifying systems with context
- 21:07Agent loop and context roles
- 24:38Graphs for intelligent context
- 28:06Context graphs and decision traces
- 29:29Best practices and closing
related talks
15mUnderstanding Context Engineering for Agentic Systems
31mBuilding the Context Engine AI Agents Need
29mHow to Structure Context to Make Your Agents Smarter
23mConnecting the Dots With Context Graphs
57mCrafting Intelligent Agents with Context Engineering
56mConnecting the Dots with Context Graphs
18mContext Engineering — Reliable AI Instead of Hallucinations
01h 03mAgentic Search for Context Engineering