Connecting the Dots with Context Graphs
summary
Argues that durable Context Graphs—combining knowledge graphs, embeddings and tool traces—fix fragmented enterprise context and improve explainability, provenance and long-term reasoning. Demonstrations use Neo4j Aura and the Neo4j Graph Builder plus MCP servers and Claude to run GraphRAG pipelines, a 300-episode memory demo (300 podcasts) and a financial case that refuses a $25,000 credit request after 7 tool calls, flagging 14 transactions in 29 minutes with 85% semantic similarity and 90% confidence; the prototypes run and return grounded, inspectable decisions.
AI-written summary · about
chapters (14)
- 0:00Introduction to context graphs
- 4:43From prompting to context engineering
- 7:10Knowledge graphs fundamentals
- 10:41Graph RAG vs vector RAG
- 18:26Graph RAG architecture
- 19:07Demo 1: Software security analysis
- 26:07Model Context Protocol basics
- 28:35Demo 2: Text-to-Cypher with MCP
- 33:12Memory types for agentic systems
- 42:44Demo 3: Lenny’s podcast memory
- 45:58What is a context graph?
- 50:13Demo 4: Financial services use case
- 53:24Building complex AI systems
- 54:08Resources and next steps
related talks
23mConnecting the Dots With Context Graphs
18mConnecting the Dots with Context Graphs
49mConnecting the Dots with Context Graphs
48mOpening Keynote: Exploring Context Graphs: From Data to Decisions - Extended Version
26mOpening Keynote: Exploring Context Graphs: From Data to Decisions
31mBuild Intelligent AI Agents With Context Engineering
29mThe AI Agent Memory Landscape
30mContext Engineering: Connecting the Dots with Graphs