Why Your Company Needs a Context Graph (and How to Build It) — Gil Feig, Merge
Source
youtube.com
Author
AI Engineer
Date
Why it matters
Explains when to cache versus call live, and why company-specific definitions and data provenance belong in the context layer, for agents that must answer questions over large datasets.
Key takeaways · AI-distilled
Feig's test question, "which customers were upset last week?", shows why connecting a few MCPThe Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.Full definition → servers falls short: live API lookups can't search across large datasets, so a synced copy is needed for semantic queries.
His architecture combines a router, synced and live context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →, a summarizer, and skills plus memory that encode company-specific definitions, such as what counts as a happy customer.
He describes four tiers of context including derived summaries, and requires every answer to track where its data came from, when, and what invalidates it.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
videoSmall Models, Big Results: Training a Finance Agent for Under $500 — Charles Dickens, Snorkel AI
videoFrom Context to Memory: Your Agents Need a Real Memory Layer — Anders Swanson, Oracle
videoGeneration Is Cheap, Review Is Expensive: How to Stop Shipping AI Slop — Gabriel Martinez, G2i
videoWhy 80% Reliability Isn't Good Enough — Felipe Blanes, Amazon AGI Lab
