Demand-Driven Context: A Methodology for Coherent Knowledge Bases Through Agent Failure

on YouTube ↗01h 08m

" summary

Claims a practical loop: give agents real incidents, capture where they fail, extract missing facts, and add those as small Markdown context blocks so agents improve. Demo uses Copilot, Cloud Code and flat-file/GitHub storage, runs a batch of past incidents and 14 incidents raises confidence 1.4→4.4 after iterative curation. Built an automated context gap scanner, prefers stores curated context in GitHub for PRs and review, and shows the approach works in the demo and scales via automation.

AI-written summary · about

" chapters (19)

1. 0:00Introduction and speaker background
2. 2:47The situation: Analogy to the movie Memento and AI's memory constraints
3. 3:55Evolution of AI: From prompt engineering to deep agents
4. 4:33Enterprise AI challenge: Why productivity isn't moving
5. 5:33The problem: Green (general), Orange (taught), and Red (institutional/tribal) knowledge
6. 10:11The Monolith: Why institutional knowledge is often outdated or missing
7. 11:24Solution introduction: Demand-driven context
8. 13:05The "Pull" strategy: Learning by doing vs. pushing information
9. 14:48The agent lifecycle: Problem to discovery to documentation
10. 17:46Demo introduction: Using a framework for context management
11. 19:12Live demo: Incident root cause analysis and context discovery
12. 24:05Scaling: 14 incidents to show confidence level improvement
13. 26:27Automated scale: Validating knowledge across the monolith
14. 33:01Storage strategy: Why GitHub is preferred for knowledge repositories
15. 34:47The Meta Model: Navigating domain relationships
16. 36:27Value proposition: Knowing the unknown and managing knowledge
17. 39:02Summary: The 80/20 rule and cache-based context blocks
18. 40:15Workshop takeaways: Repositories and scanners
19. 43:33Q&A Session: Addressing scalability, tooling, and cost

" related talks

• 29mHow to Structure Context to Make Your Agents SmarterNyah Macklin · DevWorld 2026
• 31mBuilding the Context Engine AI Agents NeedBrandon Waselnuk · QCon AI Boston 2026
• 27mContext is the new code: The context development lifecyclePatrick Debois · PlatformCon 2026 Live Day New York
• 28mContext Is the New CodePatrick Denis · AI-fokus 2026
• 31mBuild Intelligent AI Agents With Context EngineeringMichael Hunger · NODES AI 2026
• 15mContext Is the New Code for AI AgentsPatrick Debois · Background Agents Virtual Summit 2026
• 15mUnderstanding Context Engineering for Agentic SystemsRobert Shelton · Optimized AI Conference 2026
• 56mConnecting the Dots with Context GraphsStephen Chin · Devnexus 2026