Andrew Ng just released a 12-page PDF on 'Graph Engineering' for multi-agentic systems. It covers 4 design patterns: Reflection (add a critic for 10-30% quality lift), Tool Use (execute code, search web, query DB to stop hallucination), Planning (write a structured plan and reroute on tool failure), and Multi-Agent (split roles like coder, reviewer, tester). The key idea: graph architecture externalizes shared state so agent loops can persist overnight. Based on his courses and DeepLearning.AI curriculum.
1mo
Andrew Ng just released a 12-page PDF on 'Graph Engineering' for multi-agentic systems. It covers 4 design patterns: Reflection (add a critic for 10-30% quality lift), Tool Use (execute code, search web, query DB to stop hallucination), Planning (write a structured plan and reroute on tool failure), and Multi-Agent (split roles like coder, reviewer, tester). The key idea: graph architecture externalizes shared state so agent loops can persist overnight. Based on his courses and DeepLearning.AI curriculum.
1mo
No comments yet. Be the first!
Comments
No comments yet. Be the first!