Navigating the Rise of Multi-Agent AI SystemsIn this episode of 'I Snap That,' hosts Michael Rollins and Thanos discuss their experiences and insights into the development and challenges of using multi-agent AI systems versus single-core LLM agents. While Thanos shares his recent travels in Greece and upcoming vacation plans, Michael dives deep into the complexities of AI coordination, the Blackboard model, and the surprising efficiency gains realized through distributing tasks among specialized AI agents. They explore the counterintuitive nature of these improvements, the role of memory and semantic understanding in AI workflows, and touch upon future concerns about QA and ensuring trust in AI outputs. The episode balances technical discussions with light-hearted banter and real-world anecdotes, creating an informative and engaging dialogue about the evolving landscape of AI technology.00:00 Welcome to I Snap That!00:30 Vacation Plans and Travel Stories04:27 Exploring AI Agents06:42 The Blackboard Model Explained11:00 Memory and Data Coordination15:08 Building and Testing Multi-Agent Systems20:02 Performance and Efficiency Insights23:04 Token Usage and Trade-offs in Computer Science24:30 Challenges with Coding Agents and Data Access25:57 Handling Semi-Structured Data and CSV Ambiguities27:20 Improving Coding Agents with MCP Servers32:12 Ensuring Trust and QA in AI Systems39:24 Learning and Implementing Lang Graph42:13 Real-World Coding Agent Experiences46:08 Concluding Thoughts and Future Topics