.NET Rocks! Thinking Agentic AI with Seth Juarez
11 snips
Jul 3, 2025 Join Seth Juarez, a Microsoft AI expert with a knack for making complex concepts understandable, as he unpacks agentic AI. Learn about the innovative potential of MCP and NLWeb for creating collaborative AI agents. Seth discusses the underlying mechanics of these tools, emphasizing the need for governance to harness their capabilities responsibly. The conversation also explores the evolution of AI agency and the nuances of prompt politeness, which can significantly impact AI performance. It's a deep dive into the future of AI in programming!
AI Snips
Chapters
Transcript
Episode notes
Agent as Control Structure
- An agent is essentially a control structure deciding what function to run next based on input reasoning.
- The key magic is transforming natural language intent into structured function calls with parameters.
Control Agent Tool Access
- Do not trust agents blindly; restrict their tool access and supervise their tool outputs carefully.
- Always control the sequence of tool calls to maintain accuracy and safety.
Agents Communicate in Language
- Multi-agent systems communicate via natural language as their interface, treating other agents as tools.
- Agents use augmented instructions to augment prompts and orchestrate complex tasks collectively.