Agentic Al in SW Development: Evolving Patterns & Protocols
summary
Demo shows a working Nova ACT model driving a browser to search and add an Apple Pencil to cart using three English sentences, proving simple agent scripts work. Presents seven agentic patterns (RAG, self‑correcting RAG, function‑calling, MCP, agent‑to‑agent) and a concrete stack: vector store + LangGraph orchestration + MCP server registering Jira/Confluence/Git tools. Self‑correcting RAG uses a grader LLM to detect hallucinations; MCP demo reads and creates Jira tasks, and the GitHub repo contains runnable code.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:43Demo
- 3:27State of the union of AI in development
- 5:02Agentic systems: A simple mental model
- 6:19Pattern 1: Conversational clients
- 7:34Pattern 2: Contextual conversational clients
- 8:38Pattern 3: Retrieval augmented generation (RAG)
- 10:20Pattern 4: Self correlating RAG
- 17:20Pattern 5: Agent with MCP
- 21:38Pattern 6: Agent to agent integration
- 22:46Summary
- 23:25Outro
related talks
49mAgentic AI Patterns
27mEmerging Patterns Agentic AI
01h 03mAI Coding Agents and how to code them
01h 26mBuilding Agents with Amazon Nova Act and MCP
29mPairing with the machine: How AI Agents are rewriting the rules of software development
31mAgentic AI architecture patterns
46mAgentic Design Patterns Where You Are the Hero
48mAI Coding Agents and how to code them