Cut the Hype and Build Your Own AI Agents
summary
Claims that big-name agents fail for many tasks (one tool nailed 15% in a test) and that custom agents are cheaper and faster lead directly to examples: a docs-updater GitHub action built with Crew AI that reads a git diff, updates markdown and screenshots using GPT40; a workflow that generates ESLint rules and retries until tests pass (uses O4 mini and structured Pydantic outputs); and an MCP admin server queried via chatwise and open router showing 33 open orders. Tools shown: Crew AI, Langraph, Langchain, cursor, open router, MCP.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:08AI agents defined
- 3:37AI agent products overview
- 6:55Choosing agent frameworks
- 8:42Role goal backstory in Crew AI
- 11:12Demo: docs automation
- 19:59Demo: custom linting
- 28:54Model Context Protocol
- 33:13Demo: MCP server
- 40:16Recap of key concepts and tools
- 41:22Q&A
related talks
32mBeyond the Hype: A Pragmatic Look at How Developers Actually Use AI Tools
30mBeyond the Hype: Using AI Coding Agents Today
01h 23mBuild your own AI Agent
26mAI Agents as Your Co-Developer
01h 28mAI Coding Agents and how to code them
40mAgents in Action: Advancing Open Source AI Missions
01h 53mAgentic AI: Build a Multi-Agent Application with CrewAI
40mAI Agents in Practice