How to Make Useful AI Agents
summary
Claims a production-ready stack: Arcade.Dev infrastructure for OAuth, multi-tenant tool runtime, and cached user tokens. Demonstrations include ambient Slack assistants that read emails and put calendar events, GitHub-integrated tools reading private repos, and LinkedIn/Twitter posting; a live prototype at chat.arch.dev. Argument boils down to one concrete rule — agent must call tools and work as the user — and the demos show those integrations actually running rather than just returning weather.
AI-written summary · about
chapters (9)
- 0:00Weather demos frustrate me
- 1:05What makes an AI agent
- 3:16Problem with tool calling
- 6:28Missing piece in workflows
- 8:16Agents should work as you
- 9:21Security flaws in OAuth
- 10:47Arcade solves OAuth
- 12:35Real-world agent examples
- 15:25Try Arcade GPT
related talks
26mAI Agents: How Do They Work and How to Build Them
46mFullstack AI: The Building Blocks for Production Ready AI-Powered Software
40mAgents in Action: Advancing Open Source AI Missions
34mBuilding Ai Agents with Shell Scripts
22mHOW TO BUILD YOUR OWN AI AGENT: THE COMPLETE ARCHITECTURE FOR ENGINEERING TEAMS
41mGoing beyond copilot with AI Agents
40mAI Agents in Practice
52mCut the Hype and Build Your Own AI Agents