AI agents EXPOSED: TypeScript's hidden powers revealed!
summary
Demo builds a working TypeScript SMS scheduling agent: Express webhook + Surge SDK validates messages, Puppeteer pulls privacy policies for automated compliance checks, and the agent uses the OpenAI Responses API with function-calling tools (findAvailableSlots, bookAppointment, updateCustomer) to loop on tool calls, parse JSON args (ISO8601), and send replies. Claim: automating registration cuts Twilio’s 2–3 week manual review down to 2–3 days; agent runs end-to-end in live SMS tests, needs guardrails (Boundary ML, evals) to reduce but not eliminate mistakes.
AI-written summary · about
chapters (9)
- 0:00Introduction and speaker intro
- 1:18Why AI in products fails
- 3:03AI as a tool for problem-solving
- 8:01Agency and LLMs as tools
- 12:44LLM capabilities and code agents
- 15:56Python myth and TypeScript advantage
- 18:27Building a TypeScript agent demo
- 27:43Live demo and resources
- 28:47Q&A on agent reliability
related talks
01h 26mBuilding voice agents with OpenAI
40mAI Agents in Practice
30mAI Agents Unchained
23mBuilding AI Agents in TypeScript
25mAgents are not Apps: Stop building interfaces, Start building coworkers
01h 26mAI Agents of Change: Creating, Reflecting, and Monetizing
14mA Song of Types and Agents
56mAI Agents That Don’t Suck: How to Build Ones That Actually Drive Business Value