How OpenAI Builds AI Agents That Think and Act [Josh Tobin] - 730

How is OpenAI building AI agents that ACTUALLY work? Sam Charrington gets the answers from Josh Tobin, who leads AI Agents research at OpenAI. Dive deep into the strategies and tech behind AI that can do more than just chat – agents that can research, code, and operate online for you. In this episode, you'll discover: The secret sauce: Why new OpenAI models are better at complex, multi-step tasks and recovering from errors. Deep Research in action: Going beyond basic search to truly synthesize information. Codec CLI: How this open-source agent is changing coding (hello "vibe coding"!). Operator insights: The challenges and potential of AI navigating the web. The path to reliable agents: What it takes to build trust and ensure AI tools (like MCPs) are used safely. The future of your workflow: How agents will impact software development and beyond. If you're curious about how AI agents are evolving from promising concepts to powerful, practical tools, this conversation with OpenAI's Josh Tobin is a must-watch! 🗒️ For the full list of resources for this episode, visit the show notes page: https://twimlai.com/go/730. 🔔 Subscribe to our channel for more great content just like this: https://youtube.com/twimlai?sub_confirmation=1 🗣️ CONNECT WITH US! =============================== Subscribe to the TWIML AI Podcast: https://twimlai.com/podcast/twimlai/ Follow us on Twitter: https://twitter.com/twimlai Follow us on LinkedIn: https://www.linkedin.com/company/twimlai/ Join our Slack Community: https://twimlai.com/community/ Subscribe to our newsletter: https://twimlai.com/newsletter/ Want to get in touch? Send us a message: https://twimlai.com/contact/ 📖 CHAPTERS =============================== 00:00 Intro & catching up with Josh Tobin 02:33 Why agents? From Gantry to OpenAI’s agentic push 05:09 How foundation models upend ML infrastructure 10:33 Real‑world agentic examples 14:54 Building robust agents today: practical steps 20:06 Deep Research: use cases & power users 28:11 OpenAI Operator for web automation 33:49 Codex CLI and the future of developer agents 39:40 Context, tooling, and agent workflows 45:54 Cost transparency & pricing models for agents 50:40 Vibe coding and next‑gen developer UX 55:17 AI‑powered learning & developer education 57:45 MCP and tool-use 59:33 Trust and security for agents 🔗 LINKS & RESOURCES =============================== OpenAI Codex CLI - https://github.com/openai/codex Introducing deep research - https://openai.com/index/introducing-deep-research/ Introducing Operator - https://openai.com/index/introducing-operator/ Geometry-Aware Neural Rendering with Josh Tobin - 360 - https://twimlai.com/podcast/twimlai/geometry-aware-neural-rendering-josh-tobin/ 📸 Camera: https://amzn.to/3TQ3zsg 🎙️Microphone: https://amzn.to/3t5zXeV 🚦Lights: https://amzn.to/3TQlX49 🎛️ Audio Interface: https://amzn.to/3TVFAIq 🎚️ Stream Deck: https://amzn.to/3zzm7F5