

How OpenAI Builds AI Agents That Think and Act with Josh Tobin - #730
375 snips May 6, 2025
Josh Tobin, a member of the technical staff at OpenAI and co-founder of Gantry, dives into the fascinating world of AI agents. He discusses OpenAI's innovative offerings like Deep Research and Operator, highlighting their ability to manage complex tasks through advanced reasoning. The conversation also explores unexpected use cases for these agents and the future of human-AI collaboration in software development. Additionally, Josh emphasizes the challenges of ensuring trust and safety as AI systems evolve, making for an insightful and thought-provoking discussion.
AI Snips
Chapters
Transcript
Episode notes
Agent Training Unlocks Better Workflows
- Training models end to end on multi-step tasks enables them to learn strategies better than manual human designs.
- Models can learn to recover from errors and improve workflow performance through reinforcement learning.
Foundation Models Shift Business Strategy
- General purpose models like GPT-4 reduce the need for most businesses to train their own models.
- Building on commercial foundation models is faster and cheaper than creating custom domain models.
Training Agents to Self-Correct Errors
- Traditional LLMs struggle with multi-step tasks due to compounding errors and lack of agentic training.
- RL-trained agents can detect failures and self-correct, increasing reliability in complex workflows.