Summary
Host Claire Vo sits down with Nick Bowman, an engineer at OpenAI, for a live walkthrough of how he uses Codex and ChatGPT Work together to automate both professional and personal tasks. The conversation centers on features many users haven’t discovered yet: the ability to spin up and manage multiple ChatGPT threads simultaneously, have those threads communicate with each other, and now orchestrate all of it via a new voice layer that activates with a hotkey and can see your screen in real time.
Bowman demonstrates the voice interface delegating a multi-step expense report task — finding an Amazon receipt, staging a Navon report, and holding for human confirmation before submitting — all from a single spoken instruction. He also walks through ChatGPT Sites, a new capability that gives deployed artifacts a persistent SQL database, S3 file storage, and environment variable support, with access controls ranging from fully public to email-filtered private. Live on camera, they build a site that indexes tips from Claire’s YouTube channel, categorized by tool and audience role, and deploy it directly from ChatGPT Work on mobile.
The episode is a practical look at how OpenAI’s individual primitives — thread management, vision, voice, and web deployment — are converging into a unified personal AI operating layer. Viewers come away with specific techniques for voice-delegated task management, multi-thread orchestration, and publishing shareable web apps without touching a code editor.
📺 Source: How I AI · Published August 03, 2026
🏷️ Format: Interview
