Summary
Alex Finn makes the case that ChatGPT Work powered by GPT-6 Astra has crossed a capability threshold that finally makes AI-managed computer use practical for daily knowledge work. He walks through real workflows running simultaneously during filming: the agent editing a video inside DaVinci Resolve, autonomously drafting and sending a sponsorship script to HubSpot after reading his entire email history, checking his inbox hourly, and generating visual assets for a live boot camp — all at once.
The setup section is detailed and practical: connect Gmail (business and personal accounts separately), Slack, GitHub, Unity, Linear, Notion, and the OpenAI Developers integration that allows the agent to generate its own API keys autonomously. Finn explains how to create scoped projects so agents have the right context for different areas of work, and how to configure routines — effectively cron jobs — for recurring tasks like morning inbox triage and draft response generation.
A notable thread throughout is Finn’s direct comparison to his prior experience with ChatGPT Work and Claude co-worker, which he previously found unconvincing. He attributes the shift entirely to GPT-6 Astra’s computer use capability, arguing the model now operates a computer faster than a human. Privacy concerns around inbox access are addressed briefly. The video ends with guidance on how ChatGPT Work fits alongside other tools including Codex and Grok Bot for different task types.
📺 Source: Alex Finn · Published September 12, 2026
🏷️ Format: Tutorial Demo
