Summary
Paul J. Lipsky walks through his hands-on experience with GPT-6 Astra inside ChatGPT Work, OpenAI’s task-completion interface designed for end-to-end workflows rather than conversational Q&A. The video explains the practical distinction between standard ChatGPT and Work mode, and covers GPT-6’s thinking levels — a slider that trades speed for thoroughness, with OpenAI claiming even the lowest setting outperforms the previous model at its highest.
The centerpiece demo involves connecting a local folder of business documents to a ChatGPT Work project and running a full business audit. Lipsky builds a fake plumbing business with deliberately embedded problems — unauthorized financial transfers, overdue client accounts — to test GPT-6’s ability to surface them. The model produces structured tables, identifies the planted anomalies, and generates a prioritized multi-quarter improvement plan. The demo doubles as a practical template viewers can apply to their own business files.
The second half covers creating reusable Skills inside ChatGPT — persistent instruction sets that capture writing style from sample documents so the model can replicate a user’s voice without re-explaining preferences each session. Lipsky recommends temporarily raising the thinking level during skill creation for better style capture, then lowering it for day-to-day use. The video is approachable for non-technical users and provides a clear on-ramp for getting practical value from GPT-6 Astra beyond simple chat.
📺 Source: Paul J Lipsky · Published September 14, 2026
🏷️ Format: Tutorial Demo
