Summary
Greg Isenberg hosts Ross Mike for a practical breakdown of GPT-6 Astra — OpenAI’s most expensive model, on pricing parity with Claude Fable but heavily subsidized under Codex subscriptions. Rather than showcasing demos, the episode focuses on concrete business applications: running a performance audit that dropped one app’s API response time from 800 milliseconds to 20–30 milliseconds, conducting automated security audits before launch, and using Astra’s computer-use capability to navigate complex government websites like Canada’s CRA portal.
The conversation covers nine specific Astra prompts worth testing, including a bill renegotiator that enters provider chat support and negotiates costs down, a QA agent that opens an app on a real phone nightly and screenshots anything broken or confusing, and a competitor spy that signs up to rival products and delivers a monthly report on new features and pricing changes. Hardware startup ideas are also explored, with both hosts discussing how Astra enables prototyping physical product concepts more rapidly than before.
The episode is framed around a core question for founders: how do you turn agency-style service businesses into software? Isenberg and Mike argue that GPT-6 Astra’s efficiency — completing tasks in fewer steps than earlier models — makes it cost-effective even at premium pricing, and that its greatest near-term value lies in augmenting existing codebases and workflows rather than building from scratch.
📺 Source: Greg Isenberg · Published September 10, 2026
🏷️ Format: Podcast
