Sep 3, 2026 · 32 min
GPT-6 Astra turns ambitious prototypes into practical builds
GPT-6 Astra is a banger - here’s everything I’ve built
Claire Vo’s early tests suggest stronger computer use and coding could shift AI from answering questions toward operating tools and building working prototypes.
- 1Astra succeeds at product intelligence, browser QA, hardware control, 3D creation, and desktop app development where earlier models repeatedly failed.
- 2Its stronger computer use expands what one person can prototype, from manipulating Blender to developing and testing software across desktop environments.
- 3The practical tradeoff remains speed and cost, with Astra’s potential as a daily driver still weighed against its capabilities.
Don't miss
Vo’s demonstrations show Astra connecting product work, browser testing, 3D creation, hardware control, and desktop development into practical prototypes.
The brief
Claire Vo introduces GPT-6 Astra through early-access experiments, arguing that its ability to complete tasks earlier models repeatedly failed has materially expanded what she is willing to attempt.
The demonstrations move beyond chat: Astra handles product intelligence in ChatPRD, browser-based QA, hardware control, 3D asset creation in Blender, and desktop app development.
The episode’s central argument is that stronger computer use changes prototyping economics, letting a single builder connect software, browsers, creative tools, and physical devices more directly.
Vo also keeps the enthusiasm grounded, weighing Astra’s speed and cost against its capabilities and asking whether it can become a dependable daily-driver model.
The standout is the cumulative shift in ambition: tasks that once stalled across separate tools now become plausible end-to-end projects under one model’s control.
Featuring
Claire Vo
OpenAI
FloraListen to the full episode and explore every guest, topic, and moment on PodLume.