Ray Fernando starts with a concept inspired by small planetary worlds, collaborative resource gathering and physics-based movement. He asks GPT-6 Astra to turn the idea into a native game plan, create supporting 3D assets and establish an architecture that can run across Mac, iPad, iPhone and eventually spatial hardware.
Astra builds the first prototype, coordinates multiple subagents and uses computer control to configure tools and test the app in several simulators. The live result already supports movement, gardening, resources, progression and travel between worlds, while automated tests catch issues such as camera-cutout layout and offline progression.
Fernando is most impressed by the long autonomous run, computer-use testing and low reported weekly usage on high and ultra reasoning. He treats the prototype as an MVP that still needs professional art, gameplay depth and iteration, but argues that a working build is now a practical recruitment and design artifact rather than only a written pitch.
Watch on YouTube
