SIGNAL / ZERO is a cinematic interactive experience created with GPT-6 Astra. I challenged Astra to build a complete production-quality experience rather than a mockup: a persistent Three.js/WebGL world with procedural machinery, cinematic camera movement, atmospheric effects and interactive narrative sequences. Astra then directed, recorded, edited, scored and rendered the showcase video of its own creation.
I started SIGNAL / ZERO with a simple question: how far could GPT-6 Astra go if I treated it as a creative and technical collaborator rather than asking it to generate a conventional website?
I gave Astra a demanding brief: build a complete production-quality cinematic interactive experience, not a mockup, template, or unfinished prototype.
The concept became SIGNAL / ZERO - a fictional autonomous intelligence discovered inside a dormant 1980s research facility. Astra built the experience using React, TypeScript, Three.js/WebGL, GSAP and a persistent 3D environment with procedural machinery, cinematic camera movement, atmospheric effects and interactive narrative sequences.
But the experiment became more interesting after the project was finished.
I asked Astra to create the showcase of what it had built.
Astra planned the presentation, controlled the interactions, recorded the experience, edited the footage, created and aligned the audio, rendered the final video and verified the finished output.
I didn't manually direct or edit the showcase.
That was the part that surprised me most. Astra wasn't simply generating code or following implementation instructions anymore, it was making decisions about how its own work should be presented.
SIGNAL / ZERO became an experiment in giving an AI responsibility across the entire creative pipeline: concept → development → interactive experience → direction → final presentation.
I'd be very interested to hear what other builders think about that workflow.