Who needs AI psychosis? I gave myself VR psychosis

A developer’s experiments with AI-assisted game making begin as an attempt to build a better balloon-popping game for his daughter. Her verdict - that it was still boring - underscores the importance of testing with real players, but the project also reveals how quickly ideas can be turned into software. Because building and testing an iOS app slowed iteration, he shifts to web games and assembles a workflow using voice dictation, Claude’s remote-control feature, Git, Vercel, and Cloudflare. He can describe changes aloud from a pub, have an agent implement them, and push a playable update online within minutes. The piece presents this low-friction cycle as a way to sharpen game-design taste through rapid experimentation.

The author’s interest in VR grows from the exhilarating but nauseating movement of a fan-made Attack on Titan game, and from learning that Three.js, WebXR, and Vite make browser-based VR games practical to build. He argues that a game’s moment-to-moment “primary loop” matters more than its story or setting, and is especially easy to test with AI-generated prototypes. His Star Crusader experiment centers on jetpack movement and shooting across planets; repeated playtesting exposes problems such as disorientation and getting lost in space. He responds with a reorientation control, a wrist-mounted 3D map, and momentum cancellation, illustrating how quick iteration can refine both the feel of movement and the supporting mechanics.