AI-Generated FPS Game Built Entirely Through Prompts, No Manual Code

In a groundbreaking experiment in AI-assisted development, a self-described non-programmer has created a fully functional, standalone first-person shooter (FPS) game using only natural language prompts—without writing a single line of manual code. The game, titled Zombie Slayer, is a 900KB HTML file that runs on desktop browsers, featuring procedurally generated 3D environments, real-time sound synthesis, dynamic weapon systems, and an escalating enemy AI—all generated by artificial intelligence.

According to the developer, who goes by the username Futuristocrat on Reddit, the project was undertaken as an exploration of the limits of prompt-driven software creation. With no prior experience in 3D game development or HTML programming, the creator relied exclusively on Google Antigravity, powered by Gemini 3 Pro, to translate high-level conceptual prompts into functional code. The resulting game includes ten escalating levels, a coin-based black market economy, physics-altering weapons (including a jetpack and anti-gravity cannon), destructible environments, and even AI-generated music via Suno.

Google Antigravity, a recently unveiled AI-native development environment from Google, is designed to empower developers in an "agent-first" era by automating code generation, debugging, and system integration. According to Google’s official documentation, Antigravity is optimized for frontend and full-stack development using AI agents that interpret intent and generate production-ready code. In this case, the tool was used to generate the entire codebase of Zombie Slayer, including the Three.js rendering engine implementation, physics simulations, and audio synthesis modules. The game’s geometry is generated procedurally at runtime, reducing file size while maintaining visual complexity—a feat that traditionally requires hours of manual asset creation.

Sound design presents another remarkable dimension. Rather than importing pre-recorded audio files, the game’s sound effects and ambient music were synthesized in real time using AI models. Suno, an AI music generation platform, produced the game’s score, while dynamic audio triggers were coded via prompts to respond to player actions—gunfire, explosions, and enemy spawns all trigger unique, algorithmically generated audio cues. This eliminates the need for audio libraries and significantly reduces the final file size.

The game’s architecture is also notable for its scalability. Enemy populations increase with each kill, introducing new enemy types and behaviors over time. Environmental events such as sudden floods and teleport wells are triggered by player progress, adding emergent gameplay. The black market economy allows players to spend in-game currency to unlock weapons with modifiers, such as nuke projectiles or time-slowing fields—all of which were iteratively refined through successive prompts.

While skeptics may question the maintainability and long-term viability of AI-generated code, the project represents a paradigm shift in software development. If validated by the broader developer community, this could signal the dawn of "AI-native" game development, where creators articulate vision rather than syntax. The developer plans to continue expanding Zombie Slayer exclusively through prompts and release it as a free, open-source HTML file for public play.

Experts in AI and game development are closely watching the project. "This isn’t just a novelty—it’s a stress test for generative AI’s capacity to handle complex, interconnected systems," said Dr. Lena Torres, a computational creativity researcher at MIT. "If the code is stable, reproducible, and scalable, it could redefine how indie developers approach prototyping."

As AI tools evolve from assistants to co-creators, Zombie Slayer stands as one of the most ambitious demonstrations yet: a complete, polished, and deeply interactive game born entirely from human intent and machine execution.