A 9-pound cavapoo named Momo has achieved what no other dog in history has: coding a video game. Not through training, not through design—just by pressing keys on a Bluetooth keyboard while an AI interprets the chaos as creative genius. The result is Quasar Saz, a retro-futuristic arcade game featuring cosmic weapons, corrupted sound enemies, and ghost sprites that look suspiciously like Pac-Man’s classic villains.
The project, the brainchild of creator Caleb Leak, blends hardware hacking with AI experimentation. At its core, the setup is deceptively simple: Momo’s keyboard inputs are funneled through a Raspberry Pi 5, filtered by a custom Rust application called DogKeyboard, and then sent to Claude, an AI model. The twist? Instead of treating the dog’s random typing as noise, Claude is primed to interpret every keystroke as a cryptic, genius-level game design instruction.
How a Dog Learned to Code (Sort Of)
The real challenge wasn’t getting Momo to press keys—it was convincing the AI to take the inputs seriously. Leak structured the interaction around a fictional persona: an eccentric game designer who communicates in riddles and nonsense. The prompt told Claude to treat even the most nonsensical strings (like skfjhsd#$%) as deliberate, high-concept game directives. With guardrails in place—automated feedback loops, screenshot validation, and shader checks—the system could translate Momo’s keyboard mashing into functional code within 1–2 hours.
The games are built entirely in Godot 4.6 using C#, with no human intervention beyond the initial setup. A smart pet feeder dispenses treats after each burst of input, reinforcing the behavior while Claude processes the data. A chime signals when the AI is ready for more design inspiration, turning the process into an oddly satisfying game for Momo.
A Game Born from Chaos
Quasar Saz is the first (and so far only) product of this experiment. Players control Zara, wielding a cosmic saz to battle corrupted sound across six stages and a boss fight. The aesthetic is a mashup of 1980s arcade graphics and surrealism, complete with oversized, virus-like enemies and ghost sprites that evoke Pac-Man’s ghosts. The result isn’t polished—it’s raw, glitchy, and delightfully unintentional. Yet it’s fully playable, a testament to how far AI has come in interpreting abstract creative input.
Leak acknowledges the whimsy of the project, but the underlying mechanics hint at broader possibilities. If a dog’s keyboard mashing can generate functional code, what might happen if the input were more structured? Voice commands? Real-time sensor data? The experiment pushes boundaries in both AI creativity and the absurdity of human-animal collaboration.
For now, Momo’s next project remains a mystery. But one thing is clear: the line between accidental chaos and creative genius just got a lot fuzzier.
