PATCH//32

Gameplay Mechanics

PATCH//32 is a 75-second tactical survival game played across one literal 32×32 neighborhood. Corruption spreads inward from every edge. Move with WASD or arrow keys and press Space to deploy a patch, or tap any grid cell on a phone or tablet. Each patch costs scarce charge, repairs nearby corrupted cells, and may refund up to 10 charge. Keep network integrity above 48% until the timer reaches zero while building the highest score you can.

Key Features

• A complete game built around exactly 1,024 cells.

• Scarce-energy decisions: spend now on a cluster or preserve charge for the next surge.

• Large chain repairs reward accurate targeting and multiply score.

• Responsive mouse, touch, and keyboard controls with synthesized audio and no downloads or login.

• Free, browser-playable single-player experience.

What makes PATCH//32 stand out

The 32×32 constraint is the simulation itself, not decoration. Every cell is visible, vulnerable, and strategically meaningful. The tension comes from repairing a tiny network under pressure while balancing score, integrity, patch radius, and a deliberately limited energy economy.

AI Tools Used

ChatGPT and the PATCH32 Game Designer Mind from Minds by Animoca Brands.

How AI was integrated into development

The Game Designer Mind was used across multiple rounds to compare beginner, aggressive score-chaser, and cautious survival policies; identify a dominant positive-energy loop; recommend an isolated economy patch; and validate that patch with reported multi-seed policy simulations. Its recommendation reduced starting charge from 72 to 52, passive regeneration from 4.8/s to 3.0/s, and maximum refund from 24 to 10. A later isolated build raised the loss threshold from 28% to 48% after the Mind reported that all modeled policies still won 100% of runs. These were simulated policy tests rather than fresh human playtests. The Mind also reported DNS failures in its cost-curve tool, and its claimed Python simulator attachment was not delivered, so those reported results were not independently reproduced. ChatGPT supported implementation, testing, accessibility work, deployment, and submission preparation.