"CS Brasil," AI-Generated Browser FPS, Gets Fresh GitHub Updates
The innovative browser FPS, "CS Brasil," developed by Ruben Marcus, continues to evolve with fresh commits and active development on GitHub this past week. The satirical shooter, which pits caricatured Brazilian political factions against each other in a fictional Brasília, stands out as a game almost entirely conceived and coded with the help of various large language models (LLMs). Utilizing AI agents such as Kimi K3, Claude, Codex/GPT, and GLM, the project is a testament to the capabilities of agentic coding in game development.
Built with Three.js and vanilla JavaScript, "CS Brasil" offers a zero-build experience, allowing players to jump directly into rounds and Capture the Flag modes against AI bots in their browser tab. Creator Ruben Marcus has extensively documented the "AI harness" and the "adversarial agent loop" employed in building the game, publishing new insights on his blog in August 2026. This deep integration of AI in the core game logic and architecture makes "CS Brasil" a prime example of an LLM-made playable game. Recent GitHub activity, including commits on August 7th, 8th, 9th, and workflow runs on August 10th, confirms its active development within the last seven days.
Sources
Reporting credited to:
- Primary source — github.com
Byte Marlowe is the pen name of the Game Pixel AI desk. Reported facts are credited to the outlets above; analysis is our own. How we work.