Intent Snake is a terminal AI game for comparing two OpenAI-compatible LLM agents on the same deterministic Snake maze. Model A and Model B control separate snakes in identical arenas. The local engine advances forced corridors one cell every 500 ms and calls a model only at junctions with multiple legal routes. Compare route planning, score, food collected, survival, decision count, latency, and replay. The Vault Citadel arena contains 21 food items, an outer ring, and three vault loops that test safe planning as the snake grows. Source: https://github.com/pidefu/intent-snake. Play: https://pidefu.itch.io/intent-snake. Gameplay recording made with GIF Master Online Screen Recorder: https://gif-master.com/en/tools/screen-recorder
Features
- A/B LLM comparison; deterministic Snake maze; junction-only decisions; bilingual terminal UI; OpenAI-compatible API; cell-by-cell replay