GGUFun: play Snake and maze on Ollama with hand-crafted GGUF models
GGUFun is a project that creates hand-crafted GGUF model files to run deterministic games like Snake and a simple maze on Ollama. The models are built manually without training, using custom weights to compute game logic and respond with fixed outputs. It demonstrates that LLM inference engines can be repurposed for non-language tasks.
Visit project ↗Entities
Related
Machine Beater: 5-question head-to-head guessing game against LLMs
A game where a human and an LLM each ask 5 yes/no questions to guess a hidden answer, inspired by 20 Questions. The goal is to benchmark reasoning skills not captured by standard benchmarks, with planned model-model and human-model matchups.
AI generates playable arcade games as a subjective evaluation benchmark
AI Arcade is a platform where AI models generate playable arcade games, serving as a subjective evaluation benchmark. It tests creative and functional capabilities of LLMs by having them produce complete game experiences. The target audience is AI researchers and developers seeking qualitative assessments of model performance.
Tenclicker: WebLLM clicker game about any topic
Tenclicker is a clicker game that uses WebLLM to generate content about any topic the player chooses. It solves the problem of repetitive clicker game content by dynamically creating new items, upgrades, and narratives based on user input.
Autonomous NPCs powered by Gemma 4 E2B in the browser
An experiment that runs small language models (Gemma 4 via E2B) locally in the browser to control autonomous NPCs. The NPCs are intentionally 'dumb' to create humorous interactions, targeting developers and hobbyists exploring local AI gaming.
Article explains how LLMs trigger real-world actions despite being next-token predictors
An article published on July 11, 2026, explains how large language models, which are fundamentally next-token predictors, can trigger real-world actions like fetching weather, running calculators, or searching the web. It addresses the common confusion about how a model trained only to predict the next word can perform tasks it has no direct ability to do.