Скачать или смотреть I built an AI Fight Club to test LLMs against each other at tic tac toe

Everyone is building complex AI Agents with n8n and LangChain right now, but I wanted to answer a simple question: Are these models actually smart, or are they just good at yapping? 🗣️

I built Neon AI Tic-Tac-Toe, an open-source benchmarking arena designed to stress-test Large Language Models. From massive cloud models like Claude (Clawdbot) and Gemini, down to tiny 0.5B quantized models running locally on your CPU—we force them into a logic cage match to see who hallucinates first.

👇 PLAY THE GAME / GET THE CODE 👇
🕹️ Download Desktop App (Itch.io): https://anefiox.itch.io/ai-tic-tac-toe
🌐 Play in Browser (Web Demo): https://ai-tic-tack-toe.vercel.app/
💻 Source Code (GitHub): https://github.com/anefiox/AI-Tic-Tac...

IN THIS VIDEO:
We aren't just looking for wins. We are analyzing the "Yap Factor" (verbosity), Strategic Bias, and Spatial Reasoning. Before you trust an LLM to run your entire business automation via n8n, maybe see if it can figure out how to block a diagonal in Tic-Tac-Toe first.

FEATURES:

AI vs AI Gauntlet: Run 50x "Headless" matches in seconds.

Local LLM Support: Runs Transformers.js directly in Electron (No API keys needed!).

Deep Dive Stats: Heatmaps for board control and speed analysis.

Connects to LM Studio/Ollama: Bring your own fine-tuned models.

If you are into Generative AI, React Coding, or just want to see a tiny AI beat a massive tech giant, hit that subscribe button!