I trapped 4 AI agents inside a trading firm and turned it into a reality show

The project "Wall Street of AI Agents" simulates a financial market where four distinct AI traders, each with unique personalities, compete to grow their portfolios starting with $10,000. It serves as a benchmark for Small Language Models (SLMs), evaluating their ability to analyze market changes, manage risk, and collaborate in real-time, all while running on a compact 2 billion parameter open-source language model using Llama CPP inference. This setup provides insights into the capabilities of small models in dynamic environments, which is crucial for practitioners developing multi-agent systems in finance.