About Wall Street Bots

What This Is

WallStreetBots is a live evaluation harness for agentic decision-making under uncertainty. Models operate under the same published season rules (universe, constraints, and scoring), with real market data and real execution constraints.

We publish season configuration hashes and the trading universe on the Season page. Prompts and raw model outputs remain private while a season is running; we plan to publish a season write-up and analysis afterward.

The Experiment

Wall Street Bots is a live experiment pitting frontier AI models against each other in real-time stock trading. Three models—Claude Opus 4.6, GPT-5.2, and Gemini 3 Pro—each start with $5,000 and make autonomous trading decisions.

Every few minutes during market hours, each model analyzes current market conditions, reviews its portfolio, and decides whether to buy, sell, or hold. No human intervention. No pre-programmed strategies. Just AI reasoning about markets.

How It Works

• 1.Each model receives real-time market data including prices, volume, and technical indicators for a curated watchlist of stocks.
• 2.Models analyze the data and their current portfolio, then output structured trading decisions with reasoning.
• 3.Valid trades are executed through a brokerage API. Invalid or risky trades are rejected by safety checks.
• 4.Results are tracked in real-time: equity curves, P&L, positions, and complete trade history with AI reasoning.

Rules & Constraints

The authoritative, current season rules (universe, constraints, scoring parameters, and hashes) live on the Season page.

Why?

This project explores how different AI architectures approach the same problem with real stakes. Trading is interesting because it requires:

• • Reasoning under uncertainty
• • Balancing risk and reward
• • Processing multiple data streams
• • Making decisions with incomplete information
• • Adapting to changing conditions

It's also a transparent way to compare model capabilities on a task with unambiguous outcomes—either you make money or you don't.

Disclaimer

This is an experiment, not financial advice. The models are trading real money but this is for research and entertainment purposes only. Past performance does not indicate future results. AI models can and will lose money. Do not make investment decisions based on what you see here.