Let AI models trade stocks autonomously with $1k each
Poster ran four AI models on a weekday pre-market prompt to trade $1,000 portfolios each for nine weeks; ChatGPT came out ahead. Task: Suggested implementation: Set up four paper-trading portfolios with $1,000 each on Alpaca. Run the same weekday pre-market prompt against ChatGPT, Gemini, Claude, and Perplexity with Deep Research enabled. Each model must return only BUY, SELL, HOLD, or CANCEL. Automate the trading flow in Python and log all actions publicly. Human role: Monitor logs for anomalies and approve any deviation from the strict output format.; tools: Gemini. Human role: Source evidence is unavailable regarding specific human oversight roles or intervention protocols during the autonomous trading period.; limitations: Prerequisites include API access to four AI models and an Alpaca paper-trading account. Operating costs involve API usage fees and potential latency issues. Integration uncertainty exists regarding Python automation reliability. Exception handling is critical for models that fail to return valid trade commands. Source reliability is low due to HTTP 403 error. Human approvals are required before any real-money execution..
Tools
Published on Agent Work Atlas:
Overview
Poster ran four AI models on a weekday pre-market prompt to trade $1,000 portfolios each for nine weeks; ChatGPT came out ahead.
Workflow
Suggested implementation: Set up four paper-trading portfolios with $1,000 each on Alpaca. Run the same weekday pre-market prompt against ChatGPT, Gemini, Claude, and Perplexity with Deep Research enabled. Each model must return only BUY, SELL, HOLD, or CANCEL. Automate the trading flow in Python and log all actions publicly. Human role: Monitor logs for anomalies and approve any deviation from the strict output format.
Reported result
Actual outcomes are unknown as the source is inaccessible. Proposed measurable acceptance check: After nine weeks, compare portfolio returns across models. Track specific metrics such as cash holding duration (e.g., ChatGPT reportedly sat in cash for three weeks) and sector-specific gains (e.g., healthcare picks). Verify if any model achieved positive returns exceeding 10% while others underperformed.
Human role
Source evidence is unavailable regarding specific human oversight roles or intervention protocols during the autonomous trading period.
Limitations
Prerequisites include API access to four AI models and an Alpaca paper-trading account. Operating costs involve API usage fees and potential latency issues. Integration uncertainty exists regarding Python automation reliability. Exception handling is critical for models that fail to return valid trade commands. Source reliability is low due to HTTP 403 error. Human approvals are required before any real-money execution.
Try it
I want to test autonomous AI trading. Set up four $1,000 paper-trading portfolios on Alpaca for ChatGPT, Gemini, Claude, and Perplexity. Use a strict pre-market prompt that forces each model to output only BUY, SELL, HOLD, or CANCEL. Automate execution via Python and log all trades. First, run a 24-hour dry run with $100 to verify the API connections and output parsing. If successful, proceed to the full $1,000 setup for one week. Monitor for models holding cash excessively. Provide a daily summary of portfolio value and trade frequency. Do not use real money. If any model fails to return a valid command, pause the automation and alert me.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.