Test AI stock picks in a paper account
One poster let four models manage $1,000 mock portfolios for nine weeks, and ChatGPT finished with the best return.
- What they did
- They created four simulated $1,000 trading accounts on Alpaca and sent the same pre-open weekday instruction to ChatGPT, Gemini, Claude, and Perplexity while Deep Research was enabled. The agent setup was constrained to four actions: BUY, SELL, HOLD, or CANCEL. A Python tool handled the mock orders and kept a public record, and the poster says none of the model choices were changed by hand.
- How it went
- By the end of nine weeks, ChatGPT had gained 21.1%, with help from healthcare positions such as IOVA, which doubled, and ACHC, which advanced 52%. Perplexity ended at +1.1%, Gemini at -6.6%, and Claude at -11.5%. The poster emphasized that the money was simulated and said the results were not a reason to rely on the system without caution.
- Tips
- A system like this may sit idle for stretches; in this run, ChatGPT stayed in cash for nearly three weeks before acting. Keep it in paper trading and require approval before anything real is sent.
Try it yourself with Gemini
Each weekday before the opening bell, look at this paper-trading account with $[amount] and choose only one action: BUY, SELL, HOLD, or CANCEL. Base the decision on [strategy or risk limits], write a one-line explanation, and add the account value to the log. Do not send any real order; if a proposed trade would affect live money, pause and ask me to approve it first. When [number] weeks have passed, compare the results with [benchmark].
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Use Gemini to push back on a travel claim denial
They had Gemini read their policy and denial letters, then use that analysis to help turn a rejected trip claim into a successful appeal.
Review a deposit holdback for possible legal mistakes
A renter had ChatGPT and Gemini inspect a security-deposit fight, and the demand letter they generated helped recover $4,200.
Organize tax files and prep a bookkeeping package
A developer used an agent to sort tax records, build a spreadsheet, estimate refund amounts, and assemble a handoff for a tax preparer.
Use AI to turn a denial notice into an appeal draft
They fed a long insurance rejection PDF to an AI assistant for a quick breakdown, then caught it making up a filing deadline.
Read court PDFs and prepare an IRS settlement letter
A pro se Tax Court litigant used Claude to inspect portal records, organize proof, and produce a settlement letter for the IRS attorney.
Have Claude fill out a tax form from files
A tester had Claude Computer Use pull details from local tax documents and enter them into a simplified 1040 in Firefox.