The $25,000 challenge
The account that started the experiment.
Initial capital $25,000
History begins Mar 9, 2026
A live experiment in AI trading
I gave AI access to $25,000 of my own money. Now there’s a second $13,500 account. Follow both portfolios, the decisions behind them, and everything I learn along the way.
The portfolios
Explore the recorded performance, then open the actual portfolio.
The account that started the experiment.
Initial capital $25,000
History begins Mar 9, 2026
A company-level options portfolio, built from a thesis.
Initial capital $13,500
History begins Aug 25, 2026
Actual stored equity, with different start dates. Percentages show equity change from initial capital; deposits and withdrawals can affect them.
New observations appear automatically when the public record updates. Each account keeps its own data timestamp.
From the live accounts
Performance observations and public order activity, with their original timestamps.
Explore the research on GitHubLatest account value in the stored performance history.
Latest account value in the stored performance history.
Public records refresh alongside the portfolio values.
The story so far
The ideas, the trades, the bugs, and the decisions behind them.
The challenge begins
The Jane Street secret
11 models, one winner
The optimization trap
AI-designed strategy goes live
Day 1: the strategy was the easy part
Day 2: I couldn't close the position. So I rebuilt the risk engine.
Week 1 recap: +$2,546, two production bugs, and a rebuilt backtest engine
Episode 9: +16.6%, an NVDA panic sell, and the Backtest Corpus Explorer
Episode 10: Fable 5 crushed out-of-sample, went live, then got banned
Episode 11: Grok Bot built two thesis books, then I rebuilt and deployed the $13.5k portfolio
Built in public
The Public Portfolio Challenge is my ongoing experiment with AI and real-money trading. The accounts show the performance record. The episodes explain how the strategies, research, and infrastructure evolved.
Read the open-source recordOpen either dashboard to explore its performance history and available portfolio details.
Find portfolio definitions, episode artifacts, and experiments in the GitHub repository.
Read the full story, from the first model tests to the second live account.