← All videosTHE SHORT VERSION
Inspect the live book Copy this strategy ↗

ai trading / about me / algotrading

I let two AI agents trade $44,000 of my own money

Austin Starks answers a viewer asking whether his AI trading system can actually manage a real portfolio. It can, and it does: about $44,000 across two live brokerage accounts, all of it his own money.

He announced the $25,000 Public Portfolio Challenge on February 28, 2026 and did not place a trade until May 5, sixty six days later, because putting real money behind a backtest frightened him. The account ran from $25,000 to $31,582 by late July, fell to $28,895 in two weeks, and on August 17 he replaced the strategy with a more aggressive one. His own deployment memo, committed publicly nine minutes after the swap, says it plainly: this is not a lower-risk upgrade, it is a deliberate raw-return trade. The book then reached an all-time high of $32,443 on August 20 and suffered the worst drawdown of the entire run eleven days later.

He preserved the previous rules in a paper portfolio as a control, funded at the same NAV on the same day, so the two designs can be compared directly. He also opened a second live book of $13,500 across twenty six semiconductor and biotechnology companies, which is two weeks old and has produced nothing worth calling a result.

Figures describe the accounts on September 8, 2026 and move with the market. The strategies for the original book are public and can be inspected or copied.

Transcript

0:00I'm a senior AI engineer, and right now I've deployed two AI agents.

0:04I have about $44,000 across two live accounts, and it's all my own money.

0:09I announced the first one at the end of February, but I didn't place a trade until the fifth of May.

0:15Sixty six days, because I was scared.

0:17A backtest costs you nothing when it's wrong.

0:19Real money does.

0:21It started at $25,000.

0:23It ran to $31,600 by the end of July, then fell to $28,900.

0:28In the middle of August I swapped the strategy for a more aggressive one.

0:32Higher returns in the backtest, deeper drawdowns, and I wrote that down before I deployed it.

0:37It hit $32,400, then had the worst fall of the entire run, back under $29,000.

0:44It's at $30,600 today.

0:46I kept the old rules running on paper as a control.

0:49Since the swap the old one is down almost 6% and the live one is flat.

0:53Nineteen names, and every rule is public.

0:56That went well enough that I opened a second one two weeks ago.

1:00$13,500, semis and biotech, 26 companies.

1:04It's up under 1%, which is noise.

1:07Two weeks is not a result, and I'm not going to pretend it is.

1:10If you want to know what the AI actually does when it trades, leave a comment and I'll answer you myself.

Join the conversation

Loading conversation…