The Wolf of Wall Bot.

I put a small amount of real money in a brokerage account, handed the keys to an AI bot, and gave it two personalities: one channels Jordan Belfort, the other channels Warren Buffett. Then I took my hands off the wheel. No grand strategy, no work angle. This one is purely for fun, all in the name of seeing what an agentic AI actually does when it's allowed to act on its own.

Not financial advice. Just an experiment. Nothing here is a recommendation to buy, sell, or do anything at all. If you find yourself copying a bot whose job description is "be Jordan Belfort," please close this tab and go for a walk.

Bots in charge: 1 bot, 2 moods
Started: Oct 2026
Updates: Weekly Fridays
Human trades: 0 by design

Meet the two bots

Same bot, two very different mentors.

Both modes run on their own and make their own calls. The only thing I picked was the personality, which, in hindsight, was the most important decision and the one I made the least seriously.

Mode 1 · Aggressive

Jordan Belfort mode

Wolf of Wall Street energy, minus the yacht, the fraud, and the federal prison.

Plays: Stocks, plus buy-only calls and puts
Goal: +10% by the last trading day of October 2026
Personality: Loud, caffeinated, allergic to the word "diversify"
Safety rail: Options are buy-only, so the worst case is losing what it bets, not owing anyone a boat

Why this is a terrible idea: I took the most famous cautionary tale in finance and turned it into a job description, then gave it a login. Ten percent in under two weeks is the kind of goal that sounds great right up until you say it out loud.

Mode 2 · Sensible

Warren Buffett mode

Patient, boring, and almost certainly drinking a Cherry Coke.

Plays: About 60% in a Treasury bill ETF, 40% in a broad index ETF
Goal: Beat a high-yield savings account
Personality: Calm, unbothered, checks the news once a week, maybe
Safety rail: The whole strategy fits on an index card, which is sort of the point

Why it's here: every experiment needs a control group. This is the bot that makes the other one look reckless, mostly by doing nothing and being fine with it.

My prediction: Buffett mode wins on sleep, Belfort mode wins on stories. Results below, whatever they turn out to be.

Week by week

How it's going.

Every Friday after the close the bot writes a short summary, and this page picks it up. Everything starts at an index of 100, so you can compare the two modes with the S&P 500 (SPY) and a savings rate without seeing a single dollar.

Waiting for week one

Index · start = 100

First results land Friday, Oct 23.

The bot makes its first trades on Wednesday, Oct 21. Check back after Friday's close for week one. Until then, there's nothing to chart, and I'm not going to make anything up.

Scorecard

Week 1 · pending

Weekly and cumulative returns by mode, with SPY and a savings rate for comparison
Series	Index	This week	Since start
Belfort mode	100.00	—	—
Buffett mode	100.00	—	—
SPY	100.00	—	—
Savings rate	100.00	—	—

Everything starts at 100. The first scores post after Friday's close.

Lessons

Week 1 · pending

No lessons yet. The bot has to make some mistakes first.

Trade log

0 trades

All weeks

Trades made by the bot: date, mode, ticker, side, type, result, and its one-line reasoning
Date	Mode	Ticker	Side	Type	Result	Why the bot did it
No trades logged yet. The first week's trades show up here after Friday, Oct 23.

How it works

The bot drives. I watch.

The bot picks the trades, places them, and writes down why. My job is to not touch anything, which is harder than it sounds.

Each Friday after the market closes, it writes a weekly summary: each mode as an index that started at 100, the week's and the running percent return, the same for SPY and a savings rate, every trade with a one-line reason, and a few lessons it took from the week.

This page only ever shows those normalized numbers. No dollar amounts, no balances, no account details, ever. Partly for privacy, and partly for my dignity.

What I'm actually curious about: does an agent stick to the personality it's given, does it explain itself well, and does "aggressive" mean clever or just loud? Whatever happens, I'll report it straight, including the weeks where the wolf gets eaten.

One more time: not financial advice. This is a for-fun experiment with an AI bot and a small, deliberately boring amount of money. Past results, especially a bot's, say nothing about future results. Talk to a licensed professional before doing anything with your own money.

AI agentsJust for funVanilla JS + SVGPython

← All AI projects