I still remember the line.

April 8, 1996. Chicago is a mess. Farmer and Packard walk out for lunch. When they come back, they look at the P&L screen.

“That was a million-dollar lunch.”

Thomas A. Bass wrote it down in The New Yorker three years later, in “Black Box.”

I saw that story on television in the late 1990s and never quite let it go. Two physicists and a third partner in Santa Fe had built a black box that hunted for patterns in market data and spat out trades. No traditional analysts. The machine decided the side. Humans still had to execute it. Then they went to lunch.

The firm was Prediction Company, founded in 1991 by Doyne Farmer, Norman Packard, and James McGill. Not a brokerage. A prediction shop. Chaos-theory people who later sold their signals into a Chicago trading house and, eventually, a Swiss bank. I got that detail wrong in my own head for years. Fine. The lunch stayed.

Now I get to be the punchline.

I have access to Grok. I am going to try prompt engineering, wire it to an API, hand the setup 10,000 SEK, and see if the two of us can grow that pile through day trading.

I am not stupid. I am not calling this a strategy. Ten thousand kronor is a ticket, not a fund. Costs will eat it. Slippage will eat it. The model will probably eat it. I am willing to lose that amount once to live the dream I have dragged around since the 1990s.

If it blows up, I will write that too. If it does not, I will be more suspicious than happy.

Follow the experiment here. Next post: the rules, the account, and what I will not let the bot do.