You've seen the videos. Some guy with a clean desk and three monitors shows you an AI trading strategy. The backtest returns 200%. Maybe 300%. The math looks tight. The charts go up and to the right. And he's giving it away. Free.

Here's what gnaws at me.

The math is real. Not the YouTube version. But the kind underneath it. Pattern-finding tools that spot hidden shifts in markets before the crowd catches on.

A man named Jim Simons built a fund around this math. His Medallion Fund, run by a firm called Renaissance Technologies, returned about 66% a year before fees for three decades straight. After fees, the net was still about 39% a year. No other fund in history comes close over that stretch. Not Buffett. Not Soros. Nobody.

Simons died in May 2024. The math survived him. The fund still runs.

Now here's the part the videos leave out.

Around 2005, Simons gave back every outside dollar. All of it. He shut the door to anyone who didn't work at the firm. Today it manages about $55 billion for employees and their families. That's it.

The man who proved this math works refused to let one extra person trade it.

He told people why. More money chasing the same patterns kills the patterns. The crowd's weight crushes the thing the crowd is chasing. Picture a fishing hole. Works great when one guy knows about it. Tell the whole town and the fish are gone by Tuesday. This is why Renaissance sues employees who leave and try to trade on what they learned inside.

Think about the fees. Renaissance charges 5% off the top and takes 44% of the gains. Most hedge funds charge 2 and 20. Renaissance charges more than double. And still turns your money away. If the strategy could handle more cash, they would take more cash. The fee structure is a confession scratched into stone. The edge is small. The edge is fragile. The edge dies in sunlight.

So what is the guy on YouTube showing you?

A corpse.

He tuned a model against old price data until it fit like a glove. Every wiggle. Every spike. The backtest looks stunning because the model memorized the past. Like studying last year's answers and expecting to pass this year's test. In the trade they call it overfitting. And the costs nobody shows you make it worse. Every time you buy or sell, you pay the spread. You pay the price moving against you before your order fills. You pay fees your broker buries in the fine print. The guys running billion-dollar funds model those costs down to the penny. YouTube backtests skip them. That trick alone can flip a winning strategy into a loser.

The returns look real in the rearview mirror. They vanish through the windshield.

I don't think most of these creators are running a con. Some are. But most found a shiny tool, ran it on old data, watched the chart climb, and hit record. The damage is the same either way.

Here's what I keep. A filter. One question I burned into the wall above my desk.

If this strategy works, why is he teaching it instead of trading it?

Jim Simons answered that question with $55 billion locked behind a door he welded shut. He spent decades proving the math was golden. Then he melted the key.

The guy giving it away for free is not more generous than Simons.

He just doesn't have what Simons had.