07/08/2026
Thousands of trading bots made decisions in June based on a number that wasn't real.
We found out this morning.
Here's what happened. The US jobs report came out today. Everyone expected around 80,000 new jobs, the economy lost 23,000. Every bot on the market traded that within a second.
But further down the same report, June was quietly corrected. It had been published as +57,000. It was actually +20,000.
That's not even the main problem though.
You can now hook an AI agent straight into a brokerage account. No code, one prompt, it trades while you sleep. Robinhood opened up to third-party agents this summer. The barrier is gone. Anyone can set this up by Sunday.
So think about Google Maps for a second. There's a traffic jam, your app finds you a clever shortcut through a side street, great. Except everyone else has the same app. So everyone takes the shortcut at the same moment, and the shortcut becomes the traffic jam.
The app wasn't wrong. It just stopped being a shortcut the moment everybody had it.
Now be honest. How many people build their own model? Or do most of us plug into the same two or three ready-made tools, trained on the same data, reading the same signals? A thousand bots running the same logic isn't a thousand traders. It's one animal. Big, slow, extremely predictable.
Regulators already have a name for this. Herding risk.
The Bank of England looked at it last month. Hedge fund equity balances at prime brokers are up around 40% in a year, increasingly concentrated in the same sectors, semiconductors especially. In June, US lawmakers wrote to the SEC about the retail version of the same thing.
Translation: everybody heads for the same door at the same time. And if you know where the crowd will run, you don't need to outsmart it. You push price into the spot where the stop losses sit and let the machines do the rest.
The CFTC published a warning back in January 2024 titled "AI Won't Turn Trading Bots into Money Machines." In it they describe people promised automatic money who lost tens of millions instead. One case alone took nearly 30,000 bitcoin. About 1.7 billion dollars at the time. That was before any of this got easy.
The tools are real and they aren't going away. But the danger was never that AI can't trade. The danger is that it got easy enough for everyone, all at once, all pointed in the same direction.
Being fast doesn't help when you're fast in a crowd.
Which is the whole reason we teach the boring version. A portfolio you actually understand, sized so you can leave it alone, checked once a month.
Boring is slow. Boring also doesn't get front-run.
So tell us. Are AI trading strategies the future of retail investing, or the most efficient machine ever built for taking money off beginners?