A brokerage just gave every customer a button that trades for them. Over 150,000 people turned it on in four months. The agents hit the platform’s tools about 30 million times a day.

That’s not a bug report. It’s a structural change, and it matters more than the headlines about “AI trading” suggest.

Here’s the thing nobody says plainly: those 150,000 accounts aren’t running 150,000 independent strategies. They’re choosing from a handful of shared models — the same few AI systems, pointed at the same market data, optimizing toward roughly the same goals.

A market works because the people feeding it disagree. Millions of small, independent decisions — wrong, right, late, early — average out into a price you can mostly trust. Strip out the independence and you don’t have a market of individuals anymore. You have one decision, copied a few hundred thousand times.

That’s the shift to watch, and it’s the opposite of the usual fear. The worry isn’t that the AI is irrational. It’s that the AI is identical.

Correlation is the word to watch for

The people who run markets have already started saying it. Regulators asked, in writing, what happens when “large numbers of participants rely on similar models” and produce “correlated trading decisions.” A central banker put it bluntly: human crowds panic at different speeds — some hesitate, some bail early, some hold. A population of identical agents doesn’t. They all respond to the same signal the same way, at the same moment. One wobble becomes a rout before anyone can slow it down.

That’s not speculative scaremongering. It’s the mechanism — and it’s the first time the mechanism has been named this clearly by the institutions that would have to clean up the mess.

The honest counter — and why it matters

Here’s the part that keeps this from being a doom post. The strongest evidence in the other direction came years ago, and it still holds: when researchers looked at algorithmic trading across an entire currency market, they found the algorithms were correlated — but that correlation had no measurable link to higher volatility. If anything, more algorithmic trading tracked with lower volatility.

So the truthful position is narrower and more interesting than “AI crashes markets.” It’s this: correlation is real, and whether it damps or amplifies volatility depends entirely on what the shared models are optimizing toward — and right now, nobody knows.

The risk isn’t decided. The risk is unmeasured.

What this means for the solo builder

Most advice about differentiation is about standing out — the moat, the positioning, the “be different so you win.” This reframes it. When everyone ships on the same models, the sameness stops being a marketing problem and becomes a safety problem — for you, and for whatever you’re part of.

Your leverage was never going to be the model. It’s the things you own that the shared model doesn’t: your data, your rules, your specific problem, your taste. That’s what harness-not-model means by “you rent the brain, you build the rig.” But there’s a second, less obvious reason to own your context.

Owned context is what makes your output not-a-copy. When you train on your users, your failures, your niche’s edge cases, you produce decisions that are genuinely your own. They won’t correlate with the other builders on your street the way the out-of-the-box answers do. You stop being one more echo of the same model.

In a world where sameness is becoming a systemic risk, deliberate difference isn’t a vanity metric. It’s a form of resilience — yours, and the system’s.

The one-line version

If a model is cheap and everyone has it, the model isn’t a strategy — it’s a backdrop. What you actually own is the context you feed it, and the rules you set. That’s been the argument for a harness for a while now. The new part is why it’s urgent: because when everyone ships the same AI, the sameness is what breaks things — not the intelligence.

Sources

  • Robinhood HOOD Summit 2026 (Sept 29, 2026). — 150,000+ agentic trading accounts opened since May; agents hit Robinhood’s tools ~30M times/day; in-app agents choose from shared OpenAI/Anthropic models.
  • U.S. House Financial Services Committee letter to the SEC (June 23, 2026, Foster & Sherman). — flags “correlated trading decisions” and “herding behavior” among agents “trained on similar data” as a market-integrity risk.
  • Bank of England, Deputy Governor Sarah Breeden (June 30, 2026). — autonomous agents could “amplify volatility in stress,” moving in lockstep in a way no human crowd can match.
  • Federal Reserve, IFDP “Rise of the Machines” (foreign-exchange algorithmic trading). — algorithmic trades are correlated, but with no causal link to higher volatility; if anything, more algorithmic trading tracked with lower volatility.