5:04 pm - April 28, 2026Scrolling through a leaderboard that looked more like a gaming forum than a financial report on a quiet evening in December of last year was the first time I gave this much thought. For two weeks, seven of the most sophisticated AI models in the world were released onto the U.S. stock market after receiving ten thousand dollars apiece. no human oversight. There is no safety net. Only Elon Musk’s Grok-4.20, up slightly more than twelve percent, had finished higher by the end. Everyone else was defeated. Over half of the money was lost on one model. It was odd to observe, and even more odd to see how seriously the finance industry took it.
Speaking with people in the industry these days gives me the impression that something has subtly changed. Algorithms are not new; in the 1960s, the New York Stock Exchange began feeding data into computers, and by 1976, brokers were using the DOT system to route orders electronically. However, the difference between a calculator and a coworker is the difference between then and now. The models of today do more than just perform. They provide interpretations. By Wednesday morning, they evaluate sentiment from a Reddit thread, read earnings transcripts, and determine whether geopolitical noise from a Tuesday headline warrants a change in stance. Ninety-one percent of asset managers either currently use AI or intend to do so, according to Mercer’s 2024 survey. You can determine the direction of the wind just by looking at that number.
| Snapshot — Algorithmic Investing in 2026 | |
|---|---|
| Topic | The shift from human-led to algorithmic and AI-driven investment management |
| Origin | Roots in 1960s NYSE data processing; modern phase began with the 1976 DOT system |
| Robo-Advisors Launched | Around 2010 |
| Estimated Assets Under Robo Management | Up to $16 trillion projected by mid-decade (Deloitte estimates) |
| Notable Public Test | Alpha Arena 1.5 (Nov–Dec 2025), where seven major LLMs traded live equities; only Grok-4.20 finished positive at +12.11% |
| Key Risk Event | The May 6, 2010 “flash crash” — Dow fell nearly 1,000 points in minutes |
| Adoption in Asset Management | 91% of asset managers either use or plan to use AI (Mercer, 2024) |
| Regulatory Bodies | SEC and FINRA in the U.S.; MiFID II in the EU |
| Core Debate | Whether human judgment can — or should — be replaced by autonomous models |
Determining whether any of this is genuinely effective is more difficult. For more than ten years, quant hedge funds have been promising machine-driven alpha, but the majority of them don’t disclose their performance. A 2022 study that was published in the International Journal of Data Science and Analytics examined 27 scholarly articles that asserted machine-learning models could predict markets with approximately 95% accuracy. The majority of those studies ran dozens of model variations and only reported the best one, which the researchers found to be unsettling. You can’t choose which version of yourself trades in the real world. The one you ran is yours. More than most marketing decks acknowledge, that distinction is important.
The New York Stock Exchange floor is still bustling with people wearing fleece vests and badges outside the labs, but the real decisions are increasingly being made somewhere else entirely, such as in server racks in New Jersey or in code repositories at companies that hardly ever conduct interviews. Jeff Shen of BlackRock succinctly stated that technology is an enabler rather than a replacement. Matthew Lyberg of Manulife, who is in charge of AI for the company’s wealth division, described AI as “the start of our process, not the conclusion.” People who genuinely use these tools and aren’t trying to sell them use this type of language. Both caution and belief are present. Perhaps even some weariness from the hype.
Regulators are still thinking about the May 2010 flash crash, and they probably should. Because automated systems all looked at the same signal and responded in the same way, a trillion dollars momentarily disappeared in a matter of minutes. That’s the silent peril. Models begin to think similarly when they are trained on identical data. Strangely enough, disagreement is what keeps markets honest. When you remove that, you get fragility masquerading as precision rather than efficiency.
It’s difficult to ignore how cautious the industry becomes in private and how assured it sounds in public. Retail investors now pay forty dollars a month for tools that used to cost institutions millions, robo-advisors are managing trillions, and the SEC is constantly updating its rulebooks to keep up. It’s still unclear if algorithms will control or only influence investing in the future. The fact that the humans aren’t quite leaving the room seems more certain. They are merely standing a short distance away from the screen, observing the machines in action and hoping they have sufficiently trained them.
