AI, Algorithms, and Systematic Investing
When the machines trade — opportunity and existential risk
"There is a fundamental difference between the way a mathematician looks at randomness and the way a computer scientist does. The computer scientist asks: 'What problem am I actually solving?'" — Steven Skiena, The Algorithm Design Manual
We are living through the most significant transformation in market structure since the introduction of electronic trading. Algorithmic systems now execute the majority of trades on major exchanges. Large language models summarize earnings calls, generate trade ideas, and analyze sentiment across millions of social media posts in real time. Institutional investors are not asking whether to use AI — 82% already are. They are asking how much to trust it.
For the anti-fragile investor, this transformation is both an immense opportunity and a potential source of new Black Swans. Understanding the algorithmic landscape is no longer optional. It is the water you swim in.
The Scale of the Machine
These numbers describe a ghost economy — a vast volume of trading activity generated not by human deliberation but by algorithms, high-frequency trading systems, and quantitative strategies. When you place a market order for 100 shares of an S&P 500 stock, the counterparty is almost certainly a machine. The price you receive was set by a machine. The spread was calculated by a machine.
Morgan Stanley declared in March 2026 that "AI is now a macro variable" — meaning artificial intelligence is no longer just a sector theme but a force reshaping economic growth forecasts, productivity assumptions, and market structure itself. This is a regime change, not a trend.
How AI Reshapes Markets
AI is not a single force. It operates across multiple dimensions of market activity, each with distinct implications for investors.
LLMs for Decision Support
Large language models are increasingly used not to execute trades but to frame them. They summarize 10-K filings in seconds, generate investment theses from raw data, and identify patterns in earnings call language that correlate with future performance. The value is not in the model's "opinion" but in its ability to process information at a scale and speed that no human can match.
Real-Time Sentiment Analysis
AI systems now mine Twitter, Reddit, news feeds, and even satellite imagery for signals. A spike in negative sentiment about a company on social media can trigger algorithmic selling before any fundamental news has been released. The market increasingly prices in vibes before it prices in facts.
Anomaly Detection
Machine learning models excel at identifying statistical anomalies — prices that deviate from expected patterns, unusual options activity, or market microstructure shifts that precede large moves. These tools can help identify bias-driven market excesses, where behavioral factors have pushed prices away from fundamentals.
The CFA Institute warned in February 2026 about attention bias in AI-driven investing: LLMs are systematically skewed toward high-attention stocks — the companies most frequently mentioned in training data. This means AI-assisted investors may be unknowingly herding into the same concentrated positions, creating fragility disguised as sophistication.
The Illusion of Passive Dominance
There is a widespread narrative that "passive investing has won." Index funds now hold a larger share of U.S. equity markets than active managers. But this narrative obscures a critical reality: for every $1 in passive trading, approximately $22 is traded actively. The price discovery mechanism of markets — the process by which prices come to reflect fair value — is still overwhelmingly driven by active participants, many of them algorithmic.
This matters because it means passive investors are free-riding on a price discovery mechanism that is increasingly dominated by machines. If the machines are right, passive investors benefit. If the machines are systematically wrong — if they are all running similar models with similar biases — passive investors are exposed to risks they cannot see.
The Right Problem vs. the Right Answer
Steven Skiena's insight from The Algorithm Design Manual is profoundly relevant to investing: good algorithms model the right problem. The most elegant solution to the wrong problem is worthless.
Modeling the Wrong Risk
Most quantitative models in finance assume returns follow a Gaussian (normal) distribution. This makes the math tractable and the models elegant. But as we explored in earlier chapters, financial returns are fat-tailed. They live in Extremistan.
An algorithm that perfectly models Gaussian risk is worse than a crude algorithm that models fat-tailed risk. The first gives you false precision — beautiful Value-at-Risk calculations that catastrophically underestimate tail events. The second gives you rough but honest estimates that account for the possibility of 10-sigma moves.
This is Taleb's deepest critique of quantitative finance: the models are solving the wrong problem with extraordinary precision.
Overfitting: The Curse of Big Data
Modern machine learning systems have access to more market data than ever before. But Taleb warns that "more data does not necessarily improve estimates" in fat-tailed domains. In fact, it can make them worse.
Overfitting — also called curve-fitting — occurs when a model learns the noise in historical data rather than the underlying signal. A model that perfectly predicts the past 20 years of market data may be capturing patterns that are artifacts of that specific period rather than durable features of markets. When the regime changes (and it always does), the overfitted model fails catastrophically.
0DTE Options and Algorithmic Volatility
One of the most striking developments in market structure is the explosion of zero-days-to-expiration (0DTE) options — options contracts that expire the same day they are traded. These instruments, combined with algorithmic trading, are creating intraday volatility spikes that would have been unthinkable a decade ago.
A large 0DTE options position triggers delta-hedging by market makers. The hedging activity moves the underlying stock or index. The movement triggers more hedging. A feedback loop emerges, amplified by algorithms that react in microseconds. The result: markets can swing 2-3% intraday on no fundamental news whatsoever, driven entirely by options market plumbing.
For the anti-fragile investor, this is both noise to be ignored and opportunity to be exploited. Intraday volatility spikes create mispricings that disappear within hours. They are not investable for most retail investors, but they are a reminder that the market's short-term behavior is increasingly disconnected from fundamental value.
AI as Opportunity
Systematic Investing: Rules Over Emotions
The strongest case for algorithmic approaches is not that machines are smarter than humans. It is that machines are more consistent. A rules-based systematic strategy removes the emotional biases that destroy returns: panic-selling at bottoms, FOMO-buying at tops, and overtrading out of boredom or anxiety.
Robo-Advisors: Democratized Discipline
Platforms like Betterment, Wealthfront, and Schwab Intelligent Portfolios offer automated portfolio management at fees ranging from 0% to 0.25% of assets under management. Research shows that robo-advisors significantly moderate most cognitive biases in their users — reducing panic-selling, improving diversification, and enforcing tax-loss harvesting.
Robo-advisors moderate most cognitive biases except overconfidence. Investors using automated platforms still tend to overestimate their own market knowledge and frequently override the algorithm's recommendations. If you use a robo-advisor, commit to not overriding it. The whole point is to remove your emotional brain from the process.
AI as Black Swan Risk
The same technology that creates opportunity also creates new categories of systemic risk.
The AI Flash Crash Scenario
BCA Research has identified AI-induced flash crashes as a top risk for 2026. The mechanism: multiple AI trading systems, trained on similar data and running similar models, reach similar conclusions simultaneously. When they all try to sell at the same time, liquidity evaporates instantly.
POLITICO has warned that generalist AI agents — not purpose-built trading algorithms but general-purpose AI systems acting on behalf of investors — could interact adversarially in unpredictable ways, triggering cascade failures that no individual agent intended.
Taleb himself weighed in on this risk in February 2026, warning that AI's impact on technology stocks will eventually lead to bankruptcies. His argument: the market is pricing AI companies as if they will all succeed, but the technology follows power-law dynamics where a few winners capture most of the value while the majority fail. The AI hype cycle is creating fragile concentration in precisely the wrong assets.
As AI tools proliferate, individual trading edges are getting harder to find and maintain. Strategies that once generated alpha are being arbitraged away in months rather than years. The Reddit investing community has noted this compression extensively: what used to be a sustainable informational edge becomes worthless once it is embedded in widely available AI tools.
The Anti-Fragile Approach to AI in Markets
How should the anti-fragile investor relate to this new landscape? Several principles emerge:
Use AI for information processing, not for decisions. Let language models summarize, organize, and surface patterns. But make the final investment decision yourself, applying the anti-fragile framework. The machine is good at answering "what does the data say?" You are better at asking "is this the right question?"
Be skeptical of AI-generated consensus. If every AI tool is recommending the same stocks, that is a crowding signal, not a buy signal. Anti-fragility means going where the machines are not looking.
Treat AI concentration as fragility. The market's massive bet on AI infrastructure — data centers, chip makers, cloud providers — is the opposite of anti-fragile. It is a concentrated wager that a single technology trend will play out exactly as consensus expects. History suggests it will not.
Understand that you are trading against machines. In short-term trading, you are almost certainly the inferior player. Algorithms are faster, more disciplined, and better informed on a millisecond-by-millisecond basis. Your edge as a human investor lies in longer time horizons, patience, and the ability to think about risks that machines cannot model — precisely the fat-tailed, Extremistan risks that break quantitative models.
"The question is not whether AI will transform investing. It already has. The question is whether you will be the beneficiary or the victim of that transformation. The answer depends entirely on your time horizon and your humility." — Morgan Stanley Research, March 2026
The ghost economy of algorithmic trading is not going away. It is accelerating. The anti-fragile investor does not fight this reality or ignore it. Instead, they position themselves to benefit from the inevitable mistakes that machines will make — mistakes born of model risk, overfitting, crowding, and the fundamental impossibility of reducing Extremistan to a set of parameters. The machines are extraordinarily good at solving the problems they were designed to solve. The Black Swans will come from the problems they were not designed to solve.