The phrase “what has never happened before happens all the time”, originally coined by economic historian Charles Kindleberger and later popularized by Nassim Nicholas Taleb in The Black Swan,captures a fundamental blind spot in human risk assessment: confusing the absence of historical evidence with the impossibility of future occurrence.
It breaks down into three core mechanics:
The Illusion of Induction (The Turkey Problem)
Humans naturally rely on past data to predict future risk. If a system operates without failure for 1,000 days, standard statistical models conclude the system is safe and that Day 1,001 will behave the same way.
Taleb illustrates this with a turkey fed every day by a butcher. Each feeding increases the turkey’s statistical confidence that the butcher cares for its well-being—reaching maximum confidence on the Wednesday before Thanksgiving, right before its worldview is catastrophically invalidated. In complex systems, stability does not mean safety; it often just means the underlying fragility hasn’t been triggered yet.
Local Novelty vs. Systemic Frequency
For any single company, market, or piece of infrastructure, a catastrophic failure, a “Black Swan”, is unprecedented. Because it has never happened to that specific system, operators treat its probability as zero.
However, when you zoom out across the entire globe and across history, unprecedented events occur constantly. A specific supply chain route might never have been blocked, a specific server architecture might never have suffered a cascading failure, or a specific financial asset might never have lost 90% of its value overnight; yet somewhere in the world, an “impossible” tail-risk event is happening every week.
The Failure of Normal Distribution Models*
Traditional risk management relies heavily on the Gaussian bell curve, where extreme outliers (3+ standard deviations from the mean) are treated as statistically negligible.
Real-world systems; financial markets, geopolitics, software ecosystems, and pandemics; operate in “Extremistan,” a domain governed by power laws rather than bell curves. In Extremistan, a single outlier event doesn’t just slightly nudge the average; it completely dominates the outcome and wipes out decades of incremental gains.
Because you cannot predict which unprecedented event will happen next, relying on historical data to prove a system is immune to failure is a logical fallacy. Resilience requires shifting focus from predicting specific threats (which relies on the past) to mitigating exposure to catastrophic failure (assuming the unprecedented will happen, regardless of track record).