The Lindy Effect

Why "it's survived this long" is sometimes good news and sometimes means nothing at all — through two papers, twelve years apart.

The Lindy effect says: the longer something has already survived (a book, a technology, a habit), the longer it's likely to keep surviving — if it isn't decaying. It's a statistical rule about expected remaining lifetime for a population, not a promise about any one thing.

The two papers below attack the same underlying question — why does risk-of-dying change with age the way it does? — from opposite directions: one is pure abstract probability, the other is biology. The chart is a sketch of the three shapes both papers keep bumping into.

Lindy — no drift: risk keeps dropping the longer it lasts
"Sick" / negative drift: risk flattens to a constant, age stops helping
Ordinary biological aging: risk climbs, then flattens very late

Illustrative sketch of the shapes each paper describes — not fitted data.

Paper 1 — Taleb, "Lindy as Distance from an Absorbing Barrier"

Nassim Nicholas Taleb · Wilmott magazine, 2026

Like you're 5: imagine a bug walking near the edge of a table. The longer it's been walking without falling off, the farther it probably is from the edge — so the safer it probably is. That's the Lindy magic trick: surviving longer makes you seem safer. But if the bug is also slowly drifting toward the edge as it walks (getting tired, aging), that trick breaks — it stops getting safer just for having survived, and instead falls off at some steady, boring rate no matter how long it's already lasted.

Taleb models "survival" as a randomly wandering process that gets absorbed the moment it hits a barrier (death/failure). With no drift — pure random wandering, no built-in decay — the math (first-passage time of a Brownian motion) works out to exactly the Lindy shape: risk keeps declining the longer you've lasted.

Add even a small negative drift — a steady push toward the barrier on top of the randomness — and the Lindy property breaks completely. The paper's key result (Remark 2): any amount of negative drift knocks survival out of the power-law/Lindy class, and risk instead settles toward a constant rate.

What does "negative drift" actually mean?

Picture noisy up-and-down jiggling (pure luck) versus that same jiggling with a steady current layered on top, always pushing one direction. No drift means every step is just luck — and luck doesn't remember the past, so lasting a long time really does mean you've wandered safely far from danger. Negative drift means part of your risk isn't luck at all — it's a fixed countdown ticking in the background, regardless of how far away you currently are. That countdown doesn't care how long you've already survived, which is exactly why it ruins the "age = safety" logic.

Taleb applying this to a real question

Tommy Griffith (@TommyGriffith) — Jan 6, 2019
@nntaleb How does the Lindy effect apply to crappy entrenched industries? For example, yellowcabs have been around for ~50 years and Uber ~5 years, but long term I would bet on Uber. Is this just an exception?
Nassim Nicholas Taleb (@nntaleb) — 3:48 AM · Jan 6, 2019
No, Lindy is statistical. The trick is to condition on lack of sickness, just as with insurance life expectancy tables.

"Conditioning on lack of sickness" is Remark 2 in plain English. An industry being actively disrupted isn't a healthy 50-year-old — it's the "sick" one, carrying a negative drift toward failure. Its age stops being reassuring, because part of its risk is no longer random luck; it's a steady current pulling it toward the edge.

Paper 2 — Podolskiy et al., "Critical dynamics of gene networks is a mechanism behind ageing and Gompertz law"

D. Podolskiy et al. · arXiv:1502.04307, 2015

Like you're 5: think of your body as a huge team of light switches (genes) that are supposed to flip on and off in the right pattern. Over time a few switches get stuck or flip at the wrong moment, and because they're all wired together, one mistake nudges others to misfire too — like dominoes. That makes the whole switchboard wobblier as you get older, which is why the risk of dying climbs faster and faster with age. But once you're really old, so many switches are already messed up that a few more mistakes barely change anything — the wobbliness stops getting worse, and the risk of dying levels off.

The paper's mechanism is gene regulatory networks behaving like a nearly-critical, unstable dynamical system: damage accumulates, destabilizes gene expression, and mortality rises exponentially with age — the classic Gompertz law. Its distinctive prediction is the "aging" curve in the chart above: exponential rise, then a flattening at very old ages, matching the well-documented late-life mortality deceleration seen in large population data.

How the two connect

They're not connected by citation — Podolskiy et al. (2015) predates Taleb's note by over a decade, and neither references the other. They're connected by target: both are explaining why real mortality curves flatten out at extreme old age instead of doing one clean thing forever.

Podolskiy et al. get there mechanistically, bottom-up from gene-network dynamics. Taleb gets to a structurally similar flat line abstractly, top-down from drift alone — no biology required. One paper supplies a candidate cause for the plateau; the other proves that some plateau is close to unavoidable the moment any decay term enters a survival process at all.