Wisdom and philosophy
The Lindy effect
For things that do not wear out, such as ideas, books and technologies, every year they survive raises how long they can be expected to last.
Named after Lindy's delicatessen in New York. Albert Goldman's column "Lindy's Law" in The New Republic (13 June 1964) said something different: TV comedians who appeared too often used up their material and lasted less. Benoit Mandelbrot restated the law in The Fractal Geometry of Nature (1982), and Nassim Taleb named it the Lindy effect in Antifragile (2012).
A person's remaining life expectancy shrinks with age. The Lindy effect says the reverse holds for things that do not physically age. A book that has been in print for 40 years can expect about 40 more; a programming language used for 60 years is likely to be in use for decades yet. The time something has already survived is the evidence.
Surviving a long time means passing many tests: changes in fashion, new competitors, wars, economic shocks. Old things also gather dependents (readers, institutions, code built on top) that keep them alive. In statistical terms, if lifetimes follow a heavy-tailed distribution (a power law), the expected remaining life grows with current age, and Mandelbrot wrote that however long a person's past collected works, they will on average continue for an equal additional amount.
It does not apply to perishable things: people, hard drives, a particular server. It also says nothing certain about any single item; it is a statement about averages. The common misuse is reading it as "old is good" and dismissing anything new. Even the name is borrowed loosely: Goldman called his own law "more a cautionary fable than an accurate description or prognostic", and it was about comedians burning out, not ideas lasting.
Example
On 4 April 2020, with more than 362,000 New Jersey residents having filed for unemployment in two weeks and the labour department seeing a 1,600% jump in requests, Governor Phil Murphy asked for volunteers who knew COBOL (Common Business-Oriented Language). "Literally we have systems that are 40-plus years old," he said. COBOL was introduced in the early 1960s. Many newer languages had come and gone since, yet a 60-year-old language was still paying benefits in the middle of a pandemic.
Where it applies
When choosing a dependency, give real weight to how long it has been used and maintained: SQL, PostgreSQL and Unix command line tools have outlived many replacements. In AI work, techniques that have held up across several model generations, such as evaluation sets and retrieval, are a safer base than last month's agent framework. And when planning a migration, assume the old system will live longer than the plan says.
Think about it
Which tool in your stack is the newest, and what would it cost you if it were abandoned in two years?
Applied it to something real? Log it.
Test yourselfWhy does the Lindy effect apply to a programming language but not to a person?
A language does not wear out physically, so surviving longer is evidence that it is robust and its expected remaining life grows. A person's body ages, so remaining life expectancy shrinks with age.
Related models
- Via negativaImprove things by removing what does harm before adding anything new, because we usually know more about what is wrong than about what is right.
- Chesterton's fenceBefore you remove a rule, a process or a piece of code, find out why someone put it there.
- Probabilistic thinkingThink in odds rather than certainties: estimate how likely each outcome is, and judge a decision by those odds rather than by the single result that happened.
- Network effectsA product becomes more valuable to each user as more people use it, which makes leaders grow faster and makes leaving hard.
- CompoundingReturns earned on past returns: a steady rate applied to a growing base turns small repeated gains into large ones, mostly at the end.