CONCEPT
Fermi Estimation
The discipline of decomposing an unanswerable question into a chain of estimable subquestions, assembling their answers so that independent errors cancel rather than compound, and holding confidence in strict proportion to the agreement between independent lines of reasoning—the cognitive practice the AI age most urgently demands.
A Fermi estimate is an admission of ignorance organized into a usable answer. When Enrico Fermi asked how many piano tuners worked in Chicago, he did not reach for a directory; he reasoned from population to household size to piano-ownership rates to tuning frequency to the working capacity of a single tuner, and multiplied the chain. The genius is structural: decompose the unknown into the merely difficult, and the independent errors in each factor tend to cancel rather than compound. The method is not a shortcut to precision—it is a discipline for making decisions before precision is available, which is to say, almost always.
Large language models can produce the form of a Fermi estimate with striking fluency, reproducing the pattern of a chain of factors and a confident final number—and can do so while being deeply miscalibrated, because the model is completing a pattern rather than interrogating a model of