A rounding error in a toy weather model accidentally discovered chaos
In 1961, meteorologist Edward Lorenz reran a simulation using a printed value of 0.506 instead of the computer's stored 0.506127, expecting a near-identical repeat. Instead the tiny difference snowballed into a completely different simulated weather pattern within two months. He published the result in 1963; the 'butterfly' name came later, from a 1972 talk asking whether a butterfly's wings in Brazil could set off a tornado in Texas.