Edward Lorenz was a meteorologist at MIT in the early 1960s, running primitive weather simulations on a desk-sized computer. One winter morning he wanted to revisit a result. To save time, he restarted the simulation from a midpoint, typing in numbers from a printout instead of starting over.

What he expected: an identical replay. What he got: a completely different weather pattern within hours of simulated time.

The reason was buried in the computer's memory. The internal calculation kept six decimal places. The printout truncated to three. Lorenz had typed 0.506 instead of 0.506127. A difference of 0.000127.

That tiny rounding error grew exponentially. By the third simulated day, the weather had drifted noticeably. By the fifth, it was unrecognizable. A microscopic input difference produced a completely different output.

Lorenz had discovered what we now call sensitive dependence on initial conditions — the defining feature of a chaotic system. The atmosphere is one. So is a swinging double pendulum, a dripping faucet, the orbit of three gravitational bodies, the human heart at the edge of arrhythmia.

In a 1972 talk, Lorenz titled his presentation 'Predictability: Does the flap of a butterfly's wings in Brazil set off a tornado in Texas?' The metaphor stuck.

The deepest implication isn't that small things matter. It's that long-term weather prediction is fundamentally impossible. To forecast accurately ten days out, you would need to measure the temperature, pressure, and humidity at every cubic meter of the atmosphere. A butterfly's wings genuinely contribute. So does every breath you take.

Weather forecasts beyond about two weeks aren't unreliable. They're impossible.