The Weasel Game
Biology is not fun. So far, we have had some major trouble with monkeys. Now it's the weasels' turn.
There's a computer program that's been doing the rounds for decades, showing up in biology classrooms, popular articles, and countless online reenactments, always making the same simple, striking point: cumulative selection is fast. Blazingly fast. Something that would take longer than the age of the universe by pure chance takes less than a minute once you let small, retained improvements build on each other. It's called the Weasel program, and it comes from Richard Dawkins' 1986 book The Blind Watchmaker.
Here's the setup, as Dawkins describes it. Take a monkey with a typewriter — twenty-six capital letters and a space bar, nothing else — and ask how long it would take, hammering away at random, to type a specific 28-character sentence: "METHINKS IT IS LIKE A WEASEL." The odds against hitting that exact sequence in one attempt are about one in 1040. Even at millions of attempts a second, you'd wait far longer than the universe has existed. (The arithmetic: The Weasel Odds.) Then he changes the rules. Instead of a single random shot, the computer starts with one random 28-character string, makes many mutated copies of it, and keeps whichever copy most resembles the target phrase. Repeat that a few dozen times, and the target gets hit. In the three runs Dawkins reported, it took 43, 64, and 41 "generations." The first run took about half an hour, in BASIC; rewritten in Pascal, the program took 11 seconds.
Nobody has ever actually seen Dawkins' original program. It doesn't appear in his book — only a description of what it did. And when pressed, years later, Dawkins himself confirmed he no longer had it. In 2009, when two candidate reconstructions were sent to him for authentication, his reply was direct: "I cannot confirm that either of them is mine. They don't look familiar to me, but it is a long time ago. I don't see what more I can say." Every version of Weasel that exists today — the online versions, the classroom demonstrations, all of it — is somebody's reconstruction or reinterpretation of a program nobody can produce.
There is nothing about this program that is not true. It works exactly as advertised, and it does prove the point it set out to make: if small changes are held tightly by the way the program works, then yes, the target will be reached relatively quickly. Not necessarily in the forty-odd steps of Dawkins' account, but certainly far more quickly than trying to land on the entire string in one shot. A set of rules that all but forces progress toward a known, fixed target is guaranteed, by the very shape of those rules, to reach the target eventually. That's not a discovery about how search works; it's just what the rules do. The only problem is that the story doesn't match reality. Real evolution has no fixed target to check against. So the game masks problems evolution faces in the real world, and it skirts real numbers that have to be reckoned with. It looks easy. When it comes to actual evolution, it is not easy at all.
Last revised 2026-11-19