Does Dawkins' Weasel Program Prove That Genetic Information Can Increase?

What the Weasel Program Does

Dawkins begins with a question: if a monkey typed randomly, what are the odds of producing the Shakespeare phrase "METHINKS IT IS LIKE A WEASEL" (28 characters)? Using an alphabet of 27 characters (26 letters plus a space), the probability of getting all 28 positions correct in a single attempt is roughly 1 in 27²⁸ — approximately 1 in 10⁴⁰. He calls this single-step selection, and rightly notes that natural selection does not work this way.

His alternative is cumulative selection, demonstrated by the Weasel program. The algorithm works as follows:

Dawkins reports reaching the target phrase in just 43 generations. Below is the actual output.

The simulation is mathematically correct. The question is whether it models actual biological evolution. The following five points examine this.

Problem 01 Natural Selection Is Absent — Intermediate Forms Cannot Survive

Evolution operates through two mechanisms: random mutation and natural selection. The Weasel program simulates only the first, effectively eliminating the second.

Consider Generation 1:

This string contains no meaningful information. Translated into biological terms, it represents an amino acid sequence with no functional proteins whatsoever. Such a structure cannot perform any biological function. If natural selection were actually operating, this organism would be eliminated immediately — before reproducing.

Yet the Weasel program uses this string as the parent of the next generation and carries it forward. It assumes all intermediate stages survive to reproduce, without any selection pressure removing non-functional forms.

If the "completed WEASEL" organisms were eventually eliminated by natural selection, why did the non-functional intermediates survive long enough to produce them? The Weasel program offers no answer to this.

Problem 02 The Junk DNA Argument Has Collapsed

Look at Generation 30:

Meaningful fragments ("METH", "IT", "LIKE") sit alongside meaningless ones ("INGS", "ISW", "WECSEL"). If this analogy maps to biology, then every organism at an intermediate stage of evolution should contain vast quantities of non-functional DNA sequences.

Dawkins was, for many years, a prominent advocate of junk DNA — the idea that most of the genome serves no function and is simply evolutionary debris. This was a natural prediction of the Weasel model: you would expect large amounts of biological "noise" accumulating alongside functional sequences.

However, the 2012 ENCODE project found that approximately 80% of the human genome has at least one biochemical function. Subsequent research has continued to identify regulatory roles, RNA structural functions, and cell differentiation signals in sequences previously labeled as "junk." The junk DNA argument is under significant revision.

Problem 03 Where Are the Transitional Fossils?

If the Weasel analogy is valid, the fossil record should show continuous, gradual transitions between forms. Given that the program is a radical simplification of actual evolution, real species transitions would require far more intermediate stages — meaning far more transitional fossils should be discoverable.

The evolutionary trees that adorn our textbooks have data only at the tips and nodes of their branches; the rest is inference, however reasonable, not the evidence of fossils. — Stephen Jay Gould, Professor of Paleontology, Harvard University, Natural History, Vol. 86, No. 6 (1977), p. 13

Gould was not rejecting evolution. He was one of its most prominent defenders. But he openly acknowledged that the fossil record does not provide the continuous gradual evidence the standard model predicts. His proposal of punctuated equilibrium was an attempt to explain this gap.

The problem extends beyond fossils. If the Weasel logic holds, the living world should contain far more incomplete organisms than complete ones: creatures with half-formed wings, partially developed eyes, hearts not yet connected to nervous systems. We do not observe this in the natural world.

Problem 04 The Program Presupposes a Designer

Dawkins repeatedly emphasizes a central claim: evolution has no direction and no purpose. Natural selection does not "aim" to build an eye or a brain. It simply preserves what happens to work better right now.

The Weasel program contradicts this premise.

How does the program know which letters in this string are "correct" and which are not? Because the target phrase — "METHINKS IT IS LIKE A WEASEL" — was programmed in advance. Without a pre-specified goal, there is no basis for deciding which letter at which position should be preserved and which should be changed.

A process that converges on a predetermined goal is not evolution — it is optimization. Optimization requires a designer who sets the objective function. The Weasel program, paradoxically, models a goal-directed process — the very thing evolutionary theory claims to explain away.

Dawkins Asked About Genetic Information Increase

In the video below, Dawkins is asked to provide an example of a mutation that has added new genetic information — a central requirement of evolutionary theory. His response speaks for itself.

Problem 05 Mutations Are Not Random and Uniform

The Weasel program assumes mutations are applied randomly and uniformly across all positions. Real genomic mutation does not work this way.

① Mutational Bias

Substitutions between the four DNA bases (A, C, G, T) do not occur with equal probability. Transition mutations (purine↔purine or pyrimidine↔pyrimidine) are significantly more common than transversions. This bias substantially restricts the sequence space that mutation can realistically explore.

② Mutation Hot Spots

Mutations are not distributed evenly across the genome. Specific positions — "hot spots" — experience mutation at far higher rates than surrounding regions.

Applied to the Weasel analogy: the positions that can change are limited, and the changes that occur are biased toward specific substitutions. This is fundamentally different from the uniform random mutation the program assumes.

More fundamentally, this calls into question whether point mutation accumulation can account for the vast genetic differences observed between related species.

The accumulation of point mutations through mechanisms like those modeled in the Weasel program cannot plausibly account for a divergence of this magnitude.

Another Angle: The Problem of Probability

100 Coin Flips

A person can arrange 100 coins in any chosen order in a few minutes. But what if you had to achieve the same sequence by flipping coins at random?

This is for a 1/2 probability event repeated 100 times. A functional protein requires hundreds of amino acids in a precise sequence, each of the correct type at the correct position.

Can We Turn a Fly Into a Chimpanzee?

With current biotechnology, can humans synthesize a fly from scratch? No. We cannot independently create any organism even simpler than a fly.

Now suppose we start with a fly and, with the complete chimpanzee genome map in hand, attempt to engineer the fly into a chimpanzee through directed genetic editing. Could we do it? No — not with any existing technology. We know the target. We have a map. We have intelligent direction. And we still cannot do it.

If a directed, intelligent process with a complete roadmap cannot accomplish this, how should we evaluate the claim that undirected random mutation — with no target, no map, no guiding hand — transformed single-celled organisms into human beings?

DNA's Information Storage Capacity

A separate measure of biological complexity is the information density of DNA itself.

In 2001, Samsung developed the world's first 4GB DRAM chip — enough for 640 books. By 2004, an 8GB chip held ~20,000 books. By 2015, chips could store the equivalent of 2 million books. Yet a DNA strand scaled to the same physical size would hold the equivalent of roughly 20 trillion books.

The best information-storage technology humans had produced as of 2015 was over one million times less dense than DNA — which is said to have emerged through undirected mutation over hundreds of millions of years.

Human engineering has repeatedly advanced by studying and copying biological systems. This practice — biomimicry — presupposes that there is something in nature worth copying: a pre-existing design sophisticated enough to improve on what human ingenuity has produced.

Conclusion

The Weasel program is mathematically valid as a simulation. But as a model for evolutionary information generation, it fails on five counts:

What the Weasel program actually demonstrates is a goal-directed convergence process run by an intelligent agent who set the target, writes the selection rules, and runs the simulation. This is not evidence against design. If anything, it illustrates why design is a coherent inference when we encounter complex, specified information in biology.

The goal of this analysis is not to dismiss evolutionary biology wholesale, but to evaluate whether the Weasel program succeeds as an argument for undirected information increase. On the evidence examined here, the structural assumptions required for the program to work are precisely the assumptions that real evolutionary theory cannot make.