How mutations actually work

Mutation is the only source of genuinely new variation there is. Everything selection works with had to arrive this way first, as a copying error. It is also, in almost every individual case, either irrelevant or bad, which makes its role as the engine of novelty faintly absurd and worth understanding properly.

The kinds, and what each does

Not all copying errors are the same size, and the differences matter enormously for what they can produce.

Almost all of them do nothing, and the rest are mostly bad

The distribution of effects is heavily lopsided, and getting it right is the difference between understanding evolution and imagining it as a series of lucky improvements.

The large majority of mutations are neutral. They land somewhere that does not affect anything that matters, or they change something in a way nothing notices. These accumulate steadily and are the basis of molecular clocks, which date divergences by counting them.

Of those that do something, most are harmful. This is not bad luck, it is what you should expect from random changes to a working system. A random edit to functioning machinery is far more likely to break it than improve it, and the better the machinery already works, the truer that gets.

Beneficial mutations are rare. Not absent, and not so rare as to be negligible over the timescales involved, but rare enough that evolution is best pictured as a very large number of trials overwhelmingly wasted, with selection retaining the small remainder. There is no mechanism producing useful variation on demand. The usefulness is decided afterwards.

Where new things actually come from

This raises a real problem. If most changes to a working gene break it, how does anything genuinely new ever arise? A gene doing an essential job cannot be experimented on, because the intermediate stages would be broken and their carriers would die.

Gene duplication is the answer, and it is elegant. When a gene is accidentally copied, the organism has two functioning versions where one was needed. One keeps doing the original job. The other is free, and mutations to it are no longer immediately fatal, because the job is still covered.

Most spare copies simply decay into nothing. Occasionally one drifts into doing something else useful. Whole families of related genes, including the proteins that carry oxygen and the ones that detect light, arose this way: duplicated, then diverged. Almost all major evolutionary novelty traces back to a copy that was temporarily allowed to be useless.

The rate is itself under selection

Mutation rates are not fixed constants. They are the outcome of how much a lineage invests in proofreading and repair, and that investment is heritable and can evolve.

Too high a rate and offspring accumulate damage faster than selection can clear it, and the lineage degrades. Too low and there is no raw material, and the population cannot track a changing environment. Observed rates cluster in a narrow band, and organisms in unstable conditions tend toward the higher end. Evolution has, in effect, tuned its own tuning parameter.

A three-outcome model, chosen deliberately

This world has no DNA, so it does not have substitutions or frame shifts. What it has is a mutation model applied to inherited numbers, and it was written to reproduce the shape of the distribution above rather than its biochemistry.

Network weights mutate under a four-tier scheme: most connections are untouched in any given reproduction, some receive a tiny adjustment, fewer receive a moderate one, and a small fraction receive a large one. The physical genes use a comparable graded model. The effect is the same lopsided distribution biology has, with small changes common and large ones rare, and the exact probabilities and step sizes are on genome, inheritance and mutation.

Two absences are worth stating plainly. There is no duplication here, and no mechanism by which the genome or the network could grow. The number of genes and the shape of the network are fixed for the life of a universe, which means the single most important source of genuine novelty in biology is unavailable, and the space of possible organisms is bounded however long the world runs. That is the central limitation discussed on open-ended evolution, and it traces directly to this.