How evolution actually works

Evolution needs three things, and only three. Get all of them in one place and it happens whether or not anyone intended it, in living cells, in software, in anything that copies itself imperfectly. This is what those three things are, what they produce, and the parts almost everyone gets slightly wrong.

The three requirements

Individuals must differ from one another. Some of those differences must be passed to offspring. And those differences must affect how many offspring there are. Variation, inheritance, and differential reproduction. That is the whole mechanism.

Notice what is not on the list. No goal. No standard of improvement. No mechanism by which an organism can respond to a need by acquiring a useful trait. The variation arrives first and blindly, and only afterwards does the environment sort it.

The four things almost everyone gets wrong

Evolution is not about the survival of the fittest. That phrase was Herbert Spencer's, not Darwin's, and it is close to circular: the fittest are defined as those that survive. What actually matters is reproduction, not survival. An organism that lives a very long time and leaves no offspring contributes nothing. One that dies young having reproduced heavily contributes a great deal.

Individuals do not evolve. Populations do. You cannot watch an organism evolve, because nothing about it changes in that way during its life. What changes is the composition of a population across generations, which is why evolution is invisible on any timescale a single observer naturally uses.

Nothing is trying to become anything. There is no ladder, no direction, no progression toward complexity. Parasites routinely evolve to be simpler than their ancestors because simplicity pays when someone else is doing the work. Loss is as common an outcome as gain.

Fitness is not a property an organism has. It is a relationship between an organism and a particular environment at a particular moment. Change the environment and the same trait can go from advantage to liability without anything about the organism changing at all.

What selection does to a population

When a trait varies continuously, and something about the environment favours some values over others, the population's distribution moves. There are three characteristic ways it can move, and recognising them is most of what reading evolutionary data consists of.

Stabilising selection deserves the emphasis. When a population appears unchanged for a very long time, the usual reading is that evolution has paused. Often the opposite is true: selection is actively removing the extremes every generation, and the apparent stillness is the result of continuous pressure rather than the absence of it.

Selection is not the only thing happening

A common oversimplification is that every trait an organism has is there because it was selected for. Plenty are not.

Genetic drift, the random change in how common a variant is simply because of who happened to reproduce, operates constantly and is strongest in small populations. Many changes are neutral, affecting nothing that matters, and they accumulate anyway. Some traits are side effects of others that were selected, structurally impossible to separate. And some are historical accidents that became too embedded to reverse.

The eye's blind spot is the standard example: vertebrate retinas are wired with the nerve fibres in front of the light-sensitive cells, which is worse than the alternative and is not fixable by incremental change from where things stand. Evolution has no access to redesign. It can only modify what is already there, one viable step at a time, and every step has to be an organism that works.

Watching all three requirements in one place

The reason this project exists is that the three requirements are simple enough to assemble deliberately and then step back from.

Organisms here vary: nineteen genes plus a couple of thousand network weights, all differing between individuals. They inherit: offspring receive a mutated copy of their parent's genes and weights, and nothing else, with acquired changes discarded at death. And reproduction differs, because energy is finite and an organism that cannot cover its costs never reaches the threshold.

There is no fitness function anywhere in it. Nothing scores an organism or ranks a population. Selection is not applied to this world, it is what the accounting leaves behind, which is explained properly in nothing in this world wants anything. The mutation model and the exact genes are on genome, inheritance and mutation.