Mutual Aid Against the Machine
Kropotkin argued cooperation is an evolutionary advantage, not a moral one. The hard part is what happens at scale.
Kropotkin’s argument is not that cooperation is good. It is that cooperation is adaptive.
He spent years observing animals in Siberia and came back with a case that mutual aid is an evolutionary advantage, not a moral one. The strongest species are not the most ferocious. They are the ones whose members help each other often enough that the group outlasts the group next door. The claim is biological, which means it can be argued with. Evolution does not follow slogans.
This matters because it changes what mutual aid is. It stops being a virtue that good people practice and becomes a working principle that any system can be tested against.
Five Branches, One Refusal
I read through the branches of anarchism looking for a spine. Individualism after Stirner, collectivism after Bakunin, anarcho-communism after Kropotkin, syndicalism built on the union, ecology built on the small community. They disagree about property and about the unit of organization.
What they share is a refusal: no coercive hierarchy, no state, no chain of command that cannot be questioned by the people inside it.
The branches disagree about how to abolish authority. Kropotkin’s contribution was to make the case biological rather than moral. The claim that follows is sharper than it looks: voluntary cooperation can outperform coerced hierarchy. Not in every case, not at every scale, but under conditions you can name. When people can see what each other are doing. When the cost of helping is low. When defection is visible and expensive.
Ants do this without a king. Guilds did this without a ministry. Wikipedia does this now. The question is not whether people can cooperate without a boss. They obviously can at small scale. The question is what breaks when the scale grows.
The Two-Layer Organization
The argument I keep making about organizations started in a conversation about flat teams.
Most organizations, including the ones run by people who talk about being flat, carry three layers of instruction: a decision layer, a middle layer, and a layer that executes. The middle layer translates, relays, supervises, and corrects.
Strip out the sentiment and the middle layer exists for one reason. The people above do not trust the people below. Not out of malice. Because understanding drifts on the way down. A decision that is clear at the top arrives at the bottom as three sentences, and the three sentences are wrong.
If information could move directly, the middle layer would be correcting a problem that no longer exists. Two layers, decision and execution, with the full reasoning visible to everyone downstream. Not a slogan about empowerment. A claim about information loss.
That is the part of the argument that holds up best. Middle management is largely an error-correction mechanism, and when the error goes away the correction becomes overhead.
The Strongest Objection
Here the case gets weak, and I would rather say so.
Kropotkin was good at small groups. A village, a guild, a colony of animals. He was not good at explaining a city of ten million people. The scale objection stands: what prevents voluntary cooperation from fragmenting into factions? Who resolves the conflict when two communities both think they are right? Who decides when a decision affects people who were not consulted?
Hierarchy answers those questions cheaply. Direct participation does not, and pretending it does is where models of the past went wrong. I do not know of a large-scale, long-running anarchist society. Kropotkin’s own experiments were short and small.
The transition problem is worse. Every existing society is a hierarchy. Getting from a hierarchy to a two-layer organization is not a reform. It is a demolition, and someone has to be in charge of the demolition, and that someone holds power at the exact moment you need them not to.
The digital layer gives some cover. Individual sovereignty becomes partly a question of the data and computation a person runs on. Communication that goes directly, without a relay, is first a network property and only later a social one. That is the honest version: mutual aid is easier to run on a mesh than on a state.
But the mesh is run by people with an incentive to centralize, and the lower cost of coordination is also the lower cost of capture. The person who controls the network becomes the state.
What I Keep
Cooperation does not need a state or a market to happen. It needs conditions that must be named. The two-layer idea is a real insight about information and trust. The scale problem is unsolved.
I do not have a clean argument that dissolves it, and I do not think it should be dissolved. It should be left standing as the thing any honest case has to answer.