Things that were supposed to be impossible
A comprehensive encyclopedia in three hundred languages, written and maintained by unpaid strangers, was not the expected outcome of anything. The expected outcome was a mess. It was assumed that reference work at that scale required paid editorial authority, and it turned out not to.
The protocols the internet runs on were not legislated. They were proposed, argued over, and adopted because they worked, by people with no power to compel anyone to use them and no way to charge for them.
After large disasters, informal networks — neighbors with boats, volunteer kitchens, ad hoc logistics run out of spreadsheets — routinely reach people before the official response does, because they are not waiting for authorization to act.
Open-source software runs most of the world's infrastructure and was built by people who mostly were not paid to build it. Mutual aid societies provided medical care and unemployment insurance to millions of working people before either was a government function.
Each of these was, at some point, the sort of thing that obviously needed compulsory funding or central control. Each got solved by people who chose to show up.
Why it works, structurally
This is not a claim that people are nicer than assumed. There are mechanical reasons voluntary systems tend to outperform coercive ones, and they have nothing to do with virtue.
Feedback is immediate and it bites. If a voluntary school is bad, families leave; if a charity wastes money, donors stop giving. Coercive systems break that loop on purpose — funding arrives regardless of results, which means the signal that something is failing has no route back to the thing that could fix it.
Exit does work that voice cannot. Leaving is a fast, cheap, individual act that requires convincing nobody. Changing a policy requires assembling a majority, which is slow, expensive, and usually fails.
Adaptation is parallel rather than serial. A hundred groups trying a hundred approaches find a good one faster than one body deciding, because none of them has to be right the first time.
And costs land closer to whoever creates them, which is the entire difference between a decision someone has to live with and a decision someone gets to make.
The honest limits
Voluntary alternatives to most government functions are currently smaller and worse funded than their compulsory counterparts. That is true and it should be said first, not buried.
Part of the reason is crowding out — people already taxed for a service are markedly less inclined to fund a second version of it voluntarily, which is a real effect and not an excuse invented for the occasion. But part of it is simply that these things are smaller, and pointing at a mutual aid network and a national health system as though they were comparable would be silly.
The question is not whether voluntary approaches are as large today. It is whether they could grow into the space if the compulsory version were not occupying it.
That question does not have a proven answer. It has evidence pointing in a direction, and a fair amount of history where the compulsory version arrived after the voluntary one was already working and then displaced it.
What has not been tried
Notice what the examples above have in common. Everyone participating chose to. Every contribution was given rather than collected. Every provider has to keep earning the support they get. And any of it can be improved, forked, or abandoned without winning a political fight first.
The philosophy's claim is directional rather than triumphant: cooperation trends toward flourishing, and coercion trends toward conflict, stagnation, and the slow erosion of the assumption that other people are worth dealing with.
Which leaves the more interesting question. If voluntary cooperation built the encyclopedia, the network, and the disaster response — what else might it be capable of that nobody has attempted, because it was assumed from the start that force was the only way?