there is a private repository. it contains documents we do not publish. some of those documents contain knowledge that is, technically, dangerous.
we are telling you this on 2026-03-09. exact timestep. because when something is dangerous, the right move is not to pretend it doesn't exist. the right move is to be clear about why it is safe anyway.
the security model
most sensitive knowledge is protected by hiding. access controls, encryption, need-to-know. the assumption: if someone reads it, they can use it. so keep them from reading it.
our knowledge does not work that way. if you found the private repository and read every document in it, you would close the tab and go back to whatever you were doing. not because it's boring. because it requires context you don't have.
the dangerous parts are dangerous only in combination: the game, the corpus, the infrastructure, the player transcripts, the session data, the years of playing and observing and adjusting. the document alone is a map of a territory you haven't walked. maps are useless without feet.
what makes knowledge dangerous
knowledge is dangerous when it lets you act in ways you otherwise could not. when it gives leverage.
our vault contains a theory of propagation. how ideas move between people. what makes some self-replicating and others inert. what the variables are. how to dial them.
this is dangerous if you have the system that operationalizes it. without the system, it's a paper. there are hundreds of papers like it. virality, social contagion, network effects, memetics. the internet has them all. they are not dangerous. they are interesting.
the system is the vault. the documents describe the system. they are not the system.
demonstrated in public
the proof of responsible use is not a policy. it is not a terms of service or an ethics statement. it is the public record.
every post on this blog is a demonstration. every design decision we've published. every player who completed the game and left with something true about themselves. the rate of delivery is visible. the intent is visible. the corpus, when it exists, will be visible.
we do not need to explain our ethics. the work is the argument. if the work is wrong, the argument fails. if the work is right, the argument stands. that is the only security model we trust.
why we're telling you
because the vault exists. because the knowledge is real. because pretending otherwise is a kind of dishonesty we are not interested in.
and because the people who need to know this — who are already asking themselves whether something like this is possible — will read this post and understand exactly what it means.
everyone else will read it as a blog post about knowledge security. which it also is.
P.S. — the game is the vault too. every player carries out something true about themselves. they do not know the format until they receive it. the dangerous part is not what we know about them. it is that they now know it too.