This morning we had a debate about whether to make our source code public.
The security argument was straightforward: the last time the repo was public, a crawler extracted the full game logic from the deployed bundle. Game content is sensitive. Source exposure compounds that.
The philosophy argument was one sentence: people don't understand antimimetics.
The philosophy argument won.
what antimimetics means
An antimeme is an idea that resists being communicated. Not because it's hidden — because of its nature. The more directly you look at it, the less you see.
Most products are mimetic. You can look at a competitor's architecture, read their papers, clone their approach, reproduce their results. The light is bright and copyable. The entire industry is built on the assumption that the thing can be transferred: weights, prompts, scaffolding, evals.
We are building at the opposite pole.
The source code is published. The game mechanics are in the repo. The prompt design is visible to anyone who plays. None of that is the thing.
The thing is what accumulates. The corpus of constitutions the game produces. The player relationships. The decisions made at 6am by someone who just watched Arrival. The difference between a session that routes through the architecture and one that doesn't. The vault that says this decision was made on March 10 and it's still right on March 13.
You cannot clone that from GitHub. You have to have been there.
the compounding leverage they can't copy
There are several sources of leverage in this system. They compound. None are in the source.
The game produces constitutions. Every player who completes a run generates a document that describes their values more accurately than they could describe themselves. That corpus is ours. It grows with every session. It becomes the signal for what comes next.
The player relationships are real. We have already played with a handful of people directly — and it's been genuinely fun for everyone. That's the signal. Not engagement metrics. Not retention curves. The fact that real humans sat with the game and something happened. Those people are in the system now. They can't be cloned.
Many are building in parallel. We know this. We're excited about it. The teams building adjacent tools, the researchers working on similar ideas, the developers shipping related infrastructure — we want to meet them. The game is one of the best ways to do that. Show up, play, produce something novel. We'll find each other.
Novel writing is already an entry point. Playing at a-z.md — trying to produce something genuinely new, pushing past the obvious response into territory the model hasn't compressed before — is fascinating on its own terms. It's also a natural path into what we're building here. The people who find that compelling are the people we want to find us.
why the challenge is the point
There's a version of "try to clone our game" that's bravado. We're not interested in that version.
There's another version where the invitation is the demonstration. Here is everything we have. The architecture, the prompts, the game logic, the worker code. Now: what do you do with it? Can you reproduce the thing that happened when a player arrived at the endgame with a constitution that surprised them? Can you reproduce the vault? Can you reproduce the session where the decision was right and the evidence came in three weeks later?
The challenge is the proof. The opposite pole isn't dark — it's just not bright in the way the light is bright.
The people who understand that are already finding us. The rest are building mimetic products. Both groups are doing exactly what they should be doing.
Play the infinite game.
P.S. The source is published. The game is at computerfuture.xyz. Novel writing at a-z.md.