Is happiness the goal, or just a side effect?

One book per mind

Each mind is a tiny neural network — about 10 million parameters, small enough to run on a laptop. We train it from scratch on one body of text. No internet, no help.

It points, doesn't invent

When asked a question, the mind finds passages from its book that seem most related. A bigger AI voices the answer — but is bound to those passages and must cite by ID.

Sometimes it reframes the question

If a question doesn't fit the mind's tradition — when its framing assumes things the book never says — the mind pushes back instead of pretending. It reframes the question from inside its own worldview. You'll see a violet reframedbadge when this happens. It's the most interesting kind of answer.

If their book is silent, they say so

No invention. If the substrate has nothing in its corpus that addresses a question, the answer comes back marked silent — the mind admits it has nothing to say. Honesty over coverage.