What the AI never decides

It never sets an award

The Monkey and Llama awards are published equations over recorded acts — a weighted average you can reproduce with a calculator. There is no model anywhere in that computation. Which acts exist is a sourcing question; what they are worth is a published table.

It never approves a record

Nothing about a living person is published because a model produced it. Every path to the public record runs through review, and review is a human with a name and a stated reason.

It never awards a point

Karma is earned by logged, evidenced, corroborated work and verified by people. The AI pass reads the month and suggests; it does not score.

It never decides what you should think

The product's job is to make your reasoning visible and checkable, including to you. It records your positions, shows you where a chain is thin, and applies your values. It does not have a position.

It never fills a gap with a guess

This is the boundary that matters most, because it is the one most tempting to cross. Where the record is missing, the answer is this is missing — named, published, and offered as work somebody could do. A plausible-sounding interpolation is worse than a gap, because a gap is honest and an interpolation is indistinguishable from evidence.

Why the boundary is drawn here

The movement's central complaint is that citizens are told to defer to an authority that will not show its work. Replacing that authority with a model that will not show its work would be the same failure wearing a different coat — and everybody would be right to say so.

So: the AI reads, checks, and proposes. Humans decide, and the arithmetic is arithmetic.