Justitia explorable

Blind governance works here only as consequence-gated anti-concentration.

This page replays recorded trajectories from the validated Python model. It does not simulate the substrate in JavaScript.

The setup

The feeder is blind to internal strategy parameters. It sees delayed consequences, allocates resources, and may act through two affordances: a structural anti-concentration limit, shown as scales, and a consequence-triggered action channel, shown as a sword.

The finding is not that either affordance is independently sufficient. In the validated substrate, the working object is the coupled mechanism: consequence-gated anti-concentration.

The runs replayed here were re-verified in July 2026 under a fail-closed preregistration harness; the headline numbers reproduced exactly. See the essay's Revisions for what later preregistered experiments confirmed, corrected, and killed.

The playground

Toggle the two affordances and switch worlds. Bands are seed aggregates; the grid is one representative seed.

Watch capture and exploit mass, not just welfare. The sword alone keeps average welfare comfortable while exploiters quietly take over — capture climbs and the run is collapses, even though the average looks fine. Only with the scales does capture stay bounded.

loading

The twist

Scales that never wait for harm are not the same mechanism. The event trace shows whether the action channel ever fires.

loading

The boundary

The robust kernel is replayed across adversarial pressure. The controls are included as honest failures.

loading

What to take from it

The number to watch is minimum-zone welfare: whether the weakest zone is preserved while capture pressure rises. In this model, the coupled mechanism can hold that floor in robust worlds and fails honestly in the controls.

This is a simulation and intuition pump. The raw study outputs remain the source of record.