Phase 29 · the frozen law

Protection, or just slower?

AT-0223 → AT-0230. A compact state-dependent coupling law, g(r) = g0/(1 + α·|r − target|), is discovered once on a training ensemble, locked, and carried untouched into hundreds of unseen toy systems. Then we try to take it apart: dominance, matched progress, matched identity, functional-form necessity, gating triviality and closed-cycle debt.

Core objective, unchanged

The target is unchanged: substantial genuine relational-location change while preserving identity and invariants, under locality, bounded propagation, finite action, no hidden debt, robustness, scale persistence and nontriviality. A mechanism that buys identity preservation by moving less has not advanced that target.

Plain language

We found a small rule that dials a system's internal coupling down while it is still far from where it is going, and back up as it arrives. We locked its three numbers and carried it, untouched, into hundreds of systems it had never seen. It halved the disturbance to identity — and it almost never travelled as far. That is a trade, not a discovery. This page exists to make the trade impossible to hide.

Controls

SeedTraining and unseen ensembles are drawn disjointly from it.Training systems — 120Parameters are selected here, once, then frozen.Unseen systems — 500No retuning is permitted on any of them.Actuator effort — 0.375Shared by both mechanisms except in matched comparisons.

Running the frozen-law package…