Phase 28 · attack the scoreboard
Is it the rule, or is it the score?
AT-0215 → AT-0222. The current toy leader is put through thousands of randomised scoreboards, compressed to its smallest honest form, frozen and carried into unseen systems with no retuning, ablated against damping and generic-feedback controls, and pushed until mandatory coupling breaks it.
Core objective, unchanged
Lawful substantial genuine change in gauge-invariant / operational relational location, without destructive traversal, while preserving identity and invariants, respecting locality and bounded propagation, carrying no hidden debt, spending finite action, remaining robust, persisting across scale, and staying nontrivial.
Plain language
One idea keeps coming out on top. The obvious worry is that it is winning because of the scoreboard we invented rather than because of the rule it uses. So we redraw the scoreboard thousands of times at random, and we count how often each idea still wins. Then we do the harder thing: we shrink the winning rule to its smallest form, lock its numbers, and test it on systems it has never seen — different sizes, shapes, targets and definitions of identity — with no adjusting allowed. If it only works when we are allowed to retune it, or only under one definition of success, we say so and downgrade it.
Controls
Redrawing the scoreboard…
Where this sits
This page inherits every earlier constraint. See AT-0207 → AT-0214 for the matched-budget tournament that produced this candidate, where the path breaks for the gauge-invariance audit every progress number must respect, and the blocker ledger for the standing obstacles behind each requirement.