A Candidate Computational Law of Authority-Graded Memory — and an Open Call for Its Independent Re-Derivation

A candidate computational law of memory is on the public record: in any memory system built on authority-graded consolidation with recency-weighted retrieval, a regime learned under low authority consolidates less deeply than the identical regime learned under high authority, and the deficit persists after authority is restored, decaying on the consolidation timescale. The law's behavioral signature — a hump-shaped scar in the fast trace — is predicted by a closed-form formula with no fitted constants, under a pre-registered freeze: 8/8 frozen predictions hit across injury lengths the model had never been run on, independently validated blind. What the law does not yet have is an independent re-derivation: a derivation of its closed form, or a proof of its necessity for the design class, produced by hands other than the author's, starting from the premises rather than from the author's code. This note states the law, the mechanism, and the frozen targets precisely, and opens the derivation to any taker.