SURFACE / ORDEREXACT FINITE NODES

PROFILE / MASK / COHERENT MEAN · PRINCIPAL ν · S ABOVE 2,3

One common profile.
Every surviving column.

A norm determines the actual coefficient and unit phase, while several ideals can share that norm. Inspect how one complex smooth profile aligns them all, and why a sixth-power factor still deletes raw columns.

WD changes with D. Its sup norm is 1; its derivatives cost Ok(Dk). These finite exact records concern a varying-profile class. The fixed-W target remains open.

Loading the hash-bound norm groups and exact raw masks…

NEW / ONE FIXED COMMON W / ACTUAL A₆=1 ROWS

Keep the diagonal. Keep the mean.

Twenty nonprincipal rows give a positive Gram matrix. Keeping only coprime column pairs removes its diagonal and leaves a signed, indefinite slice. Explore the exact arithmetic behind its negative contribution.

One selected q, two column ideals. D=60000, U=244, H=3833; A₁=37, A₂=7, other ideal factors1 and literal unit1. This is a complete row block for that class, not the whole column family or h=1 sum.

Loading the saved forty raw character values…

WHAT THE CONSTRUCTION DOES AND DOES NOT ESTABLISH

A varying profile is a different quantifier.

The exact norm-group constancy lets WD align the original AU coefficients, rather than replace them by arbitrary weights for individual ideals. The selected rows εβv6 have the same nonprincipal order6 primitive character, so their actual principal residue is 0. Raw collision zeros are retained.

A separate classical coprime-ideal count gives a cDH1/6 lower bound in this profile-varying class. It matches the DH1/6 term when 0<θ<2/3 and H=Dθ. At θ=3/4 its exponent is 9/8, below the current 7/6 upper envelope; it does not establish7/6 sharpness.

The profiles have ||WD||∞=1 but derivatives Ok(Dk). A bound uniform in this class with Mk(W)α(HD)εH for every ε>0 needs kα≥1−5θ/6; at θ=3/4 this is 3/8. That is a necessary cost, not a sufficient estimate.

The fixed-W target, remaining A₆=1 witness sectors, signed all-row cancellation and arithmetic determinant/zero-location bridge remain open. Positivity on the selected h=1 block does not imply positivity of h=1 over other row classes. No new upper-bound improvement, training, Lean theorem, quantum advantage or GRH proof is claimed.