← Exceptional Geometry home

Protected parity, exact local decoding, and self-reconstructing W33

The missing tenth bit is coefficient parity

The 240 protected edge vectors span the unique 9D invariant layer V9. The quotient is now explicit:

π(A*b) = Σᵢ bᵢ mod 2 ker π = V9 = A*(even coefficient vectors)

Every vector in ker A* has even coefficient parity, so π is well defined. Any single line-star A*eᵢ supplies the missing tenth direction.

π=0: 1 + 135 z^16 + 240 z^20 + 135 z^24 + z^40 π=1: 40 z^12 + 432 z^20 + 40 z^28

The 240 even weight-20 vectors are exactly the protected edge shell; the 40 odd minimum vectors are the 40 line-stars.

An exact 10-sample local decoder

One local center line-star plus nine spokes from the 13-line Borel cell form a basis of all H10. Sampling ambient coordinates 0,1,2,3,4,5,7,8,10,11 gives an invertible 10×10 binary restriction. Its frozen inverse reconstructs all 40 ambient protected bits from those ten samples and was exhausted over all 1024 protected vectors.

The first decoded coordinate is π; the remaining nine are local spoke coordinates. The separate eight-state Borel quotient is spectrally complete but intentionally forgets identity within each orbit, so it is not this full decoder.

The Borel reaches the global splitting-order ceiling

All five maximal subgroup types remain nonsplit. Standard subgroup structure reduces every remaining proper subgroup class above order 162 to representatives of orders 360,324,288,216,216,192,192; exact section systems for all seven are inconsistent. Hence no subgroup of order greater than 162 splits the apartment extension, while the order-162 chamber/Borel does.

Q(5,q): a new q=7 anchor without overclaiming

An independent elliptic-quadric build of Q^-(5,7)=GQ(7,49) gives 2,752 points and 17,200 lines with exact binary ranks rank N=2451 and rank(NᵀN)=2150. Together with q=3, both anchors match

ρ(q) = (q²+1)(q²-q+1) = |L|/(q+1)

but the all-q rank formula remains explicitly conjectural. The exact general odd-q theorem proved here is A*²=0 over F₂.

Zeta data recovers the protected filtration

For W33, primitive-six degree-two Walsh data obeys M6=48(J-I)+204A*. Therefore it determines A*, and hence H10=im(A*), the 9D even-coefficient layer, and the parity quotient. The same cross-check explains why the prism→edge collapse is special to t=3: for odd t a 3-rung fan equals its (t-2)-rung complement modulo the line-graph kernel.

Three outside-the-box consequences

Fixed-vector duality. The parity functional is π(x)=B(x,1), pairing with the unique PSp-fixed protected vector, so V9=1⊥.

Self-reconstruction. Join two of the 40 odd minimum vectors when their XOR has weight 20. The resulting graph is exactly dual W33, SRG(40,12,2,4); adjacent differences reproduce the 240 protected edges.

108 local bases. In the 12 neighbors of the center line, omit three. If the omitted triple is a triangle, edge+isolated, or independent, the remaining nine columns have ranks 7, 8, or 9 respectively, with counts 4, 108, 108. Adding the center to the 108 independent cases yields exactly 108 full local H10 bases.

Boundary. These are finite GF(2), finite-group, graph-zeta and representable-matroid statements. The closed Q(5,q) rank formula is still a conjecture beyond exact q=3 and q=7 anchors. No physical charge, spacetime locality, hardware decoder threshold, or particle identification is inferred.