Follow a result back to its evidence.

SEAM.BIT.TWISTBLIND.01: the twist-state attack on the alignment bit DECIDED -- 'mark-decorated states do not see delta', an honest, decided NEGATIVE result (v508 style): the NINTH side-blind test on the v512 scoreboard (8 -> 9); the free-plus-twist class (everything Wick-computable) is EXHAUSTED (20 checks, ~80 s). [E] ALL LIFTINGS WELL-DEFINED AND BLIND: (a) sigma-gauge arc decoration (Z2): sigma o r = sigma on the swap axes makes the twisted Gram a diagonal unitary congruence D M D -- spectra IDENTICAL to untwisted (dev 0.0); (b) Kadanoff-Ceva 4-twist insertion omega(D .)/omega(D): well-defined on N = 32 (omega(D) real nonzero; the N = 16 odd-m degeneration is the C(even) = 0 checkerboard artifact); the RP SPECTRA depend on delta for the first time in the program (pairwise up to 1.05), but the INERTIA is (16, 0, 0) PD with the SAME eta = +i for every m -- the decoration sees delta, positivity does not; (c) the mu4 twist at beta = pi/2 is pure plane gauge; (c') genuine Bogoliubov defect beta = pi/4: Theta-compatibility selects exactly 2 of 16 sign patterns (NS-wrap forced), the winner genuinely non-monomial -- still PD, spectra identical. STRUCTURE THEOREM (empirical, mechanism exact): EVERY Theta-compatible quasi-free mark decoration is RP-spectrally INVISIBLE (the defect planes never straddle a cut bond; orthogonal congruence). [E] SUB-RESULTS: the defect energy E_def = 1.2752872 for EVERY m (spread 4e-40) -- delta-blind; the 4-twist Casimir ln|omega(D)| measures delta (-0.890 .. -1.348) but is exactly mirror-symmetric under m <-> 8 - m -- a family gauge, never a side selector; Kramers <Gamma>_tw = +1 for every m. [E] TWO BIT-PRESUPPOSING pi/2 STRUCTURES (the harvest): the eta flip +i -> -i on the mark-fixing axes (sigma o r = -sigma, odd sector exactly negated); the twist frustration of the mark-fixing mirror ((4, 4, 0) indefinite for both eta -- the lattice shadow of the Ising twist field's mirror-oddness) -- both formulable ONLY on the fixing axes existing iff delta = pi/2: they PRESUPPOSE the bit (v512 facet class), no selector. [E] NEGATIVE CONTROLS: the twist-free limit = v519/v521 exactly; the site cut keeps failing in every decorated state; the v512 counterwitness is arc-equivalent at N = 16 to the blind member pi/4 -- no state-level selector can exclude it; a single twist has omega(D) = 0 identically (twists exist only in mark pairs); the deck-paired FOUR-twist decoration is the minimal well-defined one. [O] the door narrows to a genuinely INTERACTING A_hol whose OS data are NOT Pfaffian-reducible to the chiral vacuum -- beyond every Wick-computable class. VERDICT: KILL exactly as preregistered; the alignment bit remains genuine discrete input. Discovery provenance: experiments/tfpt-discovery/seam_alignment_bit_twist_state_probe.py (20/20, 2026-07-23). NO marker moves.

Linked claims

SEAM.BIT.TWISTBLIND.01

Paper sections

Source section 1Source section 2Source section 3Source section 4Source section 5Source section 6Source section 7Source section 8Source section 9Source section 10Source section 11Source section 12Source section 13Source section 14Source section 15Source section 16Source section 17Source section 18Source section 19Source section 20Source section 21Source section 22Source section 23Source section 24Source section 25Source section 26Source section 27Source section 28