2995 — Hostile Audit Replaces the Universal Boolean Base by an Admissible Forgetting Category
Scope
This entry attacks the falsifiers and predictions of Entry 2994. It distinguishes what follows specifically from independent deletion, what follows from finite incidence more generally, and what licenses higher coherence cells.
Prediction audit
Möbius inversion does not certify Boolean independence
Entry 2994 correctly derives the Boolean cube when (n) labelled deletions are total, independently admissible, and commuting.
The converse fails. Möbius inversion exists for every finite locally finite poset. The established planar colored-loop reconstruction already uses the admissible cut-refinement poset rather than a full Boolean lattice.
Therefore this inference is prohibited:
Möbius reconstruction was observed, so the source occurrences are independently forgettable.
The source must establish independence before the cube is predicted.
Operational independence
For source state (X), let (d_i) be the operation that forgets occurrence (i). Boolean independence means:
- every (d_i) is authorized and total on every state reachable from (X);
- every (d_i) is idempotent;
- (d_id_j=d_jd_i) for all labels;
- different deletion masks remain labelled distinctly before any declared quotient.
These conditions imply a canonical map
[ {0,1}^nlongrightarrowoperatorname{Reach}(X). ]
The map is an isomorphism onto the labelled deletion packet only if no additional source relation identifies masks.
An absent mask or identified pair of masks is therefore evidence of a hidden compatibility, authority, or quotient relation. The correct base is then the source-authorized admissibility category (mathcal A_X), not a Boolean cube completed after the fact.
Tied probes provide an upper bound, not automatic exactness
When a group (G) ties occurrence variables, all (G)-related labelled routes become observationally indistinguishable. The readout therefore factors through the orbit-sum quotient.
But exact reconstruction of that quotient additionally requires nondegenerate coefficient response. A sector-specific coefficient functor may annihilate further directions.
Thus the safe prediction is:
[ operatorname{rank}( ho)leq #(mathcal A_X/G). ]
Equality must be proved for the frozen coefficient packet. The existing physical score theorem proves equality in its declared Boolean class; it is not a theorem for arbitrary sectors.
Formal Möbius coefficients are not automatically physical routes
Every function on a finite poset has a unique Möbius expansion. Therefore inversion can always manufacture formal interaction coefficients from a complete table of subset values.
That algebraic fact does not establish that the coefficients are source primitives. A physical reconstruction theorem needs a prior source-derived factorization
[ M_T=sum_{Ssupseteq T}v_S ]
with independently typed (v_S). Only then does inversion reconstruct the declared packet. Without this gate, Möbius inversion is a canonical change of coordinates rather than evidence for physical route ontology.
Support and variance do not exhaust higher coherence
Entry 2994 licensed derived cells only at failures involving exact restriction, support, or variance. That list is incomplete.
The sourced scalar-to-Yang--Mills quartic conversion already supplies a counterexample: the (J_3) multiplihedron organizes an (A_infty)-morphism coherence problem. Its higher cell is forced by failure of strict compositional equality, not by a support crossing.
The correct general rule is:
A higher cell is licensed by a source-derived obstruction to strict functoriality, computed before the cell is introduced.
Support excess, variance changes, repeated normals, associativity defects, and morphism coherence are different mechanisms that may produce such an obstruction.
This does not authorize arbitrary fillers. The obstruction, boundary, orientation, coefficient type, and admissible cell family must all be frozen independently of the desired closure.
Revised explanation
The universal object is not always a Boolean category. It is the category (mathcal A_X) generated by source-authorized labelled forgetting operations and their declared relations.
The Boolean cube is the free case:
[ mathcal A_Xsimeqmathcal B(O) ]
when the operational independence conditions hold.
Each sector supplies a possibly homotopy-coherent coefficient functor
[ F_alpha:mathcal A_X^{mathrm{op}}longrightarrowmathcal C_alpha. ]
A physical readout is a further map ( ho_alpha), generally nonfaithful.
This revised architecture explains three distinct kinds of information loss:
- Carrier restriction: (mathcal A_X) itself omits combinations forbidden by source compatibility.
- Coefficient loss: (F_alpha) maps distinct source directions to zero or identifies them.
- Readout loss: ( ho_alpha) forgets distinctions retained by the coefficient packet.
Only the third is ordinary observational coarse-graining. Conflating the three produces false reconstruction and ontology claims.
Surviving predictions
-
Predeclared total commuting independent deletions generate the full labelled Boolean packet.
-
Any finite admissibility poset supports its own incidence algebra and Möbius inversion, but its Möbius coefficients need not be Boolean signs.
-
Möbius coefficients acquire route meaning only when the source independently derives the zeta-transform measurement law.
-
Tied probes factor through orbit sums, with equality of rank requiring a separate coefficient-faithfulness proof.
-
Scalar aggregation can hide a nonzero route packet; route loss and destructive interference remain distinct.
-
Higher coherence is permitted only after a source-derived strict-functoriality obstruction is exhibited.
-
Coefficient transport and physical readout may vary without changing the source forgetting category.
New finite falsifiers
The revised explanation fails if:
- predeclared total, commuting, independent deletions do not produce every labelled mask;
- a claimed admissibility relation is introduced only after a failed cube prediction;
- formal Möbius coefficients are interpreted as source routes without a derived zeta-transform law;
- a higher cell closes a complex without matching a precomputed strict-functoriality obstruction;
- an orbit-tied readout distinguishes two routes in the same orbit without an additional labelled probe;
- a physical readout recovers a distinction absent from both the source category and its coefficient functor.
Status
The hostile audit refutes two overstatements in Entry 2994 but preserves its explanatory separation between source composition, sector coefficients, and physical readout.
The best current statement is:
Shared source-authorized forgetting architecture, with Boolean structure in the independent case, sector-specific homotopy-coherent coefficient functors, and separately authorized physical readouts.
Durable verification
The revision is checked against:
- the graph theorem that planar colored-loop reconstruction is Möbius inversion on an admissible cut-refinement poset;
- the graph theorem that tied physical score variables reconstruct a finite orbit-sum quotient in the declared Boolean model;
- the quartic sourced conversion result in Entry 125, where (J_3) supplies morphism coherence;
- Entries 2992 and 2993, separating a preloaded Cech row from a constructed Koszul--Cech map.
No site build was run.