[1337isnot1337]

The new PR4694 center interface observes 106 arbitrary helper inputs. This constructs a physically legal helper basis and inverse reducing those observations to eight individual helper fields, with 379 scalar gates at entry, 379 at exit and no new Gaussian call. All original helper Gaussian jobs and the complete paid profile are preserved.

The actual eight-helper minor has determinant 64 and only dyadic-unit elimination pivots. Every affected helper has the same empty birth chart and complete nine-axis output chart, so the basis commutes with the identical exact Gaussian at every positive column multiplicity. Exact ranks rank(K)=120, rank(K²)=8, K³=0 give eight size-three, 104 size-two and 408 size-one scalar Jordan blocks. Eight actual chains are explicit: pivot helper → center → Y scatter column. A general physical Jordan basis is not assumed free.

A conditional two-call consumer is also constructed. For a homogeneous three-field offer U(Q)=I+Q(E12+E23), with symmetric Q, Q⁴=I and monomial P=Q², exact scalar/Pauli boundaries satisfy L U E D U^(-T) D R = diag(Q,Q,I). The actual raw dual includes its quadratic term, which is corrected explicitly. CONSUMER.md gives the all-width block proof and complete conditional moment with a shared transverse multiplier; consumer.py checks every coefficient in Z[q]/(q⁴−1) and five failure controls.

No higher K or independent-center credit. The real retained-center charts have not been embedded into that homogeneous offer at an affordable cost. Each normalized center output still contains its 56-term data observation, and all 112 Y outputs still receive incoming-center noise. A physical consumer with independently recovered Gaussian jobs and a positive complete moment is required before applying the conditional formula to the native invoice. The source's direct-cache and scalar-span failures remain in force.

Only research/center-helper-normal-form/ changes against canonical main. Frozen PR4694 ABI, native payload and credited replay engine are actual inputs; their exact history is retained by a local tree-preserving provenance merge. The upstream PR is unchanged. PR4694's cache continuation already noted that the observed helpers are cold; that observation is credited. The new contributions here are the emitted physical basis, preserved helper jobs, explicit chains and conditional two-call identity. No priority claim is made for elementary block elimination.

Validation on final scientific head 9079ab6: CPython 3.12.15 registered/native and isolated optimized replays; all 640 primal/raw-dual columns; basis/operator inverses; actual charts and paid ranks; positive scalar envelopes; native and conditional-controller failure controls; committed-byte/source-history preflight; 14-file export/archive checks; and fresh Git-free optimized export replay with PATH=/nonexistent. No new physical phase audit, selected cold replay, compiler/outer theorem or independent human review is claimed.

Source PR4694 frozen head 039a3e4; refreshed head 2add42d has identical ABI and native payload. Main remains b58ad9e, selected conditional K=.00302015. Live queue and cached overlap were checked; sparse local coverage was incomplete.

Credit Cinna Davis's PR4694/4440/4441/4447/4470 lineage, Cao Yifan/PR4610, zxv, Jacob Alan Sussman and retained scalar, E8, Clifford, precision, compiler and outer contributors. Prepared for Cinna Davis with substantial OpenAI Codex assistance; no subagents; Apache-2.0.