[CubeJerry]

Result

Bind every actual selected raw positive PAID request to a complete current
coded direct-leaf construction. For source rank r in 1..8 and native width f,
a literal 39-port metadata program creates a LEAF-specific descriptor
with f'=r*f, preserving the original cutoff and all other 22 values. The
positive Gaussian provider acts on exactly the r source axes in every column;
all spectators, coefficient fields, rows and the original parent are retained.

The actual direct-branch condition is r*f<=cutoff. The complete child uses
388 fixed private ports and costs its full O(r*f*V+(E+1)^4) numerical bill
plus at most 1000E+10000 metadata instructions, with cleanup and historical
backing charged. It invokes no ordinary supplier. A provider-relative whole
MOVE wrapper reuses the published protected stage and retains both monomial
contracts and their ordinary premises. No new kappa; the full multiplication
bound remains conditional.

Why this is new

The published current positive leaf supplied arbitrary width but bound only
the initial rank-one source event. The published protected MOVE joined that
one event. This package materializes the missing rank-times-column header
and cutoff interface for all 3,453 actual selected PAIDs, then certifies all
2,886 source chart factorizations including absolute phase. The census itself
is not claimed as new. The first-helper native header package changes three
different fields and preserves f; its caller placement and nested token
binding are excluded here.

Canonical base is 7c49fe32c55de02c2c873655c804b16e4c3fef0a.
Exact immediate dependencies are
protected MOVE,
positive leaf,
current monomial, and
literal native header pattern.
All 128 retained predecessor files have exact public paths, Git blobs and
SHA-256 hashes. Frozen packages are unchanged.

Verification

python3 -B scripts/research_package.py --package research/current-coded-all-source-leaf

Six bounded normal/optimized gates recompute all source certificates, run
the literal metadata and interface controls, and re-execute complete
4096-record rank-two success/cutoff/parity traces. The independent Clifford
checker uses 18 Pauli images and all 512 entries of one normalized column per
chart; the proved commutant lemma fixes every matrix column and absolute
phase. Phase-only, input-Z, wrong-rank and typed source mutants reject.

Complete rank-two and rank-eight physical child traces returned respectively
481530645 and 1501302941 primitive actions; rank eight completed in 255.269 s
under a 580 s cap. Its standalone replay command and saved receipt are in the
package; the registered receipt check does not rerun rank eight. The cutoff
and late-parity controls preserve the entire field/header and retire all work.
Forty physical metadata cases and long-then-short history checks preserve
complete bytes, ancestors and origins. Separate AI lead checks of the proof,
33296 derivative cases and 32 exact C conjugations are explicitly distinguished
from human or kernel verification.

Limitations and open obligations

This is the direct leaf domain only. The changed header is never supplied to
a monomial or general native consumer. Complete normalized source placement,
signed-prefix/grid admission, inherited numeric source-to-machine contracts,
and both ordinary-facing monomial contracts remain premises. No full generic
three-factor MOVE trace, enclosing nested/product caller, nonleaf rate or new
complete multiplication bound is claimed. Rolewise materialized tape extents
are distinguished from actual visited backing; all historical cohorts remain
in the source space invoice. Whole-MOVE port counts depend on the actual
ordinary factory.

Next questions are a verified supplier-free monomial join with intermediate
phase headroom, actual enclosing caller/header binding, literal numerical
lifecycle export, and the remaining selected framework/nonleaf obligations.

Credit and assistance

Prepared for @CubeJerry. Credit the source work in
#3087, #3091,
#6293, #6901,
#6930 and selected
#6030, plus the exact published dependencies above.
Substantial OpenAI assistance supported research, proofs, implementation,
tests and documentation. All attribution and source-specific notices remain;
this is not presented as independent human or formal-kernel admission.