Classification: exact identities, proved implications with explicit hypotheses, certified computations, and named open inputs. No diagnostic is a proof input. This is a substantive local draft based on PR #20, not a merged or tagged version. No new window, tail metric, G2 or RH claim.
route_selection.tex is the complete new section, embedded verbatim in the
self-contained live manuscript. It includes the two new certified propositions,
fixed and moving prime-channel loss obstructions, the strength check on a
proposed bump-discrepancy estimate, adaptive concentration equivalence and
scalar countermodel, and complete CCM selection criteria and scoped obstructions.comparison_replays.tex gives the two certified statements and their proofs.ns43-route-assessment-2026-09-22-v1.html gives the versioned route assessment.provenance.json indexes the new author proofs, independent reviews and all
external NS-44/45 artifacts by size and SHA256. The v1.52 manifest covers this
directory; the verifier below checks the external index as well.build_report.json records the actual integrated build, not source validation.review_record.md records coordinated reviews and resolved findings.NS-44 freezes new exact dyadics on indices 17 through 128. Fresh Arb 320/448-bit gates prove both Weil energies positive, but q/D is below 7e-18 even and 12e-18 odd, contradicting the stipulated 1e-8D lower comparison.
NS-45 uses the first three blocks 26:50, 51:100, 101:200 of the existing lambda=5, N=25 partition. New pair witnesses yield exact comparison Rayleigh quotients 341/250 and 2093/1500, both greater than one. If positive smaller metrics exist, their block energies are positive and the norm obstruction applies; if they do not, the criterion fails its prerequisite. No full-block inertia claim is needed. Both parities have fresh 320/448-bit and signed-index gates.
The verifiers use the same audited analytic coefficient builder, with fresh empty
caches and enclosed infinite-series remainders. Independent parity/sign assemblies
are not a second analytic derivation of the kernel. Floats only proposed frozen
vectors and are not proof dependencies. Optimized Python that disables assertions
is rejected. These are new replacement certificates, not recovered v1.27/v1.28
artifacts. Both historical NS-1 recovery groups remain OPEN in evidence/MISSING.md.
The full coefficient-one capacity estimate remains open. The new channel theorem only excludes deleting fixed/sub-endpoint prime channels while retaining the specified Gaussian weight and the other signed terms. The restricted bump input is not proved and, for the exhibited probe, would already exclude off-line zeros.
The complete unrestricted adaptive operator hierarchy is equivalent to the same complement-floor target. Its compactness proof is at each fixed physical window; it gives no effective number of levels or new arithmetic estimate. Scalar set extrema fail in an exact nonarithmetic countermodel; no actual Paley–Wiener or zeta-field nonimplication is claimed. QG plus its separate packet input concerns bounded-complexity obstruction, not positivity.
CCM needs a complete relative separator/selection estimate plus entire-transform control. The absolute-residual countermodel is explicitly abstract and is not support-consistent arithmetic. The coarse rescaled limit theorem explicitly uses the full both-parity Weil form; finer scales remain open. CCM step (b) is the sole gold node and circle retains its exact open measurement-lane note.
From the repository root:
.venv/bin/python evidence/ns44_metric_replay/replay.py --verify
.venv/bin/python evidence/ns45_block_replay/replay.py --verify
python3 evidence/v152/verify_artifacts.py
python3 tools/verify_manifest.py v152
./manuscript/build.sh
Replay uses the frozen witnesses; do not overwrite them with new proposals. The review and build record state which checks were independently repeated. Earlier v1.50/v1.51 manifests remain unchanged. Existing PRs #18–20 remain open.