Runnable research utilities and concrete report
Ten offline utilities support the catalog. They distinguish static metadata, exact finite arithmetic, floating-point evaluation and documentary linting. None independently verifies the repository's proofs or establishes buyer demand. Python's standard library is sufficient; saved outputs use exclusive creation and need fresh filenames or directories.
| Utility | What it currently solves | Evidence and limit |
|---|---|---|
| Proof preflight | Locates checker configurations, modules and trust settings | Static scan of 416 configurations; no Lean execution |
| Withdrawal impact | Tracks explicit proof withdrawals and named dependencies | Three notices and two explicit edges; incomplete as a general theorem graph |
| Embedding audit | Measures finite all-pairs Lp distortion | Floating point; no full-space or future-query certificate |
| GAD capacity objective | Evaluates the explicit scalar channel objective and candidate signals | Floating point with grid allowance; no certified rounding bound, finite decoder or device-fit guarantee |
| Claim contract audit | Exposes source, scope, model and obligation gaps | Nine curated contracts; supplied evidence contents remain unverified |
| Periodic interface reference | Calculates the stated cubic-periodic perimeter profile | Floating-point formula; excludes additional material energies and different cell models |
| Binary waveform audit | Separates aperiodic peak, periodic correlation, merit factor and sampled spectrum | Exact integer finite correlations; spectrum sampled; no new signing constructor |
| Ramanujan graph audit | Checks a strict nonconstant spectral bound for a small regular graph | Exact rational arithmetic; 64-vertex cap; no graph constructor or formal program proof |
| Cyclic-chain reference | Computes a shortest normal-step chain with cyclic quotients above an embedded subgroup | Exact validated group table; order at most 16 and bounded work; no spectrum witness or theorem check |
| Evidence bundle | Joins source hashes, static trust metadata, withdrawals and a claim audit into a reviewable report | Fingerprints 2,933 source files; hashes identify local bytes, not proof validity or publisher authentication |
The concrete evidence-platform report is available now, with a source-hash manifest. It inventories 372 families, 719 manuscripts and 242 selected scope documents, records two external-kernel configurations and exposes two documentary gaps in the finite-electrode example. It does not constitute a completed ten-claim expert benchmark or paid pilot.
Repeatable terminal examples
Run from /Users/bradleymutemi/Documents/ultimate. These commands print results without replacing saved files.
python3 research/tools/gad_capacity.py --gamma 0.55 --nu 0.3
python3 research/tools/audit_claim_contract.py research/fixtures/finite-electrode-claim.json
python3 research/tools/periodic_interface_reference.py --volume-fraction 0.2 --side-length 1
python3 research/tools/audit_binary_waveform.py research/fixtures/barker13.json
python3 research/tools/audit_ramanujan_graph.py research/fixtures/petersen.json
python3 research/tools/cyclic_chain_reference.py research/fixtures/c4-times-c2-first-subgroup.json
The evidence bundle requires a fresh --output-directory. The embedding audit requires both --input and a fresh --output. Withdrawal scanning requires --source source/openai-math and a fresh --output. Proof preflight also requires --inventory research/snapshots/2026-10-08-baseline/inventory.json. Use --help for exact interfaces. Dated batch/validation scripts preserve completed research records; do not rerun them into an existing edition.
Saved component validation
- GAD checks: 33 analytic/direct-matrix checks and 1,500 sampled arbitrary ensembles. Random search is not an optimality proof.
- Original claim contracts: ten checks of scope drift, finite measurements, missing evidence, unknown obligations and documentary completeness without certification.
- Added claim contracts: 13 checks of normalization quantifiers, added reductions, Artin complexity/scope, ideal quantum gates and embedded-subgroup distinctions.
- Geometry and waveforms: 21 checks of transitions, scaling, complementary volume and known finite correlations.
- Exact graph audit: eight known-spectrum, malformed-input and work-budget checks.
- Cyclic-chain reference: 21 checks of cyclic and elementary-abelian groups, the paper's different embeddings, noncentral dihedral subgroups, relabeled identity, associativity and invalid inputs.
Those six component reports record 106 passed checks. Earlier static preflight, withdrawal and embedding fixtures are additional separate records. These checks cover finite components and documentary behavior, not an independently reproduced theorem, operational benefit or commercial demand.
Next implementation work
Enrich source-location records for the nine contracts, prepare the ten-claim expert comparison, and add a trusted proof-run adapter only in a suitable environment. For direct algorithm products, extract pseudocode and practical constants before extending the finite reference utilities into general constructors.