Status: MVP proof pipeline implemented for internal fixtures, with generated Lean/artifact emission and CI gates. Owner context: Grand Challenge Labs / LAXFORGE. Purpose: make LAXFORGE a gauge-aware, proof-producing discovery engine for Lax pairs, zero-curvature representations, and operator identities.
Core doctrine:
LAXFORGE proposes. LAXCERT certifies. AETHER records. MODULUS interprets.
LAXFORGE remains the search and symbolic-discovery engine. LAXCERT is the Lean4 companion that receives structured certificates and machine-checks their algebraic claims.
This pack contains:
SPEC.md— main product and research specification.docs/ARCHITECTURE.md— component diagram and data flow.docs/LEAN_API.md— Lean module contracts and theorem surface.docs/PROOF_OBLIGATIONS.md— exact proof goals and acceptance tests.docs/EMITTER_CONTRACT.md— Python/SymPy-to-Lean certificate interface.docs/AETHER_LEDGER.md— proof provenance and governance records.docs/ROADMAP.md— build phases and milestones.docs/SPEC_TRACEABILITY.md— ordered section 13 gap register and closeout evidence.docs/CONSOLE.md— local visual console for candidates, artifacts, job queue, and operator commands.schemas/laxcert_candidate.schema.json— JSON schema for emitted candidate certificates.python/laxforge_laxcert/emitter_contract.py— typed Python-side contract sketch.lean/— Lean project skeleton and illustrative module boundaries.ci/GITHUB_ACTIONS.md— CI gates and failure modes.
Phase 1 implementation status:
- field names
u,p, andq; - jet variables as
(field, order); - scalar expression AST with rational constants, jets, sums, products, negation, and powers;
- spatial derivation
Dx, including jet shifting and product rule; - lightweight scalar simplification for generated proof targets;
- toy theorem certifying
D_x(pq) = p_x q + p q_x.
Initial phase 2 implementation status:
- bounded differential operator coefficient structure with zero/add/neg/sub instances;
- coefficient extensionality theorem;
- first-order operator constructors for multiplication by a scalar and
D_x; - first-order Leibniz/Ore composition into second-order normal form;
- toy theorem certifying
D_x ∘ p = p_x + pD_x; - wrong-sign rejection theorem for the same toy identity.
MVP pipeline status:
- exact Lean toolchain pinned to
leanprover/lean4:v4.29.1; - mathlib dependency declared at tag
v4.29.1; - Python package baseline via
pyproject.toml; - strict candidate schema validation;
- deterministic Python normalization and Lean emission;
- generated artifact folders under
artifacts/<candidate_id>/; - internal positive toy candidate, internal diagonal and off-diagonal 2x2 candidates, and false-candidate failure gate;
- section-10 transport calibration for
L = [[D_x^2, -q], [p, -D_x^2]], including a committed LAXFORGE-emitted artifact; - generated Lean coefficient certificates for toy and 2x2 fixtures, replacing vacuous
Truetheorems; - GitHub Actions workflow for Lean, schema/emitter, generated positives, false negative, and pytest.
Local verification:
cd lean
lake -R build LaxCert
cd ..
pytest
python -m laxforge_laxcert.emitter_contract candidates/toy_lax_zero.json --repo-root .
python -m laxforge_laxcert.emitter_contract candidates/matrix_2x2_zero.json --repo-root .
python -m laxforge_laxcert.emitter_contract candidates/matrix_2x2_offdiag_zero.json --repo-root .
python -m laxforge_laxcert.emitter_contract candidates/akns_d2_transport_zero.json --repo-root .
python -m laxforge_laxcert.emitter_contract artifacts/LaxforgeAKNSD2TransportZero/candidate.json --repo-root .
python -m laxforge_laxcert.emitter_contract candidates/false_wrong_sign.json --repo-root . --expect-failureLocal visual console:
python -m laxforge_laxcert.console_server --repo-root .Open http://127.0.0.1:8765 to inspect candidates, proof artifacts, the command queue, Lake/Pytest runs, and certifier results.
For Lean/mathlib dependency refreshes, prefer WSL native storage rather than
building from /mnt/<drive>. A reliable workflow is:
rsync -a --exclude=.lake --exclude=artifacts --exclude=.pytest_cache /mnt/f/_codex/LAXCERT/ ~/LAXCERT-wsl/
cd ~/LAXCERT-wsl/lean
lake update
lake -R build LaxCertThe MVP targets the current LAXFORGE scope: one spatial direction x; scalar fields u or paired fields p,q; formal differential operators written as powers of D(x); self-adjoint L; skew-adjoint P; and certification of identities such as
[ L_t = [P,L] ]
or, for matrix connections,
[ U_t - V_x + [U,V] = 0. ]