Gate Matrix¶
| Gate | Purpose | Current status |
|---|---|---|
026-G00 |
Baseline / provenance | PASS |
026-G01 |
Architecture audit | PASS |
026-G02 |
V&V infrastructure | PASS |
026-G03 |
Corpus design | PASS |
026-G04 |
Linear / element robustness | PASS_WITH_LIMITATIONS |
026-G05 |
Modal / dynamic / harmonic | PASS_WITH_LIMITATIONS |
026-G06 |
J2 maturity extension | PASS_WITH_LIMITATIONS |
026-G07 |
Geometric nonlinear and arc-length review | PASS_WITH_LIMITATIONS |
026-G08 |
Buckling maturity extension | PASS_WITH_LIMITATIONS |
026-G09 |
Contact maturity extension | PASS_WITH_LIMITATIONS |
026-G10 |
Advanced Nonlinear / Research Audit | PASS_WITH_LIMITATIONS |
026-G11 |
Adversarial / failure / metamorphic | PASS_WITH_LIMITATIONS |
026-G12 |
Performance / scalability | PASS_WITH_LIMITATIONS |
026-G13 |
External correlation aggregation | PASS_WITH_LIMITATIONS |
026-G14 |
Full regression / architecture freeze | PASS_WITH_LIMITATIONS |
026-G15 |
Owner release review | PASS |
G00 is a refactor guard with an explicit dirty-worktree limitation, not a replacement for immutable 0.2.5 evidence. G01-G03 close only audit, infrastructure and corpus design. Capability-gate outcomes are not implied by this foundation.
026-G05 references the supplemental G05-B contract, registry and evidence, the all-family coverage record, the final-evidence pack and the Owner decision record. The controlled campaign reaches MOD 14/14, DYN 32/16 and HAR 12/12 across eight family rows; its bounded refinement policies are Owner-approved. The gate is closed as PASS_WITH_LIMITATIONS: external correlation is bounded and not element-family complete.
026-G06 is closed as PASS_WITH_LIMITATIONS by Owner decision. Small-strain J2 is qualified within the tested TET4/TET10/HEX8/HEX20 scope. The evidence is tied to execution source SHA 8bd0f2d8fdce7bf27ffc4c28e6aa26e69288fa63; closeout documentation is separate. Algorithmic tangent symmetry is not independently qualified, increment-refinement independence is demonstrated only on TET4, and finite-kinematic J2 remains research-only and outside the qualified scope.
026-G07 is closed by Owner decision as PASS_WITH_LIMITATIONS. The bounded
scope is Total-Lagrangian elasticity on TET4, with compatible external evidence
over the recorded 16 points, plus the declared TET4 Arc-Length research
benchmark with bounded branch, turning-point and restart/rollback evidence.
HEX8 complete-history behavior and ARC-002 refined-mesh comparability remain
explicitly excluded or deferred. TET10 and HEX20 remain research routes; no
general TL, Arc-Length, finite-kinematic or production-wide claim is made. The
machine-readable closeout is qualification/0_2_6/g07_owner_closeout.json.
026-G08 is closed as PASS_WITH_LIMITATIONS by the active Owner review in
qualification/0_2_6/g08_owner_final_review.json. The review preserves the
bounded first-factor/first-mode scope and records the later corrected evidence:
TET4 is QUALIFIED_BOUNDED, TET10 and HEX20 are
PASS_WITH_LIMITATIONS, and HEX8 is MORE_EVIDENCE_REQUIRED. The superseded
positive-load Euler screen is excluded from active metrics. No post-buckling,
multi-mode or general physical-validation claim is made.
026-G09 is closed as PASS_WITH_LIMITATIONS by the Owner closeout in
qualification/0_2_6/g09_owner_closeout.json. Lots 1 and 2 provide internal
bounded node-to-triangle evidence for mesh, load paths, rollback and failures.
Lot 3 adds Code_Aster unilateral open/close and TET4 structural-path
comparisons, plus a CalculiX pre-contact tie-breaker. The external comparison
is intentionally limited because Code_Aster uses an exact unilateral
constraint while QF uses penalty contact; it does not qualify the penalty law
itself, finite sliding, general surface-to-surface contact, friction, or a
universal conditioning cutoff. The candidate 1e4..1e6 interval remains
EXPERIMENTAL_ONLY, and the mesh-dependent transition warning is retained
explicitly.
026-G10 is PASS_WITH_LIMITATIONS. Lot 1 records a controlled audit of
existing nonlinear and research routes in
qualification/0_2_6/g10_research_audit_matrix.json and
0_2_6_g10_lot1.md; the adversarial review is recorded in
0_2_6_g10_owner_review_lot1.md. The selected external campaign and final
Owner classification are recorded in
0_2_6_g10_selected_external_campaign.md and
0_2_6_g10_owner_closeout.md. The closeout passed the applicable targeted
controls with no functional source change or numerical regression; full
regression was skipped by policy. Existing small-strain J2, transaction and
failure diagnostics retain their bounded ownership. Total-Lagrangian
elasticity and arc-length remain G07-owned and deferred; finite-kinematic J2
and the coupled routes remain experimental, deferred or not qualified under
their explicit classifications. G10 does not change G07, G08, G09, G11 or
G12.
026-G12 is closed by Owner decision as PASS_WITH_LIMITATIONS. The final
campaign records 18/18 deterministic finite route measurements across nine
bounded rows, and the optimized scaling evidence records full solves through
107 811 actual DOF plus an assembly-only 300k probe. The 1M probe is
RESOURCE_LIMITED after the controlled timeout and is not a success claim.
The requirement is fully satisfied within this declared measured scope; no
universal route-by-family scaling law, general HPC qualification or global
solver speedup is claimed. The unique full regression produced 1849 passed,
184 skipped and 18 historical failures with zero fix-only failures; those
release blockers remain outside G12. The Owner record is
qualification/0_2_6/g12_owner_closeout.json.
026-G11 is closed as PASS_WITH_LIMITATIONS by Owner decision. The bounded
closeout covers 20 route-native runtime cases across linear static, nonlinear
static, geometric nonlinear static, modal, linear buckling and linear static
contact. All cases have deterministic replay, finite diagnostics and explicit
failure classification; mutable retry cases preserve committed state. DIAG-005
and DIAG-008 remain bounded because the evidence does not claim exhaustive
coverage of every public failure class, structured nonlinear reason or route
combination. The 18 historical full-regression failures remain release
blockers outside G11.
026-G14 closes the final capability-coverage audit with limitations. The
registry contains 33 public capabilities and 44 public element-analysis
mappings, with no unregistered implementation, duplicate capability or
unbounded public maturity claim found. The audit is recorded in
qualification/0_2_6/g14_capability_coverage.json and
qualification/0_2_6/g14_owner_closeout.json. Full regression was
SKIPPED_BY_POLICY for G14; the subsequent G15 release sweep reused the
applicable 1887 passed / 184 skipped / 0 failed evidence and completed the
release review. The G14 cleanup findings remain historical records.
026-G15 is closed as PASS by the final release sweep recorded in
qualification/0_2_6/g15_release_sweep.json. The Git tag v0.2.6a0 targets
qualified source SHA 93561c2c0ae1c173deb81e47c3fa3852643275cb. The release
state does not claim a GitHub Release or PyPI publication.