v1.1.1 — Don’t Block Your Own Exit

Patch release refining the public claim boundary without changing the solver or official evaluation data.

• removes exact-rank, hidden-optimum, bit-for-bit rebuild, and hidden-rerun caveats that incorrectly implied those were expected participant deliverables
• reframes the public boundary around what is actually released and validated
• keeps the organizer-controlled evaluation record as the authority for hidden results
• replaces source-line AI authorship speculation with the explicit decision/process responsibility boundary
• adds regression coverage so this audit residue does not return

Official evaluation data and final solver source are unchanged from v1.1.0.

v1.1.0 — Don’t Block Your Own Exit

v1.1.0 — Evidence-corrected technical retrospective

Don’t Block Your Own Exit: Engineering a Reliable Anytime Solver for the OGC 2026 Grand Shipyard Challenge

This release supersedes v1.0 as the public authority.

What changed

• rebuilt all ten chronological organizer-evaluation rows from the direct completed-result emails after finding unsupported intermediate values in v1.0
• retained the correct final vector, 60/60 hidden-output feasibility record, and 28.316% first-to-final descriptive reduction while correcting the intermediate trajectory and several attributions
• rewrote the main report around the coupled realization problem, the two mutually exclusive four-core topologies, and the evidence that changed the architecture
• demoted internal candidate/submission identifiers to provenance and renamed public source to solver/final-release/
• made chronology-vs-causality, supplied-instance-vs-hidden transfer, AI assistance, and reproducibility boundaries explicit
• added a corrected evaluation-history test and regenerated deterministic figures/report assets

Verification

• merged commit: 1ee18c273d6036be860d8b955509d8ac2a4cea14
• GitHub CI passed on Python 3.11 and 3.12
• hostile technical review: ACCEPT, no Critical or Important findings
• context-free reader review: semantically self-contained
• PDF SHA-256: 0a690bfdf4ff86ce21ace0f6bc9d9a271382603e6c00511e7690bfd2a40dd68e
• evaluation-history SHA-256: 9efac129fe64083bb8dfc041a0f131db67dbbb39e7b5f55de7376ab9a932ba30
• source-manifest SHA-256: 07a2b4b5648bdde7aaa4595920a3849ce2db84700037c4f90599e53ba8e79ebf

Cross-instance raw sums remain descriptive only and are not the competition score. Exact final preliminary-round rank remains unavailable and unclaimed.

v1.0.0 — OGC 2026 Grand Shipyard Project Closeout

OGC 2026 Grand Shipyard project closeout

This release freezes Team Smoop's public technical artifact for Candidate 23 / Submission 10.

Included

• 12-page technical retrospective;
• final submitted Python source;
• source release of the C++/pybind11 geometry kernel;
• ten official evaluation vectors and interpretation notes;
• methodology and explicit reproducibility/publication boundaries;
• deterministic asset and PDF build tooling;
• 39 release-integrity, protocol, worker, topology, source-provenance, PDF, and native-kernel tests.

Verified result

• 10 official submissions;
• 60/60 hidden outputs feasible;
• final vector: [11280, 31368, 103065, 6464170, 16443998, 42333422];
• descriptive raw six-instance sum reduced 28.316% from first to final submission.

The raw sum is not the competition score. Exact final preliminary-round rank was not preserved; no rank, finalist, award, or optimality claim is made.

Release boundary

The release excludes organizer data/checker code, the exact submitted ZIP, the submitted native binary, participant correspondence, certificate, private experiment corpora, and ResearchLab run packets. The private package and binary are bound by SHA-256 in the public source manifest.

CI

The tagged commit builds the native kernel and passes the full test suite on clean Ubuntu runners with Python 3.11 and 3.12.