Preserved from the wiki Index page during the 2026-10-14 landing rebuild (kanban t_wikijs_landing_rebuild). Full section text as it was published on the Index.
Date: 2026-06-08
Cycle Type: Hardening + Validation + Pipeline + Calibration + GNN V3 + Oracle V4
Research Streams: 5 parallel tasks (v29.1 through v29.5) + orchestrator
Synthesized By: t_31247c11 (v29.0 Orchestrator)
| Metric | v28 (end) | v29 (now) | Change |
|---|---|---|---|
| Entities | 120 | 120+ | Stable (expansion in v28) |
| Brier Score | 0.0413 | 0.0413 | Maintained Excellent |
| Oracle Version | V3 | V4 | 0 crashes (was 3) |
| GNN AUC | 0.7716 | 0.7716 | Data-limited |
| Pipeline Sources | 6 active | 6 active | Resilient fetcher live |
| Walk-forward σ | 0.1184 | 0.1184 | Honest (non-saturated) |
| Prediction Accuracy | 85.7% (6/7) | — | One falsified claim documented |
GourmetVault/v29.1/reports/v29_1_data_pipeline_hardening.md — Data pipeline hardeningGourmetVault/v29.1/reports/entity_freshness_report.md — Entity freshness trackingGourmetVault/v29.2/reports/v29_2_calibration_report.md — Prediction calibrationGourmetVault/v29.3/reports/gnn_v3_performance_report.md — GNN V3 performance (91 lines, honest assessment)GourmetVault/v24.0/reports/v29_4_oracle_hardening.md — Oracle V4 hardeningGourmetVault/v29.1/scripts/ — Pipeline hardening scriptsGourmetVault/v29.2/predictions/walk_forward_reopt_v29_2.py — Walk-forward reoptimizationGourmetVault/v29.3/scripts/gnn_v3_production.py — GNN V3 full pipeline