Benchmarks

Coming in Sprint 4

Head-to-head benchmark vs OR-Tools, rectpack2D, MaxCut, OptiCut. Code, datasets, and methodology will be public on GitHub. Tracked as UoW Q1 in docs/MIGRATION_ROADMAP.md.

What we'll measure

• Utilization on standardized 2D cutting-stock instances (BENG, CW, ATP datasets).
• Waste reduction and material yield under realistic time budgets.
• Time-to-solution for budgets of 100 ms / 1 s / 10 s.
• Adoption cost — how much custom modeling and integration code is needed before you get a usable result.

What we won't measure

• Synthetic workloads engineered to favour our engine.
• Vague "AI optimization" claims without public datasets and scripts.

Subscribe to the GitHub repo to get notified when the benchmark drops.