Historical experiments#
This archive preserves dated optimization questions, rejected layouts, and hardware measurements that are useful for explaining why the current backend was chosen. These pages are evidence records, not the active roadmap or current public contract.
Open questions and TODO tables inside an archived page describe the state at that page’s recorded source revision. Before resuming one, check the current verified results, optimization ledger, source, and tests.
Historical numbers remain labeled with their original backend, source state, fixture geometry, and cache conditions. They must not be quoted as current performance without a fresh run under the benchmark methodology.