This page holds the atlas' surprises. It publishes only after the dual-check machinery has been run repeatedly, the community can re-run everything from reproduce.sh, and every claim carries its sampling metadata and content hash.
Publication here is gated on finding at least one of:
(a) a rewrite rule not recorded in the literature that preserves ZX distance in a structural
neighborhood where known rules break it; or
(b) a structural-feature combination whose prediction of Δd = 0 beats baseline significantly
and is interpretable.
Until then, this page stays deliberately empty. An empty findings page is a feature: it is the visible half of the project's honesty boundary.
| Step | Action |
|---|---|
| 1 · Re-verify | Dual-implementation cross-check re-run until fatigue; anchors + oracle re-green |
| 2 · Community re-run | Repository + dataset public; independent reproduction invited before any post |
| 3 · Technical post | "We found new distance-preserving ZX rewrites while mapping the rewrite space" — atlas link + findings page + runnable repo |
| 4 · Acknowledge | Short email to the RPK line with links — no co-authorship ask, no signature request |
| 5 · Follow-up | Only if they engage positively |