diff --git a/recoveries/excerpt-placement-audit-zephyrus-sample.md b/recoveries/excerpt-placement-audit-zephyrus-sample.md new file mode 100644 index 0000000..28b6801 --- /dev/null +++ b/recoveries/excerpt-placement-audit-zephyrus-sample.md @@ -0,0 +1,100 @@ +# Excerpt fidelity and placement audit (Q20) — 91 magazine entries, every-10th sample + +> **Screen.** Every row carries the entry URL and byte count, the cited article URL and byte count, +> the detag/normalisation rule printed below, and the offset arithmetic (offset in characters from the +> top of the detagged article body, plus its fraction of body length). Readings are from this host, +> 2026-10-11T13:09–13:10Z. Where a fact could not be verified it says `unverified`; there are none. + +## Sample rule (stated in advance) + +- **Population:** the 91 magazine entries on `https://alexanderwhitestone.com/stackchain/blocks/` + (HTTP 200 / 34,488 B / 2026-10-11T13:09:26Z). +- **Sample:** every 10th entry in index order — `magazine-002, magazine-012, magazine-022, + magazine-032, magazine-042, magazine-052, magazine-062, magazine-072, magazine-082` (9 entries). +- **Excerpt unit:** every `
` on the entry page with detagged text ≥25 chars, quote text + taken as everything before the blockquote's own ``. 26 excerpts total (3+3+3+3+3+3+3+3+2). +- **Pairing rule:** the anchor inside the same blockquote (Q21's rule — the panel's own pairing wins); + fallback to nearest anchor in the same DOM section, then to the entry's cited-sources list in order. + **All 26 rows used the panel's own pairing; no fallback was needed.** + +## Detag/normalisation rule (printed so the numbers are re-runnable) + +- Strip `