For each (query, modern_uri, legacy_uri) triple: run cupertino search "<query>" --limit 10, classify the outcome as modern_wins (modern rank < legacy rank), legacy_wins, tied, modern_only (only modern in top-10), legacy_only, or both_missing. The Class C concern is that the ranker should prefer the modern Swift form on every pair; an agent grounded on cupertino search "URL" should land on apple-docs://foundation/url (Swift struct), not apple-docs://foundation/nsurl (legacy ObjC class).
Each card opens its own page. The headline and charts above are all you need at a glance; the cards are for the why and how.
A zero "regression" column with a positive "improvement" column is the clean-win shape.
Read details →flowchart TD Q["30 (modern_query, modern_regex, legacy_regex) triples<br/>harvested from search-quality-deprecation-baseline-v1.2.0.md"]:::input Q --> H[search-quality-phase1-extended.py]
Read details →