deduplisted

# /dedup — find & merge duplicates (supersede, never delete) > 🧒 **When reporting to a non-technical operator:** end with a child-simple "In plain words" recap in their language. ## 🖥️ Dashboard first (the operator works visually) `python "$IMPORTS_ROOT/build_dedup_dashboard.py"` (it runs the scan itself) → open `$OBSIDIAN_VAULT/_Dashboards/Dedup-Dashboard.html`: candidate clusters by topic, colour = similarity (🔴 ≥0.7 almost certainly a duplicate); each cluster shows which note to keep (the newer one) and which becomes superseded. DISPLAY ONLY; merging is done by hand below (supersede, never delete). Complementary notes are left alone. ## Step 1 — Scan (deterministic; AK-47, ~free; token law) `python "$IMPORTS_ROOT/dedup_scan.py"` → scans ACTIVE notes by theme, difflib similarity → candidate clusters in `$IMPORTS_ROOT/dedup_report.txt`. For a DIFFERENT folder, copy the script and change its `OPS` path. Start at similarity ≥0.5 (high-confidence), then ≥0.4 for the next tier — **loop-until-dry** (stop when remaining clusters are complementary/noise). ## Step 2 — Judge each cluster (NOT every look-alike is a dup) - **Identical** → keep the original (lowest msg_id), mark the echo superseded. - **Same topic, different nuance, non-conflicting** → keep the NEWER; graft the older's unique nuances into it. - **Conflict** → NEWER wins (the operator's law: fresher beats older on the same topic). BUT an explicit `origin: anton` rule is overridden only by the operator or an explic