oregon-stories · semantic index over the whole OAR
The copy-paste rulebook: 654 rules in 239 near-verbatim families
At cosine ≥ 0.985 — near-verbatim territory — 654 of Oregon's 36,953 rules sit in 239 twin families, 18 of them spanning more than one agency's chapter. The largest family stretches across 28 chapters — one procedural text, adopted again and again across the government. That is not plagiarism; it is how Oregon administrative law is designed to work (agencies adopt model rules), and the interesting part is the margins: where one agency's copy quietly differs from the family.
The twelve widest-travelling texts (cross-chapter families)
| rules | chapters | members |
|---|---|---|
| 28 | 28 | 137-008-0100 · 170-001-0020 · 173-001-0015 · 257-090-0010 · 274-007-0001 · 291-001-0115 · 335-090-0010 · 339-001-0018 … +20 |
| 16 | 16 | 604-040-0030 · 607-040-0030 · 611-020-0030 · 623-020-0030 · 642-040-0030 · 645-020-0030 · 646-020-0030 · 647-020-0030 … +8 |
| 14 | 14 | 137-008-0120 · 274-007-0002 · 291-001-0110 · 340-011-0004 · 440-055-0010 · 459-060-0210 · 660-001-0410 · 690-004-0030 … +6 |
| 11 | 11 | 604-040-0050 · 607-040-0050 · 611-020-0050 · 623-020-0050 · 642-040-0050 · 646-020-0050 · 656-020-0050 · 657-040-0050 … +3 |
| 10 | 10 | 604-030-0030 · 607-030-0030 · 611-030-0030 · 617-030-0030 · 623-030-0030 · 643-030-0030 · 645-030-0030 · 656-030-0030 … +2 |
| 5 | 5 | 604-040-0020 · 607-040-0020 · 642-040-0020 · 657-040-0020 · 670-040-0020 |
| 3 | 3 | 604-030-0040 · 607-030-0040 · 643-030-0040 |
| 2 | 2 | 309-120-0080 · 416-425-0020 |
| 2 | 2 | 340-266-0050 · 603-077-0112 |
| 2 | 2 | 340-266-0075 · 603-077-0119 |
| 2 | 2 | 407-120-0130 · 943-120-0130 |
| 2 | 2 | 407-120-0140 · 943-120-0140 |
Read this before quoting the chart
Twinship is textual, not legal: two near-identical rules can have different legal force in their different contexts, and adopting model rules is standard, sanctioned practice — nothing here implies impropriety. Similarity is cosine over per-document mean embeddings (BAAI/bge-m3), threshold 0.985; a family is the transitive closure of pairs above it, so long chains can join texts that differ more end-to-end than any adjacent pair. Same-chapter families (numbering echoes within one agency) exist and are counted, but the table shows cross-chapter families — the ones where text traveled between agencies. The cache records the artifact fingerprint and rebuild method.