F-03:
    title: Cross-space coupling is adjacent ordering, strongest in Currier B, and cheap to make
    statement: >
      Last glyph to next first glyph is coupled, B above every prose manuscript but inside short Latin verse at matched size,
      with no caesura or rhyme. Any skewed edge table regenerates it, so it constrains the mechanism only through which
      table and why A and B differ.
    map: [A3, A4]
    claims: [c-307, c-325, c-327, c-308, c-350, c-387, c-372, c-300, c-358, c-368]
    status: standing
    lines:
      - {axis: data, by: "3 transliterations", ref: results_s20.txt, agrees: true}
      - {axis: impl, by: "roach r2 (marginal: B similar to Priapea)", ref: runs/roach-r1r2-raw/results_r2.txt, claim: c-350, agrees: true}
      - {axis: external, by: "Rozanova and Temerev 2026: edge information stops at line breaks", ref: data/external/papers/rozanova_temerev_2026_arxiv_2608.17096.txt, agrees: true}
    points:
      - {h: H-13, rel: for, strength: weak, why: "A couples like prose, B like metrical Latin"}