Romans 8:28–30 — what nine editions agree on, and the one slot where they split (a calibration pilot)
Illustrative — the golden chain only. A calibration pilot, not corpus data (Stage-6 coding ≈ 0 in the core database). ·
The headline number, α = 1.000, is cross-model coder agreement — it shows the negative space is reproducible. It is NOT a claim of measured cross-tradition convergence.
Source: the sandbox ensemble matrix (golden_chain_ensemble_matrix.csv), 3 cross-model coders (Claude Sonnet · Haiku · Opus) — not the core database.
The negative space is the road not taken — the reading a text could have made but didn't. At Romans 8:28 (“all things work together for good”) the move-agent splits:
seven editions keep an impersonal-collective agent (it is all things that work together), while
two recast the actor as a personal God (it is God who works). That split is the negative space — and it is
reproducible across three cross-model coders (semantic_agency α = 1.000). One slot down, at 8:29–30, the disagreement vanishes: the
golden chain (foreknew → predestined → called → justified → glorified) is stable — all personal/God across every edition. The braid splits, then re-converges.
Visual 1
The path-braid — where the editions braid, split, and re-converge
Nine editions as parallel paths, left → right, across the three positions 8:28 → 8:29 → 8:30. The braid-then-split is the whole story.
impersonal-collective (7) personal / God (2) the golden chain (stable)⚑ deponent flag
Illustrative — golden chain only; a calibration pilot, not corpus data. The braid-then-split renders the semantic_agency field from the matrix; the α = 1.000 is cross-model coder reproducibility of this split, not a convergence measurement.
How to read it. Seven editions — Greek-SBLGNT, KJV, NRSVUE, NKJV, DRA, Vulgate, Luther 1545 — enter from the left and braid into one impersonal-collective lane at 8:28. NIV and The Message split off downward into a personal/God lane: the split is the road taken differently, the negative space made visible. Then both lanes re-converge onto the golden chain, which holds steady across all nine through 8:29–30. The Vulgate carries a deponent flag (cooperantur): a passive verb-form with active sense.
Visual 2
The golden-chain DAG — the within-passage flow
Five moves, directed and acyclic: foreknew → predestined → called → justified → glorified. The stable spine the braid converged onto.
Illustrative — golden chain only; a calibration pilot, not corpus data. The annotations (personal/God · active · object-role) read directly off the matrix; the α = 1.000 is cross-model coder agreement, not convergence.
How to read it. The five moves form a one-way flow — God foreknows, then predestines, then calls, then justifies, then glorifies. No move loops back (acyclic). On the matrix all three coders code every node identically: personal/God agent, active voice, object-role (the people are the ones acted upon). This is the stable spine the braided editions all converged onto.
Visual 3
The negative-space rendering — the solid spine and its ghosts
The coded chain drawn solid; the readings not taken fan off each slot as ghosts (the paradigmatic negative space — probe #8).
solid spine — the actual coded chain ghost — the unchosen reading
Illustrative. The solid spine is documented (the coded chain in the matrix); the ghosted branches are illustrative — they sketch the values not taken (the personal-God reading; the unchosen voice values passive / middle; stative, reserved). Not corpus data; not a convergence claim.
How to read it. The solid golden spine is what the editions actually coded. Off each slot, ghosts fan up: the readings the text could have made and didn't. At 8:28 the loudest ghost is exactly the personal-God reading that NIV and The Message did take — which makes the point that a ghost in seven editions is the spine in two. The negative space is not absence; it is the set of live alternatives a tradition declined.