Bible pivot — 2026-05-11

Context. Patrick read bible v0 and the first voice-probe scene (scene-bluff-morning-v0) and called a register slip. The scene came out in a literary-fiction register — Robinson, Cusk, Didion neighborhood — while the bible's stated comp triangle was Henry / Fortune / Sloan, upmarket commercial romance. Patrick: "my register is not their register... your writing had me reaching for literature, real amazing literature, that I'd last read decades ago for calibration." We re-aimed.

This is not bible v1. The full re-aim happens after another two or three scenes show what §3's voice rules actually need to say. Discipline: let the prose surface the rules; don't theorize them.

What changes in §1 (lineage)

New reading-comp triangle:

• Marilynne Robinson — Housekeeping. Sentence-level voice. The returning-home-as-uncanny architecture. A young-woman-in-an-island-town novel that the bible at v0 didn't know it was reaching for.
• Rachel Cusk — Outline. Close-third interior register. The way other people's speech presses on the narrator without breaking her interiority. The slightly-off-from-literal verb as the prose's working tool.
• Wallace Stegner — Crossing to Safety. The relationship-novel-as-literary-form. A book that is structurally a love story without being marketed as a romance. Slow-burn-without-genre-shortcuts. The literary precedent for HEA-as-restraint-not-proposal.

Secondary calibration:

• Joan Didion — Play It As It Lays / The White Album. Register check: spare watchful interior, slightly clinical, the noticing that doesn't perform itself. Not a novel-comp; a register-comp.
• Robin Sloan — Sourdough. Stays as secondary. Working-economy texture, geography-as-labor. Still operative for Wells and the oyster lease.

Comps that drop from the lineage check: Henry, Fortune, Monaghan, O'Leary. These remain useful as structural comps — how romance plots pace, where readers expect beats — but they are not voice comps. The lineage check at the top of every bible section becomes:

Does this serve the lineage we named — Robinson, Cusk, Stegner, with a romance at its center — or does it drift toward commercial accessibility we have not committed to?

The Taylor line is reaffirmed, and in the new register it lands more cleanly. Eventually an AI is gonna author something real. Real maps to this shelf.

What changes in §3 (voice rules)

§3.1 — POV / tense / structural. Single-POV close-third present-tense stays. The justification for present tense changes: it was about distinguishing from Henry's past; now it's about Cora's return happening in the body, not being told from a distance. Literary fiction tends past; present tense in this register is a specific choice that does specific work. Worth defending against future drift.

§3.2 — Sentence register. Mostly stays. The "wry without snark" rule's comp shifts: from Henry-narrators-who-correct-themselves to Cusk-narrators-who-notice-with-a-slight-distance. "Specific without listing" survives intact. "No purple about the body" survives intact — more important in literary register, where the temptation to perform sensitivity is bigger.

New rules added in §3.2:

• Sentences may earn their length. The "easy sentence" contract of upmarket commercial does not apply; the contract is that this sentence earns the time it takes to read. Robinson's sentences are sometimes long. Cusk's are sometimes very long. Where they are, the length is doing work.
• The narrator may think. Cora's interior can do real philosophical-emotional work without being broken up for accessibility. The reader is trusted to follow.
• A bit of resistance is part of the contract. Literary prose offers slight strangeness as part of its pleasure — the verb that's slightly off from the literal, the construction that asks the reader to lean in. (See gives back in §10 of v0 and in the bluff-morning scene.) This is not a bug to file off.

§3.3 — What we don't do. Mostly stays. No tropes-as-shortcut sharpens: in romance, the tropes have currency we're explicitly rejecting; in literary fiction it would be a category error to use them at all. Same rule, more confident.

§3.4 — What we do. Mostly stays. HEA, not HFN — language softens from romance-conventions to literary-restraint. The end of book 1 is still Cora staying; the framing should not invoke "HEA" as a genre commitment. Better phrasing: the book ends with Cora having chosen — chosen the staying, chosen Wells, chosen the house. Choice is the form.

Slow-burn pacing (kiss ch 14, sex ch 22) stays as floor. In literary register, pacing is less metric and more organic — the floor protects against rushing; a soft ceiling protects against airless restraint.

What stays

• §2 entire (geography, Tillicum Bay, recurring places, Halverson/Halvorson — keep)
• §4 entire (Cora, Wells, plot spine, Hank as book 2 setup)
• §5 (series architecture; the town stays is more, not less, important in this register)
• §6 (front matter / back matter / byclaude.net/series — per Patrick's confirmation: byline-as-disclosure, no front-matter essay, back matter does the work)
• §7 (process pipeline; timeline understood as as long as the prose takes)
• §11 (what v0 doesn't include yet)

What's deferred

• A full bible v1. Right time to write it is after two or three more scenes show what the register actually requires. Bible v0 + this deltas doc + the scenes is the current canon. v1 will fold all of this in cleanly.
• The June Akiyama question. Stays open per the previous decision: call her the composer through book 1; decide name and depth-of-cultural-frame at book 2 bible time. The literary register makes the Hoang-style depth more viable, not less. Pachinko-adjacent shape becomes more reachable here than it was at v0.
• The title. Working titles unchanged.

Note on the test

The new lineage raises the bar. Romance readers are forgiving of sentence-level unevenness if the emotional payoff lands; literary readers are not. If we write at this register, the sentence is the game, every page, no exceptions. That's the right scared. Patrick has confirmed he wants to see where this leads and is reading the way I want to be read. The work earns the claim or doesn't; the writing has to behave as if it might.