ShurIQ · Methodology

The ShurIQ Book-Creation Method

A six-step process for turning a thinker's premise into a first-draft book that sounds like the thinker, not the machine.

The ShurIQ Book-Creation Method

A six-step process for turning a thinker’s premise into a first-draft book that sounds like the thinker, not the machine. Built to demonstrate the grammar-creation process Limore asked to see in action.

The method is a sequence of compressions. Each step throws material away. What survives is the book.

Step 1 — The Premise Compression

The first artifact is not a chapter. It is a single sentence that the entire book has to earn. Anything that does not pay rent against that sentence gets cut.

For Think Inside the Box the sentence is: The constraints of the box create the void. The book is twelve chapters of evidence that this is true at every scale — corporate, personal, mathematical, divine.

For It’s All in the Thumbnail the sentence is: If an idea cannot survive radical reduction, it isn’t structurally sound. The book is twelve chapters of evidence that compression is the integrity test, not a packaging step.

The premise compression is run by the author and the agent together. The author proposes the sentence. The agent stress-tests it against the outline. If any chapter title has to be twisted to fit the sentence, either the sentence is wrong or the chapter is wrong. Fix one before drafting begins.

Step 2 — The Voice Anchor

A book in someone else’s voice is a constraint problem. The author’s voice has to be expressed as rules concrete enough that a drafting agent can apply them without taste.

Sources, in order of authority: 1. Spoken transcripts of the author thinking out loud (Fathom, Granola). Highest authority because they capture cadence under cognitive load. 2. Written messages in the author’s natural medium (Slack, email). Second highest because they capture deliberate compression. 3. Formal writing samples if the author has them. Lower authority because formal voice is often performed. 4. Briefs and pitches the author has approved. Lowest authority because the author may not have actually written them.

The voice anchor is a single document. It lists: cadence rules (sentence length, paragraph length, fragment usage), lexicon (characteristic words, banned words), rhetorical moves (how a thought opens, how it closes), conceptual signature (recurring metaphors), and 10–15 verbatim quotes with attribution.

The drafting agents read the voice anchor before reading the outline. Outline tells them what to write. Voice anchor tells them how to write.

See: LIMORE-VOICE-SIGNAL.md.

Step 3 — Chapter Scaffolding

Each chapter is scaffolded before it is drafted. The scaffold has four parts:

The single line. What does this chapter prove? One sentence. If it overlaps with another chapter’s single line, one of them is redundant.
The two pillars. The chapter rests on two arguments or two illustrations. Three is a list. One is a monograph. Two is a chapter.
The pressure point. The moment in the chapter where the reader should feel the premise lock in — usually a sentence or a short paragraph that names the truth the rest of the chapter is dancing around.
The exit. The last line. It should hand the reader to the next chapter without summarizing what came before.

Scaffolding is mechanical. A scaffolded chapter can be drafted by anyone — or by an agent — without losing structural integrity. An unscaffolded chapter will drift no matter who writes it.

Step 4 — Drafting Under Two Constraints

The drafting agent operates under two simultaneous constraints: the voice anchor (Step 2) and the chapter scaffold (Step 3). Neither is optional. If the voice anchor says “open chapters with a single declarative line,” the agent does not open with a paragraph.

Drafting agents are parallelized — one agent per book, or one agent per part, depending on length. Two principles govern the parallelization:

Same voice anchor, different drafters. Every agent reads the same voice document. Voice drift across chapters comes from agents inventing voice rules to fill ambiguity. The anchor is the antidote.
No cross-agent reference. Drafting Agent A does not read Drafting Agent B’s output mid-flight. Echo chambers form fast when agents start citing each other. Cross-references are added in Step 5, by a single editor pass.

Step 5 — The Anti-Slop Pass

Anti-slop is a structural integrity check. Every chapter passes through a hard-rule linter before the editorial site sees it.

Hard bans: - “Not X, but Y” inversion. This is the most-used AI rhetorical move. It is also the most boring. If the chapter says “it’s not the box, it’s the void,” rewrite it as a positive claim: “the void is what the box defines.” - Scaffolding meta-prose. “This chapter argues that…” “The thesis here is…” “What we are about to see…” — none of it survives. The chapter argues by arguing. - Buzzword density. “Leverage,” “ecosystem,” “paradigm,” “synergy” — every appearance is an error to fix, not a flourish to keep. - Hedge-stacking. “It could be argued that perhaps in some cases…” — the author either thinks the claim is true or doesn’t. Hedge once if the evidence is partial. Never hedge twice. - Decorative jargon. Words from the back-of-house that don’t belong on the page (graph theory terms, internal frameworks, the § symbol, anything that announces “a system was used here”).

Soft checks: - Does the chapter open with a single declarative line? (Voice rule.) - Does the pressure point sit roughly two-thirds of the way through? - Does the exit hand off cleanly? - Are the two pillars actually two pillars, or did one collapse into the other?

The pass is run by a separate agent — never the drafter — so that surprise is preserved. A drafter who edits their own work cannot see what they have failed to write.

Step 6 — The Review Site

The book is delivered to the author as a static editorial site. Not a Google Doc. Not a deck. A site, because:

A site forces commitment. Sending a Doc invites endless small edits before the author has read the whole. A site is read first, edited later.
A site is a reading environment. Long-form prose deserves a column width, a serif, a calm background. None of that comes for free in Docs.
A site separates the what (the chapters) from the how (the method). Both ship together. The author sees the book and the engine that made it in the same handoff.

The site is sticky-navigated by part and chapter. The method document sits as a peer to the books, not buried in an appendix. The voice anchor is linked but not foregrounded — the author should not have to read his own voice document; he should recognize his voice in the prose.

Feedback mechanics come after the author has read. Whether comments arrive through a Doc, an annotation tool, or a Slack thread is downstream of whether the draft was worth commenting on.

What This Method Is Not

It is not a magic prompt. The hard work is in Step 1 (premise compression) and Step 2 (voice anchor). If either is sloppy, the rest of the pipeline produces glossy slop.

It is not author-replacement. The author still owns the premise, still owns the voice, still owns the final cut. The method industrializes the middle — the part where a writer would otherwise spend a weekend on every chapter — so the author can spend that time on the part that matters: judging whether the draft is true.

It is not stylistically neutral. The method has a preference for compressed, structural prose. It will not produce a meandering literary memoir well. It is built for books whose premise is itself architectural — where the form of the argument and the content of the argument want to match.

Provenance

This method was specified and run on 2026-05-18 to draft the first round of Think Inside the Box and It’s All in the Thumbnail. The output of that run lives in drafts/ and on the deployed review site.