← Back to the app

The field guide.

Everything the reader knows how to do.

^ keep this open the first time through. you'll want it.


What it is

R/W reads with you.

R/W is where a finished draft becomes finishable, the last stage in the Kaizen Suite after Weaver and Rewriter. It works in marks: passages worth attention, dismissed, taught, or saved. Every pass gets quieter as it learns your taste.


Getting started

Bring your draft.

Paste your manuscript, or upload a file (.txt, .md, or .docx). The story should already exist; R/W reads finished drafts.

paste works best for a single chapter. upload for a full manuscript (it detects chapter headings automatically).

Chapters are split by headings like "Chapter One," "CHAPTER 1," "Prologue," or markdown headings. Paste without a heading and it loads as one continuous read.

Coming from Rewriter or Weaver? The landing accepts a suite bundle. Open it and your dictionary, the marks vocabulary, and everything you taught upstream comes with you. R/W picks up where the chain left off.

The reading

It reads before it speaks.

As you read, the AI quietly tags passages it wants you to look at. Each one shows up as a color shift in the letters themselves. No boxes, no icons, nothing yelling at you. Some marks are tagged as suggestions, meaning the AI is less certain. They look the same at rest but carry a confidence level the system tracks internally. It adjusts what it flags based on what you keep and what you dismiss.

Click a mark to open its panel. Try it.

Four breaths. Four heartbeats. Four walls in my room.

I swung my feet out of bed and onto the carpet. Cool fibers pressed between my toes. Everything looked exactly the way it should on a lazy Saturday morning.

Repetition

Parallel rhythm accumulating into post-image explanation: "The walls, the ceiling, the dusty morning light" lists three fragments in the same frame, then "Everything looked exactly…" restates what those details already implied.

Saved to dictionary

Except

Pacing

The shift from domestic calm to unease could land harder with a shorter lead-in. "So steady I almost mistook it" dilutes the strangeness.

Saved to dictionary

It came from beneath the floorboards, more felt than heard.

Six categories. Click one to highlight its marks above:

RepetitionStylePacingVoiceContinuityInteriority

Your moves

Three keys. That's the whole workflow.

Open a mark above, then try these. They work here.

DGot it. Dismiss the mark. You've seen it, you've decided.TTeach. Tell the AI why it was intentional. Saved to the dictionary.FLater. Flag for review. Come back in audit mode.

D is the one you'll press most. most marks are "yes, I see it, moving on."

A few more, once you're in the flow:

TabJump to the next mark.Shift+TabJump to the previous mark.EnterOpen or close the current mark's panel.QQuick-fix. AI rewrites the passage. Edit before applying.PProtect. Skip this passage on future scans.EEdit the paragraph directly. Textarea right where you are.Ctrl+EnterApply a quick-fix from the textarea (audit mode).EscClose whatever's open.

On a phone

Marks become cards.

When a paragraph holds two or more marks, the panel becomes a deck of cards. Tap any mark below to open it. Drag the top card sideways, tap an action, or use arrow keys.

She , , the way her mother .

drag the card, tap an action, or use arrow keys

D, F, P resolve inline and advance the deck. T and Q hand off to the panel surface (the same panel you used above), since teaching and rewriting need more room than a card. Protect stays on the current card so you can decide what to do next.

desktop opens in panel by default. flip to swipable cards in Settings if you prefer.

Teaching

The dictionary is where it learns.

Press T on a mark above, type a short explanation, and it saves. Next time the AI sees that pattern, the mark won't appear. Try it, then watch the dictionary below fill in.

Dictionary
Teach a mark above, or add a voice entry below, and it shows up here.

The tool gets quieter the more you use it.

this is the whole point. first pass is noisy. third pass, only the real stuff is left.

The dictionary holds two kinds of entries. Constraints are notes the AI respects, added when you teach a mark. Voice entries describe how your prose moves so rewrites sound like you. Add a voice entry above and watch the dictionary group it under its own header. Conditional rules let you scope any entry to specific POVs, chapter ranges, or scene types. The dictionary lives in your browser.

Making changes

Two ways to revise.

AI rewrite: Press Q on any mark. The AI streams a rewrite using its own reading as context. Edit the suggestion, then apply. Try it on the demo marks above.

Manual edit: Click the pencil, double-click, or press E. Try the paragraph below:

She blinked. The fog lifted. She wiped the same spot on her hands and repeated the words exactly as before.

Both paths share the same undo stack, three levels deep. Edited paragraphs show a gold bar in the margin.

After Apply, a short summary surfaces what your fix touched: continuity, setups, POV, recurrence. Drawn from what the audit notebook already knows; no new AI read needed. Press Q on a mark above, then Apply, and watch the toast.

after a manual edit, the toast also offers Run Ripple Scan. one click, the AI re-reads the chapter and surfaces only what's new. that one is a fresh read; the post-Apply summary above is read-only.

Your marks

You can mark text too.

Select any text below. A pill appears.

The road crumbled at the edges where rain had eaten through years of neglect. She kept to the center line, stepping over the cracks out of habit more than caution.

Ask AI opens the AI reader on your selection. Later flags it for audit mode, no AI involved.

author marks override system marks on the same text. your attention takes priority.

Flag and return

Some marks aren't ready yet.

Press F on a mark and it doesn't disappear. It moves to a list you check later. Try it, then resolve it from the list.

She nodded once, the way she always did when she meant the opposite.

Voice

Narrator stepping outside the moment to label the gesture instead of staying in it. Worth a second look.

Flagged for review
Nothing flagged yet.

The list lives at the top of audit mode under History → Queue. Flagged marks stay there until you handle them, across sessions.

flag is for "I want to think about this." dismiss is for "I've decided." different muscles.

The big picture

Audit mode is a notebook.

Five groups. Click the tabs:

D/T/F/Q work on audit cards the same as reader panels. check a box to build a beta reader packet.

The heatmap is the thin rail on the right edge. Each dot is a mark. Clusters mean dense sections.

Chapter sidebar appears when you hover the left edge. A table of contents slides out with scan status per chapter and expandable mark sub-lists.

Full scan reads every chapter and returns cross-chapter findings: phrases that repeat, motifs that track, continuity slips the per-chapter scan missed.

Inline context

The manuscript talks back.

Click the pills below a chapter heading. Click the chevron at the end of the last paragraph.

Chapter One: The Storm
Tension52%
Quiet score40
Marks3

The sky had been bruised all morning. She stood at the window watching it darken.


dialogue

She sat on the porch step and pressed the phone to her ear. Nothing.

He hadn't called in three days.

2 min ago, quick-fix
He hadn't called in four days. He hadn't called in three days.

Scene breaks show the scene type and pressure shifts. Density toggles in the TOC let you turn each family on or off.

vessels mount lazily as you scroll. the tool isn't building hundreds of panels you'll never see.

Passes and runs

You read it more than once.

Pick a lens before you scan. Each one changes what the AI looks for:

Each time through, the tool logs a reading run. The rest of the toolkit builds on that history:

Snapshots save the manuscript state before a big change. Analytics compare two runs: word count deltas, density shifts, which flagged marks got addressed. Journal logs every edit chronologically. Queue promotes flagged marks into a tracked task list. Recurrence clusters the same pattern across chapters so you can handle the group at once. Summary brief is a short editorial letter about the whole book, with citations.

flag a mark in the queue, fix it, check it off in the journal.

Developmental reading

Story-level tools.

Pro unlocks a second toolkit. Every scene the AI reads gets a card. Click any field to edit it:

Scene carddialogue
WhoMara, Ellis WhereThe porch, dusk ChangeMara realizes Ellis has been lying about the phone StakesTrust between them, which the next chapter depends on
Unresolved: Whether Mara says anything about it tonight.

That's one piece. The full toolkit: tension arc curves pressure across every chapter with TOC sparklines. Continuity ledger tracks characters, objects, locations, and flags contradictions. Setup and payoff watches what you planted, what paid off, and what's still dangling. Character pressure maps external, internal, relationship, and deadline weight per character per chapter. POV drift locks a baseline scene per POV character and flags when the voice shifts.

all of these live in the audit notebook now. open audit mode, pick the Structure, Voice, or Craft group.

Getting it out

Export, share, and import.

DOCX export (gear menu, Pro). Word document with your revisions as native comments and a dictionary appendix.

Beta reader packet. Check findings in audit mode (try the checkboxes above), export a DOCX with only what you picked plus a cover page. Hand it to a reader without exposing the whole manuscript.

Beta reader feedback. Gear menu → Import → Beta reader feedback. Upload a DOCX with Word comments and they come in as marks. Preview before committing. D/T/F/Q work on them the same as AI marks.

feedback import accepts any commented DOCX, whatever tool wrote it.

Settings

The gear menu.

Bottom-right corner of the reader.

Theme. Light or dark. Follows your system by default.

Tiers. Free uses Gemini on Google's paid API. Pro ($20/mo) uses Claude on Anthropic's commercial API. Both tiers run on paid APIs that don't train on inputs. Or bring your own key.

Dictionary. Everything you've taught the tool. Conditional rules let you scope entries to specific POVs, chapter ranges, or scene types.

Categories. Six built-in, plus custom categories you can create, rename, and recolor.

Voice profile. Dictionary, categories, and POV baselines bundled as a file. Export it when you start the next book.

Import. Bring back a previous DOCX export with mark recovery, or import beta reader feedback from a commented Word file.

Recent edits. All revisions in one place, regardless of source.

Mark surface. On a phone, mark cards swipe by default. On desktop, switch from inline panel to swipable cards if you prefer.

Audit mode. Opens the diagnostic notebook. Everything that used to have its own dialog (tension arc, scene cards, continuity, pressure, recurrence, analytics, queue, journal, snapshots) now lives in audit tabs.

Where this lives

The last stop in the suite.

R/W is the final stage in the chain that started in Weaver and passed through Rewriter. Anything you taught upstream (character beats, world rules, scene intents, the voice fingerprint) traveled with the manuscript, and R/W reads with that memory in hand.

started here without weaver or rewriter? that's fine. R/W reads any finished draft on its own.


You've read the guide. One last thing.

Every revision starts the same way: somebody reads closely enough to notice.

You're ready. Go read your draft. The tool already knows how to listen. Now teach it what you meant.

Open the app →