Chapter 7 — Revise with Intent and Evidence

A chapter can be rewritten twice and come out different and no better at what it is trying to do, because neither pass named what was wrong. The Inkagai Way treats revision as a response to evidence, meaning an observable reason to believe the Work is not yet doing what it is for. A reader who consistently misunderstands a key relationship is evidence. So is a source that does not support the strength of your language, or tension that falls because consecutive scenes do the same job.

A useful revision task names four things:

The observed problem.
The effect on the reader or Work.
The intended change.
The evidence that would show improvement.

For example:

The reader learns Mara’s fear only after she refuses the offer, so the refusal reads as arbitrary. Establish the fear through an earlier choice or physical response, then confirm that the refusal feels surprising but motivated.

A note like “Improve character motivation” or “tighten the argument” gives you nothing to act on.

Revision happens at several scales.

Purpose revision

Is this still the Work you intend to make, or are you holding on to an early concept the manuscript has outgrown?

Structural revision

Does the sequence produce the movement you intended, and does every major section do a job no other section does? For a writer who discovers the book in the drafting, this may be the first time the true outline gets made. For one who plots, it is where the planned model meets what the draft became.

Developmental revision

Do characters, claims, themes, examples, and reader expectations develop with enough depth and continuity?

Scene or chapter revision

What does the reader carry out of this scene or chapter that they did not bring in?

Prose revision

Is the language precise and right for the voice? Does clarity preserve complexity rather than erase it?

Production revision

Do formatting, metadata, captions, links, permissions, and edition-specific choices accurately represent the Work?

Work from the largest unresolved decision downward. Polish the prose before the structure settles and every deletion hurts, and fix every local sentence and you may never notice that a whole chapter serves no purpose.

A reader’s confusion is real, but their proposed solution may be wrong, and an editor’s preference may reflect a market you chose not to serve. Keep what people observed apart from what they prescribe; “What happened to you while reading?” usually gets you more than “How should I fix it?” Collect the patterns, such as where attention dropped, where trust weakened and what question readers thought the book was answering. Then go back to purpose and decide.

The connected model of the Work supplies evidence too. A timeline reveals impossible movement, a source ledger can show an entire chapter resting on one weak reference, and a diff can show a revision that removed a necessary setup.

AI can help with diagnosis, but its confidence is no evidence. A model names plausible issues because it has learned common patterns, so treat what it names as hypotheses and check them against the manuscript and the sources. AI-generated revisions may solve the stated problem while bringing in a new voice or adding unsupported facts. Before you accept one, read what it actually changed.

Revision ends when the Work is ready for its intended consequence, however many improvements are left over. A Work can always be made different, so perfection can’t be the gate. Ask whether the Work fulfills its central promise, whether known factual and continuity risks are resolved, whether remaining ambiguities are intentional, and whether publication and rights conditions are satisfied. When the answers are yes, more revision is probably avoidance.

Reflection

What known need is your current revision answering?
Are you polishing below an unresolved structural decision?
Which feedback describes an observation, and which merely proposes a preference?
What evidence would allow you to declare the Work ready?