Part two ended on the structural fact underneath all of this: the plan and the draft are two documents, and only one of them is updated by writing. The other needs an act of maintenance that competes with the writing session and loses.
So a fix has to do one of two things. Either it makes maintenance nearly free, or it removes the need for it. Anything that just asks you to be more diligent is a plan to fail slowly. Here is what actually works, starting with the part that has nothing to do with which app you use.
Three properties a story plan needs
Whatever you keep your notes in, these are the properties that decide whether the drift from part one can happen. They are worth knowing on their own, because they let you evaluate any tool in about a minute.
1. It has to be reachable without leaving the scene. Not "in a tab", not "one click away" — reachable while your hands are still in the paragraph. This sounds like a convenience and it is not. Every unit of friction between the draft and the reference is what converts a lookup into a re-decision, which is cost number two in part two. If checking is cheaper than guessing, writers check.
2. Entries have to carry their own provenance. An entry that says Her sister is thirty-four is not the same object as an entry that says Her sister is thirty-four — set in ch. 4, last confirmed against ch. 9. The second one can be audited; the first can only be believed. This single habit does more than any other to fix the miscalibrated confidence that made a stale plan worse than no plan.
3. The draft has to be able to update the plan, not only the other way round. This is the hardest of the three and the one that matters most, because overtaken entries (facts the draft created that nobody ever decided) cannot be caught by any amount of careful note-taking. There was no moment at which to write them down. They can only be found by reading the manuscript, which means something has to read the manuscript.
Stop trying to keep two documents in sync by hand. Make the manuscript the source of truth, and make the plan something that is checked against it — cheaply, often, and without a separate maintenance session.
The method, in whatever you already use
You can get most of the way there today, in a notebook or a document, and I would rather you had this than a recommendation.
Date every entry and record where it came from. Two extra words per line: the chapter that established it. Nothing else in this article pays back as fast. An undated entry has no state; a dated one tells you how much draft has happened since anyone checked, which is exactly the exposure calculation from part two.
Track volatility as well as category. Organising by kind is right, and it is how anyone naturally files this: characters here, places there, timeline below. It is simply not the axis that tells you what to re-check. Cutting across those categories is a second question, which is how likely a given entry is to move. Names, ages, physical facts and dates are volatile wherever they are filed. Themes, the reason a character wants what they want, the ending you are steering toward: those are stable, and re-verifying them wastes attention the volatile entries need. Mark the volatile ones, in whatever way your notes allow, and reconcile only those.
Write the fact into the draft first, the note second. When a scene invents something — the second window — put it in the scene, finish the scene, then add the entry. It reverses the usual instinct and it matters, because the manuscript is the only document that is definitely current. Notes that lead the draft are predictions. Notes that follow it are records.
Run a reconciliation pass at fixed intervals, not by feel. Every ten thousand words, take the volatile shelf and check it against the draft. By feel means when you are anxious, which is when you have already stopped trusting the notes.
- Take only the volatile shelf — names, ages, dates, physical facts, who knows what.
- For each entry, search the draft for the term and read the most recent occurrence, not the first.
- Correct the entry, and date it with today and the chapter you checked against.
- Anything you cannot resolve in ninety seconds goes on a short open-questions list rather than being guessed at.
- Stop at twenty minutes. A partial pass with dates on it beats a complete one you never repeat.
Where the whole book living in one place changes the arithmetic
The method above works, and it is still maintenance. Better-aimed maintenance, but a task you have to keep choosing. The reason we built our own workspace the way we did is that the two costs which hurt most, reachability and overtaken facts, are not really habit problems. They are consequences of the manuscript and the notes being in different applications.
When the whole book lives in one home, the plan is not a second document you visit. Story Hub holds the world-building (World, Characters, Places, Glossary, Chapter Notes, Free Notes) beside the manuscript rather than in another window, which is property one. Its job is precisely the one part two described: keeping a decision made in chapter three true in chapter nineteen.
Property three is the one that needs something reading the draft, and that is what AI Sync does: it pulls world-building details out of your scenes and into Story Hub, so a fact the draft invented has a route into the plan that does not depend on you having noticed it at the time. That is the overtaken category, the one no discipline catches, being handled structurally instead.
And when you would rather ask than search, Robin, your co-author, works from the entire book instead of the paragraph in front of you, so where did I establish her age? is a question with an answer, drawn from your own pages. You remain the one who decides what is true. What changes is that verifying a fact stops costing an afternoon of rereading.
Not that you plan less. That the plan stops being a separate chore with its own decay curve, so the reference keeps its authority, and the confidence you write forward with stays earned rather than borrowed.
Two honest limits
Neither of these should be glossed over, because the wrong expectation is its own kind of cost.
Nothing removes judgement from continuity. Something can surface that a scene put a second window in the room. Whether the room should have two windows, whether the earlier scene was better, whether the contradiction is a mistake or the draft telling you something — that is yours, and it should be. A tool that silently reconciled your story to itself would be resolving questions you would want to answer.
Consolidating mid-book is real work. A manuscript already deep in its last revision pass should be left exactly where it sits. Everything described here repays itself over the words you have yet to write, so a book with six weeks remaining will not earn back the disruption of moving it. Wait for the next project, or for a natural break between drafts — and until then run the twenty-minute reconciliation, which asks nothing of your setup at all.
The short version
A story plan goes out of date because writing changes decisions and notes do not update themselves. The bill arrives late, is set by how long you built on the error, and is paid mostly in nerve. The fix is not more discipline. It is to date your entries, mark them by how likely they are to move, let the draft lead and the notes follow, and — where you can — stop keeping the plan and the book in two different places at all.
Then the room has two windows in both documents, and you find out on the day it happens rather than in the copy-edit.
