Last week I watched a draft get better and worse at the same time.
The brief was camp. Not camp by implication, camp by name. It asked for a grindhouse heroine played straight, for scenes played broad, for a body count that skewed toward the characters who had it coming. A specific register, written down in plain words, before anything was drafted.
What came back was tasteful.
Every death was mourned. Sixteen of them, grief spread evenly across the book like a coat of primer. Every character sheet carried a personality-type profile. The broad scenes had been adjusted into something more defensible. A word count nobody asked for had been chosen on my behalf.
Nobody rejected the brief. The brief just quietly lost.
That's a familiar failure if you have ever workshopped a genre piece with people who mostly read literary fiction. The difference this time is that the drafting process left a paper trail, so I could go back and find the exact place the register drained out.
It wasn't the writing pass. It was the critique pass.
Every editing pass has a default target
Here is what I think is actually going on, and it applies to people and tools equally.
Revision is a search for improvement, and improvement needs a target. When nobody names the target, every editing pass reaches for the same one: the general-purpose idea of good writing. Earned emotion. Restraint. Every element paying off. Depth on the page.
None of that is wrong. It is the correct advice most of the time, which is exactly what makes it dangerous. It arrives with the authority of craft consensus, and it overrules specific intent without ever announcing that it is doing so.
Camp does not survive earned emotion applied to sixteen deaths. Camp needs grief quarantined into a beat or two so the rest of the book can be fun. When a critique asks "is this death earned," it is asking a genre-blind question, and the honest answer for a slasher is "no, and that's the point."
I hit the same wall from a completely different direction this month, writing a style guide for slasher prose. Working through nine novels and a stack of films, the pattern that kept surfacing was not a technique. It was a list of standard craft advice that actively damages the genre.
Enter the scene late, leave early. Correct almost everywhere, wrong before violence. A kill scene that starts late has no established geography, so the reader gets a blur instead of a chase.
Show, don't tell. Applied to injury, this produces an inventory of damage, and a list is a static object. Tension drops.
Every scene must advance the plot. The genre needs charge-only scenes that advance nothing and raise dread.
Cut the exposition. The rules are the suspense engine. A chapter of rules delivered through conflict is load-bearing.
Four rules that would each survive a workshop unchallenged, and each one makes a slasher worse.
Structure is a shared vocabulary, not a shared spine
While I was in there I parsed a large corpus of commercial plot templates: 148 templates across 37 genres, 3,339 chapters, every structural beat tagged and counted.
I expected a universal skeleton underneath. There isn't one.
Six beats appear in all 37 genres. Hook, inciting incident, B-story, midpoint, climax, resolution. Fourteen show up in 30 or more. Below that it thins fast. Return with elixir, the beat everyone can recite from the hero's journey, appears in two genres out of thirty-seven.
So what genres share is a vocabulary, not a sequence. Two writers can agree completely on what "midpoint" means and disagree completely about whether their book needs one.
That distinction matters at revision time more than at outline time. A vocabulary lets you describe what you did. A checklist tells you what you failed to do, and a pass that treats one as the other will import a genre's assumptions into a book that never signed up for them.
The fix is boring
What we ended up shipping is unglamorous, and I would use all three of these on paper.
First, a fidelity check that runs before any other critique. It lists every register element the brief names, then verifies each one survived by name, and quotes brief against draft for anything that got softened. Not "does this feel consistent." Do the words the writer chose still exist in the work.
Second, an ordering rule. The writer's register outranks the critique's taste. When those two conflict, taste loses. And when the conflict is real and unresolvable, "funny" against "every death is mourned," it gets surfaced as a decision for the writer instead of being silently settled by whoever edits last.
Third, and this is the one worth stealing for any process: annotations for deliberate refusals. When the outline skips a beat the genre expects, it records why in one line, right there in the document. Deliberate: no reconciliation scene, the isolation is the ending.
That last one solves a problem I have had for years without naming it. Revision cannot tell the difference between a mistake and a choice.
A flat character might be a failure, or might be furniture doing its job. A planted object that never pays off might be sloppiness, or might be texture. We now cap that on purpose: for roughly every three objects planted, one is allowed to stay furniture forever. Otherwise every gun on every wall fires and the book reads as contrivance.
Without a written record of intent, every pass re-litigates decisions you already made. And since the default gravity always points toward respectability, you lose a little ground each time. Ten passes later you have a competent book with no pulse, and no single edit you can point to as the moment it died.
The takeaway
If you are working on anything with a strong voice, a specific register, a deliberate ugliness, write down what is intentional before you revise. Not in your head. On the page, next to the thing.
Then let that record outrank your own good taste on the day.
The version of you that made the choice had the whole book in view. The version of you doing line edits at eleven at night has one paragraph in view, and that version will make everything sound nicer.
Nicer is not the same as better.