This week I deleted 1,731 lines of code from a writing app I've been building. Every one of them was there to help.
The app spins up a project when you start something new. A novel, a screenplay, a short story. Until Wednesday, starting a novel handed you a full professional scaffold on arrival.
A premise document with headings for logline, concept, theme, and tone. A three-act structure file with page targets on every beat, opening image through final image, the whole spine laid out before you'd written a sentence. Character sheets. A world bible. Numbered folders: 01_development, 02_outline, 03_draft.
It looked like a serious tool. That was the problem.
An empty template is not a head start
The pitch for scaffolding is that it saves you the setup. You don't have to think about where things go, so you can get straight to the work.
What actually happens is that you open a file called premise.md and it contains four headings and no premise. Under "Theme" there's a bracketed prompt asking what your story is really about. You don't know. You've had this idea for two days.
A blank page is honest. It doesn't know anything about your project either, and it doesn't pretend to. A blank template is a different object. It asserts the shape of the thing before anyone has any idea what the shape is, and every unfilled heading reads as a small accusation. You haven't decided your theme yet. You don't have a midpoint. There is no B story.
None of that is information. It's a list of debts you took on by clicking "new project."
Then it quietly becomes the plan
The heavier cost shows up later, and it's the one I didn't see coming.
Structure that arrives before the material doesn't sit there neutrally waiting to be useful. It gets filled in. That's what files do to people. If there's a three-act file in the folder, sooner or later you write into it, and now your story has three acts because a template had three headings.
You stop asking what this thing wants to be and start asking what goes in box seven. Those are not the same question and the second one is much easier, which is exactly why it wins.
The same week, I made a second change for the same reason. Creating a project used to ask for genre and logline up front, before a word existed. Now it asks for a working title and nothing else.
Genre and logline are conclusions. Demanding them at minute zero doesn't get you clarity, it gets you a guess, and a guess typed into a form field is sticky in a way a guess in your head never is. You wrote it down, so now you're defending it. The release title, the genre, the one-line pitch: those are identity the work earns as it takes shape.
What's left
One file. A notes document called ideas.
Everything else gets created when there's a reason for it. A character file appears when there is a character. The outline folder shows up the first time something needs outlining. The structure grows out of the material instead of waiting for it, and when a folder exists, it exists because the work put it there.
The technical version of this was almost embarrassingly easy. Nothing downstream needed the scaffolded files. Every path was created on demand anyway, the empty-state screen already existed, and the only thing making a completely freeform workspace feel rigid was the furniture I'd shipped into it. Forty lines in, seventeen hundred out.
This is not an argument against structure
I want to be careful here, because "structure is bad, follow the muse" is a worse idea than anything I just deleted.
Structure is enormously useful. A beat sheet is a great diagnostic. Run a draft you've already written against one and it will tell you things about your own story you genuinely could not see, like the fact that nothing happens between pages 30 and 55, or that your midpoint is a conversation.
That's structure doing its actual job: describing work that exists so you can find what's wrong with it. The failure mode is using the same artifact as a starting condition, where it does the opposite. It replaces understanding with a checklist and lets you feel organized before you've earned it.
Same document. Fatal at the beginning, invaluable in the middle.
And this generalizes past writing apps in a way I keep bumping into. The repo scaffolded with directories for tests nobody has written. The note-taking system with a tag taxonomy built before there were notes to tag. The essay outline you assembled before you did the reporting, which is now the reason you'll ignore the better piece your research actually found.
All of them are the same move. Build the container, then go find something shaped like it.
The takeaway
Here's the test I'm using now. Look at whatever system, template, or folder tree you're about to set up and ask what it is a record of.
If it records decisions you've made, keep it. That's structure earned, and it will hold weight.
If it's a list of decisions you haven't made yet, dressed up as organization, delete it. It isn't scaffolding. Scaffolding holds up a building that's going up. This is furniture in an empty lot, and it will make you build the house around the couch.
Start with one file. Let the rest show up when it's needed.