How to build a swipe file that gets smarter every time you use it

Ask a hundred marketers whether they keep a swipe file and ninety will say yes. Ask to see it and you will be shown a folder of screenshots, a Notion page last edited in March, or a bookmarks bar with forty tabs called "inspo". The file exists. It does not do anything.

A swipe file that works is not a collection. It is an argument you are having with yourself about what good looks like, and the argument only gets sharper if you keep it going.

Save less, and save for a reason

The graveyard files fail at the front door: everything that looked good went in. A hundred posts you liked is not a swipe file, it is a feed. The rule that fixes it is one question, asked before saving: what would I take from this?

Not "is it good". Not even "would I have written it". Specifically — the opening, the way it turns in line three, the length, the ending — what is the one move you want to be able to make? If you cannot name it, do not save it. If you can, write it down beside the post, in your words, right then. That one sentence is worth more than the post, because it is the only part that is about you.

On this site that sentence is a note on the post — private, and carried into every remix built from the board it sits on. But a notebook works too. What matters is that the reason is recorded where the post is.

Sort by move, not by subject

The instinct is to file by topic — "launch posts", "hiring posts" — and it feels organised. It is the wrong axis. When you sit down to write a launch post, you do not need forty other launch posts; you need the move that will make yours land, and that move might live in a post about something else entirely.

So the useful boards are boards of moves. "Opens with a number." "Turns on line three." "Ends without a call to action." "Under 200 characters." A board like that is a tool: when you know what you are writing about, you open it to decide how. The library's creator pages sort a person's work the same way — by how they open, not what about — for the same reason.

Let the file write back

Here is the part the graveyards never reach. A swipe file is input. If it never produces output — if you never actually write from it — it does not learn anything about you, and you do not learn anything from it.

The fix is to close the loop deliberately. Pick three posts from a board. Write your own post using their moves and none of their words. Then judge the result honestly, and put that judgment back in the file: this worked, this did not, this one I would publish. Do that ten times and the file has taught you something no amount of saving could.

This is what the remix studio is for, mechanically — three sources in, drafts out, a keep or a discard on each, and the kept ones feeding the next round. But you can run the loop with a text editor. The point is that the file is read and written to, every time.

A working file, at three sizes

Ten posts. Enough to write from tomorrow. Pick them for the moves you most want to be able to make, not the ones you most admire. Write the reason beside each.

Fifty posts. Enough to see your taste. You will notice you keep saving the same three shapes; those are your shapes, and the file has just told you something a coach would charge for.

Five hundred posts. Enough that you cannot read it any more — and this is where most files die, because a file you cannot read is a file you stop opening. The ones that survive at this size are the ones organised by move and read through tools rather than by eye: filter to the shape you need, sort by what travelled, remix from the top three. That is the scale this product is built for, and it is a scale a folder of screenshots cannot reach.

Start with ten. Save for a reason. Write from it this week.