← back to work case study · pocket dystopia

Pocket Dystopia: a premise generator that hands a stuck writer somewhere to start.

A web tool for fiction writers facing a blank page. It does not try to write a good story. It tries to put something on the page specific enough that you can argue with it.

Client A personal project, made for fiction writers, myself included, facing the blank page.
Sector Creative tools, writing
My role Sole maker: design, build, and testing
Team Built solo. Tested with five fiction writers.
Duration 2024
Methods Constrained-random design, prompt-structure iteration, usability testing (n=5)
Outputs A live, mobile-first web premise generator
Permissions Personal project. Live at pocketdystopia.forhuman.ca.

A blank page does not give you anything to push against.

The hard part of starting is not a shortage of words; it is the lack of a first thing to react to. Most premise generators do not help, because they answer with noise: three random nouns and a genre, as easy to dismiss as they are to read. The writer is left where they started, now with one more tab to close.

A prompt only works if it is specific enough to argue with.

So Pocket Dystopia aimed lower and more usefully: hand the writer a place to start, a seed concrete enough that they can take part of it, throw out another part, and feel the next sentence arrive. The design question was how to make a random thing feel pointed instead of arbitrary.

Roll the whole world, then lock what you love and re-roll the rest.

The generator builds a premise from five slots: a hero, a villain, a squad, a setting, and the obstacles between them. One tap rolls all of them at once. The move that makes it usable is what happens next. You lock the pieces that spark something and re-roll only the rest, so the tool stops being a slot machine and becomes something you steer. A set of moods, from corporate dystopia to political satire, leans the whole roll toward the kind of trouble you feel like writing.

the slots · what one roll fills

Hero

a parkour expert
a barista who tastes lies

Villain

an evil robot,
weak to pigeons

Squad

a translator who
invents half the lines

Setting

a spooky lab,
lights flickering

Obstacles

the elevator only
goes sideways

the dials · moods that bias the roll
corporate dystopia post-apocalyptic AI takeover political satire
Structure A premise is assembled from five slots a writer can lock or re-roll one at a time. Moods bias the whole roll, so the randomness has a direction.
Lock and re-roll Keep the traits that spark, then roll again. The locked pieces stay put while everything else changes.

The early seeds landed flat, so I made them stranger.

I tested it with five writers and rewrote the prompt structure twice, each time chasing the same problem: a generic seed is easy to wave off. A villain who wants power is nothing. A villain who is an evil robot terrified of pigeons is an argument waiting to happen. So the voice got more satirical and the details got oddly specific, because specificity is what a writer can disagree with, and disagreeing is a way of starting.

It does not make the story. It gets you off the blank page.

The tool is live and does one job. You can expand a seed into a full logline or compress it to a single line, drop it straight into a draft, and save or share a build with a code, so a prompt you liked survives a refresh.

5
fiction writers in testing
2
rewrites of the prompt structure
4
moods that bias a roll
In use One tap rolls a full premise. Lock the parts that spark, re-roll the rest, then copy the logline straight into a draft.

What I learned: randomness is only useful when the writer can steer it.

The first instinct with a generator is to grow the word bank, as if enough variety would eventually turn up something good. That was the wrong lever. Giving the writer control moved it from toy to tool: keep the good parts, re-roll the rest. A bigger pile of nouns would not have done that. It only had to get someone moving, then get out of the way.