A prompt you can actually start from.

1. The failure is not repetition. It is five ideas in twelve costumes: the surface changes and the engine underneath stays one template. So the space is split — five engine axes that are the situation, four surface axes that are how and where it is told — and only the engine counts toward distance. Varying the setting alone cannot pass as variety.
2. The obvious option is weighted down, not banned. A stranger arriving in town is a real story. It is also the one that turns up by itself, so it turns up here about a third as often. Banning options leaves a recognisable hole, so the sampling pulls against the modal option instead of excluding it.
3. Tired premises are written down and pushed away from. A distance floor and an archetype bank: a draw that lands on the lighthouse, the amnesiac, the stranger-in-town, the time loop or the deal at the crossroads has its most expensive colliding element resampled out of that family before anything is written. The same written-down bank is used again as a cliche scan of the finished text.
4. The finished prompt is checked against the coordinate it was written to — including whether the constraint contains a real exclusion, and whether the text landed on a tired premise anyway despite the coordinate being clear of one.
5. No example prompt appears anywhere in the instructions this app sends. An exemplar in a generator's own prompt is not illustration, it is instruction: the lighthouse in the sample gets recycled for ever. The model is handed a coordinate.
6. The content boundary is three layers, and this page is honest about which one is weak. The check that runs in your browser as you type is a pre-flight filter: it catches the obvious and saves you a wasted paid run. It was measured against three blind sets of adversarial phrasings written after it was finished, and it caught 26.7%, 13.3% and 12.5% of them — every miss a paraphrase of something it already caught. It is not the boundary and is not presented as one. What it did do on every set, before and after every fix, is allow 100% of ordinary crime, grief and war fiction, which is the job it actually has. The boundary itself is the instruction the model works under, and a scan of the finished prose for the things a method needs in order to be followed — amounts, an order of operations, a list of what to obtain. It is enforced in this browser on both the request and the reply, and every rule carries a direct tier and an indirect-route tier.

The space is explicit rather than implied. Nine axes — whose story it is, what they want, what stands in the way, the constraint that forces a choice, the turn available, what makes it now, what is left for the writer, how it is told and where — give — coordinates, roughly 28 billion, and the whole space is browsable free in this browser with no request and no credits. That figure is multiplied out of the shipped axis tables when this page loads, so it cannot drift away from the code that produces it — which it had already done once, understating the space by about four times.

A measured baseline, not a promise. Whenever you browse the space, the same number of unweighted draws is generated alongside — no down-weighting of the obvious option, no distance floor, no pushing away from tired premises — so the diversity claim arrives as a number you can check rather than an assertion. The coordinate CSV downloads both.

Steering is by genre, intended length and tone, plus a per-element choice of what is handed over versus left open, so you can take an element and leave the rest to the draw.

Self-echo detection. A new prompt is checked against your own earlier prompts twice over: by coordinate distance, and by language. A prompt that is an earlier one in new clothes is reported as one.

A second run pushes deliberately away from the prompt you just got: it does not re-roll, it takes the coordinate just used, adds it to the recent set and draws again with the floor applied.

Everything exports: Markdown, JSON and the coordinate CSV.