# Knock-On

> A what-if consequence tracer drawn as a pond. Write one speculative change, pick the parts of life to follow and how fast it arrives, and see its knock-on effects spread in three rings - direct effects, their knock-ons, and the knock-ons of those. Then pick any knock-on and read found documents from inside that world, each written by someone who cannot see the whole picture. Speculative fiction for writers, game masters and the curious - not forecasts.

URL: https://knock-on.skillsafe.ai/
API tutorial: https://knock-on.skillsafe.ai/api.html

## Free, in the browser
- The premise check: empty, very short or vague premises (words like "everything", "society", "technology" with little else) are flagged; an invented world without a setting is flagged; fewer than three domains blocks a run.
- Eight domains to trace: daily life, work and money, law and power, home and family, language and culture, belief and ritual, tools and tech, land and nature. Three paces: overnight, over years, over generations.
- The pond: the change as a stone, three rings, one sector per picked domain. Knock-ons you already have in mind (up to six, one per line) sit on the inner ring in the domain the page guesses from their wording until a trace places them.
- Exports: Markdown, a CSV of knock-ons (id, ring, domain, parent, seed, tone, text, who), JSON, print; kept worlds (in the browser, or in your account when signed in).

## Metered (gpt-terra), two tasks
- ripple: the premise -> a verdict by ordered rules and the knock-ons. Verdicts, first match wins: sharpen (no concrete change - only a topic or mood; no effects, 2-3 decisions to make first), fizzles (concrete but stays inside one or two domains; 2-5 effects, why it stays small, and a bigger version to try), spreads (9-14 effects over three rings across at least three domains, 1-3 tensions between effects, 2-3 story hooks). Every effect names its ring, domain, the effect it grew from, who gains or loses, and a tone (gain, loss, mixed).
- relic: one knock-on -> one to three found documents (official notice, letter, diary entry, classified ad, product label, overheard remark), each with a placard: who wrote it (a role, never a name), what they cannot see, and what the reader learns between the lines; plus a through-line.
The model may imagine effects but not the starting world: no proper names, dates, weekdays or seasons the input does not give, no numbers the input does not contain, no real-world facts presented as true, gender-neutral role words. The page checks every reply: each outer effect follows from one ring in, effects sit in picked domains, your own knock-ons are each placed once, the verdict follows its rules, every requested document came back with a blind spot, and no stray numbers or internal labels appear in prose.

## Sources
Built on the SkillSafe skills @jwynia/systemic-worldbuilding - https://skillsafe.ai/skill/@jwynia/systemic-worldbuilding - and @jwynia/oblique-worldbuilding - https://skillsafe.ai/skill/@jwynia/oblique-worldbuilding
The kicker's news item (timekeepers voting to retire the leap second, October 2026) is as listed on Wikipedia's current-events portal - https://en.wikipedia.org/wiki/Portal:Current_events