# Treat It Like

> A playful collision-zone thinking toy for adults, at https://treat-it-like.skillsafe.ai/. Name one everyday thing that keeps going wrong (the shared fridge, the laundry pile, the sock drawer), list up to five snags, and treat it like a far-off system: one of twelve in the page's deck (air-traffic control, a library, a beehive, a post office sorting room, a railway timetable, a lighthouse, an ant colony, a lost property office, a museum, a fire station, a hotel front desk, a ship's crew) or your own with three to six parts. The page pairs each snag with a part doing the same job and draws every collision as two bumper cars in a rink. One metered run writes what carries over, a first small try and where the comparison breaks.

URL: https://treat-it-like.skillsafe.ai/
API tutorial: https://treat-it-like.skillsafe.ai/api.html

## What it does

- Inputs: the everyday thing, up to five snags (each with a job you pick or the page guesses), a system from the deck or your own (a name plus parts, each with a job and anything it needs first), and an optional list of what is already on hand.
- Jobs: sort (things get mixed up), find (things cannot be found), wait (things queue or clash), remember (things get forgotten), share (nobody knows whose or who), fit (not enough room), reset (things never go back), check (mistakes slip through). The guess counts keyword hits; ties go to the earlier job in that list.
- Free, in the browser: the deck and its parts, the job guesses, the aim (your own aims first; then the first free part doing the snag's job; then the first free part, where the jobs differ), a forecast of how many snags meet a part doing their job, the deck systems that have a part for every job your snags need, the bumper-car rink as SVG or PNG, Markdown, CSV and JSON downloads, and saved collisions with the systems each thing has already met.
- Metered (one run on gpt-terra): per snag, the idea that carries over, a first small try and where it breaks, with a verdict (first match wins: jobs differ = fizzles; the part needs something not on hand = sparks; otherwise sticks), the overall contact (every collision fizzles = clean miss; none fizzles and at least one sticks = solid hit; otherwise glancing blow), the best bump (the first that sticks, else the first that sparks) and what the collision taught. The page re-checks every verdict against its own pairing.

## Contract

POST https://api.skillsafe.ai/v1/app-api/run with string fields: task = "collide", thing, system, snags (one "s1: text" line per snag), parts (one "p1: name" line per part), on_hand, facts (every part with its job and need, every snag with its job, the part it meets and whether the jobs match, and the deck systems that cover every job). The reply is one JSON object: {task, headline, collisions: [{snag, part, same_job, needs_met, verdict, idea, first_try, breaks}], contact, best, insight}.

## Notes

Playful suggestions, not professional advice. Built on the SkillSafe skill @obra/collision-zone-thinking (https://skillsafe.ai/skill/@obra/collision-zone-thinking, MIT): force unrelated concepts together to discover emergent properties, "What if we treated X like Y?"