# Swapped or Changed?

> A playful line-up for adults about a familiar thing that seems different: name the thing, say what feels off and tick a few yes/no clues. Four suspects step up to a height chart - it changed, it was swapped, you changed, or memory drift - each as tall as its clue points. The page picks one and names the cheapest test that would settle it.

URL: https://swapped-or-changed.skillsafe.ai/
API tutorial: https://swapped-or-changed.skillsafe.ai/api.html

## What it does

- You name one familiar thing (a cafe's flat white, a spare door key, an old armchair, a radio jingle), say what feels different, and add optional notes.
- You tick any of twelve yes/no clues: it looks different too, a notice of a change, a different source or batch, only one go was off, off every time now, someone else noticed, only you think so, compared side by side, an old photo or record, a long gap, your routine or taste changed, a special first time.
- Free, in the browser: fixed clue points for four suspects, a police identity line-up drawn as SVG/PNG, the page's pick with fixed rules (fewer than two clues, or a leader on one point or fewer: no pick yet; otherwise the tallest suspect, ties to the one further left), the next suspect, the critical unknown and the one cheap test that separates the top two, and Markdown, CSV and JSON downloads.
- Metered (one run on gpt-terra): the evidence for and against each suspect from your own clues, why the page picked what it did, and the unknown and the test worded for your thing. The run never changes the pick or the points.

## Contract

POST https://api.skillsafe.ai/v1/app-api/run with string fields: task = "line-up", thing, what_changed, notes, clues (the ticked clues, each with the suspects it points to and away from), facts (points, pick, leader, next suspect, critical unknown, test and what each result would point to). The reply is one JSON object: {task, headline, suspects: [{id, for, against}], verdict, explanation, unknown, probe, closing}.

## Notes

For things only, never people or pets. Not advice. If something you eat, drink or plug in seems off, stop using it rather than testing it. Built on the SkillSafe skill @yeachan-heo/trace (https://skillsafe.ai/skill/@yeachan-heo/trace), evidence-driven tracing of competing explanations.