# Which Door

> A reversibility sorter for the small personal changes people keep putting off - a haircut, a paint
> colour, leaving a club, selling something. List up to eight changes and answer five questions about
> each: how long undoing it would take, what undoing it would cost, whether it touches anyone else,
> whether a smaller try-out exists and whether waiting closes a chance. A free in-browser engine hangs
> every change on one of four doors - revolving (decide today), swing (decide, then check back), stable
> (try the small version first) or trapdoor (slow down) - and draws a door showroom with each change as
> a tag. A metered door read names the way back, a try-out, what waiting costs and one next step for
> every change. Derived from @tjboudreaux/thinking-reversibility.

https://which-door.skillsafe.ai/

## What it does

- The doors, decided in a fixed order: undoing takes months, can't be done, or costs a lot -> stable if a smaller try-out exists, otherwise trapdoor; else undoing takes a few weeks, costs some money, effort or awkwardness, or touches someone else -> swing; otherwise revolving.
- A change where waiting closes a chance is marked: for those a small staged step beats putting it off.
- Free downloads: the showroom drawing (SVG, PNG), the list with its doors (CSV), and the whole list with the read (Markdown, JSON).
- Free check-back reminders: a calendar file (.ics) with one all-day reminder per swing door (in 14 days) and stable door (in 7 days, to judge the try-out).
- Import: a .txt with one change a line, or this page's own List .csv, which brings the five answers back with each change.
- Saved lists (opt-in) keep the answers and the latest read, and show how the doors moved between runs.

## The paid task

One task, `doors`, on gpt-terra, metered per run. The input is three string fields: `task`, `items` (one change a line with its five answers and an optional note) and `facts` (the page's door per line, the counts, and the lines where waiting closes a chance). The reply is one JSON object: headline, and items[] of {n, door (revolving | swing | stable | trapdoor), undo_path, why, try_first, if_wait, next_step, differs}. The read keeps to the page's door unless a change plainly contradicts an answer, and then says why in differs. The page re-checks the reply: one entry per change, a moved door has a reason, a stable door has a try-out, no invented day or time in a next step, and headline counts match the doors.

A thinking aid, not professional advice: the read never says whether a change is a good idea.

- API tutorial: https://which-door.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@tjboudreaux/thinking-reversibility (tjboudreaux/cc-thinking-skills, MIT)