# Nights to Release

> Plan a catch-up on a series before its next entry comes out. List what came before - films, TV seasons, games,
> books, specials - set the release date and your hours a week: the nights until release day appear as numbered doors
> and the list packs into them, free in the browser. A metered read calls each entry play, recap or skip so you finish
> by release day. https://nights-to-release.skillsafe.ai/

## What is free (in the browser)

- Reading the list: one entry per line, `Title (year) | kind | hours | note`, in any order after the title. Hours can
  be `2h`, `1h45m`, `130 min`, `1:45` or a bare number. A line without hours gets a typical length for its kind
  (film 2 h, season 8 h, game 25 h, book 10 h, special 1 h, anything else 2 h) - estimates, and the page says so.
- The nights: from the start date up to (not including) release day, a chosen number of nights per week spread
  through each seven-day stretch, each holding hours-a-week divided by nights-a-week.
- The countdown: every night as a numbered door, the plan packed into them in order; a long entry runs across nights
  as parts, a gap under 10 minutes is left free, and whatever does not fit goes to "after release day".
- The page's verdict, by ordered rules: tell me more (fewer than 2 entries) > all of it fits (the whole list fits the
  nights) > long shot (the nights hold under a quarter of the list) > trim to fit (otherwise). A recap counts as 30
  minutes.
- A flag when the same title appears twice with different years (an original and its remake).
- Call buttons on every entry (play / recap / skip) that repack the nights at once.
- Exports: Markdown, a calendar file (.ics, one all-day event per planned night plus release day) and JSON.

## The metered read (task "plan")

Input: the new one's title, a note about it, release and start dates, hours a week, nights a week, the order wanted
(release, story or listed), the list as JSON, and the page's hours and verdict. Output: the verdict, a title, a
summary, the order and why, every entry once with a call (play, recap or skip), a one-sentence reason and - for a
recap - a spoiler-safe focus; the first entry to start with, the one to keep if time runs out, pacing advice, or
(for a thin list) 2-4 questions. The read works only from the list and its notes: no plot details, no invented facts
about a release, never both versions of a remake. The page re-checks every reply against its own hours.

## Sources

- Built on the SkillSafe skill @mvanhorn/pp-movie-goat (https://skillsafe.ai/skill/@mvanhorn/pp-movie-goat): its
  franchise-marathon planning (watch order, total runtime, suggested breaks) and its remake-aware title handling.
- Model: gpt-terra, metered per run. The samples are made-up series.

## Pages

- https://nights-to-release.skillsafe.ai/ - the app
- https://nights-to-release.skillsafe.ai/api.html - the API, with a worked example per verdict in eight languages