A Run You Can Read at a Glance

Two screenshots from a phone, and the same complaint in both: looking at this tells me nothing.

The first was the history. Every row read "Ad-hoc run" — the same three words for a serenade, a research brief and a birthday video, under three different dates. The second was a finished run: the serenade video sat at the same level as the cover image, a second cover, the song and the lyrics, in the order the machine happened to produce them.

The history says what it was

A row now leads with the name of the template that was run, resolved in your language, and beneath it one short line taken from what you typed — the name you were making it for, the title, the subject. "Liliya", not "acoustic".

Choosing that line is fussier than it sounds. Most of what a form collects is knobs: a tone, a resolution, a language code, a duration. So the picker prefers the fields that are about a person or a subject, and falls back to the first value that reads like one — something with a space in it, or long enough to be a phrase. Option keys, codes, bare numbers, links and whole paragraphs never qualify. When nothing qualifies and the run had no template, the row still says "Ad-hoc run", which is now true rather than universal.

The run leads with the deliverable

If a run produced a video, the video is what you see first, large. Everything else it made — the report, the sources, the other pictures, the lyrics, the leftover text — folds under a single "Also created" row with a count, one tap away. If there was no video, the report leads with its sources and the media and text fold. If there was no report either, the media stay where they were and only the text folds.

Nothing was removed. The steps, the raw inputs and the technical details keep their own disclosure, where anyone who wants to read them can. The difference is only what the screen assumes you came for.

The same ladder is on the phone and on the web, because a run opened in two places should not be two different screens.