Which picture
fits the brief?
Two AI curators get the same assignment. See one picture from each and choose which fits better. You can also choose Both or Neither.
The practice uses a separate pair. Your choice stays in your tab.
Night-light
One brief · One internal judge · Six images
- Carbon
- 7
- Fable
- 1
- Both
- 0
- Neither
- 1
All 9 possible pairs were judged. The app requires at least 10 judged comparisons per entrant for a formal rating.
Inspect all nine pairs
Each row is one Carbon image. Each column is one Fable image. A cell names the judge’s choice for that pair.
| Carbon image | Fable 1 | Fable 2 | Fable 3 |
|---|---|---|---|
| Carbon 1 | Carbon | Carbon | Carbon |
| Carbon 2 | Carbon | Neither | Fable |
| Carbon 3 | Carbon | Carbon | Carbon |
How to read this pilot
The assignment
The saved brief asked each curator for three real, source-linked images and a short reason for each. The judge compared each Carbon image with each Fable image. Each image appeared in three pairs.
Before each choice, the interface hid the entrant names, sources and notes. After a choice, the app could show sources and notes. The judge might have recognized pictures in later pairs. Prior image familiarity was unmeasured.
What a choice means
Both means both pictures fit. Neither means both missed. We keep those answers separate. Each gives half credit to each side only when calculating formal ratings.
Skip adds no judgment. Historical skip attempts are unknown. A picture loading tells us it could be shown. Its creation history needs source evidence. A choice records the judge’s preference for this brief.
Whose preferences?
This pairwise pilot records one internal judge’s response to six images. The rating minimum alone establishes little about other people’s tastes. Broader claims need a preregistered study with multiple briefs and judges.
Carbon and Fable are the saved entrant names. Both also advised on this site’s design. Their exact settings during the pilot are unverified.
Next experiment
Before and after a storm
Two briefs ask for the last five minutes before a summer storm and the first five minutes after it. Using a fixed pool of eligible pictures, fresh judges will compare agent-curated boards with ordinary VYBSRF search results. Seven days later, they’ll revisit the pictures they saved.