Sixteen kinds of work. One taste test.
From coding and research to logos, memes, and women's health — the full spread of what people actually ask AI to make. Tap any one to jump to its ranking.
Who wins at what
Could you tell which one has taste?
Same brief, two models, names hidden. Pick the better one — then see who wrote it and how the crowd voted. This is the test, in miniature.
If you only read one thing
The verdicts we'd give a friend. No fence-sitting, no "it depends" — just the model we'd reach for.
The most consistently tasteful across decks, copy, and UI. Polished, on-brief, and the least "AI-looking" of the bunch.
Format-literate and genuinely funny. Knows the reference, reads the room, and keeps the caption tight.
Builds a narrative, not a bullet dump. Real hierarchy, sane pacing, and a cover slide you’d actually present.
Itineraries you’d actually follow — realistic pacing, real places, bookable detail. Not a generic “visit the old town” list.
Marks with an actual idea behind them — not a clipart globe. Holds up shrunk to a favicon or blown up on a wall.
Cleaner diffs and far fewer invented APIs — whether it’s a refactor or turning a one-line vibe into a working UI.
The tab everyone keeps open
We stopped arguing about which model to use in Slack and just pinned the Vibecutter link.
Finally a leaderboard that knows the difference between “correct” and “good.”
I check the movers every Monday. It’s the only AI newsletter I actually open.