Preview unavailable for this view.
Model
Completed-sites bling test
100 finished sites.
Two models. No labels.
Choose the site with stronger visual polish and wow factor. Model identity and completion method appear only after your vote.
- Finished sites
- 100
- Blind matchups
- 50
- Verdict layers
- 1
Loading the completed-sites blind test
Manifest error
The completed-sites test could not be loaded.
Check the reviewed manifest, then reload this page.
Voting room
No completed pairs are available.
The test opens when the reviewed manifest contains its first matchup.
0 of 0 decisions
Votes sync anonymously after each decision.
Which finished site has more bling?
Judge visual polish, coherence, and wow factor. Ignore which model you think made it.
Preview unavailable for this view.
Model
Which result has more visual polish?
Keyboard: 1 for A, 2 for tie, 3 for B
Your vote is locked
You chose candidate A.
Audience totals will appear here when available.
Blinded AI judge
No AI visual judge is included in this human vote.
Human visual preference
A separate leaderboard for finished-site bling.
This vote measures preference across completed artifacts only. It does not change the original reliability score.
Complete every loaded matchup to reveal the human leaderboard.
0 decisions remainingSame brief. Finished artifacts. Hidden identity.
Each matchup compares the same prompt and attempt number. Side order is balanced and randomized for each browser.
All 100 artifacts are closed, browser-loadable HTML. Twenty-nine finished in the frozen one-shot run; 68 used continuations and three used disclosed restarts after blank cutoff outputs.
Judge visual polish and wow factor only. Model identity, completion method, added calls, spend, and browser-check details appear after each vote.
The results are stored under a new experiment ID and cannot alter the original one-shot benchmark.