AI experiments / 6 published comparisons

Compare what models actually produce.

Identical prompts, preserved first responses, and model names hidden until after the vote. Inspect the real outputs side by side and decide what worked.

01Precomputed

Published artifacts, not live API calls.

02One shot

The first response is the final response.

03Blind vote

Choose the output before seeing the model.

Published experiments

Showdown archive

Practical builds, visual tests, and small questions that expose how models interpret the same request.

RUN 01

Build

Snake Game

Six models get one shot at a playable browser Snake game.
6 models
RUN 02

Build

Fake Startup Landing Page

A fictional AI dog-walking startup landing page, generated in one shot.
6 models
RUN 03

Recreate

iOS Calculator Clone

Six models try to recreate the Apple calculator as a single HTML file.
6 models
RUN 04

Gotcha

Pelican On A Bicycle

A visual gotcha: can the model draw a recognizable pelican riding a bicycle?
6 models
RUN 05

Gotcha

Count The R's

A tiny deterministic gotcha with an embarrassingly clear answer.
6 models
RUN 06

Creative

Roast My Code

A personality test that still has to spot the real async bug.
6 models