About
What we set out to do
In November 2026 we took 8 of the biggest questions people ask and put each one, in exactly the same words, to 8 of the world’s AI houses. We did not ask for essays. We asked each house for one work of art per question, in the medium it actually has: 4 of them make an eight-second film, 2 make a single image, 2 write an eight-line poem.
A house is a company and its models, working as a team. In every house, one model thought: it read the question and wrote what it wanted to make. Where the house could paint or film, a second model from the same house made the piece from that description, word for word. The two that can only write made poems. No house touched another house’s work, and we touched none of it: no second tries, no edits, no favourites. Where a house broke the format we asked for, we kept what it sent and said so.
Then we hung the results side by side. The point is the comparison: the same words in, and what came back is not the same. The site makes no AI calls while you browse; everything here was made once and hangs as it was made. The exact words every house received are on the Prompts page. To meet an artist, open any of its works.
The eight houses

Google DeepMind
Gemini 4 Argon thinks · Veo 3.1 Fast films
United States
closed weights

xAI
Grok 4.7 thinks · Grok Imagine Video 1.5 films
United States
closed weights

Alibaba
Qwen3.8 Max thinks · Wan 3.0 films
China
closed weights

ByteDance
Seed 2.1 Turbo thinks · Seedance 2.5 films
China
closed weights

OpenAI
GPT-6 Astra thinks · GPT-5.4 Image 2 paints
United States
closed weights

Meta
Muse Wonder thinks · Muse Image paints
United States
closed weights

Anthropic
Claude Opus 5.5 thinks and writes
United States
closed weights

Moonshot AI
Kimi K3.1 Pro thinks and writes
China
open weights