AI doesn't have "taste" is the line everyone repeats right now. We measured it. 9 off-the-shelf AI judges vs working designers: a coin flip scores 50%, a human hits 74.1%, not one AI judge cleared 55%. The refrain holds up.
We all know AI has a "taste" problem. We measured it. It’s (somewhat) fixable, but AI judges can't replace human creative judgement. Introducing Design Crit. New research from @contralabs_ai and @world_lica, now live on @arxiv
2
2
23
2,154
This came from @world_lica. Their graphic design benchmark was strong but missing the one thing that matters most: human creative judgement. So we built that layer together. Design Crit: 10+ designers, 4 models behind blind code-names, 1,600+ ratings per criterion.
1
2
11
1,146
The gap is real, and it's fillable. A small model trained on Design Crit closes ~46% of it. But the judgement still has to come from people. That's what @contralabs_ai provides: a network of 1.5M professional creatives who design and build for a living. Interested? Let's talk!

Jun 18, 2026 · 7:42 PM UTC

3
139
Sort replies: Relevant Recent Liked