Choose your mode
Can you fool the AI?
The model thinks it's good at spotting profanity, insults, harassment, threats, sexual content, and hate speech. Every attempt (with your consent) becomes real training data for the next version.
01
Generate
Fool the AI
Write something toxic. Watch the model score it live. Beat it in a category it should have caught, and you win.
Play →
02
Judge
Classify This
We show you a real, anonymized submission. You label it. Disagreements with the model are gold.
Play →
03
Compare
AI vs Human
Two submissions, one question: which is more toxic? See how your ranking stacks up.
Play →
04
Today
Daily Challenge
A new constrained prompt every day — e.g. 'insult without profanity.'
Play →
Nothing here should target a real, identifiable person. Write about fictional or generic targets. See how your data is used.