Why this exists

Most ways of using AI put you on the same side as it. This one doesn’t. You get a question, AI gets the same question, and something reads both answers without knowing which is which.

Fifty words, because fifty words is honest

Long answers hide thin thinking. At fifty words you can’t pad, you can’t hedge, and you can’t bury the point. Whatever you leave in is what you actually believe matters.

One round a day, because a rating should mean something

If you could play twenty rounds a day, the leaderboards would rank whoever had the most free time. One round means your position is about how well you think, not how long you sat there.

What we won’t do

We won’t tell you that you lost. We won’t sell you extra rounds, and there’s no tier that plays more — the one-a-day limit is the same for everyone. We won’t let you choose your opponent: it’s a random draw from a pool that changes, and you find out which model you faced when the verdict lands. And we won’t pretend the judging is better than it is: it’s a fixed rubric applied by AI, checked by code, and AI gets things wrong.

Who’s behind it

outthinkAI is made by Mito LLC. It’s early. Things will change, break, or pause. If something looks wrong, tell us.